Document [original]

An Integrated Approach to the Conceptual Data Modeling of an

Entire Highway Agency

Geographic Information System ( GIS )

Vorgelegt von

M. Sc. Eng. Hande Demirel

von der Fakultät VI –

Bauingenieurwesen und Angewandte Geowissenschaften

der Technische Universität Berlin

zur Erlangung des akademischen Grades

Doktor der Ingenieurwissenschaften

- Dr. Ing. -

genehmigte Dissertation

Promotionsausschuss:

Vorsitzender: Prof. Dr.-Ing. Dieter Lelgemann

Berichter: Prof. Dr.-Ing. Lothar Gründig

Berichter: Prof. Dr.-Ing. M. Orhan Altan

Berichter: Dr.-Ing. Habil. Martin Scheu

Tag der wissenschaftliche Aussprache: 15. Februar 2002

Berlin 2002

D 83

Acknowledgements i

Acknowledgements:

During the preparation of this thesis a large group of people have supported the personal

effort, whom I like to acknowledge their contributions. I would like to express sincere

appreciation to my supervisors Prof. Dr. Lothar Gründig and Prof. Dr. Orhan Altan for giving

me several excellent ideas as well as carefully examining the many manuscripts of this thesis

and overall being distinguishing “doctor father”s. I wish to thank Dr.-Ing. habil. Martin Scheu

for carrying out the official examination of this thesis.

I wish to acknowledge the intellectual contributions made by faculty colleagues, especially

Dr. Frank Gielsdorf. I appreciate his enthusiasm and interest about spatial information theory

as well as for his brilliant suggestions and discussions. A special thanks to Erik Moncrieff for

his great assistance during checking and correcting the written English of this thesis.

I would like to thank for their cooperative attitude among the various components of the

highway community-researchers, government executives, engineers and others with

commitments and enthusiasm about spatial information systems.

Last, but surely not the least, thanks and appreciation for my parents, Ayşe and İsmail

Demirel, for their support and encouragement.

Berlin, 2002

Hande Demirel

Zusammenfassung ii

Titel der Dissertation:

Integrierter Ansatz zur konzeptionellen Datenmodellierung eines geographischen

Informationssystems (GIS) für Daten der Straßenverwaltung

ZUSAMMENFASSUNG:

Um ihre Aufgaben zu verwirklichen, sind Straßenverwaltungen in aller Welt dazu

aufgefordert, neue Technologien einzuführen. Grund dafür ist die große Menge an

anfallenden Informationen des Straßennetzes und die Notwendigkeit Datenquellen effizient zu

nutzen. Geoinformationssysteme für das Transportwesen (GIS-T), welche speziell für

Straßenverwaltungen entwickelt wurden, bewirken eine erhebliche Effizienzsteigerung, da sie

am besten in der Lage sind, dem räumlichen Charakter der Daten Rechnung zu tragen. Häufig

wurde dieser räumliche Charakter der Informationen bei der Systementwicklung ungenügend

beachtet, was dazu führte, dass die Möglichkeiten solcher Systeme nicht voll ausgeschöpft

wurden. Die Implementierung eines Systems kann nur dann zu vollem Erfolg führen, wenn

eine detaillierte Informationsstrukturanalyse durchgeführt wird und wenn die

Datenmodellierung formalisierten Entwurfsmethoden folgt. Im Verlauf der Untersuchungen

wurde festgestellt, dass gebräuchliche Systeme verschiedene Anforderungen von

Straßenverwaltungen nicht erfüllen. Die Probleme können wie folgt zusammengefasst

werden: Die Beziehungen zwischen geometrischen, topologischen und Sachinformationen

wurden nicht strukturiert. Die Abbildung von geometrischen Informationen in

unterschiedlichen Referenzsystemen war nicht redundanzfrei möglich. Die Verwaltung

topologischer Informationen in unterschiedlichen Abstraktionsebenen wurde nicht realisiert.

Spezifische Funktionen der Straßenverwaltung wurden nicht in ihrer Gesamtheit abgebildet.

Nicht alle existierenden Informationen und Methoden konnten in die Systeme integriert

werden.

Es ist erforderlich, Metadaten wie Konsistenzbedingungen, Qualitätsangaben und

Historisierung im System zu berücksichtigen. Speziell für die Definition von

systemübergreifend eindeutigen Objektidentifikatoren sind neue Ansätze erforderlich.

Um die Effizienz von GIS-T zu verbessern und die beschriebenen Anforderungen zu erfüllen,

wird in der vorliegenden Arbeit schrittweise ein Ansatz für eine konzeptionelle

Datenmodellierung vorgestellt, welche den Bedürfnissen einer Straßenverwaltung Rechnung

trägt. Der Grundgedanke des vorgeschlagenen Modells besteht in der Abstraktion und der

strengen Unterscheidung von geometrischen, topologischen und Sachdaten. Um die

Integration aller Daten, die Kontrolle von Redundanz und eine Optimierung der Datenpflege

zu erreichen, wurden Trassierungselemente durch datumsinvariante Parameter abgebildet.

Das vorgeschlagene konzeptionelle Datenmodell wurde erfolgreich implementiert. Dabei kam

ein objektrelationales Datenbanksystem zum Einsatz.

Abstract iii

Title of Dissertation:

An Integrated Approach to the Conceptual Data Modeling of an Entire Highway

Agency Geographic Information System ( GIS )

ABSTRACT:

World-wide highway administrations are stressed to implement new technologies, due to the

large amount of information associated with highway networks and the necessity of using

sources efficiently in order to realize their tasks. Geographic Information Systems-

Transportation (GIS-T), which are specifically tailored for highway administrations, are

identified having the highest information technology payoff potential by the highway

administrations due to road information spatial character. Contrarily, road information spatial

character is not adequately considered during system design, as a result, many of the benefits

of GIS-T are not fully realized and efficiency of this technology is mainly under estimated.

The relative success of implemented system is not clear without a detailed information

analysis and a data model, which rely on formal data model design methodologies. It was

determined during this study that several demands of highway administrations were not

responded by means of current systems. These topics can be summarized as follows; firstly,

relationships among geometry, topology and thematic information were not structured. The

geometry information can not be mapped in various reference systems without redundancy.

Thirdly, the non-planar multi- abstraction topological information was not exist. The entire

highway administrations business rules can not be performed in the current systems. The

existing information and methods were not integrated into the system.

The metadata including consistency rules, quality specifications and history information

needed to be incorporated into the system. Especially in order to determine permanent, non-

spatial and a unique object identifier, regulations and new approaches are required.

In order to increase the efficiency of GIS-T and fulfill these requirements, this study

considered a progressive approach appropriate to the conceptual data modeling requirements

of an entire highway agency. The main approach of the proposed data model was abstraction

and decomposition of geometry, topology and non-spatial data. In order to achieve data

integration, control of redundancy and optimization of data maintenance, linear elements were

mapped by means of datum invariant parameters. The proposed conceptual data model was

successfully implemented using the integrated approach in one object-relational system and

results were discussed.

Contents iv

CONTENTS

Chapter 1 : Introduction ...................................................................................................... 1

1.1 Introduction................................................................................................................1

1.2 Overview of Problems................................................................................................ 1

1.3. Aim of the Study ........................................................................................................ 3

1.4. Contents of the Study................................................................................................. 4

Chapter 2 : Concepts ............................................................................................................ 5

2.1 General Approaches................................................................................................... 5

2.2 Information Structure Analysis..................................................................................6

2.2.1 Spatial Information – Geometry and Topology ................................................. 7

2.2.2 Non-spatial Information..................................................................................... 8

2.3 Conceptual Data Modeling......................................................................................... 9

2.3.1 Methods of Conceptual Data Modeling ........................................................... 10

2.3.1.1 Entity- Relationship Approach ............................................................................... 11

2.3.1.2 Object-Oriented Approach...................................................................................... 11

2.4 Logical Data Modeling............................................................................................. 12

2.4.1 Relational Data Model .....................................................................................13

2.4.2 Object Oriented Data Model ............................................................................ 15

2.4.3 Comparison of Relational and Object-Oriented Approaches........................... 16

2.4.4 Object-Relational Data Model ......................................................................... 17

Chapter 3 : Highway Administration and GIS................................................................ 19

3.1 Overview.................................................................................................................. 19

3.2 Analyzing Highway Administrations....................................................................... 20

3.2.1 Organization Structure ..................................................................................... 20

3.2.2 Assessing the Requirements of Highway Administrations.............................. 22

3.2.3 Data Acquisition............................................................................................... 26

3.2.4 System Architecture......................................................................................... 28

3.3 Current Status of GIS-T at Highway Administrations............................................. 29

3.3.1 Current Status of GIS Technology................................................................... 29

3.3.2 Standards Established for GIS-T...................................................................... 32

3.3.3 Data Models Evaluated .................................................................................... 35

3.3.3.1 A Sample Conceptual Data Model Applied in the United States of America ........ 35

3.3.3.2 The German Highway Administration and A Sample Conceptual Data Model..... 40

3.3.3.3 Danish Highway Administration ............................................................................ 45

3.3.3.4 The Geographic Data Files (GDF).......................................................................... 46

3.4 Overview of Problems.............................................................................................. 48

Chapter 4 : Proposal for an Integrated Conceptual Data Model................................... 50

4.1 Criteria for the GIS-T............................................................................................... 50

4.2 The Resulting Conceptual Data Model.................................................................... 50

4.2.1 Overview of Highway Information Structure .................................................. 51

4.2.2 The Data Schema ............................................................................................. 51

4.2.3 The Conceptual Data Model ............................................................................ 52

Contents v

4.2.3.1 Geometry ................................................................................................................ 52

4.2.3.2 Topology................................................................................................................. 57

4.2.3.3 Road Events............................................................................................................ 62

4.2.4 Completing Conceptual Data Model................................................................ 67

4.2.4.1 Integrity Constraints ............................................................................................... 67

4.2.4.2 History .................................................................................................................... 69

4.2.4.3 Quality Aspect ........................................................................................................ 71

4.3 Special Aspects of Highway Information System.................................................... 72

4.3.1 The Linear Referencing System....................................................................... 72

4.3.2 The Cross - Sectional Information ................................................................... 79

4.3.3 Geometrical Data Integration........................................................................... 80

4.3.4 Unique Identifier (Unique_id).......................................................................... 83

Chapter 5 : Implementation of the Proposed Conceptual Model .................................. 86

5.1 Overview.................................................................................................................. 86

5.1.1 Software Used .................................................................................................. 86

5.1.1.1 Database Management System ............................................................................... 86

5.1.1.2 GIS Software........................................................................................................... 88

5.1.2 Sample Data ..................................................................................................... 89

5.2 The Realized Physical Storage.................................................................................89

5.3 Implementation......................................................................................................... 92

Chapter 6 : Conclusion..................................................................................................... 101

6.1 Overview................................................................................................................ 101

6.2 Perspective .............................................................................................................103

Chapter 7 : Literature ...................................................................................................... 104

Annex Index:......................................................................................................................... 110

Annex A: The Organization Chart of General Directorate of Highway, Turkey ........I

Annex B: Turkish Road Administration Departments Requirements........................II

Annex C: The Questionnaire........................................................................................... V

Annex D: Completing the Psychical Data Model ......................................................VIII

Curriculum Vitae:...............................................................................................................XIII

Figure and Table Index vi

FIGURE INDEX:

FIGURE 1-1: BENEFIT – COST EXPECTATIONS............................................................................................ 2

FIGURE 2-1: DATA MODELING STEPS ........................................................................................................ 6

FIGURE 2-2: GRAPH DESCRIPTION ............................................................................................................. 7

FIGURE 2-3: SPATIAL COMPONENTS OF A ROAD NETWORK....................................................................... 8

FIGURE 2-4: THEMATIC ROAD INFORMATION ............................................................................................ 8

FIGURE 2-5: THREE-SCHEMA ARCHITECTURE ......................................................................................... 10

FIGURE 2-6: CONCEPTUAL DATA MODEL APPROACHES .......................................................................... 12

FIGURE 2-7: OBJECTIFICATION OF DATABASE MANAGEMENT SYSTEMS(DBMS) [CONNOLLY,1999] . 18

FIGURE 3-1: ROAD AND TRAFFIC INFORMATION[PORTELE,2001]......................................................... 23

FIGURE 3-2: REFERENCE SYSTEMS .......................................................................................................... 23

FIGURE 3-3: (X, Y) PLANAR ORTHOGONAL COORDINATE SYSTEM........................................................... 24

FIGURE 3-4: LINEAR REFERENCED ROAD INFORMATION ......................................................................... 24

FIGURE 3-5: LINEAR REFERENCE SYSTEMS ............................................................................................. 24

FIGURE 3-6: (H, Q) REFERENCE SYSTEM .................................................................................................. 25

FIGURE 3-7: THE ROAD INFORMATION PHASES ....................................................................................... 26

FIGURE 3-8: SAMPLE VIDEO IMAGE FROM DANISH HIGHWAY................................................................. 27

FIGURE 3-9: STATUS OF GIS [SCHILCHER,2000].................................................................................. 30

FIGURE 3-10: COMPONENTS OF DUAL ARCHITECTURE [MOLENAAR,1991]............................................ 31

FIGURE 3-11: THE BASIC ELEMENTS OF THE DUEKER AND BUTLER MODEL [DUEKER,1997] ................. 36

FIGURE 3-12: THE COMPLETE CONCEPTUAL DATA MODEL [DUEKER,1997]........................................... 37

FIGURE 3-13: IMPLEMENTATION RECOMMENDATION OF ROAD EVENTS [DUEKER,1997] ....................... 38

FIGURE 3-14: IMPLEMENTATION RECOMMENDATION OF ANCHOR SECTION [DUEKER,1997] ................ 39

FIGURE 3-15: THE NWSIB CONCEPTUAL DATA MODEL SYMBOLS ........................................................... 42

FIGURE 3-16: THE NWSIB CONCEPTUAL DATA MODEL DESCRIPTION [NWSIB,1998] ............................ 43

FIGURE 3-17: THE SYSTEM ARCHITECTURE OF NWSIB [NWSIB,1998].................................................... 44

FIGURE 3-18: THE GDF ATTRIBUTE CONCEPT [WALTER,1997].............................................................. 47

FIGURE 4-1: THE HIGHWAY INFORMATION STRUCTURE .......................................................................... 51

FIGURE 4-2: THE UML DIAGRAM DEFINITIONS....................................................................................... 52

FIGURE 4-3: THE GEOMETRY COMPONENT OF THE CONCEPTUAL DATA MODEL..................................... 53

FIGURE 4-4: RELATIONSHIPS BETWEEN A)BLUE PRINT, B)CURVATURE DIAGRAM AND C)ANGLE

DIAGRAM [MÜLLER,2000] ................................................................................................ 55

FIGURE 4-5: INTEGRATION OF DIFFERENT DATA SOURCES ...................................................................... 57

FIGURE 4-6: GEOMETRIC ELEMENTS IN THE PLANE.................................................................................. 57

FIGURE 4-7: ROAD SEGMENT IN GIS, WITH COMBINED TOPOLOGY AND GEOMETRY INFORMATION ....... 58

FIGURE 4-8: SAMPLE OF A DIVIDED HIGHWAY......................................................................................... 59

FIGURE 4-9: NON-PLANAR TOPOLOGY..................................................................................................... 60

FIGURE 4-10: THE TOPOLOGY COMPONENT............................................................................................... 61

FIGURE 4-11: THE FIRST LEVEL TOPOLOGY AND GEOMETRIC ELEMENTS ................................................. 62

FIGURE 4-12: THE THEMATIC COMPONENT OF THE CONCEPTUAL DATA MODEL....................................... 63

FIGURE 4-13: THE CONCEPTUAL MODEL FOR CROSS - SECTIONAL SPATIAL INFORMATION...................... 65

FIGURE 4-14: THE OVERVIEW OF THE PROPOSED CONCEPTUAL DATA MODEL.......................................... 66

FIGURE 4-15: A TRANSACTION WITH SAMPLE OBJECTS ............................................................................ 68

FIGURE 4-16: ATTRIBUTES OF THE OBJECT LINK I.................................................................................... 70

FIGURE 4-17: THE HISTORY COMPONENT OF THE CONCEPTUAL DATA MODEL ......................................... 70

FIGURE 4-18: THE ROAD INFORMATION AFTER RE-ALIGNMENT................................................................ 72

FIGURE 4-19: THE RELATION BETWEEN COORDINATE SYSTEMS[GIELSDORF,1998].............................. 75

FIGURE 4-20: DX, DY SHOWN FOR A LINEAR ELEMENT CLOTHOID............................................................. 76

FIGURE 4-21: ONE OF THE SAMPLE CONSTRAINTS, OVERPASS-HEIGHT..................................................... 82

FIGURE 4-22: GEODETIC DATUM TRANSFORMATION ................................................................................. 84

FIGURE 5-1: LINEAR QUAD-TREE INDEXING SCHEME .............................................................................. 87

FIGURE 5-2 SAMPLE DATA FROM THE BRANDENBURG HIGHWAY ADMINISTRATION ............................. 89

FIGURE 5-3: A SAMPLE SQL SCRIPT FOR CREATING OBJECTS [PFANNMÖLLER,2001]....................... 93

FIGURE 5-4: EDITING THE NODE USING MAINTAIN COINCIDENCE MODE................................................ 94

FIGURE 5-5: MODIFICATIONS IN THE CONCEPTUAL DATA MODEL .......................................................... 96

FIGURE 5-6: IMPLEMENTED GEOMETRY AND TOPOLOGY COMPONENTS.................................................. 97

FIGURE 5-7: THE IMPLEMENTED ROAD EVENT COMPONENT OF SAMPLE DATA ..................................... 98

FIGURE 5-8: THE VERTICAL LINEAR ELEMENTS SAMPLE ........................................................................ 98

FIGURE 5-9: THE CROSS-SECTIONAL REFERENCE SYSTEM AND ANALYZED ROAD DATA ....................... 99

FIGURE 5-10: AN EXAMPLE OF VALIDATION RULE USED IN TOPOLOGY COMPONENT............................... 99

FIGURE 5-11: THE EXAMPLE OF AVAILABLE TRIGGERS IN SYSTEM......................................................... 100

Figure and Table Index vii

TABLE INDEX:

TABLE 2-1: BASIC ELEMENTS OF THE RELATIONAL DATA MODEL......................................................... 14

TABLE 2-2: THE OBJECT ROAD SEGMENT .............................................................................................. 16

TABLE 2-3: COMPARISON OF RELATIONAL AND OBJECT-ORIENTED APPROACHES................................. 17

TABLE 3-1: STATISTICAL INFORMATION ON THE COUNTRIES STUDIED .................................................. 20

TABLE 3-2: DATA REQUIREMENTS ......................................................................................................... 25

TABLE 3-3: COMPARISON OF GIS RELATED STANDARDS [CASPARY,1998] ........................................ 34

TABLE 3-4: QUALITY PARAMETERS FOR TELEMATIC DATA [WIDMANN,2000]................................... 34

TABLE 3-5: A SAMPLE TABLE FROM INFORMATION SYSTEM ................................................................. 41

TABLE 4-1: THE MATRIX OF ROAD REFERENCE SYSTEMS AND GEOMETRY ........................................... 65

TABLE 5-1: IMPLEMENTATION OF THE FIRST ABSTRACTION LEVEL OF TOPOLOGY ................................ 90

TABLE 5-2: IMPLEMENTATION OF THE POINT GEOMETRY..................................................................... 90

TABLE 5-3: IMPLEMENTATION OF THE REFERENCE SYSTEM AND THE RELATION BETWEEN

POINTGEO / REFSYS........................................................................................................... 91

TABLE 5-4: IMPLEMENTATION OF THE POINT GEOMETRY.................................................................... 92

TABLE 5-5: ROAD EVENT IMPLEMENTATION ......................................................................................... 92

TABLE 5-6: SAMPLE IMPLEMENTATION FOR THE POINT OBJECT............................................................ 96

TABLE 5-7: CONSISTENCY CONDITIONS FOR OBJECT POINT................................................................... 97

TABLE 5-8: TRIGGER TYPES ................................................................................................................. 100

1.1 Introduction 1

Chapter 1 : Introduction

1.1 Introduction

World wide, a very large amount of information associated with the national and international

highway networks exists. Coupled with the necessity of using these sources efficiently,

highway organizations are stressed to research and implement new technologies for realizing

their tasks. The rapid development in technology has also given rise to the improvement of

decision support systems. Consequently, many highway administrations moved to Geographic

Information Systems (GIS) technology to assist their decision-making.

Geographic Information Systems tailored for highway administrations are called Geographic

Information Systems-Transportation (GIS-T). According to a study done in the USA (1992),

almost 80 % of the transportation departments responded that GIS-T technology had the

highest information technology payoff potential of any technology identified [ISTEA,1995].

Although many other solutions for project management and engineering applications exist,

there are two very important reasons why GIS-T is of interest.

1. All highway administration data, with some exceptions such as legal consultance or

accounting division, has a spatial nature.

2. GIS integrates highway administration data and methods, which other systems do not

offer. GIS-T can be used as a logical and physical data integrator of all types of data

necessary to the highway sector.

However, due to its spatial nature, many of the benefits of GIS for highway administrations

such as integrating data and methods, enforcing rules and standards, cost reduction and

quality improvements, are not fully realized. The spatial information in highway

administrations is either not recognized or not fully considered during system design. Because

of such reasons, efficiency of GIS-T technology is mainly under estimated.

1.2 Overview of Problems

Highway administration specific spatial information problems, including multi-dimensional

spatial road data and the relationship between these, integration of methodologies and

topology abstraction levels are the main focus of this study. In addition to these, common

problems of GIS, such as data integration, data inconsistencies, and non-adequate data update

are examined in this context.

Highway administrations are very large governmental organizations established in order to

support transportation requirements at a national and international level. Accordingly they are

charged with many diverse tasks. In order to fulfill these tasks, data from various sources is

acquired by road administrations in different forms and formats. Each department has

different usage requirements. Although the collected data for organization tasks is in very

large amount, full efficiency is lost due to its analogue format. In addition there are many

methodologies, enterprise-rules and terms, which are not generally common, even between

departments. Research conducted in the United States reported that thirty-eight referencing

methods were being used in a single federal state highway administration [NCHRP,1997].

Additionally due to workflow needs information needs to be transferred between many

divisions, departments and organizations.

In order to facilitate these tasks, an information technology infrastructure, especially

databases and GIS, is needed. But in some situations existing databases are not updated with

new information. One of the reasons for this is that organizations can be easily frustrated by

1.2 Overview of Problems 2

the high cost of GIS implementation and data maintenance. GIS generally have long term

profit expectations and high implementation costs. The typical variation in benefit and cost

expectations for GIS with respect to time is illustrated in Figure 1-1. It has been noted that

data costs, consisting of data acquisition, modeling and maintenance, comprise 80% of the

total cost of GIS.

COST

BENEFIT

YEARS

Figure 1-1: Benefit – Cost Expectations

Secondly in its early stages, GIS-T technology was applied to a variety of project-oriented

transportation applications. Consequently problems known as “data islands” are very

common. Available digital data is isolated within divisions and there is no data exchange

between divisions and departments, even when it is desired. With many highway

administrations there are difficulties in gathering and analyzing information. Inaccurate

information and inconsistent data persist in the information systems.

Thirdly, road information has multi-dimensional spatial character with various levels of

abstraction due to diversity in the highway agencies’ requirements. Road objects can be static

or dynamic, referenced to one, two, three or four-dimensional coordinates. The most common

information referencing method used by highway agencies is the linear referencing system.

This is based on a one-dimensional specification of the unknown point in terms of direction

and distance from a known point. Many other spatial frames are in usage such as numbering

systems, addresses, topology, administrative reference systems and road names. In GIS-T

roads are defined using two-dimensional reference systems. In order to integrate these various

dimensions, typically, geographical location by two-dimensional coordinates is used, and

linearly referenced road data is considered as attributive data. However, highway information

spatial character is continually changing through new alignments and construction, therefore

its reference system is also continuously changing. Therefore, linearly referenced data is

badly affected by such geometrical changes, requiring a new referencing for the sections after

the modification. Thus, maintenance of this data is clearly necessary. Unfortunately, because

of its attributive storage in GIS-T the practical realization is often insufficient.

Additionally, since transportation facilities and phenomena exist in three and four dimensions,

the restriction of current data models to two and 2.5 dimensional space limits the ability of

GIS to effectively model the real world [VONDEROHE ,1993]. Due to historical reasons, GIS

concepts and implementations were initially closely associated with the requirements of land

information systems. Therefore, highway administration with its complex analytical network-

based models, enterprise-wide business model and topological requirements could not be well

modeled using standard GIS.

Even when all these issues are considered, without a detailed information analysis and data

model, the relative success of implemented system is not clear. Most organizations, and

1.3 Aim of the Study 3

highway agencies are not exceptions, do not rely on formal data model design methodologies.

This is considered one of the major causes of information systems failure. Lack of a

structured approach to database design often leads to inadequacy or inefficiency in meeting

the demands of organizations.

Some open issues are:

1. In which context do current systems respond to highway administration demands?

2. Is it possible to map all road related information in an integrated manner using GIS-T

for a complete enterprise?

3. How can the above-mentioned problems be solved, particularly with respect to the

spatial nature of road information?

4. Is it possible to solve the defined problems with current technology?

1.3. Aim of the Study

This study aims to provide a progressive integrated approach to the conceptual data modeling

requirements of an entire highway agency, with the intention of addressing the above-

mentioned issues.

In order to examine the current situation, clarify problematic areas and identify the

requirements of highway administrations in a wider perspective through this study four

countries were selected for particular attention namely; Turkey, Denmark, Germany and the

USA.

During this study the current status of GIS-T usage and available information sources are

examined in order to describe the highway agency specialized view of reality. Available

conceptual data models are evaluated to identify problematic areas, especially due to the

spatial nature of information. Highway administration specific aspects such as linear

referenced data integration, cross-sectional design information, spatial data integration and

identification of uniqueness are particularly considered. In addition, implementation

possibilities of a proposed conceptual data model are examined and the problems encountered

are discussed.

