Water and heat transport
of paved surfaces
vorgelegt von:
M.Sc.
Anne Timm
ORCID: 0000-0002-5279-3642
von der Fakultät VI - Planen Bauen Umwelt
der Technischen Universität Berlin
zur Erlangung des akademischen Grades
Doktorin der Naturwissenschaften
- Dr. rer. nat.-
genehmigte Dissertation
Promotionsausschuss:
Vorsitzender: Prof. Dr. Reinhard Hinkelmann
Gutachter: Prof. Dr. Gerd Wessolek
Gutachter: Prof. Dr. Markus Weiler
Gutachterin: PD Dr. Patricia Göbel
Tag der wissenschaftlichen Aussprache: 05. April 2019
Berlin 2019
An example of urban pavement
Acknowledgements
First, I would like to thank Prof. Gerd Wessolek for many things, including the oppor-
tunity for this research and the good advice throughout these intense three years. I am
most thankful however, for the trust put into me and the freedom to develop my own
ideas and make my own mistakes.
I had the great pleasure to work on this thesis within the Graduate School Urban
Water Interfaces, a joint project by TU Berlin and IGB Berlin, funded by the German
Research Foundation DFG (GRK 2032). Everyone involved made this time interesting
and easier. I am especially grateful to Prof. Dr. Reinhard Hinkelmann, Dr. Gwendolin
Porst, the other doctoral students and the Internal Steering Committee who showed
continued engagement for the project. Thank you all.
I am grateful to everyone who is or has been at the chair of soil conservation at the
TU Berlin. Your previous work and support during my research has been instrumental
for everything. Dr. Thomas Nehls and Dr. Yong Nam Rim, for installation and gradual
improvement of the two lysimeters used in this study during previous research, which
provided a perfect foundation for my own experiments. Everyone who helped me during
the lysimeter reconstruction in April 2016, especially: Sven Glawion, Dr. Steffen Trinks
and Joachim Buchholz. Dr. Björn Kluge and Reinhild Schwartengräber for always ha-
ving an open door and good advise. Dr. Basem Aljoumani for making our office home.
Dr. Andre Peters for providing the AWAT filter routine and always being available for
questions concerning it, as well as teaching a specific course and supervising my Masters
thesis, which lead to my love for data processing and R.
For everything concerning R, the community at stackoverflow.com has been a robust
and reliable help. As is tradition by now, I am very lucky to be able to count on Sinah
Ruhnau for proof-reading and general support.
Outside of work, my friends, RPG groups and the Critical Role community were
essential. Thank you for times in which my thoughts could walk different paths.
As always, I am grateful to my family for believing that I could achieve whatever I
set my mind to.
And finally - Floh. For unconditional love and support for 12 years, including the
first two of this phase. Sometimes a dog is the best listener and taking walks at night
solves problems.
Summary
The urban soil-atmosphere is typically paved with varying materials. This alteration has
drastic effects on the urban hydrological balance. With a focus on stormwater mana-
gement, most existing models treat all types of pavements identically when estimating
runoff and neglect losses caused by evaporation and infiltration processes.
A literature review reveals that very few studies have been published that measured
all components of the hydrological balance and water transport processes of pavements.
Combining these studies with other research focussed on individual processes showed
that pavements have varying impacts.
Within this project, two common pavement types were studied: cobblestones and
concrete slabs. A combination of high-resolution weighable lysimeters and sensors mea-
suring soil water content and temperature was used to gain new insights into their hy-
drological balance, as well as water and heat transport processes. Annually, cobblestones
evaporated 25 % and rarely produced runoff (3 %). Concrete slabs tended to produce
more runoff (16 %) and less evaporation (22 %). Both surfaces led to similar infiltration
(62 %). Upward water transport from underlying soil layers led to evaporation processes
during dry periods. This effect contributed 47 % of evaporation for cobblestones and 13
% for concrete slabs. Both surfaces led to higher soil temperatures compared to natural
surface covers, with concrete slabs tending to slightly higher temperatures.
Estimating evaporation from paved surfaces may be accomplished by reducing the
common grass-reference evapotranspiration. For annual estimations, the pre-existing
TUBGR model yielded good results.
Pavements are more than a runoff generator. The urban soil-atmosphere interface is
an active system with varying impacts on the urban hydrological balance. Understanding
and utilising these differences has the potential to improve the design of urban areas.
Kurzfassung
Die urbane Grenzfläche zwischen Boden und Atmosphäre ist meist mit verschiedenen
Materialien versiegelt. Diese Veränderung beeinflusst den urbanen Wasserhaushalt in
hohem Maße. Mit einem Fokus auf starke Niederschläge und Überflutung gehen die
meisten existierenden Modelle davon aus, dass sich alle Arten von Pflasterung gleich
verhalten und Verdunstung und Versickerung vernachlässigt werden können.
Eine Analyse der Literatur ergibt dass bisher nur wenige Studien publiziert sind
in denen alle Komponenten des urbanen Wasserhaushaltes gemessen wurden. Die vor-
handenen Studien zeigen auf, dass verschiedene Arten von Versiegelung abweichende
Auswirkungen haben.
Innerhalb des Projektes wurden zwei Pflasterungen näher untersucht: Bernburger
Mosaik und Betonplatten. Dazu wurden hochauflösende wägbare Lysimeter mit zusätz-
lichen Sensoren ausgestattet, um Wassergehalt und Temperatur des Bodens zu mes-
sen. Mit dieser Kombination von Methoden wurde angestrebt, neue Erkenntnisse zum
urbanen Wasserhaushalt und zu relevanten Wasser- und Wärmetransportprozesse zu
gewinnen. Bernburger Mosaik bildete nur selten Oberflächenabfluss (3 %) und 25 %
verdunsteten jährlich. Betonplatten bildeten weniger Verdunstung (22 %) und führten
häufiger zu Oberflächenabfluss (22 %). Bei beiden Oberflächen versickerten rund 62 %.
An der Verdunstung war aufwärts gerichteter Wassertransport von Bodenschichten an
trockenen Tagen maßgeblich beteiligt. Bei Bernburger Mosaik wurden 47 % der gesam-
ten Verdunstung an Tagen ohne Niederschlag verzeichnet, bei Betonplatten waren es 13
%. Im Vergleich zu natürlichen Oberflächen führte Versiegelung zu höheren Bodentem-
peraturen, wobei Betonplatten zu leicht höheren Temperaturen neigten.
Für die Abschätzung von Verdunstung versiegelter Flächen kann eine verringer-
te Gras-Referenzverdunstung genutzt werden. Dabei lieferte das bereits vorhandene
TUBGR Modell gute Ergebnisse für jährliche Werte.
Gehweg- und Straßenpflaster sind mehr als Erzeuger von Oberflächenabfluss. Die
urbane Grenzfläche von Boden und Atmosphäre ist ein aktives System mit unterschiedli-
chen Auswirkungen auf den urbanen Wasserhaushalt. Ein verbessertes Verständnis dieser
Auswirkungen und dessen Anwendung hat Potenzial urbane Räume besser zu gestalten.
Table of contents
List of Figures iii
List of Tables v
1 Preface 1
2 Introduction 3
2.1 Pavedsurfaces ................................. 5
2.2 Water transport processes . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2.1 Elements and characteristics of pavements . . . . . . . . . . . . . . 7
2.2.2 Processes ................................ 8
2.2.2.1 Runoff ............................ 8
2.2.2.2 Infiltration .......................... 10
2.2.2.3 Groundwater recharge . . . . . . . . . . . . . . . . . . . . 10
2.2.2.4 Evaporation ......................... 11
2.3 Hydrologicalbalance.............................. 12
2.4 Application ................................... 16
2.4.1 Sealing degree (SD) . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.4.2 Runoff coefficient (RC) . . . . . . . . . . . . . . . . . . . . . . . . 18
2.4.3 Numericalmodels ........................... 18
2.4.4 Datageneration ............................ 19
2.5 Conclusion ................................... 20
3 Material & Methods 21
3.1 Lysimeterstudies................................ 21
3.2 Lysimeterstation................................ 23
3.2.1 Surfacetypes.............................. 23
3.2.2 Lysimeterset-up ............................ 25
3.2.3 Built-insensors............................. 26
3.2.4 Climatestation............................. 28
3.3 Surface wetting-drying experiment . . . . . . . . . . . . . . . . . . . . . . 28
3.4 Dataprocessing................................. 29
3.4.1 Tools .................................. 30
3.4.2 Lysimeterdata............................. 30
3.4.2.1 Runoff ............................ 30
i
3.4.2.2 Precipitation, Evaporation & Infiltration . . . . . . . . . 32
3.4.3 Sensordata............................... 36
4 Results 37
4.1 Lysimeterstudy................................. 37
4.1.1 Climatological conditions . . . . . . . . . . . . . . . . . . . . . . . 37
4.1.2 Hydrological balance & water transport . . . . . . . . . . . . . . . 38
4.1.2.1 Hydrological balance . . . . . . . . . . . . . . . . . . . . . 38
4.1.2.2 Water transport processes . . . . . . . . . . . . . . . . . . 44
4.1.3 Heat balance & transport . . . . . . . . . . . . . . . . . . . . . . . 50
4.2 Processinteractions .............................. 55
4.2.1 Coupled heat and water transport . . . . . . . . . . . . . . . . . . 55
4.2.2 Correlation of processes . . . . . . . . . . . . . . . . . . . . . . . . 60
5 Estimating evaporation of paved surfaces 64
5.1 Grass-reference evapotranspiration ET0................... 64
5.2 Relationship between Eand ET0for paved surfaces . . . . . . . . . . . . 65
5.2.1 TUBGRmodel............................. 66
5.2.2 Determining the ratio between Eand ET0.............. 67
5.2.3 Outlook................................. 72
6 Conclusion 73
A Appendices I
A.1 Extended tables & figures . . . . . . . . . . . . . . . . . . . . . . . . . . . I
A.2 Lysimeter reconstruction . . . . . . . . . . . . . . . . . . . . . . . . . . . . VII
ii
List of Figures
2.1 Examples of different pavement types . . . . . . . . . . . . . . . . . . . . 6
2.2 Pavement elements and water transport processes . . . . . . . . . . . . . . 8
2.3 Hydrological balance in summer & winter . . . . . . . . . . . . . . . . . . 16
3.1 Basic principle of vegetated and paved weighable lysimeters . . . . . . . . 23
3.2 Lysimeterstation................................ 24