The proposed conceptual data model concentrates on special aspects of highway

administrations in detail such as;

 Relationships among topology, geometry and thematic road information

 Multiple topologic abstractions

 Analyzing multi-dimensional spatial road data and realizing transformations between

dimensions.

 Modeling highway administrations business rules

 Integration of existing road information and methods

 Modeling the metadata

In methodological terms, the following procedure is adopted;

1. Defining requirements, analysis of the data flow in highway agencies, examining existing

data sources and relevant standards, regulations and systems.

2. Development of the data model.

3. Drafting the data schema.

1.4 Contents of the Study 4

4. Development of proposals for implementation.

5. Implementation of proposed concepts.

1.4. Contents of the Study

In Chapter 2 the fundamental concepts for analysis of spatial data and a general overview of

data modeling steps were provided. Various data modeling approaches were compared with

specific reference to GIS-T aspects.

In order to clarify problematic areas and the requirements within highway administrations in a

wider perspective, organization structure, needs assessment, data acquisition techniques,

existing data sources, methodologies and system architectures were evaluated. The current

status of GIS-T in highway administrations was discussed, including GIS technology state,

and standards established for GIS-T. Four different conceptual data models were evaluated.

An overview of problematic areas is also provided in Chapter 3.

After identifying problematic areas, requirements and experiencing the perspective of

highway administrations, a conceptual data model was proposed. Special aspects of highway

information systems, such as linear referencing, cross-sectional design information,

geometrical data integration, feature identification and proposals developed within the

conceptual data model, were introduced in Chapter 4.

In Chapter 5, the proposed concepts were implemented for a sample project in order to

recognize gaps and unfulfilled requirements from the conceptual data modeling.

Finally in Chapter 6 the developed concepts were discussed concluding the study.

2.1 General Approaches 5

Chapter 2 : Concepts

2.1 General Approaches

The general approach of information system development can be separated into two main

categories; data-driven and method-driven. With the data-driven approach the entire focus of

the design process is on data and its properties. After identifying user data requirements, a

conceptual data model needs to be designed and to be implemented in a database, then

applications that use the database are developed. With the alternate method-driven approach,

working activities within an enterprise are determined and then application programs are

designed according to user requirements [BATINI ,1992]. Both approaches has advantages

and disadvantages during the design of a complex system, such as GIS-T. With the data-

driven approach, a complete view of the system exists, although there is a possibility not

efficiently to response specific application requirements of single users. With method-driven

approach, due to consideration of single repositories separately, application requirements are

identified in detail. However, there is a risk of not recognizing the data exchange or common

activities, which exist due to not taking a complete view of the system.

In general when designing complex information systems, for example an entire highway

agency information system, a joint data-method driven method is preferable in order to

benefit from both approaches and to close the gaps of each. After feasibility studies have been

carried out and the decision has been taken to implement the system, according to a joint data-

method approach data structure and business data analysis are carried out separately. In order

to construct the optimal conceptual data model, overlapping time intervals should be taken.

This provides a specialized view of reality, and actually helps in the understanding of the

users point of view. During the study this goal was achieved by surveying user requirements

and analyzing existing models as well as existing systems.

Following business and data structure analysis, the conceptual data model should be defined.

It is a formal way of describing an abstraction of real world phenomena for a specialized view

of reality, which reflects decisions about features and their relationships. Without describing

these different specialized views of reality, system success could not be guaranteed. Because

of this, information design is not a one-way processes. Generally, designing a successful

system requires repetitive cycles. Although the conceptual data model is the core of GIS

system design, lack of the data modeling issues result many problems that are faced today in

GIS. Such problems mainly concludes insufficient user requirement responses, unpredicted

data integration problems and finally not efficient usage of the system. Because of above

mentioned reasons this study is mainly considered a progressive approach to the conceptual

data modeling requirements of an entire highway agency in order to accomplish such

problems and increase the efficiency.

After completing the conceptual data model, a logical design should be made which is the

description of the database structure in a formal language. The conceptual data model should

be converted into one of the database structures, which can be hierarchical, network,

relational, object-oriented or object-relational. During this study object-relational logical

design approach is preferred, due its benefits during GIS data modeling that are highlighted in

further sections.

Physical database design expresses the implementation of the developed concepts for a

selected database, mainly with respect to the description of the storage structures and data

access methods. Validation and implementation of the designed system can be conducted in

parallel.

2.2 Information Structure Analysis 6

For data maintenance, further extensions to the system or data integration the above

mentioned steps should be documented in a formal language. An overview of the data

modeling steps can be seen in Figure 2-1.

Figure 2-1: Data Modeling Steps

2.2 Information Structure Analysis

In order to analyze the data structure and to map specialized views of reality as mentioned in

the first step of Figure 2-1, fundamental concepts for the analysis of spatial data should be

clarified. Analyzing the data structure refers to all types of data used by an enterprise, where

the emphasis is on which things are of interest and descriptions of the relationships between

them. A real world phenomenon is specified in three main information categories as spatial

properties, non-spatial properties and role, behavior or method.

The first two components, being, spatial and non-spatial information, will be examined in this

chapter. The third component; role, behavior and method of road information, is studied

during the conceptual data modeling.

2.2 Information Structure Analysis 7

2.2.1 Spatial Information – Geometry and Topology

Spatial information can also be sub-divided into two main components, geometry and

topology, although these components are usually indistinctly handled in GIS. One of the

fundamental concepts for analysis of spatial data is a formal understanding of geometry and

geometrical relationships between objects.

Geometry is described as “ the study of figures in a space of a given number of dimensions

and of a given type ”[WEISSTEIN,1999] or “ the study of invariant properties of given

elements under specified groups of transformations ” [JAMES,1976]. Therefore, geometry

can be also be separated into two categories; datum dependent and datum independent. The

datum dependent indicates “given number of dimensions” in the definition, specifying that

element is referenced on spatial dimensions. The datum independent indicates “the given type

or elements”; that elements are defined using a parameterized approach. Examples of these

categories are;

 Datum dependent: Point (X, Y, Z) coordinates referenced to coordinate

system, in this case geodetic datum dependent.

 Datum independent: Straight line described by its parameter length.

Topology arises from the geometry by generalization, which means in this case by abstraction

of the metric information. This describes an explicit knowledge of the mutual connectivity

such as; order, connectivity and adjacency. Road networks are generally represented as graphs

in which intersections are nodes and roads are links, although some disadvantages arise in the

case of GIS-T, because of the complexity of real world objects which will be examined in

Section 4.2.3.2.

Generally graph, which is represented by a diagram, is defined in mathematics as follows:

[BALAKRISHNAN ,2000]

“A graph is an ordered triple G = (V(G), E(G), IG), where V(G) is a non-empty

set, E(G) is a set disjoint from V(G), and IG is an “incidence” map that associates

with each element of E(G) an unordered pair of elements (identical or distinct) of

V(G). The nodes of G, and elements of E(G) are called the links of G. If, for the

edge e of G, IG(e) = {u, v}, then IG(e) = uv “

Example: Figure 2-2 shows a graph description.

If V(G) = {v1, v2, v3}, E(G) = {e1, e2, e3} and IG is given by

IG(e1) = {v1, v2 }, IG(e2) = {v2, v3 }, IG(e3) = {v2, v3 }, representing the connectivity.

Therefore, (V(G), E(G), IG) is a graph.

Figure 2-2: Graph Description

The topological relationships between the objects are invariant with respect to their position,

orientation, transformation, shape and size. Figure 2-3 illustrates the spatial components of a

2.2 Information Structure Analysis 8

road network which has different geometrical information, but identical topological

information.

x1,y1

x3,y3

x2,y2

x4,y4

x5,y5

x6,y6

TOPOLOGY

GEOMETRY

x1,y1

x3,y3

x2,y2

x4,y4

x5,y5

x6,y6

Figure 2-3: Spatial Components of a Road Network

2.2.2 Non-spatial Information

Non-spatial information, also called thematic information, is defined as descriptive

information of a real world object, which does not contain any spatial character. In the case of

highway administrations the road type, pavement type, accident statistics and capacity

analysis information can be given as examples of thematic information. In highway agencies,

this information is generally documented in analogue semantic sketches, referencing road

information on a linear system. As this information is referenced on a linear system, it has

also spatial components. However, in GIS-T, spatial components of this information is not

considered. In Section 4.3.1, the problems encountered due to not considering this spatial

component, and solution proposals are discussed. An example of linearly referenced thematic

road information is provided in Figure 2-4.

Figure 2-4: Thematic Road Information

2.3 Conceptual Data Modeling 9

2.3 Conceptual Data Modeling

A conceptual data model, is an unique central description of real world phenomena. In the

conceptual data model real world phenomena is designed to combine various subsets of the

reality, which describes relevant information and methods in the user perspective, and

information that may be held in a database. The success of GIS is highly dependent on

information structure analysis and conceptual data modeling. Geographic Information

Systems are composed of hardware, software and data. Of these components, data, including

the data model, has the longest life span and is the most costly. However, the data component

has never had the first priority during GIS establishment, especially compared to the software

component.

The basis of a data oriented approach to database design was presented in the

ANSI/X3/SPARC (1975) report. The report described a three-schema architecture, illustrated

in Figure 2-5. It identified the following three schemas: external, conceptual and internal

[ROLLAND,1992]. Briefly;

 The external schema describes the varying views of real world phenomena according

to users

 The conceptual schema formally describes the conceptual data model

 The internal schema describes the physical storage structure of conceptual schema that

may be required at any given time.

The external schema is constituted after completing data structure analyses and identifying

business methods. For highway administrations the most difficult aspect of database design is

the definition of the external schema, due to the differing abstraction levels and scales of

user’s expectations and requirements. Some highway administration applications can require

road network information for whole countries, while others, such as tunnel design are only

related to a specific road section.

Three approaches exists in order to perform a conceptual data model; being top-down,

bottom-up and integrated. Primitives are objectified in two groups, top-down and bottom-up.

Top-down primitives correspond to pure refinements which apply to a single concept, the

starting schema, and produce a more detailed description of that concept; the resulting schema

[LAURINI, 1994]. The bottom-up approach begins with low-level programs and develops the

system to a high level gradually, successively integrating the user’s need’s as much as

possible. There is always a risk of the top-down approach missing important details, which

should be included in the conceptual data model, and of the bottom-up approach losing the

wider view. A good methodology for conceptual design should ideally be a compromise

between the two contrasting approaches [BATINI ,1992].

For highway agencies the data modeling process is mostly realized top-down. Since the

conceptual data model is designed for an entire agency and highway administrations are

hierarchical organizations, using a top-down approach permits a more comprehensive and

generalized data model. However, during this study it is realized that, in order to fulfil some

specific requirements such as cross-sectional design information, the bottom-up approach is

also required due to perform better information data structure analysis.

The conceptual schema typically includes conceptual entity objects, relationships, attributes

and methods which express the system behavior required by the user’s community. The

conceptual schema should be free from the physical structure of the database. This means it

should be independent of the software and object storage techniques. This makes possible a

change at the physical data level without involving any modification of the conceptual

schema. Three important characteristics of a conceptual schema, which must be satisfied, are:

2.3 Conceptual Data Modeling 10

 Consistency with the business infrastructure and be valid across all application

areas.

 Extensible, such that new data can be defined without altering previously defined

data.

 Transformable to both the required user views and to a variety of databases and

system architectures.

Actor

Actor Actor Actor

Physical View

Interna

Leve

Physical

Schema

Databases

Conceptua

Leve

External

View A

Conceptual View

External

View C Externa

View

Leve

External

View B

User A1

User A2

User C

Conceptual

Schema

User B

Figure 2-5: Three-Schema Architecture

2.3.1 Methods of Conceptual Data Modeling

Many methodologies exist for conceptual data modeling including semantic, functional,

Entity-Relationship (ER) and object-oriented. During this study, the two main methods will

be considered, namely ER and object-oriented. Currently both two approaches are highly used

during conceptual data modeling. Conceptual modeling in relational database design often

makes use of a formal approach known as ER modeling, first represented comprehensively in

1976 (Chen,1976) [LAURINI, 1994]. With the advent of object-oriented technology, an other

modeling technique, object-oriented approach has emerged.

2.3 Conceptual Data Modeling 11

2.3.1.1 Entity- Relationship Approach

The Entity – Relationship (ER) approach adopts the view that the real world consists of

entities and relationships between them which are characterized by properties.

In this formalism the basic components are:

 Entities, which are defined as clearly distinguishable real world phenomena and atomic.

 Relationships between entities are defined as associations between two or more entities.

 Attributes for both entities and relationships, where attributes are properties of these with

specific meaning with respect to the conceptual data model.

 Cardinalities describing possible relationships for each participating entity.

 Integrity constraints, which are functional relationships between entities.

ER generally contains groups of entity types, which are high-level classifications of a major

topic of interest. It is formally defined, using a very expressive language. ER data modeling

approach is well known, easy to read and wide-spread. However, ER does not provide

adequate concepts for representing methods designed in the conceptual data model.

Additionally due to required atomic entities, expressing composite data structures it is

difficult.

2.3.1.2 Object-Oriented Approach

The Object-Oriented approach is a method of design encompassing the process of object-

oriented decomposition and a notation for depicting both logical and physical as well as static

and dynamic models of the system under design [OMG, 1999].The formal language of the

object-oriented approach is the Unified Modeling Language (UML), a general-purpose

notational language for specifying and visualizing complex, object-oriented software or

projects.

The four major elements of this approach are abstraction, encapsulation, modularity and

hierarchy. These concepts will be examined in Section 2.4.2. An object is an entity with a

well-defined boundary and identity that encapsulates state and behavior. State is represented

by attributes and relationships, behavior is represented by operations and methods.

Relationships in UML can be classified as dependency, association, generalization and

realization [BOOCH,1994].

With object-oriented technology entities and relationships are still used, generally with an

extended notation. The main advantage of object-oriented approach is ability to define and

represent methods of objects, which will save many development and maintenance efforts and

provide clear understanding of the conceptual data model. Additionally the object-oriented

approach allows objects having complex structure and therefore model the reality in a better

way, whereas with ER the entity is atomic. However, a disadvantage of object-oriented

approach could appear because of complex structure concept, since the modeled complex

objects representation must satisfy the required detail level for further purposes such as data

maintenance.

As an example, geometrical element point and topological element node entities and their

relationship with respect to each other are presented. These entities are purposely chosen as

atomic, in order to highlight differences between the two approaches.

Nodes are generalizations of points. Therefore, they cannot be represented without points. In

order to simplify the example, it is assumed that the point entity attributes are the point

identifier and (X,Y,Z) coordinates, and the node attributes are the node identifier and node

2.4 Logical Data Modeling 12

number provided by the agency. There is 0..1:1 relationship between node and point defining

that every node should be assigned to a point, but a point may or may not be assigned to a

node. In Figure 2-6 these entities, their attributes and relationships are illustrated according to

both the ER and object-oriented approaches

Point NodeRepresents

( 0, 1 ) ( 1, 1 )

Point_ID

X- Coordinate

Y- Coordinate

Z- Coordinate

Node_ID

Node num

Point

Point_ID

X-Coordinate

Y-Coordinate

Z-Coordinate

+ calcStndev( )

Node

Node_ID

Nodenum

10..1

Represents

ER-Diagram

UML-Diagram

Figure 2-6: Conceptual Data Model Approaches

Between these two very common approaches the advantages of UML can be easily

recognized. The conceptual data model documentation is clearer through definition of

methods such as calcStndev( ) which stands for calculating standard deviation values of

(X, Y, Z) coordinates. In addition the UML notation is more expressive. For example the

relation between a point and a node can be presented as aggregation which also provides the

information that Node cannot be represented without Point.

2.4 Logical Data Modeling

The goal of logical design is to translate the conceptual data schema into a logical schema

tailored to the specified database’s management system. A logical schema is a description of

the structure of the database that can be processed by the database management software.

Until recent past two decades, the traditional approach to information system design mainly

concentrated on conventional systems, including the flat file, the hierarchical and the network

data model. In GIS-T point of view, these are important, since existing data sources are still

partly stored in these systems in highway agencies.

According to information technology historical developments, firstly flat file approach was

promoted. The flat file approach is based on application programs, where each program

defines and manages its own data. All data could be entered into one large table or a flat file.

They were often developed individually to meet the requirements set by a particular

department of the organization. With its simple data structure and rapid data access, the flat

file approach was highly used. A tabular model clearly allows association of entity instance

attributes, but is not effective for different levels of aggregation or for complex situations with

2.4 Logical Data Modeling 13

many entity types. Consequently some of the data in the file was either duplicated (data

redundancy) or inconsistent, since no coordination existed between files belonging to

different groups of people. Data sharing was limited and there was no enforcement of

standards, requiring extremely careful data maintenance.

Due to above described deficits, database systems had been promoted. Database systems can

be objectified according to the following types:

 Hierarchical

 Network

 Relational

 Object-oriented data model

 Object-relational data model

The hierarchical model has two basic data structuring types; records and parent-child

relationships. Record types describe the structure of a group of records. They were stored in a

general tree structure with one record type, which has zero or more dependent record types.

Hierarchical data models were widely used between 1960 and 1980 due to the widespread use

of IBM’s Information Management System (IMS), so it is possible to come across

hierarchical databases or applications, including GIS, also today in highway administrations.

The advantages of hierarchical models were easy usage and high speed of data access.

However, many-to-many relationships can not be realized with the hierarchical data model

because the parent and child object structure enforce one-to-many relationships. A child

object can not exist without a parent object. Additionally every new relation in records

resulted redundancies in the system. Difficulties appeared especially during modeling many to

many relations and non-hierarchical structures.

Problems of flat file approach, progress the development of the network data model, also

known as the CODASYL model, in parallel with the hierarchical approach. The network data

model has two basic structuring type, record and set. While record types are defined similarly

to the hierarchical data model, set types define one-to-many relationship between record

types. Network data models support many-to-many relations. Redundancies in the system was

highly decreased. However, updates were limited. Additionally it was not feasible to solve

spatial queries. The network data model has limited flexibility for changing data and access

requirements, which is necessary for GIS.

These systems were highly desired compared with relational database initially, because of

their high performance. However, the limited data exchange, unsatisfactory access

requirements and expensive maintenance, promoted the development of the relational data

model.

2.4.1 Relational Data Model

The relational data model is based on mathematical relations, where data is logically

structured in tables. The relational model was first defined in 1970, when E.F. Codd

introduced the idea of using the mathematical concept of relations (in the set theory) as the

means to the data model. [PAPAZOGLOU, 1989]

Principally, set theory and predicate logic is used. Supposing two sets, 1

S and 2

S, where

},{ 111 yxS = and },,{ 2222 zyxS =, the cartesian product is the set of all ordered pairs.

()()

(

)

(

)

(

)

(

)

},,,,,,,,,,,{ 21212121212121 zyyyxyzxyxxxSS =× (2.1)

2.4 Logical Data Modeling 14

Any subset of this cartesian product is a relation. In order to define a general relation between

n domains 1

S, 2

S, 3

S,..... n

S, the cartesian product is defined as

(

)

},...,,,...,{... 221121321 nnnn SsSsSssssSSSS

∈

×××× (2.2)

The Entity-Relationship (E-R) approach, described in Section 2.3.1.1 is the basic approach

used in the relational databases. Entities and relationships among them are stored in tables.

They are tabulated into rows and columns. With the relational data model, each element in the

n-tuple consists of an attribute-value pair ,where tuple is a row of a relation. An attribute is a

named column of a relation. Degree of a relation is the number of attributes it contains.

Cardinality is the number of tuples it contains. Relations between entities are implemented

through foreign keys. The foreign key is a code, which is stored in a table and refers to rows

in another table. The primary key of the linked table is stored in the foreign code column of

the other table. Relational databases can also be defined as normalized relations. Normal

forms are guidelines for relational database design that increase the consistency of data. In the

relational database systems methods and integrity constrains, which are defined in the

conceptual data model are realized through external transactions.

An example, in order to clarify the relational database terms is provided in Table 2-1,

showing the relationship between point and linear elements. Two 1: 0..N relationships can be

defined between Point and Linear Element, where every linear element must have at least

one beginning point and one ending point, and every point may be assigned to none or many

linear elements. In order to realize this, the primary key of the point, named point_ID, is used

as the foreign key in the linear element table.

Table 2-1: Basic Elements of the Relational Data Model

Point_ID

(Primary key)

Reference

Sys._ID

X Y Z ......

74520144

(Point P)

12 547248,790 4400809,172 154,725 ........

74520145

(Point R)

....... ........ ........ ........ ........

74520146

(Point S)

........ ........ ........ ........ ........

Linear

Element_ID

(Primary key)

Beginning

Point_ID

(Foreign key)

Ending

Point_ID

(Foreign key)

Description ......

14001 74520144 74520145 ...... ......

14002 74520145 74520146 ...... ......

14003 ........ ...... ...... ......

Relation

Degree

Cardinality

(

)

()

RRR zyx ,,

()

SSS zyx ,,

()

TTT zyx ,,

()

UUU zyx ,,

14001

14002

Linear

Geometry

Attributes

2.4 Logical Data Modeling 15

The relational database model presents operations on algebraic expression language in order

to realize data query and manipulation. With data manipulation languages the types of

operations allowed on the data is defined through relational algebra. It was originally

proposed eight operations in relational algebra, but several others have been developed. These

operations are selection, projection, cartesian product, union and set difference. In addition,

there are also join, intersection and division operations, which can be expressed in terms of

the five basic operations [CONNOLLY,1999]. These operations are also the basis for other

data manipulation languages, where the Structured Query Language (SQL) is the international

standard query language of relational databases.

2.4.2 Object Oriented Data Model

The object model captures the static structure of a system by showing the objects in the

system, relationships between the objects, and the attributes and operations that characterize

each object.[RUMBAUGH, 1991] Communication between objects is realized using a

message passing system. A message is a request sent from one object to another in order to

execute one of the objects’ methods.

One of the advantages of an object oriented data management system is that real-world

phenomena can be modeled closer to reality with non-atomic objects. Object models could be

structured around conceptual objects rather than geometrical properties. With the object-

oriented approach, geometry can be modeled like other information. Explicit stored spatial

information, which is required in GIS, can be realized without redundancy. For example

geographical objects can be defined using methods, which can automatically generate these

with the use of their parameters. In addition it is possible to generate user-defined types

without being limited to vendor specific data types.

Another advantage is the encapsulation concept. This permits possibilities such as;

 Integration of different abstraction levels for one object.

 Designing spatial, non-spatial and methods in one object.

 Expressing business rules as methods belonging to the object.

In Table 2-2 a road segment is presented as an example to illustrate the object-oriented

concepts. Spatial information, non-spatial information and business rules are encapsulated in

one road segment object, preserving relationships between each other. Spatial data for the

provided example includes; coordinates of the middle axis, reference coordinate system,

parameters, spatial accuracy obtained via defined methods and topological elements.

Thematic data belonging to the segment is represented by road type. Finally, business

methods are introduced such as distance( ) and calalignment( ) which is defined for the

calculation of distance of the road segment and the alignment parameters.

With reference to the relational database example in Section 2.4.1, as this example is not an

atomic object, some additional information is required for this example. However, with

additional tables and their assigned relationships, it is possible to map the same information

using relational databases. Additionally methods and integrity rules needed to be provided in

the relational database in order to acquire the same result and to maintain the data integrity.

2.4 Logical Data Modeling 16

Table 2-2: The Object Road Segment

Road Segment

RoadSeg_ID

(Object

Identifier)

Link

_ID

Measured

Length

Road

Type

BPoint

_ID

EPoint

_ID

LinEl

_ID

LinEl

Type

…

10015 5001 42,521 08 74520144 74520145 14001 line …

74520145 74520146 14002 arc …

… … … … … … … …

()

zyx ,,

()

RRR zyx ,,

()

SSS zyx ,,

()

TTT zyx ,,

()

UUU zyx ,,

+distance( )

+calalignment( )

+accuracy ( )

+Move ( )

+Animate ( )

2.4.3 Comparison of Relational and Object-Oriented Approaches

In order to find the appropriate approach for designing GIS-T conceptual data models, the

advantages and disadvantages of both approaches needed to be highlighted.

One of the benefits of relational database technology considering GIS-T is that; it offers

solutions to the issues of security, versioning and referential integrity. In addition, it is mature

and available across a wider variety of platforms. Established standards such as the Structured

Query Language (SQL) for querying large databases are available. The main drawback lies in

not providing adequate facilities for specifying constraints on the data. Additionally,

relational data models are built upon elementary elements, with only atomic features being

permitted, which leads some limitations during complex features. In the areas of data

semantics, model extensions, object identity and programming interface, weak points of the

relational model can be found.

Compared to the relational approach, object-oriented methods provide a model for integrating

data with business rules. To achieve the same semantics, relational database management

systems usually require complex control methods, generated through a combination of third-

and fourth-generation languages. [BOOCH,1994] With object oriented modeling complex

objects can be generated. Concepts of abstraction, user defined data types, encapsulation of

object properties with business rules are the main benefits introduced by means of the object-

oriented approach. In addition, it provides a more semantic substance by allowing the user to

explicitly specify constraints on the data. However, the object-oriented modeling approach

has several disadvantages. The object oriented languages run slower than procedural ones due

to message expression, degrading query performance. Additionally, very few GIS-software

vendors have been successful in using an object-oriented database for storing and retrieving

2.4 Logical Data Modeling 17

spatial data. Object oriented databases do not provide a standard query language. Another

bottleneck is that the amount of data in the GIS database component is very large compared

with other systems, which makes the performance considerably low.

A comparison between the two approaches with respect to GIS is provided in Table 2-3.

Table 2-3: Comparison of Relational and Object-Oriented Approaches

Advantages:

 Standard query language,

SQL

 Enhanced spatial query

opportunities

 Versioning

 Wide-spread, mature

 Security

Disadvantages:

 Weak support in the

integration of business rules

 Pre-defined data types

 Atomic entities

Advantages:

 Integration of business rules

with spatial information

 User defined types and functions

 Enhanced abstraction concept

 Simplicity in interfacing to other

sources

 Providing solutions for

generalization problems

Disadvantages:

 Management weakness with

large amounts of data due to

messaging

 No standards for query

languages

 Not widespread, many concepts

are still at a test stage

Relational Approach Object – Oriented Approach

In GIS, it is necessary to be able to use enhanced spatial query opportunities, versioning of the

relational concept. Additionally, existing digital information is generally stored in relational

databases. Contrarily, the object-oriented approach propose solutions to problematic areas of

GIS such as multi-dimensionality, various abstraction levels and integration of business

methods. When both approaches are evaluated in the GIS context with respect to their

respective advantages and disadvantages, an hybrid approach is emerged, namely the object-

relational approach.

2.4.4 Object-Relational Data Model

The object-relational data model combines the concepts of objects and methods from the

object-oriented model with the concept of relations from the entity-relationship model. The

object-relational database management system is an extended relational database supporting

abstract data types, procedures, encapsulation and complex objects of the object-oriented

concept. Limited operations of relational databases can be extended, defining new operations

and methods. The advantages of relational databases such as standardized query languages,

security, versioning and referential integrity facilities can still be used.

The object-relational data model shows its strength in performing queries of complex

structured data. Since with the object-relational data model, it is possible to define complex

2.4 Logical Data Modeling 18

structured data in the same manner as relational database. The object-oriented messaging

approach is not required between objects, which causes lack of performance. Additionally, the

enhanced spatial query possibilities of relational databases are easily performed, since the

complex objects are stored in tables associated with their object identifiers.

The main requirements of GIS such as searching capabilities, multi-user support, handling

complex data and extensibility of systems are best satisfied with object-relational database

management systems. Stonebraker proposed a classification between database approaches

with a four-quadrant view as illustrated in Figure 2-7. [CONNOLLY,1999]

Relational Database

Management System

Object - Relational Database

Management System

File Systems Object - Oriented Database

Management Syste

capabilities/

multi-user

support

Data complexity/ extensibility

Figure 2-7: Classification of Database Management Systems(DBMS) [CONNOLLY,1999]

During this study, since in highway administrations large amount of information are mainly

stored in relational data bases and the object-oriented concepts provide solutions to identified

problematic areas, the object-relational data model was considered as the most suitable

approach and applied.

3.1 Overview 19

Chapter 3 : Highway Administration and GIS

3.1 Overview

Highway administrations, being one of the major governmental organizations, provide

services and coordination at a national and international level to ensure the mobility of people

and goods. Examples of highway administration tasks are:

 Planning new constructions

 Maintenance of existing roads

 Traffic information

 Regulations about road usage and safety

 Installing facilities on highways

 Conducting projects

 Program evaluation

In order to fulfill these various tasks at a technical and administrative level, information is

needed from diverse sources.

In addition to the above mentioned tasks, due to an increase of environmental awareness,

other tasks have arisen such as:

1. Protection of environment, referencing existing road net information during the

planning phase.

2. Rising qualitative requirements in terms of the traffic route network, e.g. increase of

road safety and separation of the traffic method.

3. Coordination of road, rail and waterways, which can only be achieved through new

technologies and uniform networks.

4. Evaluation of:

a. the improvement or adjustment of design elements in planar and vertical

positions.

b. cross-section and node design is only possible through new standard software

and the development of new conceptions.[BMVBW,1998]

Technological developments, especially GIS technology, are considered to be inevitable by

highway agencies for the realization of their tasks, since GIS provides rapid and accurate

decision support tools in a more integrated and economical manner. GIS capabilities, which

have major significance to highway administrations, are [ALTAN, 1996]:

 Ability to share spatial information, which often leads to better cooperation and

decreases inconsistency and redundancy.

 Ability to integrate loosely related data, with respect to usage of spatial references,

which can lead to discovering new properties of data and cooperation at many

levels.

 The ability to aggregate data into larger geographic units that are more appropriate

for the large scale applications, which can lead to far better understanding,

cooperation and management of the operational units within an organization.

3.2 Analyzing Highway Administrations 20

Due to the necessity of close coordination between departments and divisions, the efficiency

of GIS-T technology in highway administrations, can only be achieved using an integrated

approach. Additionally, the integrated approach has many economic benefits such as the

minimization of data collection.

3.2 Analyzing Highway Administrations

In this chapter, an evaluation of the current GIS-T situation within highway agencies is

presented for four countries in order to achieve a wider perspective through this study.

Concentration is mainly given to the following;

 Organization structure.

 User requirements.

 Data acquisition techniques.

 System architectures.

 Existing information systems and the conceptual data models.

The usage of GIS-T technology differs between the mentioned countries, which helps in

concentrating on more generalized problems of highway agencies. Some statistical

information on the countries evaluated is provided in Table 3-1:

Table 3-1: Statistical Information on the Countries Studied

Country Area

( km2 )

Road Nets( km) Number of

Motor Vehicles.

Motor-

ways

National Regional Total

Denmark 43.094 880 3.690 7.090 71.600 2.040.000

Germany 357.022 11.300 41.600 75.800 633.000 43.350.974

Turkey 774.815 1.405 31.412 28.813 381.631 4.327.885

USA 9.363.520 88.400 727.000 694.000 6.420.000 203.659.000

3.2.1 Organization Structure

The organizational structure of the highway administrations studied is grouped into two main

categories according to governmental type: federal and centralized.

The organizational structure of the German Highway Administration can be given as an

example of a federal governmental type. State highway authorities are responsible for the

administration of interstate roads and highways on behalf of the federal government. With

respect to GIS, this will lessen the amount of data managed compared to a centralized

structure.

However, problems arise with the transfer of information, standardization and data formats.

This is due to there being different conceptual data models and different systems in use. This

issue needs to be examined in detail, since this is one of the core problems of GIS-T and valid

for both centralized and federal structures.

3.2 Analyzing Highway Administrations 21

Highway administrations scope and business methods are same, therefore it is expected that

conceptual data models are similar. However, there are huge differences between existing

conceptual data models. The reason of using various conceptual data models includes;

 No formal conceptual data model has been designed, including spatial and non-spatial

data. Software vendors’ proprietary databases are used for storing spatial information.

Other information sources, which are unstructured, are then linked as non-spatial

information. Software vendors have various “black-box” data models and systems, which

are not generally fully available for third-party developers. Therefore, the designed

conceptual data model is unknown and dependents on selected GIS software.

 The conceptual data models are designed in order to respond various user requirements

separately, such as pavement information system, tunnel information system or traffic

safety information system.

 User requirements are rapidly changing due to increase of the GIS usage and

technological developments.

 The identification of user requirements can be inadequate, due to complexity of data-flow

and information structure.

If data integration is intended in federal structure, it is needed to integrate the various

conceptual data models together. Additionally, other organizations or governmental bodies,

with the task of increasing the information transfer between highway administrations, are

required.

In a centralized governmental structure these problems will tend to decrease. However, some

similar problems will be apparent concerning data management, integration of data and the

coordination between regions and headquarters. The variety in conceptual data models are

also observed. In order to give an overview of the size of highway organizations in a

centralized governmental structure, the organization chart for Turkey’s General Directorate of

Highways (KGM) is provided in Annex A.

In this chart several departments have administrative or financial duties involving spatial

information. However, these departments have a different level of interest in the spatial data.

During user surveys, it has been noted that there is a necessity to structure, model and

maintain this information, considering the information’s level of detail and spatial

characteristics. Facility management systems are recommended for the management of

information such as building management, equipment stock and personnel, since for such

requirements the efficiency need to be higher. The information concerning existing hardware,

software and peripherals, such as plotters and scanners should also be included in facility

management systems.

In contrast, GIS-T user requirements are generally concentrated more on technical aspects of

road information and upper-management administrative aspects, such as alternative routes in

emergency locations. The inclusion of other requirements, which are more adequate for

facility management, would require additional data modeling efforts and enlarge the system

with inefficient methods in order to integrate both different levels of information. These

additional information and methods will definitely results to performance problems in queries

and additional costs.

Since there is more interaction between the information users, the effectiveness of both GIS

and facility management systems can be increased in this way. Mutual agreement on

requirements will increase the information quality and reduce the costs of development and

implementation.

3.2 Analyzing Highway Administrations 22

3.2.2 Assessing the Requirements of Highway Administrations

In order to design the conceptual data model, it is necessary to identify the external views.

The external view is composed of user duties, expectations from the system and information

related aspects such as content, source, usage and methods. However, identifying the external

views is not an very easy task, since the conceptual data model must be kept abstract in order

to be able to extend it where required, and to fulfill all the diverse user requirements. The

candidate objects should be identified which represent diverse user views at the required level

of abstraction.

A variety of information sources, including documents describing the systems, were studied

in order to determine user requirements. For example, the user requirements as defined by a

special working group of the Turkish highway administration are given in Annex B. In

addition a questionnaire was prepared which specifically focused on conceptual data models,

existing data sources and user duties. During this study the questionnaire was used during

interviews with professionals within highway administrations in Germany, Denmark and

Turkey. This questionnaire can be found in Annex C.

Several requirements were mentioned during these interviews, which are of importance.

These include crisis management, public travel security, emergency vehicle management,

information exchange with other organizations, internet roadway condition map and “What

if” reports. Additionally, the information requirements of international organizations, such as

the European Union (EU), should be considered concerning future transport information

systems in Europe. These are partially listed below [EU-APAS,1996];

1. Road traffic of passengers.

2. Long-distance traffic of passengers.

3. Traffic of goods for all modes.

4. Road, rail and inland waterways transport infrastructures and networks.

5. Improve the inter-regional flow data between countries of the EU.

6. Transport of dangerous goods, for all modes.

7. Environmental variables required by the integrated scenarios.

8. Qualitative variables required by European scenarios: border effect, development of

information technologies, “just-in-time” techniques.

Generally, these requirements highlight different views of reality listed such as design

information, environmental impact analyses, accident data, cross-section design, road

inventories and project monitoring, but they have a common denominator: road.

The road information can be categorized into several various groups including; new

construction, existing data, traffic data and history, which is illustrated in Figure 3-1.Due to

simplicity of data maintenance and topology analysis, in figure geometry and topology should

be distinct. The main approach to separation of geometry and topology was in Section 2.2.1

defined and other requirements promoting this separation can be found in further sections of

this study.

3.2 Analyzing Highway Administrations 23

Road and Traffic information

New Construction

- Surveying

- Design

- Environmental

- Construction

- Controlling

- Furnishing

Existing Data

- Network Data

o Road net

o Administration

o Traffic related

- Inventory

o Construction related

o Road facilities

- Road status data

- Construction-site data

Traffic Data

- Traffic data

o Counting

stations/dynamic

traffic data

o Peripherals

o Accident data

- Signs

o Dynamic signs

o Static signs

o Light signs

History Geometry/ Topology

Figure 3-1: Road and Traffic Information[PORTELE,2001]

In order to incorporate all these views in the data model and to integrate them, the data

structure needed to be examined. It was important to determine whether data has a spatial or

thematic nature. In order to represent the variety of spatial information used in highway

administrations, three application requirements of highway administrations were selected.

These were; the generation of accident black spot maps, integration of inventory records into

the system and the analysis of existing pavement layers. These requirements are associated

with three different referencing systems illustrated in Figure 3-2.

h q

Figure 3-2: Reference Systems

Accident black spot maps are produced using three dimensional reference systems (X, Y, Z).

These are generally represented as (x, y) planar orthogonal coordinate systems, although

integration of third dimension will increase the efficiency of analysis.

3.2 Analyzing Highway Administrations 24

(0, 0)

Figure 3-3: (x, y) Planar Orthogonal Coordinate System

The second example is the integration of inventory records into the system. The inventory

data collected for maintenance purposes is one of the main information source. It shows the

physical properties of roads and the current state. The road inventory contains information

such as control section identification, road surface, intersection, cut-fill, road lighting,

geometric elements, altitude, slope and critical section.

Due to the information variety, several other departments, such as planning and the traffic

division also emphasized the importance of road inventory information being present in the

system. Road inventory data has a one dimensional reference system, generally called linearly

referenced. The linear referencing approach is used, due to its simplicity and low costs, in

nearly all application areas that are based upon networks, such as infrastructure management,

utilities management and hydrological analysis. Additionally, this method is the most natural

method in order to collect linear objects. This methodology is shown in Figure 3-4.

45+97 km

Tunnel

Traffic sign

45+00 km

010-02

200-13

108,93 m

Figure 3-4: Linear Referenced Road Information

For integrating road inventory data into the system another reference system needs to be

defined. We can derive from the simple linearly referenced system two planar systems

namely; (l, q) and (l, h). These systems are illustrated in Figure 3-5. In both reference

systems, measured length along the linear element defined by l. In the (l, q) horizontal system

the q-coordinate represents distance normal to the linear element. In the basic linearly

referenced system it would be regarded as being zero. With the (l, h) vertical coordinate

system, h is the height of points along the route.

(0, 0)

Figure 3.4.a: Horizontal System Figure 3.4.b: Vertical System

Figure 3-5: Linear Reference Systems

3.2 Analyzing Highway Administrations 25

In order to analysis of existing pavement layers, due to historical convention, another

reference system is used in highway administrations. In this work this system is termed as the

cross-sectional reference system and represented as (h, q) reference system. The coordinate

system origin is defined to lie in the middle of the road axis. This reference system is

illustrated in Figure 3-6. Further details on referencing system issues will be discussed in

Chapter 4.

(0,0)

Figure 3-6: (h, q) Reference System

Several other requirements such as; project information, traffic monitoring and capacity

analysis are selected, in order to present an overview of spatial data and the reference systems

used by highway agencies. For these requirements spatial characteristics, the analysis cycle in

the agency, data collection frequency and methods are provided below in Table 3-2.

Table 3-2: Data Requirements

Spatial Data Data Collection User

Requirements

Linear

Referenced

Referenced By

Coordinates

Analysis Cycle

Time Method

Project

Information One-time/ annual As Built Contract

Road and

Motorway

Information

Annual Daily/

Monthly

Inventory

Traffic

Monitoring Annual Hour/Daily Counting

Information

Maintenance

Costs Annual/ Multi-year Current Record

Accident

Information Annual Take Place Accident

Report/

Inventory

Pavement

Maintenance Annual Seasonal Report

Bridge

Information Annual As Built/

Monthly

Inventory

Planning

Applications Annual Monthly Report

Capacity

Analysis Annual Daily/

Monthly

Record

Work Program Annual/Year Seasonal Report

3.2 Analyzing Highway Administrations 26

In addition to these information variety in highway organizations, there is an enormous

exchange of data between departments and divisions. Since methods and definitions vary,

there are many information gaps in the technical data-flow, making the analysis of

information flow very difficult. The existing data in the highway administration is examined,

in order to achieve better understanding of external view. In highway administrations

commercial GIS and database software are in widespread use. These are generally supported

by in-house and requirement customized programs.

3.2.3 Data Acquisition

In order to realize above described tasks, the first priority is given to data acquisition in

highway agencies. The data acquisition is a continuos process, since various phases of road

information is needed in order to fulfill the requirements. These various stages are planning,

construction, current information such as traffic data and maintenance, which are shown in

Figure 3-7.

Planning

Design

ConstructionCurrent

Maintenance

Figure 3-7: The Road Information Phases

These diverse requirements are stressing agencies to collect the required information by

means of various methods, depending upon required generalization level and quality. In

addition to these, collected data have generally spatial character. The spatial data acquisition

requires the highest efforts and costs, due to the high level necessity of completeness,

actuality, correctness and well-defined data structure [BILL-I,1999]. In the past few decades,

conventional data acquisition approaches were mainly used in highway agencies and the

information was stored in analogue format. Main reasons for this situation were, digital data

acquisition techniques either were not available or economic. In addition to this, new

approaches require generally a change in the traditional way of thinking. However, analogue

data collection techniques are very expensive and not rapid. Additionally, the analogue

information could not be easily exchanged or shared with other users, therefore efforts needed

be paid only on one side.

In highway administrations, digital data requirements have increased in importance due to the

needs for increased efficiency with possible low costs. Developments in information

technology, especially in the areas of database and GIS-T, now provide the means to realize

such optimization. In parallel to this, digital data acquisition techniques, such as Global

Positioning Systems (GPS), photogrammetry and automatic counting machines, became

commonly used in highway administrations, due to their data collection speed and cost

reduction. Although the new techniques reduces the costs of conventional techniques, data

acquisition from sketch is still not affordable for the entire highway information. As a result,

although there is an increasing trend towards data acquisition with the new technologies,

3.2 Analyzing Highway Administrations 27

conventional techniques are also in usage. Consequently, in highway administrations both

analogue and digital data is available.

The road inventory data can be given as an example. It is mainly acquired using vehicle

odometers and documented in analogue semantic sketches. This existing information is

entered manually, if it is needed to be used in a system such as; Computer Aided Design

(CAD), road database or GIS-T. In addition to these conventional approach, other

technological possibilities are being considered and applied successfully. In the case of

Denmark the road information related to network profile, surface conditions, length and cross-

section profile is collected by gyro and laser combined vehicles. The data collected with this

vehicle is automatically sent via a network for further processing. It is planned to combine

this system with the Global Positioning System (GPS) in the near future. This system,

developed by the Danish Highway agency, is being used for roughness measurements in other

countries such as Sweden and Greece. In other countries also similar integrated techniques are

in usage. Especially in the USA and Germany, GPS integration was already realized.

Additionally, evaluated countries are investigating systems for obtaining geometric

information in an accurate manner, especially road widths and geometric elements. In

particular the use of differential global positioning system (DGPS) in this field is being

developed. In Germany GPS measurements have been made especially at network nodes in

order to improve the node location accuracy. Also in Germany, a research group was founded

with the task of studying the relationships between road alignment parameters and traffic

safety [NWR,1995].In particular the correlation between possible accident causes and short

and long curve parameters is being considered.

Monitoring road information is another popular data source for highway administrations,

especially for the determination of current road condition and in the clarification of legal

issues. Video image includes time, date, speed, route number, location on the road and the

direction of measurement at Danish highway administration.

Figure 3-8: Sample Video Image from Danish Highway

Another image database source is maintained by the Federal Highway Research Institute of

Germany to document the direction signs of the German federal motorway network, including

main lanes as well as connecting, exit and entrance ramps. Access to images is realized via a

search method either through selection of the motorway number and the direction or the name

of junction. After displaying an image, it is possible to maneuver by jumping from image to

image. The database is marketed in CD-ROM format. It can also be considered to use this

information for automatic extraction of data such as road numbers, distance measured or the

location of traffic facilities.

Photogrammetric data acquisition is carried out by highway agencies, especially during the

planning phase. Within the Turkish highway administration, digital terrain models (DTM) are

available for areas where photogrammetric data acquisition has been realized. In Germany,

3.2 Analyzing Highway Administrations 28

experimental three-dimensional GIS applications have been made using Digital Terrain

Models with 50 m resolution.

However, although these new techniques has many benefits and successfully applied, these

applications are specific to requirements or applications, therefore are not fulfilling the entire

highway agency requirements, especially in case of GIS-T. In order to fulfill entire highway

agency requirements, the information in various sources, formats and generally in analogue

form, is needed to be structured and integrated.

The main requirement of GIS-T is digital base-maps. Indirect methods, which are digitizing

and scanning, are commonly used in order to acquire digital base-maps. In some cases these

are provided from other organizations, such as the federal governmental land registry agencies

in Germany. However, generally base-maps are digitized by the highway agencies, as is the

case in Denmark. In base-map production, direct methods including the remote sensing and

photogrammetric data acquisition techniques are still not widely in usage as it was expected to

be. Since costs of these techniques depends highly on production scale and highway

organizations requirements varies form very large scales to very small scales. However, these

techniques provides more rapid and up-to-date solutions compared with the digitization,

therefore it is expected that, they are more commonly for base-map production in the near

future. Additionally, with the increased usage of GIS-T, other requirements such as digital

divided highway information is expected to increase.

Several other requirements appear during GIS-T establishment. Other road related

information, such as linearly referenced data or cross-sectional design data needed to be

integrated into the system. However, the available digital information is not efficiently

applicable or requires enormous efforts before it can be used in GIS-T. This is because, due to

organizational separation of the tasks, similar or identical data was collected with different

data descriptions, formats and various accuracy. Additionally, there is lack of common

terminology and various methods are used. In spite of such complications, reusability of

existing data and integration within the system is essential because of obtaining high costs of

digital information. This problem promotes the usage of integrated conceptual data modeling

and adjustment techniques, since the efforts required for data integration can be reduced and

integration can be automated. Additionally, such methods increase the quality of available

data in the system. In the further chapters of this study, various examples of such methods

were presented.

3.2.4 System Architecture

The entire highway administration GIS-T design is accelerating due to the integration of

methodologies, provision of common basics, decreasing data acquisition efforts and the

integration of existing data. In this context, as the efficiency of the system can only be

realized with up-to-date information, an appropriate system architecture should also be

considered.

In highway administrations, due to distributed user locations and various interests, a client-

server architecture, intranet and internet facilities needed be designed in order to realize

contribution of all users. The importance of the client-server architecture and communication

infrastructure have been already appreciated and realized in the examined countries. However,

in practice there are various implementations of client-server architectures. One of them is

storing information in distributed databases, as an example on basis of regional divisions of

highway administrations, and integrating them in the main server on a regular defined time

interval. Another approach is storing information on a main server. In this case, it is also

possible to distribute the information on several databases but user data retrieval and queries

are realized on the server, without a secondary storage on the user side. Additionally, in both

3.3 Current Status of GIS-T at Highway Administrations 29

approaches security rules and user rights for updating the system needed be defined and

established.

The Danish highway administration provides a good example for the establishment of system

architectures in highway administrations, since both approaches were applied and evaluated.

In the Danish highway administration, on the basis of regional districts, initially the first

approach was implemented. Every region had been stored a copy of the server information

related with its own region. The maintenance of this information was under responsibility of

regional divisions and within certain time intervals it was sent back to the main server in

Copenhagen. The advantages of this approach is quick responses to user queries, as amount of

data was reduced. Additionally, the amount of information, which needed to be maintained in

the region database was extremely decreased. However, this architecture did not provide the

expected efficiency. Firstly, the coordination ability of highway administration headquarter

was reduced, since regions and headquarters have different versions of the same information.

Secondly, the main server information was not completely updated. Thirdly, there were

redundancies considering integration of various data sources, since road information is

generally shared and used by different regions and headquarter, but different data sources

were processed separately. Due to inherited difficulties the second approach was performed.

Within the second approach, on the basis of regional districts, clients update their particular

portion of the server database. Each client has pre-defined rights and responsibilities. In order

to guarantee correct system functionality, these rights and responsibilities are designed to be

non-overlapping. Users are connected to a main server via network facilities. The

communication between the central and local systems is ensured by Integrated Services

Digital Network (ISDN). Their queries are submitted to the server where it is processed and

the response returned to the client. Database update is performed by server. Fourteen federal

countries are connected to the server database, and updates are realized on daily-bases. The

central data processing division undertakes database integrity and general maintenance. With

this approach, first approaches disadvantages are reduced to minimum. Only one version of

information exists, which highly increase the coordination in the agency. Additionally,

redundancy is avoided. However, performance is decreased, due to data retrieval from a very

large database. The performance plays a secondary role in this case, when the advantages of

this approach is considered. Additionally, the performance issue can be lessen using new

technologies, such as; Broadband-ISDN(B-ISDN), intelligent infrastructure or optimizing

data model. In general, the system architecture established satisfy the user requirements in

topics of up-to-dateness, obtaining complete data overview.

Since only database vendors existing pre-defined data maintenance facilities are in usage, in

topics of security and up-date-rights with respect to data maintenance, more attention is

needed. Consistency and data integration controls are not performed within the system,

especially with respect to spatial data. Considering these issues, several proposals will be

made in further chapters.

3.3 Current Status of GIS-T at Highway Administrations

3.3.1 Current Status of GIS Technology

Before evaluating established systems in selected highway administrations, since some

problems such as data integration issues are correlated with GIS technological development,

developments and current status of GIS technology needed to be discussed.

With the development of spatial theory and computer technology, in order to analyze data

collected by Canadian Land inventory data and to produce statistics for land management, in

the mid 60's the Canada Geographic Information System had been established. With the speed

3.3 Current Status of GIS-T at Highway Administrations 30

of computer technology today, IV Generation GIS offering many benefits especially in the

area of worldwide web technologies, data modeling techniques, with their client-server

architectures. The main concern of this study, the data modeling techniques, had been

enormously developed from geometrical data modeling into object oriented (OO), recently

into object-relational(OR) data models. Development of GIS technology is illustrated in the

Figure 3-9.

Figure 3-9: Status of GIS [SCHILCHER,2000]

Although it is not possible to classify approaches used data modeling definitely in GIS, the

file processing, dual database and integrated database approaches are the main groups. With

the file process approach, data files are stored in various file and many additional programs

are provided in order to fulfil diverse applications. Although data is handled at proprietary

databases, it is possible to access standard commercial databases. [HELOKUNNAS,1995]

Environmental Systems Research Institute’s (ESRI) ArcView and MapInfo can be given as an

example of software using this approach.

The dual database approach is the most common approach used in GIS. Each geographic

entity is decomposed into its respective spatial and thematic components and stored in

separate "dual" databases. ArcInfo, Intergraph MGE and GeoMedia Professional GIS

software can be given as an example for this approach. The thematic component is stored

entirely within a commercial Relational Database Management System (RDBMS). The

geographic component, conversely, is stored in a proprietary database with its own unique

internal access, through feature identifier, and storage methods illustrated in Figure 3-10. The

stability of system is achieved generally through feature identifiers, which are automatically

assigned or integrity rules provided by software vendor. The decomposition of spatial an non-

spatial data was inevitable in, since it was not efficient to store spatial information and non-

spatial information in the same database. The reason for this non-efficiency was the lack of

appropriate spatial indexing mechanisms for queries in databases. Additionally, visualization

could not be realized. Due to these reasons, the proprietary storage mechanisms are used for

the geographical data, in order to increase system performance and to be able to visualize the

spatial information.