3.3 Lysimetersurfaces ............................... 24
3.4 Scheme of lysimeter set-up . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.5 Sensorplacement................................ 26
3.6 Surfacewetting................................. 28
3.7 Dataprocessing................................. 29
3.8 Example of raw runoff data . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.9 Principle of AWAT filter routine . . . . . . . . . . . . . . . . . . . . . . . 33
3.10 Precipitation comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.1 Rainfalldistribution .............................. 38
4.2 Hydrologicalbalance.............................. 39
4.3 Cumulative hydrological balance . . . . . . . . . . . . . . . . . . . . . . . 41
4.4 Hydrological balance event intensities . . . . . . . . . . . . . . . . . . . . 42
4.5 Hourly hydrological balance – period with highest E . . . . . . . . . . . . 45
4.6 Hourly hydrological balance – period with highest P . . . . . . . . . . . . 46
4.7 Evaporation on dry days following rainfall . . . . . . . . . . . . . . . . . . 49
4.8 Annual mean temperature profile . . . . . . . . . . . . . . . . . . . . . . . 50
4.9 Daily temperature profiles . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.10 Differences in monthly mean temperatures . . . . . . . . . . . . . . . . . . 52
4.11 Hourly air, surface & soil temperatures – period with highest E . . . . . . 53
4.12 Monthly air & soil temperatures for different surface covers . . . . . . . . 55
4.13 Example hourly soil water content and temperatures . . . . . . . . . . . . 56
4.14 Surface wetting-drying process (timeseries) . . . . . . . . . . . . . . . . . 58
4.15 Surface wetting-drying process (thermal) . . . . . . . . . . . . . . . . . . . 59
4.16 Correlation matrix for hourly data . . . . . . . . . . . . . . . . . . . . . . 62
4.17 Correlation matrix for daily data . . . . . . . . . . . . . . . . . . . . . . . 63
5.1 Relationship between daily Eand ET0.................... 68
5.2 Monthly estimation of evaporation (E) with TUBGR model . . . . . . . . 69
iii
5.3 Deriving monthly κfrom Tair ......................... 70
5.4 Comparison of monthly estimations of E................... 72
A.1 Hourly air, surface & soil temperatures – period with highest P . . . . . . V
A.2 Deriving monthly κwand κdfrom Tair .................... VI
A.3 Impressions from lysimeter reconstruction in April 2016 . . . . . . . . . . VII
A.4 Sensorinstallation ...............................VIII
iv
List of Tables
2.1 Scientific communities researching pavement . . . . . . . . . . . . . . . . . 5
2.2 Pavement surface properties . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3 Previous studies of annual hydrological balances . . . . . . . . . . . . . . 13
2.4 Short period or single component observations of hydrological balances . . 14
2.5 Sealing degrees of paved surfaces . . . . . . . . . . . . . . . . . . . . . . . 17
3.1 Soil and paver properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.1 Climatological conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.2 Measured annual hydrological balance . . . . . . . . . . . . . . . . . . . . 40
4.3 Dayswithoutevents .............................. 43
4.4 Soilwatercontent ............................... 48
4.5 Evaporation on days with and without precipitation . . . . . . . . . . . . 50
5.1 TUBGR model coefficients for paved surfaces . . . . . . . . . . . . . . . . 67
5.2 Annual hydrological balance estimated by TUBGR model . . . . . . . . . 67
5.3 Monthly and annual estimated evaporation . . . . . . . . . . . . . . . . . 71
A.1 Monthly mean volumetric water content (θ)................. I
A.2 Hourly surface & soil temperatures . . . . . . . . . . . . . . . . . . . . . . II
A.3 Monthly mean air, surface & soil temperatures . . . . . . . . . . . . . . . III
A.4 Ratio evaporation to grass-reference evapotranspiration . . . . . . . . . . IV
v
List of Abbreviations
Symbols and units
CdConstant used in ET0calculation [s m−1]
CnConstant used in ET0calculation [K mm s3Mg−1d−1]
eaActual vapour pressure [kPa]
esSaturation vapour pressure [kPa]
EActual evaporation [mm]
ECOB Evaporation from cobblestone surface [mm]
ECON Evaporation from concrete slab surface [mm]
ET Actual evapotranspiration [mm]
ET0Grass-reference evapotranspiration estimated after Penman-Monteith [mm d−1]
E(T)Actual evaporation or evapotranspiration, as case may be [mm]
GSoil heat flux density at soil surface [MJ m−2d−1]
IInfiltration [mm]
ICOB Infiltration from cobblestone surface [mm]
ICON Infiltration from concrete slab surface [mm]
MMass
Mlys Mass of lysimeter [kg]
Mout Mass of outflow [g] or [kg]
Mro Mass of runoff [g] or [kg]
PPrecipitation [mm]
P∗Net precipitation (P−RO) [mm]
PsPrecipitation in summer [mm]
RnNet radiation at crop surface [MJ m−2d−1]
RsIncoming solar radiation [MJ m−2d−1]
RH Relative humidity [%]
RO Runoff [mm]
ROCOB Runoff from cobblestone surface [mm]
ROCON Runoff from concrete slab surface [mm]
ROmRunoff at 1-minute resolution [mm min−1]
R2Coefficient of determination based on Pearson correlation coefficient
∆SChange in soil water storage [mm]
Tair Air temperature [◦C]
Tdew Dew point temperature [◦C]
vi
Tpav Temperature at underside of paver [◦C]
Tsoil Soil temperature [◦C]
Tsurf Surface temperature [◦C]
T5Temperature at 5 cm depth below paver [◦C]
T15 Temperature at 15 cm depth below paver [◦C]
T25 Temperature at 25 cm depth below paver [◦C]
u2Wind speed at 2 m height [m s−1]
u10 Wind speed at 10 m height [m s−1]
βsInfiltration coefficient for summer [-]
βwInfiltration coefficient for winter [-]
∆Slope of saturation vapour pressure-temperature curve [kPa ◦C−1]
γPsychrometric constant [kPa ◦C−1]
κReduction coefficient (=E/ET0) [-]
κdReduction coefficient for dry days [-]
κwReduction coefficient for wet days [-]
κTReduction coefficient derived from air temperature [-]
κCOB Reduction coefficient for cobblestones [-]
κCON Reduction coefficient for concrete slabs [-]
ρSpearman correlation [-]
θVolumetric water content (VWC) of soil [Vol.-%]
θCOB VWC of soil under cobblestones [Vol.-%]
θCON VWC of soil under concrete slabs [Vol.-%]
θ5VWC of soil at 5 cm depth below paver [Vol.-%]
θ15 VWC of soil at 15 cm depth below paver [Vol.-%]
θ25 VWC of soil at 25 cm depth below paver [Vol.-%]
vii
Abbreviations
AWAT Adaptive Window and Adaptive Threshold – filter routine
COB Cobblestones
CON Concrete Slabs
CPP Classic Permeable Pavement
DPP Designed Permeable Pavement
GSV Google Street View
LPP Low Permeability Pavement
OSM Open Street Map
RC Runoff Coefficient
SD Sealing Degree
VWC Volumetric Water Content (θ)
viii
1 Preface
Urbanisation has been a central topic across numerous scientific disciplines. One main
aspect linked to urban areas are streets and sidewalks, covered in different kinds of
paving materials. These paved surfaces are a necessary infrastructure of cities. At
the same time, their characteristics lead to an altered hydrological balance compared
to rural counterparts. With increasing flood risks, runoff processes and storm water
management have become an important topic and subject of numerous models. While
these extreme events are clearly relevant given their consequences, a research focus on
them also led to missing data for regular conditions and their corresponding hydrological
balance. However, the hydrological balance of different pavement types is essential to
assess the impact of urbanisation in general and urban design in particular. An improved
understanding of underlying processes would enable urban planners to better adapt cities
to current and future challenges.
Therefore, central questions are:
•What is the hydrological balance of paved surfaces?
•Which water and heat transport processes take place?
•Are these adequately reflected in common models?
•How can evaporation from paved surfaces be estimated?
This thesis aims to answer all these questions, combining an extensive literature
review, own measurements, and an analysis of what this means for estimating evapora-
tion. The review considers all kinds of pavement, own research takes a closer look at
two pavement types commonly used for sidewalks.
Chapter 2, a detailed introduction into the hydrological balance of paved surfaces,
was published in Timm et al. (2018). It contains a summary of previous studies on this
subject and collects the state of the art knowledge of related processes. This literature
review revealed that very few measurements of water transport processes are available. It
concludes that additional research of these would be valuable valuable for both, process
understanding and modelling approaches. Therefore, a detailed study of the hydrological
balance, as well as water and heat transport processes, was conducted for two paved
surfaces: cobblestones and concrete slabs. A combination of high-resolution weighable
lysimeters and sensors installed within the soil column of the lysimeters was utilised
to gain new insights. Chapter 3 details the materials and methods of the lysimeter
1
study and an additional experiment. The results of these are illustrated and analysed in
chapter 4. They confirm and enhance many of the observations from previous studies.
Additionally, previously disregarded processes are re-evaluated. Resulting from the high
resolution and combination of methods, it can be shown that these overlooked processes
take place and play an important role for the hydrological balance of these two paved
surfaces. Finally, the results of the measurements are used in chapter 5 to evaluate
an existing model and develop a new approach for estimating evaporation from paved
surfaces. A conclusion is drawn in chapter 6, summarising the most important results
of this study and providing an outlook for future research and practice.
2
2 Introduction
Rapid urbanisation is a process attracting worldwide attention in many different scientific
research areas. As of 2014, 54 % of the global population was living in cities, with 59
countries exceeding 80 % urbanisation and some countries like Belgium reaching up to
98 % (UN, 2014). This trend is expected to continue with two thirds of the world
population becoming urban by 2050 (UN, 2014). Linked to urbanisation is a significant
alteration of our environment which creates many challenges, such as flooding (Haase,
2009; Pistocchi et al., 2015; Qin et al., 2013), air pollution (Rodríguez et al., 2016) and
altered, heterogeneous microclimates (Chatzidimitriou and Yannas, 2015).