3.3 Current Status of GIS-T at Highway Administrations 31

Thematic Attribute ....

Feature Identifier

Thematic Attribute A

Spatial Component

( x, y )

Thematic Attribute F

Thematic Attribute X

Thematic Attribute B

Figure 3-10: Components of Dual Architecture [MOLENAAR,1991]

Within this approach, the thematic aspects of data have the first priority in GIS. Structure and

processing of spatial data was secondary. [MOLENAAR,1998] The spatial component of an

entity, in this context, mainly refers to its geometric representation (i.e. point, line, polygon)

and required only for visualization of thematic data. Generally at practice geometry is

redundantly stored in the system. Although the concepts behind the design of these spatial

database management systems are well documented, the actual systems are ultimately black

boxes. [CHURCH et al,1994] Due to this non-open architecture, redundancy of geometry can

not be easily detected. In addition relational databases can not handle required amount of

integrity rules, including spatial and non-spatial information integrity, as they are not tailored

for this purpose.

Because of data maintenance problems and data integrity issues several other approaches such

as shell architecture and integrated architecture were developed. The shell approach, stores all

data within one database with a spatial support "shell" on top of a RDBMS for geographic

queries. In this architecture, spatial data is separated into its basic elements (i.e. points, lines,

polygons) and stored separately in related tables. To retrieve the information, relational joins

are performed to reconstruct the required geographic entities. Smallworld GIS software can be

given as an example. The efficiency of this approach, therefore, highly correlated with the

design of the spatial shell appropriation. In many ways the proprietary shell between the user

and the underlying RDBMS has the same drawbacks as the dual architecture. The data model

and the shell is designed and established by the GIS vendor, and the system is essentially non-

extensible. [CHURCH et al,1994]

After developments in database technology, especially enhanced spatial indexing systems and

object-oriented technology, the integrated approach has emerged. With the integrated

approach, it is possible to store spatial and non-spatial information in one database

management system with the required performance in spatial queries with the spatial indexing

mechanism. With this approach, an extended DBMS with spatial data storage capabilities or

object-oriented DBMS is required. With this approach users can extend the system through

user defined objects and methods, contrary to other approaches, where only software vendor

can decide which type of objects should be used. The Oracle8i with Spatial Option is an

example where this approach is applied. There are software vendors also supporting this

approach through their own spatial data storage concepts, such as ArcInfo with Spatial Data

Engine (ArcSDE). With this approach unlimited amount of real world object can be modeled.

Especially with its integrating approach and provided possibilities of storing spatial data in

relational database systems, this category is taking very much attention. The redundancy is

lessen compared with dual approach, since data is integrated in one database. However, due to

user-defined objects, there is an risk in data model integration of diverse sources with respect

3.3 Current Status of GIS-T at Highway Administrations 32

to compatibility. This risk needed to be minimized with complete, formally described

conceptual data models.

The evaluated GIS-T applications were realized using dual and shell architecture, that spatial

and non-spatial data are stored in separate databases and then linked back to together, using

unique identifiers. A special aspect of highway spatial information systems, which is linearly

referenced data was non-considered in studied countries, with an exception of the USA. They

are considered as non-spatial information, which abolishes advantages of dual architecture. In

the USA, these problems are already recognized and there are some efforts of integrating

linear referenced data with three dimensional reference system but these efforts are yet not

satisfactory in many cases.

3.3.2 Standards Established for GIS-T

With the prevalence of GIS usage and the resulting raise in data sharing issues,

standardization efforts are increasing worldwide.

At the international level the International Standardization Organization (ISO) has three

technical committees (TC) for the definition of data standards relevant to the topics of this

research. These are TC 211 which deals with general GIS, and committees TC 204 and TC 22

which relate to road information. The scope of TC 211 is to establish a structured set of

standards for geographic information. The standard proposal aims to standardize general

aspects of GIS, such as methods, tools and services for data management. The TC 204 is

responsible for the standardization of information, communication and control systems in the

field of urban and rural surface transportation, including inter-modal and multi-modal aspects,

traveler information, traffic management, public transport, commercial transport, emergency

services as well as commercial services in the transport information and control systems

(TICS) field.[ISO,2001] The TC 22 is concerned with compatibility, interchangeability and

safety, with particular reference to terminology and test procedures (including the

characteristics of instrumentation) for evaluating the performance of road vehicles and their

equipment.

Another international standard is Digital Information Geographic Exchange Standards

(DIGEST). This was defined by the Digital Geographic Information Working Group

(DGIWG), an initiative established by eight NATO countries. DIGEST was planned for

civilian in addition to military purposes [BILL-II,1999]. The aim of DIGEST is to assure

compatibility between multi-national agreements for digital data standards with respect to

supported data structures, feature and attribute coding scheme, exchange media, format and

administrative procedures [DIGEST,2001].

The International Cartographic Association (ICA) is an established working group on digital

cartographic data exchange standards. It aims to identify research needs arising from the

standards process and information exchange at the international level.

There are also activities by the Open GIS Consortium, which is an non-governmental

consensus-based association of public and private sector organizations. The standard

proposed by the consortium, Open Geodata Interoperability Specification (OGIS), has the

scope of providing a single 'universal' spatio-temporal data and process model. The standard

will cover all existing and potential spatio-temporal applications, a specification for each of

the major database languages to implement the OGIS data model and a specification for each

of the major distributed computing environments to implement the OGIS process model.

The Geographic Data File (GDF) has been developed as part of a European Community (EC)-

sponsored project to develop a European Digital Road Map (EDRM). Since 2000 the ISO

version of the GDF (ISO TC 204) format has become an international standard. The primary

3.3 Current Status of GIS-T at Highway Administrations 33

usage of GDF is for car navigation systems. It can also be used for many other transport and

traffic applications including, fleet management, dispatch management, traffic analysis, traffic

management, automatic vehicle location, road maintenance and public transportation. TC 278

and the GDF standards correspond to the needs of vehicle manufacturers in the area of routing

systems.

In the USA another standard has been developed, namely the Spatial Data Transfer Standard

(SDTS). This allows U.S. Federal Agencies to share spatial data between applications which

use differing hardware, software and operating systems. SDTS was designed for use with data

such as topology, raster data, hydrographic and topographic data. Also SDTS forms part of

the DIGEST specification set used by European members of NATO.

At the European level, in the field of spatial data, European Committee on Standards (CEN) is

represented by TC 287, which has the scope of identifying and defining a structured set of

concepts and components in general terms. The standard includes defining, structuring,

querying, updating geographic data and metadata with respect to geographic information.

Metadata can be defined as data about data. This definition leads to the evaluation of current

data according to various criteria such as logical consistency and completeness. Consequently

an evaluation framework is determined, which extend beyond the required quality criteria

within the field of geodesy and both suitable for spatial and non-spatial data.[SCHEU,1995]

However, several evaluation criteria including positional accuracy, spatial attribute accuracy

and logical consistency can be simply ensured with the help of approaches in the field of

geodesy, especially with the use of adjustment techniques.

Additionally CEN TC 278, which was established specifically for road, transport and thematic

data standards, is concerned with standardization in the field of telematics. The focus of the

standard is the application of telematics to road traffic and transport, including those elements

which need technical harmonization for inter-modal operation. Additionally, the standard

seeks to support communication between vehicles and road infrastructure, in-vehicle human

machine interfacing, traffic and parking management, public transport management and user

information [CEN,2001].

In Germany the Object Catalogue for Road and Traffic Information (OKSTRA) has been

initiated [OKSTRA,2000]. This is an on-going effort in the standardization of road

information exchange between state highway administrations and the divisions of these

agencies. OKSTRA is a conceptual data model, established in 1998, which is especially

focused on the modeling of the external data schemas in the highway administration divisions.

OKSTRA coverage ranges from the road design to documentation of traffic data storage.

The quality, conformance and metadata aspects are also considered by the above-mentioned

organizations. Although there is diversities between established standards in several aspects,

considering the data and the data model quality presence; a general acceptance has been

reached. Positional and semantic accuracy of data should be provided and the conceptual data

models should be complete and logically consistent. Consequently, conceptual data models

designed for GIS-T should ensure data modeling aspects and certificate data accuracy. A

comparison of GIS standards is shown in the Table 3-3.

There are many national and international organizations and communities seeking to establish

standards for GIS, some of which are referred to above, these organizations have different

scopes. Consequently, each concentrates on a particular aspect of GIS. These diverse scopes

are one of the reasons why none of these standards can be applied alone to GIS-T without

adaptation. Specifically there is no standard yet available for GIS-T, which includes all of

topology, linear referencing systems, graphical representation of road segments and

abstraction levels of information issues.

3.3 Current Status of GIS-T at Highway Administrations 34

Table 3-3: Comparison of GIS Related Standards [CASPARY,1998]

FIP

173

SDTS

1988

DIGEST

1997 Ed.2

ICA

1995

CEN

278

1996

CEN

287

1996

ISO

TC 211

1997

lineage

purpose

usage

Meta Data

Sources

Temporal

information/

accuracy

resolution

precision

Clipping

indicator

Meta Data

Model

Abstraction

modifier

Conformance

Data

Specification

Format

Specification

Position

Accuracy

Attribute

accuracy

Semantic

accuracy

Quality

Elements

Accuracy

Correctness

Completeness

Quality

Elements

Model

Logical

consistency

According to established standards, some aspects can be used in GIS-T as a framework.

Among the established standards, the closest to GIS-T are the intelligent transportation

system standards (telematic). This is because the main interest is digital road information and

data exchange between databases. Standards established for the telematic data are presented in

Table 3-4.

Table 3-4: Quality Parameters for Telematic Data [WIDMANN,2000]

Contents criteria Spatial criteria Criteria related with time

Completeness

Example:

Thematic attributes

Example:

Coverage of relevant

area

Example:

Capturing of dynamic

information or time interval

of data capturing

Precision

Scale of

information

structure

Accuracy of spatial

data

Actuality and accuracy of

date and time

Consistence

Independence of

spatial and time

related information

Independence of

contents and time

related information

Independence of spatial and

contents related information

3.3 Current Status of GIS-T at Highway Administrations 35

According to the GDF concept a "single hypothetical database" should have the following

characteristics and components [GRÜNDIG,1989]:

 Positional accuracy

 Be transformable in order to be usable in applications requiring routes and linear

referencing systems.

 Be available in a common, useful data format.

 Be subject to accuracy certification. The status of certification should be an attribute

for each feature.

 Have a permanent feature identifier.

 Have a minimum set of attributes, including permanent feature identifiers, route

numbers, names, road type and metadata

 Maintain topological connectivity

 Have maintenance standards.

3.3.3 Data Models Evaluated

The objective of examining current conceptual data models was to understand the capabilities

and advantages of the established systems. Additionally, the problematic areas facing

highway administrations were determined. As any conceptual data model requires some

degree of interpretation, comparison of these conceptual models is out the scope of this study.

Examples from the USA, Germany, Denmark and the international conceptual data model

GDF will be discussed.

3.3.3.1 A Sample Conceptual Data Model Applied in the United States of America

GIS-T technology is widely used in the USA, especially by federal state highway

administrations. Between 1996 and 1999 the number of GIS applications used by departments

increased enormously (268%). These applications were mainly concerned with base map

production, data management, linear referencing, GPS, information flow between

departments, data distribution and internet technology [GIS-T/ISTEA,2001].

National Cooperative Research Program (NCRP) was created in 1962, which is administrated

by the Transportation Research Board (TRB) and sponsored by the member departments of

the American Association of State Highway and Transportation Officials (AASHTO), in

cooperation with the Federal Highway Administration (FHWA). NCRP project 20-27 was

initiated in response to the need to define the basic structure of GIS-T based on current and

anticipated needs and the characteristics of transportation agencies. During the studies carried

out for that project, it was noted that the two major problems in defining transportation

objects were [NCHRP,1998]:

 Various definitions of road

 Varieties of criteria used to break roads into logical segments.

In order to cover these issues two proposals were made. In their proposal, Dueker and Butler

designed a GIS-T data model which defines relations between transportation data elements. It

is based on a feature (object) database approach with legacy data of varying spatial accuracy.

An alternative approach is a location (geometry) approach as suggested by Sutton (1999).

This alternative was designed to work in an environment where the location of all

transportation features would be re-collected using high precision GPS. This approach focuses

on enabling the linking of spatially accurate tracking or events to a spatially accurate map

3.3 Current Status of GIS-T at Highway Administrations 36

base [TRB, 1992]. Because the Dueker-Butler conceptual data model is designed, not only to

solve problems of linear referencing specifically, but also for the entire agencies, it will be

examined in this study. In addition, various case studies in several agencies were pointed out

in literature with respect to implementation of Dueker-Butler conceptual data

model.[SUTTON, 2000] [VONDEROHE,1997] [BENDER, 1999]

The Dueker-Butler data model is designed for all modes of transport such as highways,

railways, maritime lines, and airlines at universal enterprise level. The proposed data model is

based on the independence of geographic datum, transportation system events, geometrical

system representation and the topology of links and nodes comprising the transportation

system.[DUEKER,1997] The model is constructed through a series of steps. These begin with

the basic elements to which elements accommodating more complex needs are successively

added. The entire database is designed to eliminate the need for complex GIS software for all

methods except to display events and their derivatives using vendors’ dynamic segmentation

concepts. The basic data model is shown in Figure 3-11.

Figure 3-11: The Basic Elements of the Dueker and Butler Model [DUEKER,1997]

In the basic part of the conceptual data model, Jurisdiction is mainly used to represent

political districts. Transportation Feature is an identifiable element of the transportation

system. Within a jurisdiction, the transportation feature is identified uniquely

[DUEKER,1997]. In order to define a transportation feature, the jurisdiction identifier and the

transportation identifier are required. The location where an event occurs is called an Event

Point. The Event Point is defined initially as an offset distance from the beginning of the

transportation feature event, which is a similar concept with a linear referencing system. A

road event is described as an attribute, occurrence, or physical component of a transportation

feature. There are three event subtypes; namely Point Event, Linear Event and Area Event.

Point Event occurs at a single event point location. A Linear Event is a component or

attribute defined by two event points, representing the beginning and ending points. An Area

Event is a transportation feature component or non-transportation event related to a

Transportation Feature.

3.3 Current Status of GIS-T at Highway Administrations 37

Although Transportation Feature possesses a subordinate point, and a linear and area event,

the transportation feature to event relationships go through the event point entity. The use of

only the event point entity is described in Dueker and Butler model thus; “An event is

`owned´ by a transportation feature by virtue of its location on that feature; i.e.., its defining

event point(s)” [DUEKER,1997] .This is because road events are defined using two event

points, being beginning and ending event point similar to linear referencing concept.

In Figure 3-121 the completed basic GIS-T conceptual data model is presented. Further

detailed requirements are realized by introducing entities to represent topology, cartography

and linear datum.

* *

Figure 3-12: The Complete Conceptual Data Model [DUEKER,1997]

Topology’s main element is Traversal which is a part or route consisting of one or more

links. The atomic component of a traversal is the Traversal Segment defined by a link and its

attributes. Additionally, Link and Node topological elements were introduced into the

system.

The concept of the linear datum used by this conceptual data model had been developed by

Vonderohe and Hepworth, 1996. [VONDEROHE,1997] The linear datum was used to reduce

the impact of error sources including multiple network models, defining one linear

referencing system and thereby improve the accuracy of location data collected according to a

linear referencing system (LRS). The proposed linear datum is based on a set of well defined

and precisely located anchor points and anchor sections, to which the linear reference

1 In Figure 3-12 entities with (*)sign are discussed in this study

3.3 Current Status of GIS-T at Highway Administrations 38

measurement methods may be calibrated. Anchor points are one dimensional control points,

which their position can be determined and recovered in the field. Anchor sections are non-

branching linear features, connecting two anchor points, whose real-world length (in distance

metrics) can be determined in the field. Anchor sections are directed by specifying a “from”

anchor point and a “to” anchor point. Anchor sections have a distance attribute, which is the

length measured on the ground. [VONDEROHE,1997] In order to realize this, entities

Anchor Point, Anchor Section were introduced into the Dueker and Butler model.

Cartographic and geometric elements were introduced into the model. Several new entities

were added to support the typical cartographic elements found on transportation maps,

including those created using the dynamic segmentation methods in GIS software. Some of

the introduced entities are: Geographic Datum, Cartographic Datum, Point Feature,

Geographic Point, Cartographic Point, Real-world Location, Area Point, and Polygon. In

order to realize linear referencing requirements, new entities were introduced into the system,

including Line Segment, Base Map String, and Linear Event String.

The reason for using various criteria in order to break roads into logical segments is discussed

as; “In some ways, Traversal Segment is a resolution entity between Traversal, Link and

Linear Event. A Traversal Segment may be used to define super-links by relating non-

contiguous nodes located on the same transportation feature (Butler, 1995)”

[DUEKER,1997].

Within the proposed conceptual data model, the implementation recommendations are also

provided, which will clarify the concepts introduced. Some of them are illustrated here in

Figure 3-13 and Figure 3-14.

Figure 3-13: Implementation Recommendation of Road Events [DUEKER,1997]

3.3 Current Status of GIS-T at Highway Administrations 39

In Figure 3-13 an event_ID is used to identify records in the Event tables, which are unique

for only a given transportation feature. In the first sight, it can be seen that there are some

differences between the conceptual data model and the implemented data model. For

example, the feature Junction was not introduced in the conceptual data model and the Event

feature was initially introduced as Point, Linear and Area Event. This can lead to confusions

during implementation. Additionally, for example, in the implementation proposal adjacent

node identities are stored as attributes within the Node entity. These spatial relationships

could instead have been defined using Link entity and spatial relationships.

Although in the conceptual data model the Link feature was present, in the implementation

tables Node is directly related to the Traversal Segment feature. The usage and

implementation of Link feature is not clear. In the implementation recommendations below

Figure 3-14, the length and direction of measurements are stored in two different tables,

namely Trans Feature/Anchor Sect and Transportation Feature. This is redundant and

will definitely lead to data maintenance complications.

Figure 3-14: Implementation Recommendation of Anchor Section

[DUEKER,1997]

In the Transportation Feature table, the beginning and ending event points are those of the

entire transportation feature. In the Transportation Feature/Anchor Section table, the

beginning and ending event points are for the limits of the subject Anchor Section. The

Anchor Section table includes the beginning and ending anchor point_ID’ s. This redundancy

is required in order to perform maintenance of linear LRS locations of anchor points

separately from the datum entities of anchor points and anchor sections. However, this needs

to be carefully implemented and controlled by integrity constraints.

The main advantages of the proposed conceptual data model are; firstly, it is designed for an

entire agency where all external transportation agency schemas are integrated. Secondly, the

independence of road events, being thematic data, from topology is proposed. Thirdly, the

topology is considered independent of geometry, which in many other data models is ignored.

This approach will definitely bring many advantages for database maintenance and the

simplification of analyses. Additionally, linear referenced transportation information is

conceptually modeled. In the other data models examined, with some exceptions, this issue is

generally ignored.

Open issues concerning the conceptual data model are:

3.3 Current Status of GIS-T at Highway Administrations 40

 Efforts of integrating linear referenced data with other information in the conceptual data

model is provided through another linearly referenced method, although this proposal

results redundancy in both conceptual data model and the database.

 The linear referencing system is based on anchor point-section system. There are several

disadvantages to use only one linear reference system in order to relate road events. This

would require a change in the use of all other custom methodologies in highway

administrations and reduce them into one method. During this conversion process, it is

possible that some data will be lost. This issue will be discussed in Section 4.3.1.

 Anchor points must be complete and their field locations need be determined. This will

incur extra costs for the maintenance of these points.

 The negative impacts of dynamic segmentation, such as the creation of additional entities,

are also carried to the model. Additional entities introduced such as; Line Segment,

Basemap String causes redundancy. Due to the requirements of dynamic segmentation

software, Dueker and Butler conceptual data model illustrate the relationship of base map

strings and the transportation features as many-to-one.[SUTTON, 2000] The many-to-

many relationships can not be implemented.

 The proposal to establish unique feature identifiers, such as identifying them according to

jurisdictions, can cause problems. This is because jurisdiction borders are non-persistent

spatial frames.

 Definitions such as “Event Point, Point Event, Point Feature, Reference Point,

Anchor Point, Node, Geographic Point, Cartographic Point, Point Symbol”, and the

differences between these are confusing. This will cause many problems and redundancy

at the implementation stage.

 The introduction of the Traversal Segment and Anchor Section entities define

topological element Link with different properties. However during implementation a

clearer description is required in order to perform integrity constraints.

 The requirement for transferring topology, imposed by the use of links and nodes to

connect traversals (a.k.a., linear transportation features) to the datum and thence to the

cartography, places a still greater burden on data sharing [DUEKER,2000].

3.3.3.2 The German Highway Administration and A Sample Conceptual Data

Model

Through the introduction of the "Instruction of Road Data Base (ASB)" in September 1968,

the federally uniform structure of road data in Germany was ensured [BMVBW,1998]. ASB is

a very useful and detailed document for the describing of data collection regulations in

highway administrations. Within the ASB, roads are described in terms of construction

information, geometrical elements, type, facilities, length, vertical sections, cross-sections,

materials, width, thickness, and current road condition information. According to the ASB

instructions the “Road Information Database (SIB)” was constituted. Research projects,

including “Revision of structure and contents of ASB(2000) (09.120)”, “Federal-Information

System-Road (98.192)” have continued with the target of extending or restructuring the ASB

according to changes in the basic technical conditions and the growth of new fields like

telematics or information systems [BMVBW,1998].

The general concept of the ASB was developed using road networks as frames, where the

node-link oriented linear referencing system is used. The link-node method uses node

numbers, identified with physical features in the field as an intersection. Links are defined in

terms of beginning and ending nodes, and are the logical connections between nodes. In the

3.3 Current Status of GIS-T at Highway Administrations 41

link-node method, attribute locations along each link of the roadway system are specified and

an event that are recorded are measured as an offset from the nearest node number along a

link. A node map is required in order to be able to identify the location of the nearest node

and in order to track node number records. A link-node method has the advantage of not

needing to maintain reference posts in the field. The disadvantage appears in usage, due to

requirement of a node-map, since there is a lack of communication between third-party users.

Within ASB, firstly, nodes are determined and uniquely identified using the numbering

system. Each link, defined by two nodes, is measured and referenced with a sequential

stationing method, which begins and terminates with a node. There exists four types of nodes

being; planar node (without height difference), non-planar node (with height difference

information), node partly independent of plan and fictitious node.

In highway administrations road data is stored within a relational databases, according to the

ASB regulations. Approximately 600 features are present in databases. In addition to common

information such as pavement type, road length, road type, other detailed information such as

bikeway, tree age and tree type is available. The databases generally used by highway agency

are Oracle and MSAccess. The desktop GIS software MapInfo is commonly used in order to

visualize these databases.

Integration of the road database (ASB) and GIS is mainly realized using links, which are

uniquely defined using assigned nodes. As a digital base-map, 1:10.000 scaled governmental

topographical land registration maps (ATKIS) are used.

Since the ASB is designed as a regulation for data collection, not as a conceptual data model,

the spatial information characteristics were not considered. However, the ASB is often used

by highway agencies instead of a conceptual data model for GIS. Due to this, it is very

common that, information also includes spatial character assigned to links. The spatial

information is not modeled, instead it is directly stored in the database. It is typical that dual

architecture GIS software is used. Because of this, spatial information in the relational

database is generally also redundantly stored in software vendors’ proprietary repositories.

The geometrical information, such as alignment elements, relative height information are

stored in database tables such as the sample shown in Table 3-5. Additionally, although data

is stored in a relational database, relational data modeling techniques are unconsidered. An

example can be given again in Table 3-5, which contradicts to the second normal form.

Table 3-5: A Sample Table From Information System

Road Class Nr.

(LFD)

From Node

(VNK)

To Node

(NNK)

From

Station

Station Length Align-

ment Type Radius

Clothoid

Radius

B 2 B 46 3347012 3347010 0 84 3028 G 0 0

B 2 B 46 3347012 3347010 84 208 3028 R R 600 0

B 2 B 46 3347012 3347010 208 455 3028 G 0 0

B 2 B 46 3347012 3347010 455 575 3028 R R 350 0

B 2 B 46 3347012 3347010 575 597 3028 R A 80 350

B 2 B 46 3347012 3347010 597 617 3028 G 0 0

…. …. …. …. …. …. …. …. …. …. …. ….

3.3 Current Status of GIS-T at Highway Administrations 42

Due to the spatial character of information, GIS-T can not be used in order to integrate

diverse information sources. Although there are efforts to integrate these different sources in

GIS using existing network link information, which is available both in databases and in

digital base-map of GIS, the required efficiency cannot be reached because of redundancy. As

a result, generally both GIS and road databases are used in parallel by highway

administration.

In order to overcome such problems, several projects at the federal state level have been

initiated. One of these projects is the “Road Information Database Nordrhine-Westfalen

(NWSIB)”. It links all data with a highly detailed digital road map and visualize them. The

NWSIB builds on available databases and telecommunications networks. [NWSIB,1998].

The basic component is road, which is ensured by consistent geographical referencing and

monitoring of information. The object model of the NWSIB in its current stage of

development is divided into four basic components: Road network, inventory, geometry and

access rights. The network component does not depend on the inventory component and

forms a basic part of the information system. During the design and implementation phases,

Smallworld GIS system, which has a shell-architecture, is used. The advantages and

disadvantages of this architecture was discussed in Section 3.3.1. The object model’s

geometry relates explicitly to the abilities of the Smallworld software. Relation of objects to

the road system is made by point, line and area characteristics. The history concept, which is

one of the most required information by highway agencies, is modeled in the conceptual data

model. Since in the highway agency GIS applications are implemented using MapInfo, direct

connection of MapInfo to NWSIB is planned. In order to supply data for other applications

with road system references, Smallworld database is connected to an external Oracle

database.