Of these alterations, soil sealing is often seen as the main driver for the challenges
attributed to urbanisation, with many studies from numerous fields focusing on the
hydrological impact of paved areas (e.g. Bhaduri et al., 2001; EC, 2012; Scalenghe
and Marsan, 2009). Depending on the discipline, pavements may be seen in many
different ways, e.g. as necessity for urban life, runoff generator and pollution source,
storage for heat and water, or a product to be optimized. Table 2.1 gives an overview
of communities researching and designing pavements. While some of these have been
dealing with pavements for a long time (e.g. road construction and urban planning), the
subject might be relatively new to others (urban water ecology and climatology). Based
on different priorities (e.g. safety, function, ecological impact, drainage), they make use
of a wide range of measurement scales and methods. From a soil science perspective,
soil sealing is considered to irreversibly destroy natural soil functions (Morel et al., 2014)
and to reduce the soils ecological functionality (Lehmann and Stahr, 2007). Similarly,
urban hydrologists perceive soil sealing as the cause for drastically altered hydrological
balances, which is most visible in storm water and flood generation after heavy rainfalls
(Hibbs and Sharp, 2012; Salvadore et al., 2015). These impacts illustrate why paved
surfaces are seen as the key urban water interface, determining transformation and
transport processes of water, matter, and energy between the soil and atmosphere in
urban areas (Gessner et al., 2014).
With a focus on addressing safety issues such as flood prevention and increased
water pollution, pavements are mostly seen as an impermeable runoff generator (Fletcher
et al., 2013; Jacobson, 2011) and pollutant source (Göbel, Dierkes and Coldewey, 2007).
Accordingly, over the last few decades models have been developed for water transport
at the surface, their input to streams and sewer networks, and the effectiveness of storm
This chapter is the accepted version of an article published in Landscape and Urban Planning,
(Timm et al., 2018), doi: http://dx.doi.org/10.1016/j.landurbplan.2018.03.014. Alterations with new
results are marked.
3
water remediation measures such as localised recharge areas. Soil sealing with any
material is often defined as the prevention of any infiltration and evaporation, resulting
in very little or no losses in the rainfall-runoff relationship (Fletcher et al., 2013; Mansell
and Wang, 2010). However, it is well known in construction and building material science
that asphalt and concrete mixtures do take up moisture, which leads to damage to the
material and shortening of the service life (Kakar et al., 2015; Liu and Hansen, 2016b;
López-Montero and Miró, 2016; Penttala, 2006; Xu et al., 2016). Further evidence for
the permeability of pavements is provided by studies assessing the overall hydrological
balance of paved surfaces in Europe. Under moist mid-latitude (Cfb) climate (updated
Köppen-Geiger classification after Kottek et al., 2006), studies (see section 4) have shown
that infiltration and evaporation take place for all types of pavement, including asphalt.
Knowledge about these processes could also be used to actively design and utilize
pavements for specific purposes. For example, water may be applied to streets in order
to increase evaporation and thereby cool cities that struggle with the urban heat island
effect (Daniel et al., 2018; Hendel et al., 2016). For storm water mitigation, specifically
designed pervious paving materials are used to increase infiltration and reduce runoff
volumes (Booth and Leavitt, 1999; Brattebo and Booth, 2003; Haselbach et al., 2006;
Solpuker et al., 2014).
Due to increased contaminant loads, water quality of urban runoff has been the
subject of hundreds of studies (Göbel, Dierkes and Coldewey, 2007). Kayhanian et al.
(2012) analysed storm water runoff from highways and observed an increase of mean
concentration of most contaminants with traffic density. While few data is available for
sidewalks and bicycle paths, the available studies indicate considerably smaller concen-
tration of most contaminants for these surfaces (Göbel, Dierkes and Coldewey, 2007).
Additionally, a large fraction of contaminants is bound to particulate matter (Kayhanian
et al., 2012) which can be partly retained by the material filling joints between paving
stones (Nehls et al., 2008).
An examination of recent reviews about urban hydrology from the last 15 years,
whether focusing on runoff (e.g. Fletcher et al., 2013), catchment modelling (e.g. Sal-
vadore et al., 2015), groundwater recharge (e.g. Lerner, 2002) or the overall impact of
urbanisation (Gessner et al., 2014; Jacobson, 2011; Shuster et al., 2005), reveals that they
all reach the same conclusion: urban water cycle as well as its basic physical processes
are not yet well understood.
4
Field “Paving is...” Elements & Keywords Priority
Urban Necessity for Land-use, Urban life, Design, Function,
Planning [1] urban life Habitat quality, Safety,
and functions Streetscape Public health
Road Product to Material properties, Safety,
Construction [2] be optimized Freeze-Thaw-Cycles, Durability,
Deterioration Cost efficiency
Urban Water Runoff generator, Urban water cycle, Water quality
Management [3] Pollution source Water supply, Water quantity,
Drainage, Sanitation, Decision tools,
Storm water management, Management
Rainwater harvesting
Urban Runoff generator, Catchment hydrology, Water regime,
Hydrology [4] Pollution source Pollution of surface Flood protection,
& groundwater, Hydrological cycle
Flooding
Urban Heat & Urban heat island, Cooling
Climatology [5] Water storage Albedo, Radiation, (Urban heat
Evapotranspiration, island mitigation)
Heat stress
Soil Science [6] Soil sealing, Water & solute transport, Urban ecosystem
Loss of function Ecosystem functions services
[1] Fukahori and Kubota, 2003; Jung et al., 2017; Kaparias et al., 2015 [2] Corazza et al.,
2016; Kardos and Durham, 2015; Kelly et al., 2016 [3] Gogate et al., 2017; Willuweit and
O'Sullivan, 2013 [4] Chen et al., 2017; Jacobson, 2011 [5] Qin, 2015 [6] Morel et al., 2014;
Scalenghe and Marsan, 2009
Table 2.1: Disciplines dealing with water transport processes of paved surfaces
2.1 Paved surfaces
Surfaces, such as streets and sidewalks, can be paved using a wide range of materials.
Paving can consist of a single continuous cover (e.g. asphalt or concrete) or an assembly
of individual pavers (made of e.g. stone, concrete, or brick). In the latter case, the paved
surface will also feature numerous joints of varying width between the pavers. These
joints are filled with seam material and may facilitate growth of vegetation. Further-
more, the joints allow water to infiltrate into the underlying soil. Paving with a uniform
cover is denoted as Low Permeability Pavement (LPP) and paving consisting of pavers
and joints as Classic Permeable Pavement (CPP). CPPs differ from Designed Perme-
able Pavement (DPP), which refers to (super) porous concrete or asphalt, and other
new materials specifically designed as storm water remediation methods allowing more
infiltration (Andersen et al., 1999; Bonicelli et al., 2015; Carbone et al., 2014; Haselbach
et al., 2006; Yong et al., 2013). In this article, pavement is always defined as surface con-
sisting of paving material and (if present) joints between individual pavers. Examples of
various common pavements are shown in figure 2.1. Surfaces with all types of paving are
considered sealed soils. The definition of soil sealing often includes a complete absence
5
Figure 2.1: Examples of different pavement types
6
of infiltration (Salvadore et al., 2015; Scalenghe and Marsan, 2009), or generally refers
to any sealing material used for street, parking lots and pavement (Fletcher et al., 2013;
Hibbs and Sharp, 2012; Jacobson, 2011) including bricks, concrete and asphalt (Mansell
and Rollet, 2009; Yao et al., 2016). While CPPs are examples of pervious materials in
some studies (Mansell and Rollet, 2009; Nehls et al., 2006, 2008), they are often grouped
together with LPPs. The term pervious paving mostly refers to DPPs.
2.2 Water transport processes
The studies introduced in the previous section are commonly cited to point out that
infiltration and evaporation are processes that should not be neglected when assessing
urban ‘impermeable’ surfaces (Dupont et al., 2006; Fletcher et al., 2013; Nakayama and
Fujita, 2010; Rodriguez et al., 2008; Salvadore et al., 2015). Due to complex interac-
tions between pavement properties, aging processes (cracks and dust) and climatological
conditions, as well as difficulties in measurements, not all processes governing the water
transport of paved areas are well understood and confirmed. For example, Mansell and
Rollet (2009) pointed out that a dynamic interaction between microclimatic conditions
above the surface and the distribution of water within the paver influences three vital
processes at once: surface wetting, evaporation and infiltration. Figure 2.2 illustrates
relevant elements and water and vapour transport processes of paved surfaces, which
will be explained in detail in the following sections with the corresponding processes in
the schematic figure indicated in square brackets.
2.2.1 Elements and characteristics of pavements
As can be seen in figure 2.2a, the soil-atmosphere interface of paved soils can consist
of numerous elements. Most apparent is the paving material itself, which can have a
wide array of physical properties affecting its hydrological balance and water storage
capacities. Depressions leading to ponding (Mansell and Rollet, 2009; Nehls et al., 2015)
and slopes affecting runoff generation (Hollis and Ovenden, 1988a; Ramier et al., 2006)
also affect infiltration and evaporation. In both cases, cracks can develop over time,
altering flow regimes (Hibbs and Sharp, 2012; Hollis and Ovenden, 1988a; Ramier et al.,
2006). Additionally, the sublayer of the pavement (Starke et al., 2011) as well as the
microclimate above the pavement (Nehls et al., 2015) influence exchange processes. At
a larger scale, the connection of the pavement to drainage systems, vegetated areas and
water bodies has to be considered.
7
a) Elements
vegetation
sidewalk
drainage
street
water body
cracks
depression
sublayer
sublayer
slope
paver
joints
retention
pipe
seam material
seam material
sublayer
gutter
b) Processes
water table
1
2
3
4
5
6?
7
8?