The conceptual data model, which is partly provided by the software vendor, is shown in

Figure 3-161. The symbols used are described in Figure 3-15.

All linear referenced data in the NWSIB has implicit geometry in a planar coordinate system,

indicating position and linear geometry to the road network. Object Link or Shoulder is used

to calculate relationships between stationing and planar coordinate systems [NWSIB,1998].

Current Stationing, Operation Kilometer objects are introduced to the proposed conceptual

data model in order to integrate the linearly referenced data of the ASB with the two

dimensional coordinate system. This is related to the Road Point and Road entities. Road

Point represents point geometry.

Figure 3-15: The NWSIB Conceptual Data Model Symbols

1 In Figure 3-16 entities with (*)sign are discussed in this study

3.3 Current Status of GIS-T at Highway Administrations 43

Figure 3-16: The NWSIB Conceptual Data Model Description [NWSIB,1998]

For the integration of linear referenced data, the data segmentation concept is used. The

advantages and disadvantages of this are discussed in Section 4.3.1. Since a direct

transformation is not available in the system, these information needed to be modeled

separately and integration of these district information must be performed with the use of

additional methods. Additionally, data maintenance must be performed separately. The Route

entity is a compound road element. In the proposed system, there exists integration

possibilities the GDF international standard and the Object Catalogue for Road and Traffic

Information (OSKTRA). Two additional object GDF Junction and GDF Road Element are

* *

* * *

3.3 Current Status of GIS-T at Highway Administrations 44

also incorporated into the data model in order to provide an interface to the GDF topological

structure, since the GDF topological structure is not automatically derivable from available

data. The Road Element entity is related with the GDF Road Element.

Non-spatial and spatial objects are modeled composite. Examples from the provided

conceptual data model, which are designed following this concept, are Segment and Segment

Properties, NetArea and Area Properties and Road Point and Point Properties.

Topological elements are represented in the conceptual data model with various objects such

as Link, Shoulder, Node Point, Null Point, Net Node and Road Node.

The system architecture of the NWSIB is shown in Figure 3-17. It can be described as

follows: Firstly, the ASB conformed road database is imported into the system. In the figure

the ASB conforming database is called database A. It is stored in the software vendors

proprietary database. This holds both geometrical and thematic data, which are labeled here

G(Geometry) and T(Thematic). In this database, A is stored without any decomposition of its

spatial character. It is possible to import other information sources defined as C in the

diagram, which can be both spatial and thematic. Planar geometrical features (G) and

Thematic data (T), are extracted in order to be used for visualization and presentation. Finally,

ASB database (A) is combined with other databases T and G, if additional information is

required. During this process, if there is an update by the user, the updated information is

stored in a GIS database, and decomposed into G and T.

Figure 3-17: The System Architecture of NWSIB [NWSIB,1998]

The main advantage of the NWSIB conceptual data model lie in its integration of data sources

and consideration of history information. Existing information and additional required sources

are integrated into one GIS-T, in order to fulfill entire highway agency requirements. Using

the history concept it is possible to model the current, past and future situations, which is

highly required by all highway administrations. Additionally, compatibility possibilities with

standards used in Germany such as GDF, ASB and OKSTRA, which was in other evaluated

conceptual data models generally neglected, are existing.

The main disadvantage of this proposed data model and system architecture is the lack of

decomposition between spatial and non-spatial information. Although geometrical,

topological and thematic information is modeled in the conceptual data model, without

3.3 Current Status of GIS-T at Highway Administrations 45

conceptual separation of these components, redundancy cannot be controlled. Since the ASB

is the basic information source, the linearly referenced data is stored as thematic information.

With respect to the linear referencing issue, it differs little from the Dueker-Butler conceptual

data model. Only one methodology, the node-link linear referencing method, is selected and

spatial aspects are not considered.

It is expected that several problems will be encountered due to structural differences between

the NWSIB and GDF. In the GDF, two topology abstraction levels were defined. However, in

the described NWSIB conceptual data model, there is no topology abstraction level.

Therefore, the information from one of GDF ’s abstraction levels will be lost. Similar to the

GDF, OKSTRA proposes a conceptual data model which is more detailed than the NWSIB.

For example OKSTRA introduces alignment parameters, which are not included by NWSIB.

In the conceptual data model, network components are described as not dependent on the

inventory component. However, in ASB, inventory components are dependent on network

components due to usage of link-node linear referencing system. Consequently, a very careful

implementation and data maintenance is required with appropriate integrity rules in order to

control the redundancy.

Another problem arises in the conceptual data model during the description of geometry

objects. Although a conceptual data model should be independent of any software vendor,

realization of geometry objects is fully dependent on the software vendor definitions in

NWSIB.

Referring to the system architecture proposed, and illustrated in Figure 3-17, the data flow can

reveal problems. This is because, all information collected by highway agencies is considered

as thematic data. Additionally, in practice the relational database management system

(RDBMS) described in Figure 3-17 as T and A, are generally different databases. Therefore

many various databases and parallel applications such as MapInfo and Smallworld need to be

maintained at the same time.

3.3.3.3 Danish Highway Administration

A nationwide road information system, consisting of interconnected databases, is located at

the Danish Road Directorate. The system holds the information about road and traffic

conditions on national and regional roads. The Road Sector Information System (VIS)

contains tools for extracting and presenting the information. [VIS,1998]

Various databases, each containing specialized information within a particular area, supply

VIS information. The VIS provides map indices from which raster maps can be displayed

according to submitted queries. Generally, 1:200,000 scaled maps are used in GIS

applications. Due to this very large scale, VIS differentiates form the other evaluated

conceptual data models. It is generally used for planning purposes. The road information in

VIS is linearly referenced, based on the stationing linear referencing method. This method

uses the white marker poles along roads. The linear reference system has been standardized

throughout the agency. Measurements are taken by specially developed measuring vehicles,

and data is subsequently transferred from the Danish Institute to VIS. GeoVIS is based on

Mapinfo, which is widely used in the visualization of road databases in Denmark.

In the database, in order to authorize user update-rights, the users’ Internet Protocol (IP)

addresses are available, which was discussed in Section 3.2.4. Some other available

information in the database includes; length, pavement, accident, industry areas, road

alignment, route number, emergency line, date and statistic counting stations.

Road data is one-dimensionally considered and stored in the database using its road identifier

and offset value. In a one dimensional reference system, a road is defined by:

3.3 Current Status of GIS-T at Highway Administrations 46

 Road identifier (ID)

 Links

 Connections to other roads

In order to identify a road in the database, a road identifier naming regulation is used by the

agency. For example:

000/ 0412/ 0 defines Owner / Road number/ Side of the road

Due to the its superior update and system architecture concepts, the Danish system is an

effective system. The Danish GIS-T shares many similarities with above introduced German

system, both positive and negative. It was noted that several databases, GIS systems and other

engineering software are maintained in parallel. These include VIS, GeoVIS, Intergraph MGE

and MS Geographic. Linearly referenced information is stored as non-spatial information.

The naming regulations used in Denmark, will lead to non-persistency, due to identifier’s

spatial frames such as; owner federal country and side of a road, although the road identifiers

used in GIS needs to be unique and persistent. Naming regulations for identifiers are

discussed and a proposal in order to prevent this problem is made in Section 4.4.4.

3.3.3.4 The Geographic Data Files (GDF)

The European Community (EC) has sponsored several research programs to develop

computer-assisted systems for route guidance, travel information, fleet management, traffic

monitoring and many other applications. One of these research programs, the Drive Program,

is concerned with the road infrastructure in Europe. The main object of Drive is to find a

uniform way of collecting and exchanging traffic information between vehicles and the

roadside. Drive has resulted in the establishment of the Geographic Data Files (GDF) standard

which is now extensively used by both manufacturers and users of geographic information

[ERTICO,1996]. The GDF, which was especially designed for vehicle navigation purposes,

not only road objects but also various themes were incorporated in the basic data modeling

concept.

The GDF conceptual data model defines three levels: Level 0 contains all the fundamental

geometric and topological information. The topological information supports planar graph

theory. The form and position of the Level 0 objects are defined in the three atomic units;

Node, Link and Face. Level-I and Level-II hold additional information such as features and

attributes. The objects in these two levels differ in their level of generalization. The simple

features: point, line and area form the basic elements of the Level-I. The Level-II contains

more complex features, which are constructed in terms of the simple features.

In the GDF, link is defined as a multi-abstraction level complex element. For example, in the

case of divided highways it is possible to have eight nodes on one side and seven node on the

other. Although the GDF is still restricted to two-dimensional applications, the possibilities of

combining height information have been investigated. Metainformation is present in the data

model, including areas such as; data acquisition date, accuracy or geodetic information. In this

way quality aspects are covered.

The attribute concept of GDF is powerful, permitting segmentation as well as composite and

time dependent attributes to be built up.[GDF,1996] Four category of attributes are

introduced in the concepts of GDF as follows:

 Simple attribute

 Composite attribute

 Simple attribute with restrictive sub-attribute

3.3 Current Status of GIS-T at Highway Administrations 47

 Composite attribute with restrictive sub-attribute

With respect to this concept, in GDF a hierarchical structure is proposed between attributes.

An example of composite attribute concept GDF feature is Divider, which is defined with

sub-attributes Divided Road Element, Divider Type and Divider Width. This attribute

indicates the existence of a physical or legal divider (solid painted (double) line) along the

center line of a single bi-directional Road Element. All sub-attributes together form the

composite attribute Divider. [GDF,1996]

Attributes, which are called Segmented Attributes, are related to a feature in such a way that

they reference a certain part of it. In the case of line features the particular segment is defined by

a position From and a position To value. These positions represent the curvimetric distance,

expressed in meters. Conversely, the position from and position to values may be left blank to

indicate that the entire feature is subject to the associated sub-attribute. The GDF attribute

concept is presented in Figure 3-18.

Type

Value

Feature

From

Position belongs

to begins at

belongs

to ends at belongs

has

belongs

has

belongs

to has

is a part

consists

Attribute

Position

[Data

type]

Entity type “A” has data type

“Data type” and domain

“Domain”

“A” plays role “x” to “B”

Uniqueness constraint

Exclusivity constraint

Mandatory role

“A” is subtype of “B”

[Domain]

Figure 3-18: The GDF Attribute Concept [WALTER,1997]

The international standard GDF’s data modeling concept is mainly adapted from other GIS-T

data models. With its abstraction level concept and support for the representation of divided

roads, many of the requirements of highway administrations can be fulfilled. Multiple-

dimensions of road geometry information was designed, with the introduced complex

elements. Metainformation was only considered in GDF, which has a vital importance in GIS.

However, there are several problematic areas. Some of these problems have already been

pointed out in the literature. [WALTER,1997] [DUEKER,2000].With respect to GIS-T, some

disadvantages are apparent, mainly;

3.4 Overview of Problems 48

 The GDF is based on planar topological concepts. This implies that every intersecting link

of the planar graph defines a node. The planar topological basis has been widely used

because of its simplicity. However, for applications where road network navigation is

essential, such as the GDF, planar topology has many disadvantages. One of these

disadvantages is that planar topology does not always represent real world phenomena

appropriately. In particular constructions such as bridges, tunnels, overpasses are

problematic. In order to alleviate this problem virtual nodes have been introduced into the

system. These are termed “Brunnel”s from the combination of the words “Bridge” and

“Tunnel”. Such virtual nodes do not provide a complete solution. This topic will be

further discussed in Section 4.2.3.2.

 In the GDF every simple feature is linked with a topology and corresponding geometry.

Topology and geometry are not independent concepts in the GDF. Composition of these

concepts requires additional efforts in data maintenance, in cases such as geometrical

alignments in highway administration, although topology is invariant of geometry.

 In the terminology used by the GDF several descriptions can cause confusions, such as the

terms Node and Link as used for geometrical features. For example in the GDF

segmentation, a node must be inserted at each position where the road width changes

where node term in this case is used instead of event points.

 In the GDF segmentation concept sub-attribute types are introduced, indicating an

hierarchical order. In the case of changes in the main attribute, sub-attribute information

will be disconnected. Additional efforts are required in order to connect this information

again.

 Attributes are assigned using linear referencing methods in one dimension using To

Position and From Position entities along a road segment. Several problems were

encountered, especially during data maintenance, since multi-spatial dimensions and their

integration is not considered. This will be discussed further in Chapter 4.3.1.

 The standard external feature identifiers do not exist for most of the features defined in the

Feature Catalogue, apart from Administrative Areas.

3.4 Overview of Problems

The evaluated countries and conceptual data models present different aspects of the GIS-T

current situation. Although levels of implementation and GIS-T usage vary; some common

problems are:

 Due to diversity of highway administrations tasks, various abstraction levels of road

information needed be coordinated, which was not fully realized.

 Road information data structure analyze is not generally performed. There are various

definitions of road and road logical segments. However, highway administrations

requirements were not fully satisfied.

 A huge amount of data needs to be collected. The existing data is mostly in analogue

format. Data produced by one department can often not be shared with other users, due to

lack of integrated approaches and adequate tools.

 There is insufficient, in some cases no, connection between GIS-T and databases. This

results in;

 Data acquisition and updates were realized separately.

 Many different systems often need be used.

3.4 Overview of Problems 49

 Systems were maintained in parallel.

 No complete overview of the entire agency exists.

 Road databases are not tailored for GIS-T. Road databases were designed and

implemented historically before GIS-T, without consideration of the spatial properties of

road data.

 The necessary attention were not given to GIS data modeling. There is lack of formal

description and documentation of designed data models.

 The difficulties in combining existing data in the database with GIS systems are

mentioned by all highway administrations.

 The definitions and terms used in conceptual data models do not generally match.

 In evaluated countries, linear referenced road information is stored as attributive data. The

integration of the various linear referencing systems used throughout an agency is

generally not considered. Multi-spatial dimensions and their integration is neglected.

 The integration of linearly referenced data with other spatial data is provided using

another linearly referenced method. Additional objects are introduced into conceptual data

models, leading to redundancy. These solutions bring extra costs to implementation and

maintenance.

 Representing road network topology by planar graphs causes many problems due to the

inadequacy of the topological model for representing reality.

 Continuous update involving data entry is required. However, required update

mechanisms and consistency controls have not been established.

 No standard is established with respect to GIS-T.

 For GIS-T purposes several regulations and definitions need to be established. Others

must be redefined such as road identifiers.

The above listed technical issues will be discussed in further chapters of this study with

proposed solutions. In addition to the technical problems, there are organizational problems.

In particular there is widespread ignorance throughout organizations with respect to the

fundamental principles and capabilities of the GIS-T. This issue has shown its effect,

especially during user requirement assessments. However, this problem is correlated with

technical issues, such as:

 Solutions were not provided for the entire agency. Therefore, this technology was not

fully available to all users.

 In system architecture design, other technological possibilities, which can distribute the

information efficiently, such as internet and intranet was not considered.

 In highway administration, user has various backgrounds. Therefore there exists various

level of knowledge with respect to the GIS. This issue could be easily minimized with

successive system designs, including user interfaces, which will lead minimum

requirement of fundamental knowledge. Afterwards, the required minimum basic

knowledge could be supported by means of workshops or other educational activities.

4.1 Criteria for the GIS-T 50

Chapter 4 : Proposal for an Integrated Conceptual Data Model

4.1 Criteria for the GIS-T

Based on the examined standards and conceptual data models evaluated in Chapter 3 an

essential criteria list for GIS-T is constituted. This criteria list covers the problematic areas

detected during the study, and aims to clarify the necessary properties of an successive

highway administration conceptual data model. GIS-T data models should ensure;

 Topological, geometric and thematic information should be conceptually independent.

 Support for multiple topological representations and for various abstraction levels needed

be realized.

 Non-planar topological model should be used.

 Spatial characteristics of road data, such as data collected using the linear referencing

system, should be modeled with independent geometric and thematic character. An

integration method is required.

 Thematic road data should not have a spatial nature.

 Highway administration business-rules should be modeled.

 Existing road information databases should be integrated into the system.

 Metadata, such as consistency rules, quality specifications and history information, must

be incorporated.

 Interfaces to existing standards should be implemented.

 Permanent non-spatial unique feature identifier is required.

 Quick response to queries is needed.

 The conceptual data model should be designed independently of consideration of the

specific software with which it is going to be implemented.

4.2 The Resulting Conceptual Data Model

The advantages of separating geometric and thematic data have been appreciated by the GIS

community from the very earliest developments, which also remains the basic philosophy of

most GIS approaches. However, such decomposition has not been implemented satisfactorily

with the result that the problems associated with data maintenance have not been dealt with in

an appropriate way. This is due to a lack of clarity in the separation of geometric and thematic

data, as well as the widespread practice of integrating geometry with topology. Therefore, the

main proposal of this data model is the conceptual independence of topology, geometry and

road thematic data.

Due to the above observations, the data model is designed with three distinct components;

topology, geometry and thematic road data. The basic component of the proposed data model

is geometry, which is not usual in GIS. In most GIS, geometry plays a secondary role

compared to thematic data. It is conventionally used for graphical visualization and is

redundantly implemented. By considering geometry to be the basic component, many of the

problems noted in Chapter 3 are avoided. Topology component is separately modeled in the

proposed data model. Separating geometry and topology also leads to a consistent data model

and increased topological analyses efficiency. In the proposed data model the thematic road

data has no spatial character in the third-dimension, where today GIS are only capable of

4.2 The Resulting Conceptual Data Model 51

handling two and at most two and half dimension. The geometrical properties of the thematic

data is provided by referencing the fully three dimensional geometry component of the model.

4.2.1 Overview of Highway Information Structure

In order to develop an appropriate highway information structure, a mind mapping diagram

was used to identify clearly the required main components of the conceptual model. The main

information used in the diagram refers to the analysis presented in Chapter 2 and Chapter 3.

Highway information can be structured into four categories; road events, topology, geometry

and metadata. The road events describe the thematic information collected or required by

highway administrations, which is of a non-spatial character. The geometry component is

subdivided into three categories; point geometry, linear geometry and area geometry. The

point geometry is defined in terms of a three-dimensional coordinate system, including height

information. The linear geometry, which is again subdivided as; datum dependent and datum

independent. It is defined in terms its parameterized geometrical elements which are either

horizontal or vertical planes. Area geometry may be either planar or non-planar. According to

the requirements of highway agencies, topology should be defined with two abstraction levels

composed of the topological elements; link and node.

The metadata needed to be included with consideration of the database and data model

standards defined in Chapter 3. Metadata is composed of: integrity constraints, history

components, quality aspects and catalogue. The history component is used by highway

agencies to fulfill the queries for past, current and future information. The catalogue is used to

document the definitions and descriptions of the conceptual data model objects such as road

event. An overview of the highway information structure identified is shown in Figure 4-1.

Topology

Road Events

Geometry

Meta Data

Level I Link I

Node I

Level II Link II

Node II

Point Geometry

Point coordinates

Point height

Linear Geometry

Datum dependent

Horizontal

Vertical

Datum Independent

Line

Arc

Clothoid

.....

Area Geometry

Planar

Non-Planar

Integrity Constrains

History

insert

delete

update

Catalogue

Quality

Highway Information System

Ver. 12 15.09.99

Figure 4-1: The Highway Information Structure

It is shown that all information required by highway administrations can be structured into

four separate categories. The basic idea of the proposed conceptual data model depends on

decomposition of these categorizes, in order to achieve the improved conceptual data model

with well defined data structure and preventing redundancies.

4.2.2 The Data Schema

The proposed conceptual data model, with its relevant objects, is described using the Unified

Modeling Language (UML). Every object is composed of attributes, instances and methods.

According to the UML notation, relations between objects are defined through associations. A

special relationship introduced by UML is aggregation. Aggregation differs from association

4.2 The Resulting Conceptual Data Model 52

in that it introduces a stronger relationship between objects, namely that one object is part of

another object or has a relationship to an entire object. A multiplicity is a constraint on the

number of objects that can be associated with another object. In Figure 4-2, the notation used

is presented.

Aggregation

Association

Multiplicities

exactly n

n to m

minimum n

n...m

n...*

Object Name

Attribute

Methods,Operations,

Roles

Figure 4-2: The UML Diagram Definitions.

4.2.3 The Conceptual Data Model

The proposed conceptual data model is introduced here step by step. Since the geometry is

regarded as the basic element of the conceptual data model, the geometry component will be

considered first.

4.2.3.1 Geometry

The geometry in the data model is composed of three main objects;

 Point geometry object.

 Linear geometry object.

 Area geometry object.

The conceptual data model designates points and linear elements in order to describe the road

geometry. In GIS-T data models, the linear geometry is composed of either polygons or line

strings in the horizontal plane. However, the requirements reported by many highway agency

divisions indicated that it is necessary to model linear geometry by both planar and vertical

elements. Due to this requirement, in the proposed conceptual data model, linear geometry is

defined as being either planar or vertical parameterized linear elements. The vertical linear

elements are defined in an (l, h) system as defined in Chapter 2. The reference system and

transformation object, with its transformation parameters, are introduced. Attributes assigned

to objects will be briefly described in this chapter. More information is also provided in the

implementation explanations in Chapter 5.

Point Geometry is defined using a unique identifier shown as Point_ID in Figure 4-3. Other

attributes shown in the figure are related to topology. These will be explained in the following

section. The other geometrical elements Linear Geometry and Area Geometry are also

defined using their identifiers as LinGeo_ID and AreaGeo_ID.

In GIS data models, information in different coordinate systems are transferred into one

selected coordinate system. As a result the relationship between the reference system and

point geometry is pre-defined as being 1:N. This 1:N relationship indicates that one reference

system can have many points, but every point must be defined in only one reference system.

4.2 The Resulting Conceptual Data Model 53

However, coordinates are properties of geometrical point elements capable of being defined in

any reference system or any projection. It is therefore unrealistic to model points as being

dependent on only one reference system. The conceptual data model should be independent of

pre-definitions by software vendors. More importantly, in order to provide logical consistency

in the conceptual data model, the relationship between Point Geometry and Reference

System is modeled as being N:M.

Figure 4-3: The Geometry Component of the Conceptual Data Model

Reference System is identified by a unique identifier RefSys_ID. Points can be assigned to

many different reference systems and every reference system can have many points. The N:M

relationships can not be directly implemented using current database system technology. This

bottleneck was overcome in practice by the introduction of additional association objects

between related objects. In order to establish a relationship between Point Geometry and

Reference System, an additional table PointGeo/RefSys was introduced. The

PointGeo/RefSys contains; Point identifier (Point_ID), Reference system identifier

(RefSys_ID), point coordinates (X, Y, Z), as assigned by the reference system, and the

standard deviation values of the point´s planar coordinates ( zyx

σσσ

,, ).

4.2 The Resulting Conceptual Data Model 54

In order to transform from one referencing system to another, two separate 1: 0..*

relationships are defined between Reference System and Transformation. These

relationships describe the initial reference system relation and the target reference system

relation. A reference system may have many transformation type and every transformation

must be assigned to one reference system. These relationships are shown in Figure 4-3.

Transformations between reference systems requires the specification of transformation

parameters in the model. This is achieved by assigning a 1: 0..* relation between

Transformation and Transformation Parameters.

In the proposed conceptual data model, each linear element is defined using its parameters

e.g. length, radius. Three types of planar linear elements are defined: line, arc and clothoid.

Roads are represented in the vertical plane by linear elements line and arc. These are pre-

defined by highway agencies as the geometrical road design elements. Each linear element is

specified by references to a beginning and an ending point, as well as a fixed number of

additional parameters. An N:M relationship is described between Linear Geometry and the

both Linear Element Vertical and Linear Element Planar. This N:M relationship means

that a linear geometry may consistent of many linear elements and a linear element may

belong to many linear geometry objects. In order to realize the N:M relationship

LinearGeo/LinVer and LinearGeo/LinPlan, association objects are introduced. The

additional attribute Linear Element Order is defined in order to store sequences of linear

elements in the linear geometry.

Between Linear Element Vertical and Point Geometry two aggregated 1: 0..* relationships

are defined. Every vertical linear element must have one beginning point and one ending

point and every point may be assigned to none or many vertical linear elements. Relationships

between Linear Element Planar and Point Geometry are the same as those for Linear

Element Vertical and Point Geometry.

The planar linear elements have three parameter types; line, arc and clothoid. These parameter

values are introduced by defining the object Parameter. There is association between the

Linear Element Planar and Parameter objects. This 1: 0..* aggregation means that a linear

element planar may have many parameters and every parameter must be assigned to only one

linear element. However, the Linear Element Vertical has only line and arc parameter types.

Considering this, it is not required to model an additional parameter object for a vertical linear

element. These are defined within the linear vertical element using the curvature value.

The geometrical element information is essential, since a parameterized solution using these

elements is proposed in the conceptual data model. The geometrical elements of highway are

defined uniquely with respect to their horizontal and vertical alignment elements. Although

this information is widely used in highway administrations, in many cases it is not in digital

form. Occasionally this information is available in digital form as Computer Aided Design

(CAD) drawings. Although these can be exported into GIS using software vendor provided

interfaces, user defined interfaces and Open DataBase Connectivity (ODBC), these methods

are not yet fully satisfactory. In GIS-T, alignment elements are generalized, depending on

map scales. Therefore, a reverse process can be performed in order to detect the alignment

elements, with the help of adjustment techniques. Consequently, the usage of an automatic

calculation technique, namely Detecting Alignment Elements, is proposed. In the conceptual

data model, in order to detect alignment elements the Detecting Alignment Elements method

of the Linear Geometry object is defined.

This method is based on the detection of alignment elements realized over significant

parameters and elements, with the help of curvature diagram. If the beginning point and

beginning tangent angle is known, with an approximation, alignment element parameters and

their sequence can be uniquely defined. [GRÜNDIG-I,1988]

4.2 The Resulting Conceptual Data Model 55

The curvature diagram is a graphical representation of the curvature ( k ), where ( k ) is

defined with respect to stationing length ( l ) as;

= (4.1)

Therefore, alignment elements in this diagram can be identified with simpler functions being;

straight lines parallel to axis, straight lines not parallel to axis and quadratic parabola.

Relationships between geometrical design blue prints, curvature and angle diagram are

illustrated in Figure 4-4.

lUA

lUPi lUE

lUA

U E

UPI

lUPilUE

Quadratic Parabola

b) c)

Straight

Line

Clothoid Arc

xUA

xUPi

yUPi

U E

ζi

yUE

yUA

xUE

Figure 4-4: Relationships Between a)Blue Print, b)Curvature Diagram and c)Angle Diagram

[MÜLLER,2000]

For each point of the alignment, where the bearing angle is a function of stationing value, the

cartesian coordinates ( x, y, z ) of points of the horizontal (or vertical) alignment results

detecting the alignment elements with the use of simple functions integration. Adjacent

elements have to fulfil conditions of transition in order to enforce the smoothness of the

alignment and of its first derivative. The approximation in the diagram of the first derivative

of the alignment, corresponds to a spline analysis using parabolic curves of second order.

[GRÜNDIG,1992]

The approximation of a sequence of points in the diagram leads to another task. It is necessary

to find out the parabolic curve element to which the point has the closest distance. For every

point, the bearing angle and the distance is required, where these information can be obtained

automatically by means of adjacent points. Since the unavoidable very small distinguishing

errors results undesired dispersions in the curvature diagram, during this process

generalization effect is used. [GRÜNDIG-II,1988] In order to realize this task adjusted spline

analyze with predetermined restrictions is used. Additional constraints for geometrical and

driving dynamics are considered in the adjustment model as observations, in order to achieved

the optimum result.

4.2 The Resulting Conceptual Data Model 56

Obtained results needed to be optimized. In order to perform this optimization, parameters

must be identified. The unique parameters for the straight line, arc and clothoid elements are;

Straight line: l: Length of a straight line

Arc: lR ,: Radius and arc length

Clothoid: lRR fP ,, : Radii of the arcs of the preceding and

following element, the clothoid length

Any linear element between point a and point b can be uniquely identified with coordinates

xa, ya, xb, yb and tangents ta, tb:

ba xparametersfx

)( (4.2)

ba yparametersfy

)( (4.3)

ba tparametersft

)( (4.4)

With the use of available initial values, the functions of ( f ) can be linearized.[BAHNDORF,

1994]. A linear substitute system results which can be solved minimizing a weighted squared

sum of residuals of the parameters in a least squares way. As result of analysis :

 Sequence of alignment elements, element type, radius, stationing values

 Coordinate list of alignment main points with approximation values, tangent

bearings and stationing values for mentioned points

were obtained.

Points sequence, which are representing the road, is required in order to implement this

method. This information is available in digitized maps and can be efficiently used.

Beginning point and beginning tangent angle can be even obtained from large scaled digitized

maps. However, the frequency of digitized points can be insufficient in some places, in order

to detect main points of alignment elements. The appropriate frequency can be determined,

with sampling theorem. A calculation according to Nyquist-frequency ( fN ) is possible.

fN∆

1 (4.5)

(t∆)can only contain full information over the process, if no frequency is bigger than

Nyquist-frequency ( fN ) (TAUBENHEIM 1969) [KULMANN,1996]. In our case instead of

time variable ( t∆), the length variable ( l

∆

) needed to be considered. The required

frequency can be successfully achieved using integration of highway agencies rich data

sources by means of adjustment techniques. This additional sources include; GPS data,

geometrical road design regulations, digital terrain model and orthophotos. The Figure 4-5

illustrates an example where planar linear elements have been obtained by integrating

different sources such as digitized highway network and geometrical design regulations, using

the Detecting Alignment Elements method.

4.2 The Resulting Conceptual Data Model 57

Figure 4-5: Integration of Different Data Sources

There is an N:M relationship between Area Geometry and Linear Element Planar, which is

defined using AreaGeo/LinPlan. AreaGeo/LinPlan also contains the sequence of linear

elements for additional requirements. A similar relationship is also defined between Area

Geometry and Linear Element Vertical using the AreaGeo/LinVer association object.

Geometrical elements defined in the conceptual data model are shown in Figure 4-6.

Linear Geometry

Point

Linear Element Planar

Area Geometry

Figure 4-6: Geometric Elements in the Plane

4.2.3.2 Topology

Three main issues concerning the topology need to be considered during the conceptual data

model design. These are:

 The relationship between topology and geometry.

 Topological abstraction levels.

 Non-planar topology.

With respect to the first issue, although topology is an abstraction of geometry, in many GIS

data models topology and geometry are combined, such as in the example shown in Figure

4-7. In this figure only the red line is available in GIS-T systems, representing both geometry

and topology. In spite of some advantages, such as the reduction of total data storage

requirements and the increasing of query processing performance, many disadvantages arise

with data maintenance. This is especially the case when there are geometrical changes, a

situation which is very common in highway administrations. Since the geometry and topology

are combined, displacements in geometry also effects topology. Although, topology is

invariant under position, orientation, transformation, shape and size, data maintenance is

Obtained

linear

elements at

plain

4.2 The Resulting Conceptual Data Model 58

required both for topology and geometry after every geometrical displacement. Additionally,

the topology needs to be related to one-dimensional, two-dimensional and three-dimensional

space. In order to avoid the above mentioned problems, topology and geometry are separated

in the proposed model. Topology is considered as a logical abstraction of geometry.

Figure 4-7: Road Segment in GIS, with Combined Topology and Geometry

Information

Topology abstraction levels, which is the second issue, has not yet been considered in GIS-T

data models. However, since there are many diverse applications in highway agencies,

multiple representations of topology are required. Two topology abstraction levels are defined

in the conceptual data model. These two topology abstraction levels are shown in Figure 4-7.

Figure a: An Air-photo, Representing Reality

4.2 The Resulting Conceptual Data Model 59

Figure b: I-Level Topology Figure c: II-Level Topology

Figure 4-7: Different Abstraction Levels of Topology

In order to highlight the necessity of multiple topology abstraction levels, divided highways

provide a good example. Since topology is a logical abstraction of reality represented by

geometrical elements, in the case of divided highways in many cases two different geometry

exist for each.

Without a second abstraction level, the divided highway illustrated in Figure 4-8 would be

modeled with one link, representing geometry a and b. Additionally, every piece of thematic

information belonging to geometry a and b should be combined and assigned to this one link.

As a result, the geometrical characteristics of thematic information, such as traffic accident

location can not be identified. This results in a loss and/or mismatching of information in the

data model. In GIS-T, divided highway information is not considered to be a basic

requirement. However, it is certainly true that the modeling of divided highways is inevitably

required by all highway administrations, therefore, the conceptual data model must fully

support multiple topological abstraction levels. When this becomes the case, the second level

topology can be implemented without any radical change in the data model. This is also

required for analysis such as; route planning and intelligent transportation systems (ITS).

Link

ode

Geometry

Figure 4-8: Sample of a Divided Highway

4.2 The Resulting Conceptual Data Model 60

The third issue is non-planar topology. In the conceptual data models evaluated, all topologies

were planar. With planar topologies links cannot cross each other without creating an

intersection. The crossing links must therefore be split into several individual links. Planar

networks have many advantages, principally that they are common and simple. However, this

does not reflect the reality for road networks where links can cross without creating

intersection. This is the case for example with bridges, tunnels, overpasses, underpasses and

viaducts.

Highway administrations commonly use linear referencing methods including the link-node

method. Therefore, there is a necessity to differentiate “real” nodes from “virtual” nodes

which are generated due to the usage of planar topology. Due to this, as well as other reasons

which were described earlier, virtual nodes were introduced by all the conceptual data models.

However, this solution is inefficient in practice due to the high level of data collection

requirement. In the conceptual data model developed, non-planar topology was implemented

in order to avoid these problems.

In order to implement such non-planarity, use is made of the third dimension namely height

information. With the designed objects Linear Element Vertical and Point Geometry, and

Linear Geometry object methodology Detecting Alignment Elements, non-planar topology

third-dimension is achieved in the conceptual data model. Since the geometrical vertical

alignment elements were detected using the Detecting Alignment Elements, vertical

alignment geometry information is available as “build-in”, where these are illustrated in

Figure 4-9.

ode

Point

Link

Linear Element

Vertical

Figure 4-9: Non-planar Topology

In order to implement this information, there are diverse height information sources, due to

departments’ usage purposes. Since the relative height information is generally measured by

means of surveying techniques, mainly leveling and real time GPS, obtained accuracy is

adequate to fulfill user requirements such as; freight transportation and ITS technology.

Several other solutions, which adjustment techniques can be applied, in order to achieve non-

planar topology was introduced in Section 4.3.3.

The topology component of the proposed conceptual data model is shown in Figure 4-10.

The main elements of topology are Node and Link as described in Chapter 2. According to

these definitions two different relationships between Node and Link are defined. Firstly, in

order to define a link, beginning and ending nodes are required. Secondly, a node can be

4.2 The Resulting Conceptual Data Model 61

assigned to many links. These relations are shown in the data model with two 1: 0..*

aggregation associations. Associations between Node and Link are applied to both

abstraction levels, using the same aggregated associations.

Figure 4-10: The Topology Component

Relationships between two abstraction levels are modeled as follows:

1. Node I, being a higher abstraction level may be composed of Link II and Node II.

Between the first level topology element node (Node I) and second level topology

element node (Node II) a 1: 0..* relationship is assigned. This relationship maps

reality adequately because Node I may be composed of many Node II and every

Node II is assigned to the first level topological object Node I.

2. The relationship between Node I and Link II is modeled as 0..1: 0..*, where Node I

may be composed of second level links (Link II) and a second level link (Link II)

may be assigned to Node I. Road junctions are examples of such situations.

GEOMETRY

TOPOLOGY I TOPOLOGY II

4.2 The Resulting Conceptual Data Model 62

3. The relationship between Link I and Link II is modeled as 0..1: 0..*, where Link I

may be composed of second level link’s (Link II) and a second level link (Link II)

may be assigned to Link I. The merging and subsequent separation of divided

highways is an example of this.

Using the above defined relations, other required information can be extracted. Relationships

between topological and geometrical components are described using the following

associations.

The Node object is represented at the geometrical level by a point. Node I and Node II have a

0..1: 1 relationship with respect to Point Geometry. A node must be represented with a point,

but a point need not be a node. This relationship is valid for both abstraction levels. Linear

Geometry has a relationship between the two abstraction levels; Link I and Link II.

Between Link (Link I and Link II) and Linear Geometry a 0..1: 1..* relationship is

assigned. A link may be composed of many linear geometry and a linear geometry may be

assigned to one link. The N:M association between topological element Link and Area

Geometry is realized using association tables LinkI/AreaGeo for the first level, and Link

II/AreaGeo for the second.

External objects, administrative areas and road objects, they are assigned to the topology

component Link I. The object Administrative Area include governmental borders, highway

division borders or national borders. The Road object only stores road names or codes

provided by highway administrations, as attribute. Both objects are only associated with first

level topology. The Administrative Area and Link I objects have an N:M association,

realized using the AdministrativeArea/LinkI table. In the data model links are not

dependent on administrative areas. A link may be assigned to many administrative areas and

an administrative area may contain many links. The Road object has an N:M association with

Link I. A link may have many names and/or codes. Many roads may be identified by one

link. One road may reference many links. The association table Link I/Road is introduced in

order to implement this N:M association. It also stores the sequence of links along the road in

order to simplify possible queries.

The topology and geometry components introduced are illustrated in Figure 4-11.

Linear Geometry

Point

Linear Element Planar

Area Geometry

ode I

Link I

Figure 4-11: The First Level Topology and Geometric Elements

4.2.3.3 Road Events

The Road Event object stores all non-spatial information including the attributes, occurrence

and physical components of the road. The Road Event is assigned to the geometrical objects;

4.2 The Resulting Conceptual Data Model 63

Point, Linear Geometry or Area Geometry. Examples of typical road events are: accident,

project, road facility, video image, pavement type and road type. Road events could be

assigned to:

 A Point Geometry such as a traffic sign, a traffic accident or a maintenance

workstation location.

 A Linear Geometry such as project information.

In addition to these, events can occur at the same time and in the same location. In this

component of the conceptual data model Road Event, Road Event Properties and

associations between the geometry components are introduced. These are illustrated in Figure

4-12.

Figure 4-12: The Thematic Component of the Conceptual Data Model

In the Road Event object all non-spatial information needed to be stored according to the

provided road event identifiers in the event catalogue. All the road events with their

associated geometrical types, needed to be identified and documented in this catalogue in a

standard way agreed throughout the entire highway agency. Having agreed on the standards

for identification, associations between road events and geometrical elements can be made.

THEMATIC

GEOMETRY

4.2 The Resulting Conceptual Data Model 64

There is a 0..1: 0..* relationship between the Geometry (Point Geometry, Linear Geometry

and Area Geometry) and Road Event objects. These relationships mean that every road

event should be assigned to one of the geometry types and many road events may be assigned

to one geometrical object.

The Road Event object has a very important method, called Dynamic Reference

Transformation. This method is defined to transform linearly referenced road data into three-

dimensional coordinates, as well as to re-transform it back into one of the user linear

referencing methods, if required. Since this information is stored in a three-dimensional

system, it can be transformed to any one dimension system. Usually, in highway agencies this

will be a linear referencing system. This methodology is explained in Section 4.3.1.

All non-spatial properties of Road Event are stored in the object Road Event Properties.

This object has a tree structure with which it is possible to expand properties of road events to

any desired level. An association between Road Event and Road Event Properties is

introduced here as 1:0..1. This means that every road event may have many road event

properties, or the necessary information is provided in the catalogue and every road event

property must be assigned to only one road event.

With the objects introduced above, the requirements of highway agencies are satisfied in

multi-dimensional space, including planar and vertical sections. Additionally, non-spatiality

of road data is achieved.

In Section 3.2.2 four spatial reference systems were described. The remaining system which

has not yet been introduced to the conceptual model is cross-sectional system (h, q). This can

not be realized using just the introduced objects. The cross-section reference system is

discussed in Section 4.3.2 in more detail. In order to realize the cross-sectional spatial

information, five other objects are introduced into the data model. These are shown in Figure

4-13.

In order to realize this task, the Road Event object is divided using a generalization

relationship into two objects. Generalization relationships declare that objects are fully

consistent with the super-class. These are Dimensional and Non-dimensional.

In the object Dimensional, the cross sectional reference system and the reference system

parameters are identified. After the determination of the reference system, the existing three

dimensions are introduced. These are named; Zero Dimensional, One Dimensional and Two

Dimensional. Between the object Dimensional and the objects Zero Dimensional, One

Dimensional and Two Dimensional there exist 1: 0..* associations. These associations mean

that every Dimensional object may have many Zero Dimensional, One Dimensional or

Two Dimensional objects, and every object must be associated with a dimensional object.

Other associations are defined:

a. Two 1: 0..* aggregated associations between Zero Dimensional and One

Dimensional objects exist. This means that every One Dimensional object is defined

by a beginning and ending Zero Dimensional object, and a Zero Dimensional object

may be assigned to many One Dimensional objects.

b. The association between the One Dimensional and Two Dimensional objects is

N:M, realized using the association table Onedim/Twodim.

4.2 The Resulting Conceptual Data Model 65

Figure 4-13: The Conceptual Model For Cross - Sectional Spatial Information

After having introduced the above objects to the conceptual data model, the reference systems

described in the information analyses Section 3.2.2 can now be fully mapped. This is shown

in Table 4-1 .The two empty boxes are obtained using other associations.

Table 4-1: The Matrix of Road Reference Systems and Geometry

X, Y, Z h, l h, q

Point

Line

Area

An overview of the proposed conceptual data model is shown in Figure 4-14. In order to

simplify the diagram, only those specifically generated methods, Dynamic Reference

Transformation and Detecting Alignment Elements, are shown in the diagram.

4.2 The Resulting Conceptual Data Model 66

Figure 4-14: The Overview of the Proposed Conceptual Data Model

In the proposed conceptual data model; topology, geometry and thematic data are modeled

independently. Using two abstraction topology levels, multiple topologic representations may

be realized. In addition, with relationships assigned between objects, multiple geometry can

GEOMETRY

THEMATIC TOPOLOGY

I. LEVEL

II. LEVEL

4.2 The Resulting Conceptual Data Model 67

be associated with a single topology. Although planar topology is simpler to both create and

maintain, a non-planar topological model was implemented with newly introduced height

information and required method in order to better model reality and to appropriately satisfy

user requirements. Various spatial characteristics of road data, such as linear referencing

systems, are non-redundantly modeled using the introduced methods. Agency business rules

are introduced into the conceptual data model. Thematic information is modeled separately

from the geometry and topology on the basis of a detailed examination of the spatial

characteristics of the road data structure. The modeling process is carried out independently of

software vendors proprietary systems.

By applying the proposed conceptual data model:

 All the varied highway agency views were implemented.

 Geometrical, topological and thematic information was modeled transparently.

 Spatial information was referenced using the geometry element: point.

 Multiple topological levels are designed.

 Non-planar topology is implemented.

 A modular structure was designed.

 A dynamic transformation between three dimensional and one-dimensional coordinate

systems was defined.

4.2.4 Completing Conceptual Data Model

In order to simplify the conceptual data model diagram, the metadata components of; integrity

constraints, history and quality will be considered separately.

4.2.4.1 Integrity Constraints

Several agency rules, for instance the relationship between Administrative Area and the

topological element Link, have already been modeled in the conceptual data model, others

have not been defined yet. In order to implement agency rules in databases, a variety of

techniques are provided based on consistency conditions and integrity constraints.

Constraints are functional relationships between objects, object attributes and associations.

[RUMBAUGH, 1991] In general, two types of integrity constraints can be distinguished.

These are static integrity constraints which define the valid state of a database, and dynamic

integrity constraints which are the conditions on the allowable transitions from one database

state to another. [MOLENAAR,1998] While static integrity constraints validate completeness

and correctness of the current data in the database, dynamic integrity constraints are used

during local updates.

In GIS, due to redundancy, integrity constraints are required wherever geometry interacts.

Due to the limitations and requirements of current GIS software for data modeling, many

additional integrity constraints are required in order to validate geometry and topology. For

example, defining geometrical elements via their parameters would avoid many redundancies.

This is because only beginning point, ending point and parameter values are required to

generate the linear element. However, these parameterized elements cannot be visualized in

GIS. Since, every visible feature in the system has to be stored with its geometrical

coordinates. In this case, for visualization, the point coordinates of the polygon need to be

accessed to provide the geometry of the linear element. Therefore, additional methods need to

be designed in order to satisfy user requirements and control the redundancy.

4.2 The Resulting Conceptual Data Model 68

In Chapter 2, encapsulation concept was described, as permitting the design of spatial and

non-spatial data as well as methods within each object and importance for GIS was

highlighted. However, with relational and object-relational databases the encapsulation

concept is not yet fully realized. As a result, these required methods need to be defined

separately and controlled by the user.

Especially in the case of highway administrations, the maintenance of the spatial aspect of the

data in the conceptual data model is difficult. This is due to the complex relationships

between objects, the complex business rules and other prerequisites. The spatial aspects of

information must be validated whenever objects are updated. This task is automated when the

defined methods are available. Relationships between geometry and topology can be used to

formulate consistency constraints for spatial databases.(Hadzilacos and Tryfona 1992, Kainz

1995 and Plümer 1996). [MOLENAAR,1998] [GRÖGER,2000] During the transaction

process for topological updates, three actions are possible. These transactions are insertion,

deletion of a link and changing the geometry which the link is assigned to.

In Figure 4-15 an example of a simple operation is shown. A node has been displaced due to

data quality improvements or correction of errors. Therefore, the possible discontinuities and

errors in the topological and geometrical elements must be controlled.

ode

Link

Point

(Geometry Element)

Linear Elemen

Figure 4-15: A Transaction With Sample Objects

The transaction of inserting a link is as follows:

Transaction: Insert a link

Conditions:

a. Two nodes, the beginning node and ending node, are required.

According to the conceptual proposed data model it is not possible to

insert a link without the prior existence of the two nodes.

b. The beginning point of the first geometrical element and the ending

point of the last geometrical element should correspond with the

beginning and ending nodes.

Action: The link is inserted as a logical connection between both the two nodes

and the corresponding geometrical element points.

4.2 The Resulting Conceptual Data Model 69

The transaction of deleting a link is as follows:

Transaction: Delete a link

Conditions:

a. Two nodes are required in order to define a link.

b. A node may be assigned to many links.

c. A link has an association between geometrical elements.

Action: Search for the beginning and the ending nodes of the link. If either of

these nodes are assigned to only this link then delete the node in

question. No action is taken for nodes which are assigned to more than

just this link. Search for geometrical elements assigned to this link and

delete them.

The transaction of changing the geometry which the link is assigned to;

Transaction: Change geometry of link

Conditions:

a. Topology is invariant of geometry.

b. The beginning point of the first geometric element and the ending

point of the last geometric element must correspond with the

beginning node and the ending node respectively.

Action: Identify in which element geometrical change takes place.

 If the geometrical displacement takes place at the first or the

last element, determine whether the beginning point of the

first element or the ending point of the last element is

changed.

i. In the case of a displacement which modifies a point

assigned to a node, the user must be warned of the

necessity of running other consistency checks, such

as geometrical element consistency.

ii. In case where the displacement is of a point which is

not assigned to a node, no action is taken.

 If geometrical displacement is not in the first or in the last

geometrical element take no action.

Some other examples of integrity constraints used in the implementation are provided in

Chapter 5.

4.2.4.2 History

Highway administrations need to be able to track the changes in the system over time. This is

important for the preparation of documents such as annual reports or yearly plans. It is

additionally necessary to be able to reconstruct the state of the system at any specified point in

time. The visualization of such “rolled-back” system states is essential.

In the conceptual data model, it is proposed to model history as an object. This means the

history for all objects, including relationships between each other, are stored in one history

object. The transaction log approach is adopted. [DORSEY,1999]. By using the history object

it is possible to report or re-create a required transaction.

4.2 The Resulting Conceptual Data Model 70

In order to track the object history in the conceptual data model, the additional objects;

History, Value Range, Event Properties and Event Type were introduced. In order to

illustrate the approach used, the Link I object from the conceptual data model is presented as

an example.

The object Link I has three attributes:

 Link I identifier, data type declared as long and name LinkI_ID, since link identifier

is not automatically generated, but externally identified.

 The beginning node identifier (Node I), which is assigned to Link I, data type

declared as long and name as NodeI_IDBegin

 The ending node identifier (Node I), which is assigned to Link I, data type declared

as long and name as NodeI_IDEnd

The Link I object is represented in Figure 4-16.

LinkI_ID

NodeI_IDBegin

NodeI_IDEnd

Link I

Figure 4-16: Attributes of the Object Link I

Objects, History, Value Range, Event Properties, Event Type and their relationships are

shown in Figure 4-17.

Link I History Event Type

*11*

0..1

Event

Properties

Value Range

0..1

Attributes

(LinkI_ID

NodeI_IDBegin

NodeI_IDEnd)

Figure 4-17: The History Component of the Conceptual Data Model

The Link I object has a 1:N relationship between its attributes. A 1:N relationship between

Link I and its attributes ( LinkI_ID, NodeI_IDBegin, NodeI_IDEnd) means that every object

must have at least one or many attributes, and every attribute must belong to only one object.

The Value Range object contains a pre-defined set of attributes. In the case of Link I, these

are the set of identifier values defined by the highway agency. A N:0..1 association exists

between Attributes and Value Range. This means one attribute may have at most one pre-

defined value range, such as link identifiers, and a value range can be defined for many

attributes. The Event Properties object stores only the altered values. Additionally other

4.2 The Resulting Conceptual Data Model 71

required information such as operator details and time records, which are required in order to

control sequence of history event and reconstruct it when is it necessary, is kept in this object.

The Event Properties object has a tree structure which can be expanded or collapsed as

required. The association between Range Values and Event Properties are mapped as N:

0..1, the same as the association between Attributes and Range Values. This is because the

character of the information has not changed, only the attribute value was altered. In the

History object, the history identifier, which should be automatically uniquely generated, and

the object identifier, in this case Link I, are stored as the attributes. The association between

History and Event Properties is 1:N. This means that a History has many event properties,

and every Event Property must be assigned to only one History. In order to track the

changes, the Event Type defines which type of action has been performed, such as updating

or deleting. A 1:N association exists between Event Type and History. One event type can

have many history objects, for instance an operator can update many objects at one time

(before the transaction), and every history object is associated with only one event type. All

these processes are controlled by transaction rules.

4.2.4.3 Quality Aspect

Due to the ever increasing requirements for data integration and data exchange, data quality

aspects have become more important. For all data and applications, a minimum required

quality criteria exist, such as positional accuracy or specific semantic attributes. These criteria

are generally dependent on the use of data and defined by the administrative regulations, data

collection techniques, data structure characteristics as well as standards. However, during

evaluation of the current situation in highway agencies, it was identified that information

stored in the system can be inaccurate; in many cases attributes are missing or incorrect.

There are many reasons for this situation, such as:

 Lack of data maintenance and an appropriate maintenance tools.

 The high economic cost for data acquisition.

 Information obtained from varied sources can differ when it has not been structured

and integrated.

 Automation has not been fully applied, especially in the areas of data entry and

acquisition.

In addition, due to variations in user requirements and in the generalization level of reality,

the required data quality also differs considerably. Therefore, it is necessary to provide

existing information quality, in order to evaluate the appropriateness of analyses and to guide

other possible users. As a result, data quality certification is highly attractive to highway

agencies, especially with respect to data integration and data exchange. This needs to be

solved in an integrated GIS-T approach for each entire agency. Only by doing so will the

user’s expectations be fulfilled.

In the conceptual data model the quality aspect was implemented using member methods of

the individual objects. The quality aspect was not modeled in the proposed conceptual data

model in the same manner as was done for history. This is because it is very unlikely that

errors or poor quality data needs to be regenerated. Using this approach the quality of the

current data is reproducible at any time in the form of documents or tables. As an example, in

the proposed conceptual data model, the standard deviation values ( zyx

,, ) were stored in

object Point as attributive information and appropriate methods were provided in order to

regenerate positional accuracy using adjustment techniques.