9
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
1 Evapotranspiration from vegetated area
2 Infiltration of precipitation and irrigation
2 Infiltration of precipitation and irrigation
2 Infiltration of precipitation and irrigation
2 Infiltration of precipitation and irrigation
2 Infiltration of precipitation and irrigation
3 Runoff to vegetated areas
4 Evaporation from paver pores
5 Evaporation from paver surface storage
6 Evaporation from sublayer through paver
7 Evaporation from pavement joints
8 Evaporation from sublayer through joints
9 Runoff to drainage
10 Evaporation from sublayer through cracks
11 Filling of depressions
12 Evaporation from filled depressions
13 Runoff to water bodies
14 Evaporation from water bodies
15 Capillary rise from sublayer to vegetation
16 Direct (rain) & artificial (irrigation) recharge
17 Infiltration through paver pores
18 Infiltration through paver joints
19 Vapour transport from sublayer to pavement
20 Infiltration through cracks
21 Infiltration from filled depressions
22 Artificial recharge by leakages
23 Localised recharge
24 Indirect recharge
Figure 2.2: a) Elements of the urban soil-atmosphere interface affecting water transport
processes, b) water transport processes of paved surfaces. Processes not yet
verified are marked with a ‘?’∗
2.2.2 Processes
2.2.2.1 Runoff
The generation of runoff [Number 11 in figure 2.2b] depends on rainfall intensities (Flöter,
2006; Rim, 2011; Ragab et al., 2003; Ramier et al., 2006, 2011), the slope (Ramier
et al., 2006) and properties of the pavement surface affecting surface wetting, filling of
depressions, infiltration and evaporation (Mansell and Rollet, 2009; Nehls et al., 2015;
Ragab et al., 2003). Mansell and Rollet (2009) describe and model four stages of runoff
generation depending on surface texture and micro topography of paved surfaces: 1)
∗The two processes marked with a ’?’ were observed during own measurements and are subject of
chapter 4
8
surface wetting, 2) filling of depressions, 3) overflow and runoff generation, and 4) surface
drying. Surface wetting describes upper layers of paver and seam material absorbing
and detaining water. The duration of the surface being wetted depends on the number,
duration and intensity of precipitation events (Wiles and Sharp, 2008). Once this layer
is saturated, runoff will start to flow towards local depressions, where some of the water
will be lost to infiltration [21] and evaporation [12] (Mansell and Rollet, 2009; Nehls
et al., 2015). If rainfall continues, these local depressions will overflow and fill larger
depression areas, finally leading to runoff (Mansell and Rollet, 2009). Clearly, slopes
will affect the rerouting of water flows on the surface. Further, steep slopes can decrease
the surface storage capacities of pavements and hence increase the runoff (Ramier et al.,
2006). Additionally, some water will infiltrate through the joints [18] or through the
porous paver material [17]. Evaporation losses and hence runoff delay are not only
caused by evaporation from filled depressions, but also from heated up paver surfaces
[5, 7] (Wessolek and Facklam, 1997).
Numerous works have determined minimum rainfall intensity for runoff generation,
surface storage capacity, and the depression storage capacity of different paving mate-
rial (table 2.2). Minimum rainfall intensities for runoff formation range from 0.01 to
0.2 mm/min, pavement surface storage capacities from 0.05 to 2.0 mm and depression
storage capacities from 0.07 to 2.0 mm. Differences for the same material may be caused
by the method of determination or by changed properties due to aging processes (Flöter,
2006; Nehls et al., 2006; Wessolek and Facklam, 1997). For example, while granite paving
blocks retained 0.4 mm when tested in the lab, under field conditions runoff started to
form after 1.2 and 1.8 mm were applied during cold and warm weather, respectively
Minimum rainfall intensity Pavement surface Depression
for runoff formation storage capacity storage capacity
[mm/min] [mm] [mm]
Low Permeability Pavement (LPP)
Asphalt 0.1 [6] 0.05 – 1.0 [2][5][6] 0.08 – 2.0 [4][7]
Classic Permeable Pavement (CPP)
Small concrete pavers 0.16 – 0.20 [1] 0.24 [1]
Large concrete pavers 0.01 – 0.18 [1][3][6] 0.17 – 1.4 [1][3][5][6] 0.09 [4]
Small natural stone pavers 0.02-0.09 [3][6] 0.93-2.0 [3][6] 0.08 – 0.22 [4]
Brick paving blocks 0.13 [6] 0.8 – 1.3 [5][6] 0.58 [4]
Granite paving blocks 0.4 – 1.8 [5] 0.07 [4]
Rubber paving blocks 0.19 [4]
Designed Permeable Pavement (DPP)
Porous asphalt 0.43 – 0.45 [2]
Porous concrete 1.41 [4]
[1] Flöter, 2006, [2] Ramier et al., 2004, [3] Rim, 2011, [4] Nehls et al., 2015, [5] Wessolek
and Facklam, 1997, [6] Wessolek, 1994, [7] Hollis and Ovenden, 1988b
Table 2.2: Minimum rainfall intensities for the generation of runoff, surface and depres-
sion storage capacities for different pavement types
9
(Wessolek and Facklam, 1997). On a larger scale, Hollis and Ovenden (1988a) investi-
gated the relationship between precipitation and runoff for concrete and macadam roads,
and observed initial losses averaging 0.8 mm, but reaching up to 8.8 mm.
2.2.2.2 Infiltration
Infiltration mechanisms have not been studied directly, but infiltration has been mostly
attributed to cracks [20] (Wiles and Sharp, 2008) and joints between pavers [18] (Flöter,
2006; Nehls et al., 2006, 2008; Wessolek, 2001). Since infiltration is concentrated at these
points, less water and time is necessary to saturate these flow paths, creating preferential
flow (Wiles and Sharp, 2008). After passing through the joints, water will be distributed
laterally beneath the paver to a certain extent, fully saturating the ‘paving bed’ for small
paving stones but not reaching the mid-section of larger paving stones (Flöter, 2006).
However, a weak correlation between share of joints and infiltration rates indicates that
infiltration occurs not only through cracks and joints (Wessolek and Facklam, 1997),
but also through the porous system of the paver [17] (Liu and Hansen, 2016a,b; Ramier
et al., 2004; Xu et al., 2016).
It has been shown that moisture uptake and hydraulic conductivity of these materi-
als increase in the presence of salt (e.g. introduced as de-icer) and in case of freeze-thaw
or other damage to the material (Liu and Hansen, 2016a,b; Xu et al., 2016; Zhang et al.,
2016). Furthermore, depressions play an important role in infiltrating processes, as
they can store significant amounts of water which will mostly evaporate or infiltrate,
depending on the hydraulic properties of the pavement, its initial moisture and weather
conditions (Nehls et al., 2015). Pavement aging influences infiltration rates, which are
reduced because of fine dust particles entering pavers (clogging the pores) and joints
(changing seam material properties), as well as the increased evaporation through veg-
etation growth (Bonicelli et al., 2015; Flöter, 2006; Nehls et al., 2006; Wessolek and
Facklam, 1997).
When studying infiltration, drip infiltrometers are used to determine maximum in-
filtration rates (Wessolek and Facklam, 1997; Wiles and Sharp, 2008). Depending on
the material, share of joints and age, infiltration rates of 2 to 288 cm/day have been
reported for different paved surfaces (Gilbert and Clausen, 2006; Hollis and Ovenden,
1988a; Wessolek, 1993; Wessolek and Facklam, 1997; Wiles and Sharp, 2008).
2.2.2.3 Groundwater recharge
In contrast to natural soils, water that infiltrated paved soils are not stored and later
evaporated and transpired through plants and capillary uprising, but instead contribute
to groundwater recharge (Flöter, 2006; Hibbs and Sharp, 2012; Wessolek, 2001) if it is
not intercepted by subsurface structures or drained into leaking sewer pipes (Bricker
et al., 2017; Dirckx et al., 2016). Natural direct recharge, where water infiltrates and
recharges at the point of precipitation [16] (Wiles and Sharp, 2008), is limited to adja-
10
cent vegetated areas. Though direct recharge may be decreased, storm water detention
ponds as well as cracks and joints of pavements can act as preferential flow paths, in-
creasing localised recharge [23] (Hibbs and Sharp, 2012; Lerner, 2002; Wiles and Sharp,
2008). Additionally, artificial recharge [22] caused by leaking water supply and drainage
networks, as well as over irrigation of parks, further compensates for decreased direct
recharge (Lerner, 2002; Wiles and Sharp, 2008). Rates of leakage between 20 and 25 %
are common, but can reach up to 50 % in some water mains (Lerner, 2002). Finally,
recharge can occur as indirect recharge from losing streams [24] (Hibbs and Sharp, 2012).
Flow paths in the sublayer of pavements are governed by numerous factors affecting per-
meability and hydraulic conductivity, such as the sublayer material, buried structures,
fractures and utility line trenches (Wiles and Sharp, 2008).
2.2.2.4 Evaporation
Evaporation from paved areas occurs from the upper surface storage of the pavers [5]
(Flöter, 2006; Hassn et al., 2016; Mansell and Rollet, 2009; Ragab et al., 2003; Ramier
et al., 2004, 2011), the porous network of the whole paver [4] (Garcia, Hassn, Chiarelli
and Dawson, 2015; Hassn et al., 2016; Ramier et al., 2004) the joints between pavers
[7] (Flöter, 2006; Wessolek, 2001; Wessolek, Kluge, Nehls and Kocher, 2009), from free
standing water in filled depressions [12] (Mansell and Rollet, 2009; Mansell and Wang,
2010; Nehls et al., 2015) and through cracks [10].
Only a small portion of the water infiltrating through the joints will be held against
gravity by the material and subsequently evaporate [7] (Flöter, 2006). Once the topmost
few centimetres of the seam material have dried, its hydraulic conductivity is reduced
significantly. This leads to almost no upward transportation of water through the joints,
which effectively inhibits evaporation from the sublayers through the joints [8] (Flöter,
2006; Wessolek, 2001). The same concept might apply for evaporation through cracks
[10]. However, pavers themselves can play a significant role in evaporation processes,
sometimes even surpassing the evaporation from joints (Flöter, 2006).
Depending on the material, asphalt pavements can have an extensive network of air
voids which are very similar to well-studied soil structures, so that the same principles
of water movement in saturated and unsaturated soils can be applied (Garcia, Hassn,
Chiarelli and Dawson, 2015; Hassn et al., 2016).Under a constant energy source, evap-
oration of water stored in the pore network of asphalt [4] follows three stages (Garcia,
Hassn, Chiarelli and Dawson, 2015). In the first stage, water evaporates directly from the
pores to the atmosphere, during which temperature and heat flux through the asphalt
increase. Next, the evaporation rate increases until it reaches its peak, absorbing latent
heat and reducing the temperature of the asphalt. During this stage, the surface of the
asphalt already appears to be dry, as the waterfront is lower in the asphalt and water
moves upward by diffusion. In the last stage, the evaporation rate decreases until the
asphalt is dry and no more water is available, leading to increasing surface temperature.
11
In general, only the topmost layer, consisting of the pavement material and joints, is
considered to evaporate, with water that infiltrated to the sublayer being shielded from
evaporation [6] (Berthier et al., 2006; Flöter, 2006; Wessolek, 2001; Wiles and Sharp,
2008).
Determined by temperature gradients, water vapour will move downward in sum-
mer, when the upper layers are warmer than the lower ones, and upward in winter when
the lower levels are warmer [19] (Flöter, 2006). However, this water will not evaporate
through the pavement layer but condense at the pavement underside, increasing the wa-
ter content of the soil directly beneath the pavers, making the pavement an evaporation
barrier (Flöter, 2006; Wessolek, 2001). This process together with different hydraulic
conductivities between the sublayer and paver indicate that no evaporation from water
in the sublayer through the paver [6] will take place.
2.3 Hydrological balance
Investigations of pavement water storage and fluxes gained popularity after studies sug-
gested that some general concepts do not fit observations. For example, groundwater
recharge in urban areas was considered to have been declining as an effect of increased
soil sealing, preventing infiltration (Berlekamp, 1987; Jacobson, 2011; Scalenghe and
Marsan, 2009). However, some studies showed that urban groundwater tables were in-
creasing instead of decreasing (Hibbs and Sharp, 2012; Lerner, 2002). This effect has
been attributed to leakage from water supply networks (Fletcher et al., 2013; Lerner,
2002; Wiles and Sharp, 2008), altered flow regimes of semi-permeable pavements (CPPs)
allowing (reduced) infiltration while decreasing evaporation (Nehls et al., 2008), or storm
water remediation practices involving localised infiltration of runoff (Göbel, Coldewey,
Dierkes, Kories, Meßer and Meißner, 2007). Today, it is well established that pavement,
including asphalt, is not truly impervious. In the following, studies providing measure-
ments of urban hydrological balance are introduced. All studies except for one were
conducted in Europe under moist mid-latitude (Cfb) climate, which is characterised as
warm temperate, fully humid, and with warm summers (Kottek et al., 2006). One study
by Dreelin et al. (2006) was conducted in Georgia, USA, under Cfa (warm temperate,
fully humid, with hot summers) climate.
In order to measure the urban hydrological balance (infiltration, runoff and evapo-
ration), different methods are applied. Most commonly, the well-known application of
lysimeters for natural surfaces is transferred to paved ones. These lysimeters mostly
consider only a small area and can consist of only the paving material (e.g. Mansell and
Rollet, 2006; Ramier et al., 2004) or also include a soil sublayer (e.g. Flöter, 2006; Rim,
2011. One might also determine the runoff for a certain area by measuring the flow in the
connected drainage system, and combine this with precipitation measurements and soil
moisture profiles for monitoring infiltration processes (Ragab et al., 2003). The resulting
hydrological balances of studies that measured at least two of the three processes over
12
Study Location & Pavement SD Annual Summer Winter
climate zone P I RO E P I RO E P I RO E
mm mm mm mm mm mm mm mm mm mm mm mm
(%) (%) (%) (%) (%) (%) (%) (%) (%) (%) (%) (%)
Wessolek Berlin, Road (asphalt) IV 610 50 441 119 325 25 226 74 285 25 215 45
2001 Germany; (100) (8) (72) (20) (100) (8) (69) (23) (100) (9) (75) (16)
Cfb Sidewalk (concrete n.a. 610 233 191 186 325 110 77 138 285 123 114 48
and cobblestones) (100) (38) (31) (31) (100) (34) (24) (42) (100) (43) (40) (17)
Diestel & Berlin, Small granite stones II 575 426 40 109
Schmidt Germany; (100) (74) (7) (19)
2001* Cfb Small cobblestones II 575 385 52 138
(100) (67) (9) (24)
Interlocking II 575 449 57 69
concrete pavers (100) (78) (10) (12)
Rubber pavers II 575 495 63 17
(100) (86) (11) (3)
Grass pavers I 575 391 35 149
(100) (68) (6) (26)
Brick pavers II 575 437 57 81
(100) (76) (10) (14)
Ragab Crowmarsh Asphalt car park IV 501 40 351 110
et al. Gifford, (100) (8) (70) (22)
2003 UK; Asphalt car park IV 501 45 351 105
Cfb (100) (9) (70) (21)
Block paving car park IV 501 30 351 120
(concrete sublayer) (100) (6) (70) (24)
Asphalt road IV 501 30 351 120
(100) (6) (70) (24)
Flöter Hamburg, Small concrete II 852 670 111 71 391 266 71 54 461 404 40 17
2006 Germany; pavers road (100) (79) (13) (8) (100) (68) (18) (14) (100) (89) (8) (3)
Cfb Large concrete pavers III 852 447 360 45 391 179 175 37 461 268 185 8
sidewalk (100) (53) (42) (5) (100) (47) (44) (9) (100) (59) (39) (2)
Gravel parking lot I 852 573 55 221 391 177 25 186 461 396 30 35
(100) (67) (7) (26) (100) (44) (6) (50) (100) (86) (6) (8)
Rim Berlin, Small cobblestones II 552 383 85 84 303 181 67 55 249 202 18 29
2011 Germany; (100) (70) (15) (15) (100) (60) (22) (18) (100) (81) (7) (12)
Cfb Large concrete pavers III 549 348 145 56 302 163 106 33 247 185 39 23
(100) (64) (26) (10) (100) (54) (35) (11) (100) (75) (16) (9)
* selected materials (without barrier layer)
Table 2.3: Observed annual hydrological balances of paved surfaces as absolute values and with percentage of precipitation in brackets; SD =
sealing degree (see table 2.5); Cfb = warm temperate, fully humid with warm summer
13
Study Location & Pavement SD P I RO E Measurement
climate zone mm mm mm mm Period &
(%) (%) (%) (%) Remarks
Ramier Nantes, Old asphalt IV 433 9 316 108 03/09/2002 to
et al., France; concrete (100) (2) (73) (25) 06/01/2003
2004 Cfb New asphalt IV 252 8 186 58 08/01/2002 to
concrete (100) (3) (74) (23) 16/05/2002
Porous asphalt II 197 114 32 51 06/09/2001 to
(100) (58) (16) (26) 08/01/2002
lysimeter, no soil
Mansell & Paisley, Flat III n.a. n.a. n.a. n.a. 9 weeks in winter
Rollet, UK; concrete slabs (100) (1) (69) (30) lysimeter, no soil
2006 Cfb Inclined III n.a. n.a. n.a. n.a.
concrete slabs (100) (2) (93) (5)
Hot-rolled IV n.a. n.a. n.a. n.a.
asphalt (100) (0) (56) (44)
Dense Bitumen IV n.a. n.a. n.a. n.a.
Macadam (100) (0) (36) (64)
Bricks II n.a. n.a. n.a. n.a.
(100) (54) (9) (37)
Dreelin Athens, Asphalt IV 46 33 9 storm events
et al., Georgia, (100) (72) between February
2006 USA; Grass pavers I 46 5 and April 2003
Cfa (100) (1)
Hollis & Redbourn, Bitumen Mac., IV 392 11658 1 year (1983)
Ovenden, UK; 1-2.4 % slope (100) (3-8) storm events
1988b Cfb Bitumen Mac., IV 392 16 (57% of annual rain)
3.6-3.9 % slope (100) (4) catchment size
Bitumen Mac., IV 392 16-106 100-3500 m2
4-5.4 % slope (100) (4-27) gully buckets
Hot-rolled IV 392 71 & rain gauges
asphalt (100) (18)
Ramier Nantes, Asphalt road IV 2141 1606 3 years
et al., France; (100) (75) catchment size
2006 Cfb Asphalt road IV 2315 1458 300-500 m2
(100) (63) gully buckets
& rain gauges
Table 2.4: Short period or single component observations of hydrological balances of
paved surfaces; precipitation (P), infiltration (I), runoff (RO) and evaporation
(E) as absolute values and with percentage of precipitation in brackets; SD =
sealing degree ranging from low (I) to severe (IV) (see table 5); climate zones
after Kottek et al. (2006): Cfb = warm temperate, fully humid with warm
summer, Cfa = warm temperate, fully humid with hot summer
14
a period of at least a year are summarised in table 2.3. To our best knowledge, there
are only five studies providing this long-term information. Studies that measured only
one component and used models to derive the other two, or only offer data for shorter
periods, are not considered but may be found in table 2.4.
While asphalt materials (LPPs) exhibit very similar results, with high runoff values
(70-72 %), low infiltration (6-9 %) and moderate evaporation (20-24 %), CPP materials
such as concrete pavers are showing a wide spread, with lower runoff values (7-42 %),
higher infiltration (38-86 %) and low to moderate evaporation (3-24 %). Only one long-
term measurement for a DPP material is available from Diestel and Schmidt (2001),
where grass pavers show low runoff (6 %), high infiltration (68 %) and high evaporation
(26 %). For porous asphalt (DPP), data is available from Ramier et al. (2004) of a
four month measurement campaign in winter, which yielded low runoff (16 %), medium
infiltration (58 %) and high evaporation (26 %). Nevertheless, this winter measurement
cannot be transferred to an annual hydrological balance, as there are distinct seasonal
differences between summer and winter periods. Figure 2.3 compares the hydrological
balances of five materials for summer and winter graphically. As can be seen, the pro-
portion of infiltration is higher during winter, rising from 44-68 % in summer to 59-89
% in winter. The difference is highest for gravel, where the proportion of infiltration in
winter is almost double that in summer. Evaporation and runoff tend to be higher in
summer, increasing from 2-12 % to 9-50% and 6-39 % to 6-44 %, respectively. Again,
the difference in evaporation is most distinct for gravel, which is more than six times
larger in summer. The highest increase in runoff has been observed for small cobble-
stones, which triples from 7 % in winter to 22 % in summer. This increase of runoff
and evaporation and decrease of infiltration from winter to summer can be attributed
to different precipitation patterns and other climatic conditions, such as the amount of
available energy. Despite these general trends, it also becomes apparent that there are
very large variations between different materials.