4.3 Special Aspects of Highway Information System 72

4.3 Special Aspects of Highway Information System

4.3.1 The Linear Referencing System

Depending on the required use, the real-world phenomena road is multi-dimensionally

defined:

 one-dimensional, linear referencing system

 two dimensional, planar coordinates

 three dimensional, planar coordinates and height information

 four dimensional, time in the case of dynamic objects

However, in many cases this variety is not fully supported by GIS-T. This is due to the

fundamental problems already highlighted including the lack of an adequate analysis of the

road information structure or limitations in the GIS software. One of the main aspects

discussed in this study is the integration of one-dimensional data with three-dimensional data.

Some of the main problems in the management of linearly referenced data include;

 The geometrical state of highways is subject to progressive refinement. During this

process all linearly referenced data changes its reference due to the new geometry and

it is possible for existing data connections to be lost. An example is illustrated in

Figure 4-18. After realignment, the traffic accident at kilometer (2.2+33.314) would

appear to have occurred along the new stretch of road. This is completely incorrect,

and because the location of the accident is now not part of the road network, the road

event must be deleted from the current system. Similarly, the traffic accident at

kilometer (3.4+63.886) would “move” if its linear reference were not to be updated. It

must therefore be corrected according to the new kilometer value. Additionally, other

situations such as; corrections in length due to improved accuracy following re-survey,

or topological modification also cause similar data maintenance complications.

2.2 + 0.00

2.3 + 0.00

??? ???

Traffic Accident

Figure 4-18: The Road Information After Re-alignment

4.3 Special Aspects of Highway Information System 73

 There is no agreement or common linear referencing methodology for road events.

The linear referencing systems used varies from country to country, even from road

department to road department. This results in communication and data exchange

problems within highway agencies, due to not integrated various methods.

 Reference points in the field are often poorly maintained due to high costs.

 The more accurate, rapid and economic solutions provided by techniques such as;

GPS and photogrammetry, cannot be easily combined with linear referencing methods

due to their three-dimensional nature.

 Enormous effort and planning is required in order to keep the road data consistent.

Efficient usage of GIS-T can only be realized when the different methods applied for linear

referencing are integrated. These include; route-mile-point, stationing, link-node, control-

section, route-reference-post offset, and route-reference point-offset. Users unfamiliar with

linear referencing systems have difficulties using the provided data, leading to many problems

if data is planned to be shared with other organizations such as the ministry of health, national

police, cadastre and the private sector. However, road administrations point out that even if

other data acquisition techniques such as GPS and remote sensing become widespread,

linearly referenced data collection will remain common. This is because the linear referencing

method is the most “natural” way of collecting linear object road information.

Several solutions exist in order to overcome this situation. The current solutions include:

1. Generic road data models

- Beginning with the initial NCHRP 20-27 project, the research evolved into the

GIS-T Linear Referencing Pooled-Fund Study (Fletcher et al., 1995), and then

to the current NCHRP 20-27(3) project which proposes guidelines for the

implementation of multi-modal transportation location referencing systems.

2. Framework road data models

- The NSDI framework is intended to enable integration of disparate data sets. A

more formal model for classifying road features has been proposed by Dueker

and Butler.

3. Location approach

- An alternative approach can be characterized as a location (geometry)

approach [SUTTON, 2000]. This approach embraces the use of two and three-

dimensions, such as those that might be derived from a series of GPS-derived

coordinates.

With reference to the solutions proposed above, GIS vendors have developed proprietary data

models covering linear referencing issues. The two most well known models are ESRI’s

Arc/Info Dynamic Segmentation module, and Intergraph’s MGE Segment Manager method.

Both software products use different methods to manage and query linearly referenced data,

reflecting their underlying data models for managing spatial data and topology. [SUTTON,

2000]. Additionally, Oracle have developed its own methods of linear referencing.

Dynamic segmentation associates network attribute databases that are linearly referenced with

topologically structured spatial databases (network models) whose reference frameworks are

coordinate-based. To avoid the need for explicit representation of all point features and

segment boundaries within the spatial database, dynamic segmentation computes coordinates

from linear references “on-the-fly”. Dynamic segmentation and network overlay enable

spatial analysis and integration of highway inventory databases and any other databases that

are linearly referenced.[VONDEROHE ,1993] Using dynamic segmentation, the basic

edge/node structure is preserved, but a structure is added above it. This represents the two

types of discrete entities, zero dimensional and one-dimensional, located at arbitrary locations

4.3 Special Aspects of Highway Information System 74

on the network.[GOODCHILD,1996] Although these methods provide partial solutions, some

of problems remain unresolved, and additional requirements arise. These issues include

 All approaches, with the exception of the location approach, do not provide solutions for

the integration of one-dimensional data with two or three-dimensional data. In these

approaches, one linear referencing method is selected. The different linear referencing

methodologies used by highway administrations are not considered. By standardizing on

a single referencing method within an agency, the data held in the obsolete formats will

either be lost or else must be converted.

 The linearly referenced data is stored as attributive data or in an explicit object without

any spatial character ,such as operation kilometer or anchor section, in the database.

Although dynamic segmentation is realized “on-the-fly”, for representation purposes,

efforts of maintaining non-integrated spatial information sources do not lessen. Different

data sources must be maintained parallel.

 In addition, the solutions require very high levels of data acquisition activity. During this

research it was discovered that, at least in Turkey and Germany, many of the referencing

points have been lost through time. The implementation of GIS-T with the proposed

solutions can only be realized after the determination of every referencing point on the

highway network. Application of this method will lead to extra cost and a delay in

implementation.

 The location approach requires a full re-digitizing of the road features using GPS. This is

needed to enable linkage by coordinate snapping of spatially accurate tracking, or events,

to a spatially accurate map base. “However, this approach has not been formally stated or

tested. Issues, such as repeatability of GPS positions, how to abstract networks, how to

relate to other location referencing systems, and representation at smaller scales have not

been adequately addressed”. [DUEKER,2000]

 The dynamic segmentation solutions offered by GIS vendors are dependent on software

platforms. In order to use one of these solutions, the core modules of the appropriate GIS

vendor’s software needs to be deployed at the agency. Additionally, GIS vendors have

prerequisites with respect to the conceptual data model design, which can limit the data

modeling scope. Existing databases therefore need to be modified to take into account the

prerequisites of the software data model.

According to research carried out in the USA, there is on average a 50m error (800 m in the

worst case) in translating a two dimensional coordinate into a linearly referenced

system.[NORONHA,1999] This is caused by alignment errors and generalization in the

databases. Due to disagreement in linear measurements, there is a further 50m error caused by

communicating the location to a linear reference in a different database. In other words, the

offset of a fixed point on the ground could be 800m with respect to one database, and 850m

with respect to another.

Due to these problems, the usage of a three-phase transformation method is proposed for the

integration of linearly referenced road data and three-dimensional coordinate system.

[GIELSDORF,1998] This transformation was introduced to the conceptual data model as the

member method Dynamic Reference Transformation of the Road Event object.

When using the Dynamic Reference Transformation methodology, spatial data, including

linear referenced data, is integrated into the system in three dimensions. All data can be

modeled without redundant storage of geometrical information. As all linear reference

systems are based on the specification of the direction and distance from a known point to the

unknown point, every linear reference system can be identified using ( l, q ), where ( l ) is the

distance between the origin and the beginning point of the road event, and ( q ) is the normal

4.3 Special Aspects of Highway Information System 75

to the linear element. Since highway administrations consider the use of linearly referenced

methods to be inevitable, transformation from ( x, y, z ) to ( l, q ) systems is required. Another

very important point is that during this transformation, linearly referenced data is not stored in

one-dimension, but rather, in three-dimensions. Transformation needs to be realized

dynamically, in order to prevent data inconsistencies and redundancy. In the proposed

conceptual data model, the geometrical properties of a road axis from the ( x, y ) plane are

described using the linear elements; line, arc and clothoid in three dimensions. For the

transformation between the ( l, q ) system and ),( yx

′

, the value of curvature ( k ) and the

bearing angle (

), which are both functions of the length, need to be defined.

The curvature ( k ) of the linear element is;

Line:

k (4.6)

Arc:

constant

k== 1 (4.7)

Clothoid:

LRRR

AEA

1111 ⋅











−+= (4.8)

Bearing angle (

) for linear elements;

Line:

τ (4.9)

Arc:

⋅

(4.10)

Clothoid:

()

kklkl AEA 2

)(

⋅−+⋅=

(4.11)

In Figure 4-19 the relationship between the three-dimensional and the linearly referenced

coordinate systems is shown.

x′

y′

l p

Figure 4-19: The Relation Between Coordinate Systems[GIELSDORF,1998]

The transformation can be realized in three steps:

STEP 1:

4.3 Special Aspects of Highway Information System 76

( x, y ) point coordinates in a three-dimensional coordinate system are transformed into a local

two-dimensional cartesian coordinate system ),( yx

′

using a four parameter transformation

[BAUMANN,1995]. The local origin is defined to be the linear element’s beginning point. The

x axis of the local system is defined to be parallel to the element’s tangent at the origin. For

transformation, the identical points in both systems are the element’s beginning and ending

points. With reference to Figure 4-19;

PAP xAxx

′

. (4.12)

)(.

APP xxAx −=

′− (4.13)

Transformation matrix: 









−











−

εε

cossin

sincos

A (4.14)

where,

The scale factor is 22 oam += (4.15)

and the rotation is 











arctan

(4.16)

STEP 2:

Transformation of an ( l, q ) system into an ),( yx

′

system or the retransformation of an

),( yx ′′ system into an ( l, q ) system.

a) In order to transform a linearly referenced system ( l, q ) into an ),( yx ′′ system,

firstly the ),( yx ′

′

coordinates of point F should be calculated. F is on the linear

element and perpendicular to point P. The point P is defined by ( l, q ) coordinates

with respect to the linear reference system.

dllxd )(cos

= and dllyd )(sin

(4.17)

The coordinates of ),( yxF ′′ are given by:

dllX

F∫

)(cos

and dllY

F∫

)(sin

(4.18)

This is illustrated in Figure 4-20:

τ dy

Figure 4-20: dx, dy Shown for a Linear Element Clothoid

Depending on the linear element type, integral solutions are:

For line lXF= and 0=

Y (4.19)

4.3 Special Aspects of Highway Information System 77

For the arc R

RXFsin.= and R

RYFcos.= (4.20)

For the clothoid a direct solution is not available. Two alternative solutions exist.

i) Series expansion: [MÜLLER,1984]

......

345640 8

lXF+−= and ......

422403366 10

YF+−= . (4.21)

where lRA ⋅=

2 and the origin is defined by RA = 0.

ii) Numerical integration:

The approach taken here is to solve the problem using by discretising the continuous

method )(xf and directly integrating the discrete known methods thus:

∫

dxxf )( (4.22)

The most straightforward numerical integration technique uses the Newton-Cotes

formulae (also called Quadrature formulae). Other similar techniques are the

Trapezoidal Rule and Simpson’s Rule. Numerical integration is more suitable for

computational implementation.

The final part of Step 2a is to determine ),( yx

′

coordinates for point P:

)(sin lqXX PFp

⋅−= and )(cos lqYY PFp

⋅+= (4.23)

b) In order to transform ),( yx

′

′ into a linear referencing system ( l, q ) of the linear

element, The origin of ),( yx

′

′ system is situated on the origin of the linear element.

The l-coordinate corresponds to the distance of the point along the element and the q-

coordinate represents normal to the linear element.

F is on the linear element and is perpendicular to P as illustrated in Figure 4-19. The

distance between P and F must be minimized.

min)( !

=lq (4.24)

Therefore

min)()( !

2== lflq (4.25)

0)( =

′

=lf

df (4.26)

Then,

222 )()()( FPFP YYXXflq −+−== (4.27)

FPFFFPFF YYYYXXXXf ′

−

′

−

′

′ (4.28)

In general

∫

FdllX

)(cos

and ∫

FdllY

)(sin

(4.29)

4.3 Special Aspects of Highway Information System 78

)(cos PF lX

′ and )(sin PF lY

′ (4.30)

For linear elements;

Line:

FP Xl = (4.31)

PP Yq = (4.32)

Arc:











−=

Rl arctan. (4.33)

sin

−= (4.34)

For the clothoid, because a direct solution for the integral is not available, P

l can be

solved using Newton’s method. The general formula for Newton’s method is

)(

mm lf

ll ′′

′

−=

+ (4.35)

FPFFFFPFFF yyyyyxxxxxf ′′

−

′′

′

′′

−

′′

′

′′ 22 (4.36)

)(sin)( lτlkxF

⋅

−

′

(4.37)

)(cos)( lτlkyF

⋅

′

(4.38)

Applying the differentials;

))()(sin)()((cos)(1

)()(sin)()(cos

PFPF

mm xxlτyylτlk

yylτxxlτ

ll −⋅−−⋅⋅+

−⋅+−⋅

−=

+ (4.39)

where m is the iteration number.

The distance qp for a clothoid is given by;

)(sin lτ

qPF

−

= (4.40)

STEP 3:

The l-coordinate in the linear element’s system is then transformed to the linear referencing

system. The origin is determined according to whatever linear referencing method is being

used. For example, in the case of the link-node method; the link’s beginning node.

In the proposed conceptual data model, the stochastic properties of the linear elements and the

( l, q ) parameters are determined from the standard deviations of the point coordinates and

the significance test’s parameter ( q ). The transformations need to be performed dynamically,

which means without storage of linear referencing information. Therefore, stochastic

properties are required in order to control consistency. By using the significance test the

points in an ( l, q ) reference system can be checked to see if they belong to a specific linear

element. Within the implementation of the conceptual data model, the method Dynamic

4.3 Special Aspects of Highway Information System 79

Reference Transformation was implemented using a Visual Basic program in order to

automate the process.

With the implemented Dynamic Reference Transformation method;

 Decomposition of spatial and non-spatial information is realized. No additional

objects must be defined for linear referencing data. By using the existing objects

proposed in the conceptual data model, information is modeled. This increases the

stability of the GIS-T and simplify the data maintenance.

 Thematic data is independent of geometrical displacements such as; realignment and

error corrections.

 Multi-dimensional road information can be mapped into the conceptual data model

without redundancy.

 No pre-defined methodology is required, users are free to apply the most appropriate

methodology from their point of view. This is because every one-dimensional

reference system is transformed dynamically “on-the-fly”, and stored in a three-

dimensional coordinate system.

 Re-transformation into linearly referenced system is modeled and supported by means

of interfaces.

 Stochastic properties of linear elements and ( l, q ) parameters are available.

 Full integration is realized with other data acquisition techniques where information is

referenced in three-dimensions, such as; photogrammetry and GPS

 Existing data with multiple referencing systems can be fully integrated.

 Due to the minimal data acquisition requirements, a more economical solution is

provided compared with other techniques.

 The proposed technique is independent of any software vendor or platform.

Although this approach provides solutions to the problems identified earlier, there is currently

a performance problem. It has been noted in the literature that extensive algorithmic

processing for techniques to integrate linearly referenced data, is not desired. This is due to

performance issues and is especially so in the case of ITS applications [GOODWIN, 1996].

However, ITS technology requirements differ from those of highway administrations with

respect to the speed and relevancy of processing linearly referenced road information.

Linearly referenced data is of interest to the ITS community, but it is essential for highway

administrations. Additionally, with respect to the rate of developments in information

technology the performance problem can be expected to diminish.

4.3.2 The Cross - Sectional Information

The final reference system to be considered is the ( h, q ) reference system, named in this

study as the cross-sectional system. Other highway administration requirements can be

fulfilled using the defined referencing system. These include; the maintenance of road

facilities, and the determination of suitable advertising panel locations from a traffic safety

point of view.

In one of Germany’s federal states, approximately 100,000 cross sectional data objects are

maintained yearly. [NWR,1995] Although there is a high demand and a source of the spatial

information, in the conceptual data models this information is modeled without consideration

of its spatial structure. One of the reasons this spatial information is neglected in GIS-T is that

4.3 Special Aspects of Highway Information System 80

other systems are used, such as road design software or CAD, to process this information.

Unfortunately, such neglect results in the spatial character of the information being wasted.

In order to model this spatial information in GIS-T three approaches can be followed;

 Consider the information as thematic information

 Consider the information like any other information. Then, according to the proposed

conceptual data model concept, decompose this into spatial and non-spatial

components.

 Consider this information as thematic information with spatial character.

The first approach is the one mainly used in current GIS-T. The cross-sectional design

information is linearly referenced in road databases. Therefore it is assigned as thematic

information to proprietary GIS databases using road network links. This information can be

easily queried and the results can be presented in tables, just like any other thematic

information. Other systems, such as CAD, are required for the processing of geometrical

information, due to the degree of detail of this information. As a result, current systems need

to deal with a continuous data exchange between GIS-T and CAD, and avoid data

redundancies in both systems. However, although interfaces between both systems exist,

automation of the data exchange over time is not fully supported. Both systems are

accordingly maintained in parallel. Additionally, CAD systems have limited capabilities for

querying spatial information. Therefore using this approach will not be efficient.

The second approach would be a complete solution, but it cannot be realized. Cross sectional

design data also has spatial and non-spatial character, and can be handled similarly to other

road events. The information provided using this reference system could be decomposed into

geometrical and thematic parts, where geometry is stored in the geometrical component of the

proposed conceptual data model, and non-spatial data in the thematic part. However, GIS

architectures currently only support at most two and a half dimensions with roads represented

by lines or polygons. As a result, reference system objects could not be visualized in GIS, due

to limited abstraction levels and required information detail degree in cross-sectional design.

In the conceptual data model, in order not to lose the spatial character the cross sectional

reference system was modeled as a road event with a dimension, following the third approach.

This was described in Section 4.2.3.3. With the development of object-oriented databases in

the future it will be possible to model such situations in an optimized way. This is due to;

 The layer philosophy will be superceded by the adoption of object-oriented concepts.

 Coordinates and reference systems can be considered as attributes of objects.

 An object can be presented at various abstraction levels.

 Encapsulation of methods and data can be realized, such as reference systems and

transformations.

 Objects can be identified without their coordinates.

 Objects may have various geometries.

4.3.3 Geometrical Data Integration

Data integration is highly desired due to the decreasing long-term costs of obtaining and

maintaining data, as well as its beneficial effects on data consistency and accuracy within

agencies. Paradoxically, complete and integrated databases are very rare. It is very common to

encounter situations where multiple sources of the same information exist. In addition, as user

requirements vary considerably, information is also of varying quality.

4.3 Special Aspects of Highway Information System 81

Since an important aim of GIS-T is the integration of data sources, in particular spatial data,

adjustment techniques have proven to be the most effective tool. This is due to the technique’s

provision of consistent data and the ability to determine accuracy.

During this study the data integration approach was widely used, especially for the detection

of alignment elements in the horizontal and vertical planes. Non-planar topology was also

implemented using the following objects and techniques;

- Linear Element Vertical

- Linear Geometry member method Detecting Alignment Elements

- Implemented relative height information

These were described in Section 4.2.3.2. In order to implement non-planar topology, height

information is required. Since relative height information is sufficient to fulfill the

requirements of highway administrations, during the implementation of the proposed model

relative height information was used. This is because the relative height information can

provide concrete height information for geometrical elements and is accurate enough for

applications such as; freight management or ITS technology.

However, in some cases relative height information is incomplete or absent (at least in digital

form). It is very common in the conceptual data models that this information was considered

as thematic information. As a result highway linear vertical elements can not be introduced

into the system. In this case, there are other information sources available were height

information can be determined by highway administrations.

Examples of such sources are;

 The road gradient and road inclination values.

 Geometrical design regulations, for example minimum overpass height, driving

dynamics and safety regulations.

 Digital Terrain Models (DTM).

In order to obtain non-planarity, for the entire highway network, concrete height information

for geometrical elements is required. These are only fully available in Digital Terrain Models

among the introduced sources. However, several problems are apparent. Firstly, the required

accuracy in applications which need non-planar topology, such as ITS and freight

management, can not be achieved economically. Because DTM accuracy is tightly correlated

with economical aspects which depend on the data collection method and map scale. Costs

increase as the provided accuracy increases. Additionally, having high accuracy DTM data

will not alone fulfill this requirement. This is because DTM data does not match with road

structures such as; bridges, tunnels and overpasses in a one-to-one manner.

In order to solve this problem, other data sources, such as; geometrical design regulations,

driving dynamics and traffic safety regulations, needed to be introduced to the system.

However, in this case the solution is not unique, leading to redundancy. By means of

adjustment techniques, the required accuracy can be achieved and the redundancy can be

controlled.

The adjustment theory is an established optimization technique used to determine unknown

parameters based on given observations. It provides a straightforward solution to the above

described problems. The aim of least squares adjustment is to optimize the solution of a

functional model by minimizing the residuals of the observations.

∑

=min

ii Pv (4.41)

4.3 Special Aspects of Highway Information System 82

where, v are the residuals and P is the weight matrix containing values corresponding to the

observation accuracy.

The unknown parameters

can be solved according to the following equation:

))(()( 0

1xflPAAPAx TT −= − (4.42)

where A is the Jacobean matrix of the function derivatives with respect to the unknowns, l are

the observations and )(0

xf is the value of the function calculated with approximate values.

This optimization problem can be solved in one of two ways; direct and indirect. The direct

approach is introduced into the system using conditional equations. The indirect approach is

generally preferred due to its better suitability for computation and error estimation. With the

indirect approach, two options are available;

- Introduce conditional equations between the unknowns.

- Enforce specific observations as being more accurate in the stochastic model.

Since conditional equations produce large normal equation systems, the second approach is

preferable.

In order to help clarify the proposed second solution approach, an example illustrating the

interpolated DTM height information and one of the constraints, minimum overpass height, is

presented. This information is shown in Figure 4-21.

4,70 m

Figure 4-21: One of the Sample Constraints, Overpass-Height

The unknowns in this example are the relative height differences ( h

∆

)

BA HHh

−

∆

(4.43)

The interpolated height observations, representing the relative height differences between

points along a road, can be defined in the functional model as;

BAh HHvh

−

∆

∆ (4.44)

Constraints, in this case overpass height, is introduced into the system, using the same

functional model;

mHHvh LUh 70.4

−

+∆ ∆ (4.45)

In the stochastic model, each observation’s corresponding weight is stored in the diagonal

elements of P matrix as:

∆

(4.46)

For this example, the standard deviation of point heights obtained from the interpolated DTM

is assumed to be m

h5±

∆

, and the observation variance is m1

0±=

. The overpass-

height observation is introduced with a standard deviation of cm

∆

. These introduced

4.3 Special Aspects of Highway Information System 83

source, which is acquired more accurately, define conditions in the stochastic model. Since

∑ii Pv2must be a minimum and every observation must completely fulfil the conditions, the

introduced overpass-height constraint enforces the model in order to fulfill its condition.

Since the standard deviation information is “fixed”, other parameters must change, including

the observations. This process is performed iteratively, until

∑

ii Pv2 is minimum and all

conditions are fulfilled. Consequently, when the proposed method is applied to a system

where the height differences are interpolated from a DTM of lower accuracy, the non-planar

topology can be optimized for the entire highway network. Therefore, high accuracy

expectations are fulfilled at low cost. However, with this approach there is a risk of

introducing very “strong” constraints, which results in undesired deformations of other

observations in the system. This is due to, an adjustment approach allows for a change of all

parameters while simultaneously enforcing constraints together with the tolerance of the

constraining points. This problem can be solved by loosening the “strong” constraints until

the appropriate solution is achieved.

4.3.4 Unique Identifier (Unique_id)

The necessity for data integration highlights another problem, being identification of the

object in the system. Although object identifiers could be defined “temporary or persistent,

locally or globally unique ” [BISHR,1999], GIS feature/object identifiers lose their meaning

during data integration and data exchange between systems if they are locally unique and

temporary. In GIS, a feature or an object can not exist without an identifier. Since every

spatial and non-spatial information object is assigned to this identifier, the object identifier in

GIS is the only information assigned which should be kept unchanged during the object’s life-

span in the system. It must therefore be persistent.

There are two types of object identification approaches;

 System automatic identifier generation.

 User defined identifiers.

The system automatic identifier generation is the main approach followed by GIS software

vendors. Using this approach objects in GIS are uniquely defined by identifiers which are

automatically generated by the system. This identified object is only unique in the defined

system, database or table, not outwith the system. The automatic generated keys solution will

have problems if there is a need for data integration and data exchange with other systems. If

information is exported from a system, processed in the target system and re-imported to the

original system with additional information, the use of automatically generated identifiers will

result in two different and independent objects. This will definitely lead to redundancy and an

effective loss of information from the database.

Secondly, the system automatic identifiers are assigned to the planar coordinates ( x, y ) of the

object. Due to this, persistency of the system cannot be ensured. Object coordinates can easily

change during the life-span of an object. For example, data obtained by means of a more

accurate data acquisition technique or integrating additional information will lead to a change

in these coordinates. Additionally, these assigned coordinates are two-dimensional while

highway information is multidimensional.

The limitations in using system generated identifiers promote the use of user defined

identifiers in GIS. When choosing an appropriate object identifier for highway agencies, there

are many candidates including; road numbering systems and administrative area codes.

However, persistency and uniqueness can not be ensured using these as identifiers due to

involved spatial frame.

4.3 Special Aspects of Highway Information System 84

At first sight highway numbering systems, which are mainly used to provide guidance for

drivers and by the technical applications of highway administrations, can be used as

identifiers. This is because they are complete and uniquely identified for all road networks.

However, during this study it was discovered that these identifiers are not permanent. This is

due to geometrical frames being contained in the numbering systems, Therefore they cannot

fulfil this requirement.

The highway numbering system of Germany is a good example in order to show the usage of

spatial information as a frame. In this system, nodes are named uniquely and every net node

consists of seven units. The first four digits are defined using the name of the 1:25000 scaled

Gauss -Krüger projection map. The remaining three digits are determined from the node

numbers for example [….499]. Every net point is numbered by the highway agency without

consideration of which road they belong to (national, district..etc).