While few studies provide annual observations for all hydrological balance compo-
nents, some only offer data for shorter periods or for one component (table 2.4). For
example, the rainfall-runoff relationship of urban catchments covered in paving material
has been measured using gully buckets and rain gauges (e.g. Hollis and Ovenden, 1988b;
Ramier et al., 2006). Some insights into evaporation processes of paved surfaces have
been won by using a tunnel evaporation gauge to study evaporation independent of the
other hydrological balance components (Göbel et al., 2008; Starke et al., 2010). These
studies compare evaporation rates of a CPP consisting of concrete pavers and a DPP
made of pervious concrete. Göbel et al. (2008) concluded that in summer, the CPP
results in higher evaporation rates than the DPP for dry days, but the relationship is
reversed after precipitation events, for which the DPP exhibits higher evaporation rates.
In winter, the DPP has higher evaporation rates. Additionally, the pattern of evapo-
ration differs. While the CPP are characterised by short and high evaporation rates
directly following precipitation events, the DPP continued to evaporate even after four
15
days of dry and sunny weather after a precipitation event (Starke et al., 2010). These
results support the concept of significant losses due to infiltration and evaporation. Fur-
thermore, they again illustrate the diverse behaviour of very similar materials. This
might be influenced by rainfall intensity and distribution (Rim, 2011).
infiltration runoff evaporation
small concrete pavers
small natural stone pavers
large concrete pavers
large concrete pavers
gravel
winter summer
Hydrological balance of paved
surfaces in summer & winter
Figure 2.3: Hydrological balances for summer and winter for different paved surfaces
after data from Flöter (2006) and Rim (2011). The area of the circle reflects
the percentage of the component (infiltration, runoff or evaporation) of the
total precipitation during the period. For gravel, percentage of runoff for
summer and winter is equal, leading to overlapping circles
2.4 Application
Many methods have been developed to assess the urban hydrological balance. Depending
on their intended application, they range from rather rough estimations to complex and
detailed numerical models. They are useful tools that allow classification and evaluation
of different urban developments. In the following, a few of these methods are introduced.
In case of the sealing degree approach, an adapted classification based on the previously
presented hydrological balance measurements is proposed.
16
2.4.1 Sealing degree (SD)
Most models assessing the overall hydrological balance focus on annual or monthly esti-
mations and utilize easy to use parameters and readily available data, such as the sealing
degree (SD), which represents the degree of imperviousness and is provided for many
cities (e.g. Berlekamp, 1987; Glugla et al., 1987; Haase, 2009; Jacqueminet et al., 2013;
Montzka et al., 2008). To that end, certain paving materials or land uses are combined
in classes for which the SD ranges from low (open concrete stones with grass in between)
to severe (asphalt street) (Glugla et al., 1999). The SD attributed to the same land use
can vary significantly depending on the process used, e.g. ranging from 40 to 100 % for
streets and 20 to 50 % for single-family houses (Wessolek, 2001). An adapted classifica-
tion of SDs based on surface cover and the resulting hydrological balance is presented in
table 2.5.
There are some empirical, user-friendly models allowing for more or less rough es-
timations of urban hydrological balances depending on the SD. For example, Wessolek
et al. (2008) proposed the Hydro-Pedo-Transfer-Functions (HPTFs) to predict annual
percolation rates for numerous surface covers, including urban areas. In this model,
infiltration coefficients for predicting infiltration and runoff in summer and winter are
provided for four different SDs. Furthermore, they introduce an empirical method for
estimating the annual actual evaporation of (partially) sealed surfaces that adjusts the
annual potential evapotranspiration as described by Allen et al. (1994) by introducing a
reduction coefficient based on lysimeter measurements.
Sealing Degree I – low II – medium III – high IV – severe
Surface cover <70 % 70 – 94 % 94 – 98 % >98 %
Examples Grass pavers, Small pavers of Large concrete Asphalt
gravel stone/concrete/brick pavers
Properties Focus on joints Many narrow to Very few and No joints or vegetation,
and vegetation wide joints, narrow joints, cracks caused by
vegetation possible very little vegetation damage
Hydrological balance:
Runoff None Low Medium High
0 – 7 % 3 – 15 % 16 – 42 % 70 – 72 %
Evaporation High Low to high Low to medium Medium
26 % 8 – 35 % 8 – 22 % 20 – 24 %
Infiltration Medium to high High Medium Low
58 – 68 % 63 – 79 % 53 – 64 % 6 – 9 %
Table 2.5: Sealing degrees of paved surfaces based on different cover and observed hy-
drological balance (annual means, not applicable for heavy rainfall events).
Table has been altered compared to published version to include results from
own conducted measurements as detailed in chapters 3 & 4
17
2.4.2 Runoff coefficient (RC)
In urban hydrology, runoff coefficients (RC) are used to describe how much of the precip-
itation reaching a (partially) sealed urban surface will turn into surface runoff and leave
the area. Many models use one RC describing any kind of sealed surface that is often
set to 1.0 (e.g. Pistocchi et al., 2015). Still other models distinguish between materials
or application (e.g. Angrill et al., 2017) and might differentiate summer and winter
RC (Hollis and Ovenden, 1988b; Ragab et al., 2003). For the typical sealing materials
asphalt and concrete, annual RC of 0.4 to 0.9 have been reported (Angrill et al., 2017;
Eckley and Branfireun, 2009; Ramier et al., 2006). The percentages of runoff provided for
different materials in tables 2.3 and 2.4 can be translated into annual and seasonal RCs.
Other factors such as climate and slope may further affect the RC (Grodek et al., 2011).
Recently, studies have suggested that individual rainfall events have to be analysed in
order to get specific RCs related to the rainfall-intensity (e.g. Diestel and Schmidt,
2001; Dreelin et al., 2006; Gilbert and Clausen, 2006; Rim, 2011). While RC are widely
used, not least due to their practicability and simplicity, they are also viewed critically
as they do not consider all relevant processes determining the hydrological behaviour of
paved surfaces (e.g. Ramier et al., 2011). Hence, some studies additionally consider the
micro-topography of sealed surfaces, which may include depressions which reroute and
store water and have to overflow to form runoff (e.g. Mansell and Rollet, 2009; Mitchell
et al., 2001; Nehls et al., 2015).
2.4.3 Numerical models
Urban water and vapour transport processes are simulated with numerical models by
many different disciplines on various scales, with different components and outcomes.
One might include water supply and drainage infrastructure to assess the overall water
cycle (Bach et al., 2014; Delleur, 2003; García, Barreiro-Gomez, Escobar, Téllez, Quijano
and Ocampo-Martinez, 2015; Mitchell et al., 2001; Niemczynowicz, 1999), simulate the
generation of storm water flows and their impact on urban rivers and lakes, as well as
flooding events (Miller et al., 2014; Revitt et al., 2014; Zoppou, 2001), or estimate urban
evapotranspiration (Berthier et al., 2006; Grimmond and Oke, 1991).
In general, the majority of studies and models focus solely on storm water events and
neglect smaller rainfall events. As Zoppou (2001) shows, there are hundreds of storm
water models developed by academic institutions, regulatory authorities, government de-
partments and engineering consultants, which range from simple conceptual to complex
hydraulic models. Commonly used models might not be applicable to sealed areas, as
they have not been designed for this purpose. For example, Kodešová et al. (2014) sim-
ulated water and heat transfer in soils with different surface covers like grass, gravel and
concrete, using the widely applied hydrological model software HYDRUS-1D (Šimůnek
et al., 2016). While they achieved a good fit for most surfaces, soil water contents below
concrete paving could not be simulated in a satisfactory way, which was attributed by the
18
authors to the 3D water flows under the pavers. Additionally, it has been pointed out,
that water and heat transport processes in urban areas are influenced by more extreme
temperature variations, in which case commonly neglected vapour transport processes
have to be considered when modelling such areas (Kodešová et al., 2014).
Urban hydrological balance models also often use ‘tank’-systems, in which the dif-
ferent surfaces and their underlying soil layers are seen as tanks with different storage
capacities from which the water is redistributed (e.g. Mansell and Wang, 2010; Mitchell
et al., 2001). There are only few models estimating evaporation from (partially) sealed
surfaces, a lack that has been pointed out over the last 20 years in the field of hydrology
(Grimmond and Oke, 1991; Mansell and Wang, 2010). Existing models may be based
on the calculation of available water storage and the filling level of that tank. For exam-
ple, the model AQUACYCLE by Mitchell et al. (2001) considers surface depressions of
impervious surfaces (including paving) as water storage. Any precipitation higher than
the volume of the depression will form runoff, while the remaining water will completely
evaporate without any infiltration. Other models originate from urban climatology; they
usually consider larger urban sites with grass as well as paved surfaces, and are based on
interception and surface energy balances (e.g. Grimmond and Oke, 1991; Ward et al.,
2016).
2.4.4 Data generation
Readily available data and maps of urban areas often contain land-use, area covered by
road networks, or SDs of larger areas, but do not standardly provide information on
paving material. While images obtained via remote sensing can be used to gain insight
into built-up area and different land-uses such as industrial, housing and infrastructure
(García and Pérez, 2016; Morabito et al., 2016; Smith et al., 2010), their resolution is
too low to distinguish different paving materials.
Detailed information about paving material might be obtained from OpenStreetMap
(OSM) (Open Street Map Contributors, 2017) which provides free geographical infor-
mation collected by local contributors. One attribute assigned to road and sidewalk
areas is the surface, which by default is considered to be paved. The term paved is seen
as non-specific and includes all types of paving materials including asphalt, sett stones,
and wood. However, the surface cover can be further specified with a wide range of
materials (e.g. asphalt, sett, paving stones, cobblestones, grass pavers, gravel) allowing
for detailed pavement surface data (Open Street Map Wiki Contributors, 2017). Data
availability and level of detail depend on the amount of contributors in the area. For
the city of Tehran, Forghani and Delavar (2014) compared road maps from OSM and
official (municipally produced) maps and found considerably varying quality of OSM
maps. They point out that heterogeneity of completeness of OSM data compared to
the reference map is the main factor influencing the quality. If the paving material is
specified, OSM provides high-detail data that is not easily available from other sources.
19
Another possibility is the usage of Google Street View (GSV) images (Google, 2017)
to replace or support field surveys. GSV is used in research for numerous applications,
e.g. the estimation of urban tree canopy cover (Richards and Edwards, 2017; Seiferling
et al., 2017) and assessment of urban space quality (Chiang et al., 2017). Similarly, GSV
can be used to determine the paving material of streets and sidewalks that are available
in the data base.
2.5 Conclusion
Pavement, as a soil-atmosphere interface of urban surfaces, is more than a runoff genera-
tor. For nearly all paving materials, infiltration and evaporation takes place, significantly
reducing runoff volume. Infiltration takes place through the porous system of the paving
material, joints of pavers and cracks. Evaporation occurs from the total surface storage
capacity of paver and seam material and from water standing in depressions. While
these processes take place for all materials, their extent and interactions vary consider-
ably. The range of paving materials is too large to provide empirical measurements for
all of them, especially concerning the complex interaction of many different conditions
(e.g. age of material, slope and climatic conditions). For now, the presented studies
allow a rough estimation of the hydrological balance of different paved surfaces based
on their sealing degree.
The drastic differences between materials, the importance of their condition (e.g.
frost damage, cracks) and surface micro topography indicate that in order to accurately
simulate the hydrological balance of urban areas, it is crucial to assess a much smaller
scale than the commonly used parameterisation over large areas. New data sources, such
as Google Street View and OpenStreetMap, can provide detailed, small-scale information
about paving material used in a study area.
Overall, a better understanding of the hydrological processes of pavements is essential
to improve our model concepts and planning choices, making cities more resilient against
the impacts of rapid urbanisation, climate change, and its manifold effects on urban
residents.
20
3 Material & Methods
The aim of this study is to measure the hydrological balance of two different kinds
of paved surfaces, as well as underlying processes of heat and water transport. Over
a measurement period of one year, a combination of weighable lysimeters and built-
in temperature and soil water content sensors are used to achieve this. After noise
filtering, data aggregation and combination, hourly and daily results could be obtained.
The following sections offer detailed information about the lysimeter station and data
processing.
3.1 Lysimeter studies
Lysimeters are a well established and widely used method to study water movement
across a soil boundary (Howell et al., 1991), the soil-atmosphere interface. They are
used to measure actual evapotranspiration, rainfall and drainage (Meissner et al., 2010).
In 2004, Lanthaler (2004) conducted a survey on European lysimeters and found 117
institutions operating lysimeters or sewage water samples at over 178 sites. According
to the survey, there were 2440 lysimeters of which 84 % were non-weighable. Two thirds
of all lysimeters were used to research arable land, one fourth for grassland and 1 % for
forests. As of 2010, there were more than 1700 publications using lysimeters, with a
gradual increase in more recent years (Meissner et al., 2010). Despite being cost and
maintenance intensive, fixed to a site and requiring extensive data processing (Hannes
et al., 2015; Rana and Katerji, 2000), lysimeters are a popular and valuable method to
study hydrological processes.
There are two basic types of lysimeters: weighable and non-weighable. While weigh-
able lysimeters are a direct measurement approach using the mass balance of the system,
non-weighable lysimeters are indirect and rely on the volume balance (Hirschi et al., 2017;
Howell et al., 1991). Non-weighable lysimeters are suitable for long-term (annual) water
balances for which the change in soil water storage (∆S), which is only measured by
weighable lysimeters, can be neglected (DVWK, 1996). Precision weighing lysimeters
(Unold and Fank, 2007) allow highly precise measurements of these processes, as well as
dewfall (Meissner et al., 2007). Due to their improved temporal resolution, newer weigh-
able lysimeter installations can be used to develop and test models of soil hydrological
processes (Meissner et al., 2010).
Both types of lysimeter make use of the general water balance (DVWK, 1996):
21
0 = P+E(T) + RO + ∆S(3.1)
PPrecipitation [mm]
E(T)Evapo(transpi)ration [mm]
RO Runoff [mm]
∆SChange in soil water storage [mm]
Using this general water balance equation and assuming that precipitation and
evapo(transpi)ration do not take place at the same time, the mass balance used for
weighable lysimeters, both vegetated and paved, can be written as (based on Schrader
et al. (2013); Peters et al. (2014)):
M=⎧
⎨
⎩
Mlys +Mout if vegetated
Mlys +Mout +Mro if paved (3.2)
∆P=⎧
⎨
⎩
∆Mfor ∆M > 0
0for ∆M≤0
(3.3)
∆E(T) = ⎧
⎨
⎩
∆Mfor ∆M < 0
0for ∆M≥0
(3.4)
where
MTotal mass of system [kg]
Mlys Mass of lysimeter [kg]
Mout Mass of outflow (infiltration water) [kg]
Mro Mass of runoff [kg]
∆MChange in total mass of system [kg]
∆PPrecipitation [kg]
∆E(T)Evaporation (paved) or evapotranspiration (vegetated) [kg]
Figure 3.1 illustrates the water movement processes and measurement principle of
weighable lysimeters. The difference between vegetated and paved lysimeters is the
additional collection and weighing of runoff water generated by most paved surfaces.
This additional runoff water has to be considered when determining the precipitation
volume. For vegetated lysimeters, a common measurement error of up to 20 % results
from the difference in vegetative area and the inner dimensions of the container when
determining the evaporating area (Rana and Katerji, 2000). This source of error does
not apply to the sealed lysimeters used in this study, as the evaporating area consisting
of pavers and joint material is more fixed and corresponds to the inner dimensions of
the container.
22
Δ
S
( )
M
lys
I
( )
M
out
P ET
infiltration
water
container
lysimeter
soil column
Δ
S
( )
M
lys
P E
RO
( )
M
ro
tipping
bucket
I
( )
M
out
a) vegetated lysimeter b) paved lysimeter
Figure 3.1: Basic principle of vegetated and paved weighable lysimeters. ∆S=change of
water storage, P=precipitation, I=infiltration, ET =evapotranspiration,
E=evaporation, RO =runoff. Weighted components are Mlys =mass of
lysimeter, Mout =mass of outflow (infiltration), and Mro =mass of runoff
3.2 Lysimeter station
All measurements were conducted on the southern outskirts of Berlin, Germany. There,
the two lysimeters used in this study are situated as part of a lysimeter station operated
jointly by the Federal Environmental Agency (Umweltbundesamt) and the Technical
University Berlin (N 52.3967◦, E 13.3673◦, climate zone: Cfb; warm temperate, fully
humid). The overall station consists of twelve lysimeters and a climate station situated
next to the lysimeters. Figure 3.2 shows the two lysimeters with paved surfaces, two
grass-reference lysimeters used in other studies, and the climate station.
3.2.1 Surface types
The two lysimeters used in this study are covered in two pavement sealing types com-
monly used for sidewalks in Berlin (Fig 3.3). The first one (type “cobblestones”) uses
“Bernburger Mosaic” paving which consists of second hand cobblestones of different
sizes. It is characterised by large a joint area (joints are 20 % of total surface area)
and is often used at the edges of pavements to better include trees, rain pipes, drainage
holes and other structures that have to be circumvented. The other one (type “concrete
slabs”) uses concrete slabs (30×30×4.4 cm), resulting in narrow joints (joints are 6 % of
total surface area). While both surfaces can be classified as classic permeable pavements
(CPPs), their sealing degree (table 2.5 differs. Due to the large joint area, cobblestones
23
belong to the sealing degree II (medium) and concrete slabs to sealing degree III (high).
Both pavings are installed with a slight slope to prevent the formation of puddles. The
immediate surrounding is covered in the same concrete slabs to prevent island effects.
Both material types have been used in previous studies (Rim, 2011), resulting in slightly
weathered surfaces.
Figure 3.2: Research site – lysimeter station at Berlin-Marienfelde
Figure 3.3: Lysimeter surfaces: left – cobblestones, right – concrete slabs
24
3.2.2 Lysimeter set-up
scale
scalescale
data
logger
PC
tipping
bucket
infiltration
water
automatic
drainage
runoff
drainage
data transfer
soil &
surface
temperature
soil water content
soil temperature
pavement layer
seam material
construction
sand
layered
sand & gravel
drainage layer
scalescale
scale
scale
scalescale
rain
gauge
slope
Figure 3.4: Scheme of lysimeter set-up
A schematic set-up of the lysimeters is illustrated in Fig 3.4. They have a circular surface
area of 1 m2and a depth of 50 cm. The surface layer is 4 to 8 cm deep consists of the
paving material and fine sand seam material. Beneath, the soil layer consists of common
construction sand. The last 10 cm are a drainage layer consisting of gravel and very
coarse sand whose grain size increases towards the bottom. Substrate properties can be
found in table 3.1. The drainage layer ensures the transport of infiltrated water to the
bottom of the lysimeter where it drains into a mesh-covered pipe. The infiltrating water
is collected in a canister standing on top of a scale (resolution 0.1 g = 0.0001 mm). In
order to reduce maintenance effort, the canister collecting infiltration water is emptied
automatically using pumps once a high degree of filling is reached. The lysimeters
themselves are weighted using three weighing cells each (resolution 10g = 0.01 mm). A
covered drain surrounding the surface area collects runoff and empties it into a weighable
tipping bucket based on the device introduced by Nehls et al. (2011). When runoff takes
place, a first tipping bucket fills up gradually until the maximum volume of the bucket is
reached, resulting in a tipping which drains the bucket and positions the second bucket
to be filled up. This one will then tip over in the other direction when full, bringing the
first bucket back in position, and so forth. Normally, only the event of buckets tipping is
recorded, so that events not resulting in a tipping leads to difficulties in aligning runoff
volumes and the corresponding time periods. Additionally, runoff collected in a bucket
may evaporate if no tipping occurred, which leads to an underestimation of the overall
runoff volume. In this study, the weight of the tipping buckets is logged (resolution
25
0.1 g = 0.0001 mm) to enable the registration of even small amounts which would not
result in a tipping and evaporate before the next rainfall event leading to runoff. Data
from all scales is logged simultaneously; lysimeter weight and infiltration water at 1-
minute, runoff at 1-second intervals. Maintenance consists of regular removal of weeds
growing in the seems and leaves from the runoff drain, replacement of logger batteries,
and general check ups.