With this approach, since the first four digits are taken from the 1:25000 map, the numbering

system is framed with spatial information, in this case the map projection. If another map

projection needs to be used, these nodes change their spatial frame, concluding their name and

cannot be found at their original locations. Indeed they may be in a completely different

quadrangle. As a result their uniquely identified name will begin with a different number.

This is shown in Figure 4-22.

Figure 4-22: Geodetic Datum Transformation

Additionally, it is possible to introduce new naming systems, as in the case of the United

States where all nodes names were newly defined according to an other naming regulation.

Another proposal is to use jurisdiction or administrative areas as the reference frame, and to

identify features uniquely within this frame. This will also not provide the required solution. It

is again a geometrical property which can change over time due to governmental or

organizational requirements.

Consequently, the candidate user-defined object identifiers should ensure;

 Uniqueness

 Persistency

 Independence of any spatial information and frame

 Independence of any hardware or software vendor

 Simplicity of implementation

4.3 Special Aspects of Highway Information System 85

In addition to these requirements, there is currently an on-going discussion on the

establishment of “intelligent (compound) identifiers” in order to identify data source or the

owner of information in the case of data exchange between departments or other

organizations.[EAN,2001], [OPENGROUP,2001] [ANSI,2001] [INTERLIS,2001]

For highway agencies, the proposed solution is based on the naming of links, which have no

geometric character. This is because the topological element link is a logical connection and

invariant with respect to geometry. It is therefore independent of transformations, scale and

other factors which can affect identifier permanency. All other identifiers such as, name,

number and code need to be stored as thematic data in the database. The use of these should

be avoided as they are in many cases redundant, incomplete, non-persistent and non-unique.

The required persistent and unique link identifiers can be named according to the following

existing naming regulation steps. This is true even in the case of maps. These frames can be

used in order to assist highway agencies, but including this information in the identifier’s

name must be avoided. The appropriate naming regulation for link identifiers must be

independent of any spatial frame including, geometrical structure, location, coordinate

system, map projection or administrative area.

The unique identifier’s structure should be defined as follows. The number of digits should be

determined according to the length of the road network and with consideration of the database

and GIS capacities. If it is decided by highway administrations to establish intelligent

identifiers, country codes, which are internationally unique, can be integrated into the

identifier, in order to exchange data at international level. Additionally, a code identifying the

national highway organization can be included in the system, in other to avoid data exchange

problems with other national organizations such as; national police and health ministry. Other

codes, including administrative area, highway administration division or any related to

software or hardware peripherals should be omitted in order to preserve persistency.

5.1 Overview 86

Chapter 5 : Implementation of the Proposed Conceptual Model

5.1 Overview

During the conceptual data modeling stage, implementation details and system requirements

were not considered. These topics will be considered in the context of the physical

implementation of the data model. The physical storage proposals, as well as constraints and

triggers, which needed be applied in order to realize methods were explained in this chapter.

The developed data model and concepts were implemented on a sample project in order to

recognize gaps and requirements which are not fulfilled during conceptual data modeling.

Before implementing the proposed concepts in a pilot project, it is necessary to decide which

approach is the most appropriate to be followed. From the currently available approaches, the

integrated architecture approach is selected, due to the expectation that this approach will

promote the success of implementation. The benefits of the integrated approach were

described in Section 3.3.1.

In order to support the usage of this approach, an appropriate software needed to be selected.

To date a large number of geographic information system software packages exist. However,

the GIS systems available on the market are not specifically tailored for road administration

organizations, nor for the proposed conceptual data model.

Currently available GIS and database software was examined with respect to the developed

conceptual data model in order to determine physical data modeling requirements, potential

problems and limitations. The appropriate software selection criteria can be listed as follows;

 Open architecture.

 Integration possibilities between databases.

 Spatial data support.

 Storing spatial and non-spatial information in the same database.

 Ability to introduce user defined types.

 Ability to introduce user defined methods.

5.1.1 Software Used

5.1.1.1 Database Management System

During this study, Oracle8i, object-relational database management system (DBMS) with its

Spatial Data Option extension was used for implementation. The Oracle8i provides a high

level of functionality with user administration, multiple user access, distributed data retention,

backup/recovery and replication.[ORACLE8i, 2001] The system provides many benefits for

GIS architectures, especially through its use of Structured Query Language (SQL), spatial

data storage and processing capabilities, as well as support of client-server environments.

With Oracle Spatial it is possible to store and process spatial data in databases through the use

of indexing mechanisms. Storage of spatial and non-spatial data in a common database with

integrated administration is possible. Currently only two-dimensional data is fully supported.

There is however the possibility of calculating indices for higher-dimension points and for

storing these in partitioned tables. Oracle Spatial defines three different two-dimensional

elementary graphical objects being: point and point cluster, n-point polygons, as well as line

strings and arc line strings.

5.1 Overview 87

The data model of the Oracle Spatial is oriented towards the simple features standardization

proposal of the OpenGIS. Oracle8i extends the database model to include an SDO_Geometry

object.

In Oracle Spatial, a spatial index provides a mechanism to limit searches based on spatial

criteria. Three types of indexing mechanisms are used; R-tree indexing (the default), Quadtree

indexing or Quadtree Hybrid indexing, which is a combination of both R-tree and Quadtree.

An R-tree index approximates each geometry by a single rectangle that minimally encloses

the geometry, which is called the minimum bounding rectangle (MBR). An R-tree index

consists of an hierarchical index with respect to MBR´s tree structure. In the linear quadtree

indexing scheme, the coordinate space is decomposed into tessellations, completely defining

tiles for every stored geometry, until the termination criteria is reached. This is illustrated in

Figure 5-1. Fixed or variable-sized tiles can be used to cover the geometry. Success of the

query depends on the level and size of the defined tiles. If very small tiles are used the

number of tiles needed to cover large areas would be very large. If large tiles are used, index

selectivity will be unsuccessful due to the inability of approximating small geometries.

Quadtree Hybrid indexing uses a combination of fixed-size and variable-sized tiles for

spatially indexing a layer.

Figure 5-1: Linear Quad-tree Indexing Scheme

A two-tier query model with two distinct operations is used to resolve spatial queries and

spatial joins. These are the primary and secondary filter operations. With the primary filter

pre-selection is realized, these results are then transferred to the secondary filter. Oracle

Spatial uses the secondary filters to determine the spatial relationship between entities in the

database. The most common spatial relations are based on topology and distance. Filter

operations provide spatial sorting and operations for objects, such as; locate, fence locate,

spatial join, touch, equal or any interact.

The importance of using the Oracle Spatial Cartridge during implementation of the proposed

data model include;

 One database management system including SQL

 Non-proprietary database.

 Integrated data management of spatial and non-spatial data.

 Storage of vector geometries.

 Query spatial relationships between geometries.

 Extensible component of network computing architecture, flexible application support.

5.1 Overview 88

 Combining standard components, lower cost of implementation, and standard

protocols SQL, CORBA, HTTP.

5.1.1.2 GIS Software

The proposed conceptual data model was implemented using the Geomedia Professional 3.0

and ArcInfo8TM. Both software are fulfilling above listed criteria for implementation purposes

with their concepts of distributed databases, open architecture, possibility to introduce user-

defined objects and methods.

Intergraph’s is a software solution for the entry, administration, modeling, analysis and output

of spatial data. With the software, it is possible to use the standard Microsoft database

application Access, or to connect to another distributed database. Read/Write database access

for Oracle is already provided by the software vendor. GeoMedia Professional 3.0 also

provides extensive methods for data acquisition and processing, such as raster data

registration and processing and geometry analyses. GeoMedia Professional 3.0 does not

provide geometric network possibilities for the storage of topological information, but another

concept is introduced, called “On the Fly” topology. This concept treats points, which are

situated within a certain distance from each other as topological nodes. The consequence of

this is that with geometrical manipulations these points are shifted together. GeoMedia

requires that features stored in databases have a primary key, in order to realize the linkage

between geometry components and the database. System recommends a maximum of eight

pairs of coordinates for the Oracle Spatial module. Within GeoMedia Professional, it is

possible to introduce user-defined objects since it is based on the Component Object Model

(COM) technology developed by Microsoft, which allows components to be reused at binary

level.

Second software used during implementation was ArcInfo8TM. It is a powerful GIS software

developed by Environmental Systems Research Institute (ESRI), provides solutions for the

entry, administration, representation, analysis and output of geographical and thematic data.

ArcInfo8TM is based on three applications; ArcMap, ArcCatalog and ArcTool . The ArcMap

is the environment providing the graphical interface for working with map data. The

ArcCatalog is the data manager, which enables the administration of local directories and the

relational data bases available in the system network. ArcTool is used for performing

operations, such as data conversion and geographic datum transformations. Implementation of

the data model is provided by means of the Computer Aided Software Engineering (CASE)

tools. CASE tools also support the creation of COM classes. ArcInfo8TM supplies various

possibilities for the data management and integrity such as; attributes domains, relationship

objects with relationship rules, geometric network with connectivity rules, user-defined

objects and user-defined rules for objects.

ArcInfo8TM is ArcSDE, which manages the physical storage of feature geometries in order to

use simple standard data types within the host RDBMS. [ESRI,1999] Similarities of ArcSDE

with the above described Oracle Spatial Cartridge can be found in the way that both of them

manage geometrical storage of geometrical features in relational databases. When using

Oracle Spatial Cartridge with ArcSDE, which was the case in the pilot project, geometry,

spatial index and spatial searches are performed by Oracle Spatial.

The ArcSDE gateway enables the ArcInfo8 Geodatabase to leverage Oracle Spatial to store

and manage feature geometry. With the ArcSDE, it is possible to realize;

 Versioning ,which is long transaction editing

 Gateways to the ArcIMS, the internet web-server of ESRI,

5.2 The Realized Physical Storage 89

 Gateways to the Dynamic Segmentation module, which covers linear referencing

issues

With the ArcObjects object-component model, users can extend the data model using exactly

the same COM technology which ESRI used to build ArcInfo8TM. ObjectExtensions are an

ActiveX DLL from ESRI which equips a feature object with additional behavior. With the

implementation of the IobjectObjectValidation interface, it is possible to program user-

defined consistency conditions.

5.1.2 Sample Data

The Brandenburg State Office for Traffic and Roads (BLVS) of Germany provided data for

the pilot project. The provided data is located in an area to the northeast of Berlin around road

B2. It was received in MapInfo (*.mif) and .dbf format shown in Figure 5-2. The data is

created according to ASB regulations.

Figure 5-2 Sample Data from the Brandenburg Highway Administration

Both data sources were successfully imported into the system according to below defined

physical storage descriptions.

5.2 The Realized Physical Storage

After determination of implementation environment the physical storage was performed. The

proposed data model has been implemented according to the provided tables. The tables

topology, geometry and thematic components were structured according to the proposed

conceptual data model concepts. Complete implementation proposal tables can be found at

Annex D.

In the conceptual data model, the topological component consists of 12 objects with

corresponding relationships between each other. Object were uniquely identified by their

identifiers. In Table 5-1 the physical storage for several first level topological objects are

shown. Second level topology can be implemented in the same way for the specified objects.

5.2 The Realized Physical Storage 90

Table 5-1: Implementation of the First Abstraction Level of Topology

Object Attribute Data type Description

Node I NodeI_ID Long Identifier of Node

LinkI_ID Long Identifier of Link

NodeI_IDBegin Long Identifier of Beginning

Node of Link

Link I

NodeI_IDEnd Long Identifier of Ending

Node of Link

Road_ID Long Identifier of Road Road

RoadName String Name of Road

LinkI_ID Long Identifier of Link

RoadName String Name of Road

LinkI/Road

LinkIDOrder Integer Order of Links for Road

The geometrical component consists of 13 objects. In Table 5-2 and Table 5-3 the geometrical

elements Point Geometry, Reference System and PointGeo/RefSys is provided according

to the conceptual data model design.

Table 5-2: Implementation of the Point Geometry

Object Attribute Data type Description

Point_ID Long Identifier of Point

NodeI_ID Long Identifier of Node I

NodeII_ID Long Identifier of Node II

LinkI_ID Long Identifier of Link I

Point

Geometry

LinkII_ID Long Identifier of Link II

5.2 The Realized Physical Storage 91

Table 5-3: Implementation of the Reference System and the Relation Between

PointGeo / RefSys

Object Attribute Data type Description

RefSys_ID Integer Identifier of Reference

System

Sysname String Reference System Name

Projection

Algorithm

String Projection Algorithm

Name

Projection

Parameters

Long Projection Parameters

Storage Center,

Resolution

Geodetic Datum String Geodetic Datum

Reference

Ellipsoid

String Reference Ellipsoid

Name

Reference

System

Ellipsoid

Parameters

Long Ellipsoid Parameters

Point_ID Long Identifier of Link

RefSys_ID Long Identifier of Beginning

Node of Link

X Long X – Coordinate

Y Long Y – Coordinate

Z Long Z – Coordinate

SigmaX Double Standard Deviation of X

Value

PointGeo/

RefSys

SigmaY Double Standard Deviation of Y

Value

The implementation recommendation as described above overcomes any redundancy, which

can occur if different referencing systems are used. Unfortunately it is not possible to realize

this practically with conventional GIS software. With relational or object-relational GIS, each

geometrical feature or object is assigned to its geometrical planar coordinates ( x, y) with

feature identifiers for the visualization of these objects. If the point geometry does not have

( x, y) coordinates and an object identifier assigned to its coordinates, it is not possible to

visualize point geometry with GIS. Therefore, the Point Geometry object was reconsidered

in order to implement the proposed conceptual data model, as follows;

5.3 Implementation 92

Table 5-4: Implementation of the Point Geometry

Object Attribute Data type Description

Point_ID Long Identifier of Point

NodeI_ID Long Identifier of Node I

NodeII_ID Long Identifier of Node II

LinkI_ID Long Identifier of Link I

LinkII_ID Long Identifier of Link II

X Long X – Coordinate

Y Long Y – Coordinate

Height Long Z – Coordinate

SigmaX Double Standard Deviation of X

Value

Point

Geometry

SigmaY Double Standard Deviation of Y

Value

Clearly, after this revision X, Y, Z, SigmaX and SigmaY should be deleted from the

PointGeo / RefSys object.

Road Event, the third component of the conceptual data model was implemented in the

following way, illustrated in Table 5-5;

Table 5-5: Road Event Implementation

Object Attribute Data type Description

Roadevent_ID Long Identifier of Road EventRoad Event

Geometry_ID Long Identifier of Geometrical

Type of Road Event

Road event objects are assigned to one of the highway administration pre-defined geometrical

elements through the geometry identifier. Road event identifiers are listed and assigned to

road event objects.

5.3 Implementation

Within the pilot project, the first abstraction level topology, geometry and thematic

components were implemented as presented in Chapter 4. Implementation was realized using

SQL scripts directly in Oracle8i. The implementation was realized according to the above

described physical storage. Using the SQL scripts, implementation of the proposed conceptual

5.3 Implementation 93

data model was simplified and automated. A sample SQL script generated for the linear

geometry object is illustrated in Figure 5-3.

………

CREATE TABLE "OSC"."LINEARGEO"(

"LINEARGEO_ID" NUMBER(10) NOT NULL CONSTRAINT plineargeo PRIMARY KEY,

"POINTGEO_BEGIN_ID" NUMBER(10) NOT NULL CONSTRAINT flineargeo_1

REFERENCES "OSC"."POINTGEO"("POINTGEO_ID"),

"POINTGEO_END_ID" NUMBER(10) NOT NULL CONSTRAINT flineargeo_2

REFERENCES "OSC"."POINTGEO"("POINTGEO_ID"),

"LINK1_ID" NUMBER(10) NOT NULL CONSTRAINT flineargeo_3

REFERENCES "OSC"."LINK1"("LINK1_ID"),

"GDO_GID" NUMBER(38)

)

TABLESPACE "USERS"

…….

Figure 5-3: A Sample SQL Script for Creating Objects [PFANNMÖLLER,2001].

With Oracle’s enchanted spatial indexing concept it was possible to store spatial and non-

spatial information in one database. The SQL scripts were directly implemented in Oracle and

access to GIS software was realized with software vendor provided interfaces.

In GeoMedia Professional 3.0, the conceptual data model’s thematic and geometry

components was fully realized. Two methods, which were introduced in Chapter 4, the

Detecting Alignment Elements and Dynamic Reference Transformation were successfully

implemented into system. In order to implement Detecting Alignment Elements AXTRAN

software was used. The Dynamic Reference Transformation implementation was realized

using a Visual Basic program.[PFANNMÖLLER,2001] Using these methods, the

parameterized approach of the proposed conceptual data model was realized and automated.

These methods were implemented into GeoMedia Professional 3.0.

Within the proposed conceptual data model, it was assumed that linear elements are defined

by means of their parameters and the beginning and ending point and then generated without

storage of any other geometrical information. However, although it was possible to

parameterize defined geometric elements and introduce them to the system;

 It was not possible to generate these geometrical elements automatically using their

parameters, without the storage of any other geometrical information. Because,

geometrical features must be identified by their planar coordinates in the software.

 It was not possible to visualize the clothoid.

The first mentioned problem was solved by implementing the geometrical elements

redundantly and controlling the redundancy with the developed methods. The visualization of

clothoid was realized with the help of other geometrical elements, since it was not possible to

define geometrical objects in the system.

Unfortunately the topology component can not be implemented in the manner proposed in the

conceptual data model. The main problem encountered was the lack of user-defined types,

5.3 Implementation 94

meaning implementation opportunities were limited due to software vendor defined features,

specifically spatial features. Topology information is considered differently in GeoMedia.

Only in the “Maintain Coincidence Mode” (“On-the-Fly” topology instruction), displacement

of one point has the consequence that, points which are situated within a determined distance

will follow this movement. Additionally, it is not considered whether these points are

assigned to topological nodes or not. In Figure 5-4 displacement of a node by means of

“maintain coincidence mode” is shown.

Figure 5-4: Editing the Node Using Maintain Coincidence Mode

There are some additional reasons, which affects the implementation of topology. The

available mechanisms were insufficient in controlling the redundancies. The versioning

concept is unavailable. A transaction is always automatically terminated in the system after a

new object is created. This requires consistency checks to be realized immediately. However,

in the conceptual data model consistency checks are pre-required for some entries, such as

Link and Node, which were discussed in Section 4.2.4.1. There are also situations where this

is reversed, such as determining Point Geometry and Node relations. In this case control

mechanisms did not provide expected results. Additionally, it is not certain whether the

quantity of the consistency conditions required are realizable using triggers. A further

problem results from the fact that the user does not have influence on the end of a transaction

[PFANNMÖLLER,2001].

The second selected software ArcInfo8TM has a very important benefit. Users can define

feature types and are not limited to the software vendor’s concepts and definitions due to

provided ArcObject concept. With the “geometry network” possibility, topological elements

can be realized according to the descriptions in the conceptual data model. In ArcInfo8TM

integrity constraints called “validation rules” are also available and can be defined by the user.

Additionally, the versioning concept is available. The conceptual data model was successfully

implemented through ArcInfo8TM , although during the implementation some problems

needed to be solved. These are as follows;

1. Geometrical elements can not be generated using their parameters.

The main concept is similar to that used by GeoMedia Professional, where

geometrical features are assigned to their planar coordinates for visualization

purposes. In order to solve this issue, geometrical features and parameters were stored

5.3 Implementation 95

by using the user-defined objects and methods. The occurred redundancies was

controlled using user-defined methods and validation rules.

2. Features having geometrical characteristics, must be assigned to a pre-defined

geometrical feature of ArcInfo8TM.

Regardless whether an object is defined as a geometrical component or not, every

object which should be visualized in GIS must have planar coordinates assigned to it,

including the topological elements: node and link. During the implementation, an

additional column was added to the physical data model table in order to define the

geometry of feature. Consequently, objects such as linear elements were stored

redundantly in the shape column with their geometrical information, in this case the

coordinates of the polygon’s points. These redundancies were controlled by

consistency rules.

3. The linear element clothoid is not supported in ArcInfo8TM

Although clothoids were stored with their parameters in the database, automatic

generation of parameterized elements was not possible. In order to be able to visualize

them, clothoids were simplified as line and arc objects during visualization.

4. Object can only be characterized by their object identifier (Object_ID), which are

automatically assigned by the software. Object_ID´s, with respect to object- identifier

principal requirements, are not unique throughout the system, only unique within a

table.

With respect to the unique identifier problem mentioned in Section 4.3.4, a significant

problem became apparent. An individual column or a combination of columns can not

be uniquely defined in ArcInfo8TM. However, it is possible to implement unique

attributes which can be realized through user programmed ObjectExtensions. This

uniqueness of attributes can only be realized using the ArcMapTM Editor tool, Validate

Selection. Additionally, this can only be executed manually by the user. In this case,

the importance of establishing regulations for unique identifiers and the automation of

this process gains more importance.

5. By using the “geometric network” of ArcInfo8TM, topological relations can only be

modeled between geometrical objects.

With the “geometric network” of ArcInfo8TM, topological objects and relationships

between each other could be realized. However, due to non-separation between

geometry and topology, this realization can only be made between geometrical

objects. Due to this limitation, implementation of a 0..1:1 relationship between Node

and Point could not be realized.

In order to implement the proposed conceptual model, several modifications were made with

respect to fifth item. The geometric element point was subdivided into three parts being; node,

element point and intermediate point. Node was preserved as node. From the other new

defined objects Element Point defines the beginning or ending point of a linear element. The

Intermediate Point was introduced in order to indicate any location along a link, which is

neither the beginning nor the ending point of the linear element. The modifications are

presented in Figure 5-5.

5.3 Implementation 96

Figure 5-5: Modifications in the Conceptual Data Model

After these, implementation of the conceptual data model in ArcInfo8TM was realized. A

sample implementation for Point object is provided below in Table 5-6.

Table 5-6: Sample Implementation for the Point Object

Object Class Object Data type Description

ObjectID Long ArcInfo8TM assigned object

Identifier

Shape Geometry (Point) Geometry type for all

features assigned in Point

S_Type Integer Division of Node, Element

point, intermediate point

LinkI_OID Long Identifier of Link I

(ArcInfo8TM assigned)

LinkII_OID Long Identifier of Link II

(ArcInfo8TM assigned)

Point_ID Long Identifier of Point

NodeI_ID Long Identifier of Node I

NodeII_ID Long Identifier of Node II

Height Long Z - Coordinate

SigmaX Double Standard deviation of X

Point

SigmaY Double Standard deviation of Y

5.3 Implementation 97

In the table Table 5-6, S_Type column indicates the feature geometry type. In addition to the

above mentioned Point object, consistency conditions were implemented in order to satisfy

realization of the required rules, which are illustrated in Table 5-7.

Table 5-7: Consistency Conditions for Object Point

Relation Attribute Condition

Referenced

primary key

column

Cardinality

Object_ID Positive integer, NOT NULL,

Unique

__ __

S_Type NOT NULL, 0 - 2 __ __

LinkI_OID NOT NULL, Positive integer Link I.Object_ID 0..1 to 0..*

LinkII_OID NOT NULL, Positive integer Link II.Object_ID 0..1 to 0..*

The conceptual data model thematic component Road Event was also implemented subject to

similar modifications.

With the defined method Detecting Alignment Elements of Linear Geometry geometrical

component of the proposed conceptual data model was generated. Topology component was

implemented as proposed in the conceptual data model, which is shown in Figure 5-6.

Figure 5-6: Implemented Geometry and Topology Components

The defined method Dynamic Reference Transformation of Road Event was generated

using ArcObjects. Through compatible programming languages such as Visual Basic, Visual

C++ and Delphi, it is possible to create or add user-defined objects to a database, which

operate directly with ArcObjects. As the sample data for this pilot project was dependent on

the link-node linear referencing system as proposed by the ASB, this methodology was used

5.3 Implementation 98

for re-transformation during designed interfaces. Road Events, described in the conceptual

data model, were obtained using the program and existing interfaces shown in Figure 5-7.

[PFANNMÖLLER,2001]

Figure 5-7: The Implemented Road Event Component of Sample Data

In addition, Road Event object information identified at vertical sections were also realized

using the methods “Detecting Alignment Element” and “Dynamic Reference

Transformation”. The height information was already existed in the provided data set. This

was linearly referenced using the link-node method, which is illustrated in Figure 5-8.

Figure 5-8: The Vertical Linear Elements Sample

The cross-sectional design information, which was predefined as the (q , h) reference system,

was also realized. Within this example, the stationing data of the road layer structure was

connected to various types of layers, coded by the entries for Art1, with the coordinates of the

(q, h) coordinate system describing the layers, such as Art2, Art3 [ASB 1998]. The cross-

5.3 Implementation 99

sectional design reference system origin was identified using the linear-referencing system

data from the given node. The sample example is shown in Figure 5-9.

Figure 5-9: The Cross-Sectional Reference System and Analyzed Road Data

Some of the defined integrity constraints in the proposed conceptual data model was

performed using existing “validation rules” in “geometry network” of ArcInfo8TM. Others

were introduced into the system by means of triggers, which will be introduced below. An

example of the validation rules used in the system is given in Figure 5-10.

Figure 5-10: An Example of Validation Rule Used in Topology Component

The maintenance of complex relationships and validation of complex rules are often needed

to be defined externally, due to lack of realization of the encapsulation concept. In order to

define these, triggers are in the Oracle database management system. A trigger is a method

which is invoked whenever a specified object or attribute is inserted, updated or deleted.

Ideally, it should be possible to invoke the full range of GIS methods within a trigger and it

5.3 Implementation 100

should be possible to cause the current transaction to be rolled back if an invalid condition is

found within a trigger. Procedures are started explicitly by the user, by an application or also a

trigger. These are procedures written in PL/SQL, Java, or C that execute ("fire") implicitly.

With the releasing event it concerns one or more Data Manipulation Language (DML)

operations (insert, update and delete), illustrated in Table 5-8.

Table 5-8: Trigger Types