3.2.3 Built-in sensors
Supplementing the more traditional lysimeter measurements, sensors were installed in
each of the lysimeters to measure:
•surface temperature (Tsurf )[◦C]
•temperature below paver (Tpav)[◦C]
•soil temperature (Tsoil) at 5, 15, and 25 cm below the underside of the paver [◦C]
•volumetric water content (θ) at 5, 15, and 25 cm below the underside of the paver
[%]
Soil temperature and θare measured simultaneously by the same sensors (sensor
type 5TM by Decagon), which make use of the time-domain reflectometry principle.
These sensors are installed at 5, 15, and 25 cm below the underside of the paver, with
two sensors at each of these depths. The distance between lysimeter wall and sensors is
roughly 25−35 cm, the distance between the sensor pair is approximately 50 cm (Figure
3.5). In order to prevent cables running directly over another sensor, which would affect
the water flow and hence θ, sensors at the three different depths are shifted slightly so
that they do not align vertically. For each lysimeter, two temperature sensors (one on
the surface, another directly below a paver) are installed. Additionally, a rain gauge on
ground level between the two lysimeters provides precipitation data. All data is logged
at 5-minute intervals.
(a) schematic placement (b) installed: placement indicated by stones
Figure 3.5: Horizontal placement of the build-in sensors
26
Cobblestones Concrete slabs
Paving layer
Paver material Natural stone Concrete
Paver dimension [cm]2−6×4−6×4−8 30 ×30 ×4.4
Seam share [%] 20 6
Paver porosity [Vol−%]∗3.9 5.1
Field capacity [Vol−%]∗<0.51
Surface storage capacity [mm]∗0.4 0.8
Seam material
Depth [cm]0−8 0 −5
Soil type Sand
Dry bulk density ρb[g cm−3]†1.15 −1.67
Particle size distribution
Clay [%] 0.8
Silt [%] 2.3
Fine sand [%] 20.4
Medium sand [%] 70.6
Coarse sand [%] 6.0
Sublayer material
Depth [cm]8−40 5 −40
Soil type Sand
Dry bulk density ρb[g cm−3] 1.59
Particle size distribution
Clay [%] 0.4
Silt [%] 1.2
Fine sand [%] 9.7
Medium sand [%] 76.3
Coarse sand [%] 12.4
Van Genuchten parameters
Scale parameter α[cm−1] 0.0496
Saturated water content θs[cm3cm−3] 0.287
Residual water content θr[cm3cm−3] 0.037
Shape parameter n[−]5.518
Drainage layer
Particle size [mm] at depth 40 −42 [cm]1.00 −2.00
Particle size [mm] at depth 42 −43 [cm]2.00 −3.15
Particle size [mm] at depth 43 −46 [cm]3.00 −5.00
Particle size [mm] at depth 46 −50 [cm]5.60 −8.00
∗(Wessolek and Facklam, 1997)
†(Nehls et al., 2006)
Table 3.1: Properties of lysimeter soils and paving materials. Particle size distribution
determined using DIN/ISO 11277
27
3.2.4 Climate station
The climate station which is situated right next to the two paved lysimeters is operated
by the Federal Environmental Agency (Umweltbundesamt) and provides hourly and
daily data of:
•precipitation (P)[mm]
•air temperature (Tair)[◦C]
•wind speed (u)[m s−1]
•relative humidity (RH)[%]
•dew temperature (Tdew)[◦C]
•solar radiation (Rs)[W m−2]
•daily sunshine hours (N)[h]
3.3 Surface wetting-drying experiment
Figure 3.6: Example of surface wetting
In order to further research the paved layer as urban soil-atmosphere interface, a
wetting-drying experiment was undertaken. Its focus was the interaction of surface
temperature and surface water storage for the two paved surfaces used in the lysimeter
study.
The experiment was carried out at the lysimeter site on July, 25th, 2016, which
was characterised by very high air temperatures and low cloud cover, offering optimal
conditions for evaporation (mean Tair = 23.87 ◦C, daily Rs= 270 W m−2).
A thermal camera (IR-TCM 384 IR – Jenoptik, resolution 384 ×288 pixel and
<0.05 K) was temporarily installed above the lysimeter using a mobile tripod. It was
set up to take an individual picture every 4 seconds, using the IRBIS®remote 3 software
28
(InfraTec, 2012). After starting the automatic image taking, the surface was watered
using a hand pump pressure sprayer with low pressure setting (figure 3.6). This allowed
to produce an even water layer completely covering the lysimeter surface. This wetting
process took 2 minutes. After that, the entire drying process (until no visible water left
on surface) was recorded. The thermal recording took place in the early afternoon, when
Tair reached its daily maximum of 30 −31 ◦C,Rswas high (670 −763 W m−2h−1), RH
was low (33 −35 %), and wind speed was moderate with 1.7 m s−1.
Before thermal pictures were taken, the tripod was used to record the wetting-drying
process with a regular camera (Olympus E-510, resolution 3648×2736 pixels at 314 dpi).
Pictures were taken automatically every 10 seconds.
3.4 Data processing
Figure 3.7: Overview of origin and processing of data
Measurement period in this study started on June 3rd, 2016 and ended June 2nd, 2017.
Within this period, data is available for 333 out of 365 days (91 %). Data gaps resulted
from conducting additional experiments (e.g. wetting-drying experiment) resulting in
artificial data (n= 4) and from technical disruptions such as power shortage or logging
problems (n= 28). A larger gap resulted from a power shortage on site from October
4th to October 20th 2016. Raw data was logged at different time-intervals ranging from
29
1-second (runoff) to 1-hour (climate). After data processing, the highest temporal reso-
lution is 1-hour. This results from data availability and requirements of used filters. An
overview of data processing can be found in figure 3.7 and details can be found in the
following sections. All data has been converted to CET without daylight saving.
3.4.1 Tools
Thermal pictures obtained during the wetting-drying experiment were processed using
the Infrared Thermography Software IRBIS®3 Professional (InfraTec, 2008). For appli-
cation of the AWAT filter routine software was provided by Andre Peters. Non-linear
regression was conducted using the Datafit 9.1.32 software by Oakdale Engineering (Oak-
dale Engineering, 2014). All other data processing was done using R version 3.3.1 (R
Core Team, 2016). Apart from packages included in the standard edition of R, the
following additional packages have been used for data processing and visualisation:
•data.table (Dowle and Srinivasan, 2017)
•ggcorrplot (Kassambara, 2016)
•ggplot2 (Wickham, 2009)
•lubridate (Grolemund and Wickham, 2011)
•padr (Thoen, 2017)
•plyr (Wickham, 2011)
•zoo (Zeileis and Grothendieck, 2005)
For selection of accessible (colour-blind safe and photocopy friendly) colours for the
plots, the Color Brewer tool at colorbrewer2.org (Brewer et al., 2013) has been used.
3.4.2 Lysimeter data
Raw lysimeter data contains the measured weight of the lysimeters (1-minute interval),
the weight of the infiltration water container (1-minute interval) and the weight of runoff
tipping bucket (1-second interval). This section provides details on how this raw data is
processes to obtain hourly results.
3.4.2.1 Runoff
Runoff is recorded by measuring the weight of the tipping buckets in which the runoff
water is collected at 1-second intervals. Once a certain amount of water is reached,
the tipping device will tilt, emptying the collected water and collecting water in the
second bucket. Depending on the runoff rate, this tipping process occurs at a weight
of 480 −500 g. For heavy rainfall, tipping can occur in quick succession, so that a high
30
temporal resolution of 1-second is necessary to record weight changes before a tipping
occurs. If a tipping occurs, the tipping bucket will sometimes sway from side to side
before coming to rest after 4−5seconds. Figure 3.8 contains one month of raw runoff
data for concrete slab surface. As can be seen, runoff may occur between tipping events
(bucket filling period between two tipping events). Weighing of the tipping buckets allows
for better temporal assignment of events. The figure also illustrates evaporation from
the buckets, which would result in reduced runoff if only tipping events were recorded.
Figure 3.8: Example of raw runoff data (concrete slabs surface)
In order to process and evaluate the runoff data, these tipping events have to be
adjusted. At 1-second- resolution, a value of ROs<−1 g is used as tipping indicator, as
neither noise nor evaporation would cause this change of weight. Since the bucket needs
4−5seconds to come to rest, the next five data points after a tipping was triggered
are considered invalid as well. At 1-second-resolution, the runoff rate can be assumed
stable for small periods of time. Hence, the six values associated with a tipping event are
replaced by the mean runoff rate immediately before and after the event. The tipping
event function used to adjust the 1-second-resolution raw runoff data is:
if ∆ROi<−1g:
∆ROi,∆ROi+1, ..., ∆ROi+5 =∆ROi−1+ ∆ROi+6
2(3.5)
where
31
∆RO Recorded change of weight (runoff) at 1-second-resolution [g].
Once the data series has been adjusted for tipping events, the 1-second-resolution
data is aggregated to 1-minute values. At this resolution, the overall change of weight in
case of runoff is larger than the recorded noise of the scale. The 1-minute values include
evaporation processes which lead to a decrease of weight (RO ≤0 mm min−1). Only
values RO > 0 mm min−1are used to obtain hourly runoff data, so that:
ROm=⎧
⎨
⎩
ROmif ROm>0
0if ROm≤0
(3.6)
where
ROmRunoff at 1-minute-resolution [mm min−1].
3.4.2.2 Precipitation, Evaporation & Infiltration
Lysimeter measurements are affected by numerous disturbances resulting in noise, which
has to be filtered from the data in order to successfully separate and quantify the indi-
vidual fluxes (Hannes et al., 2015; Schrader et al., 2013). In this study, lysimeter mass
(Mlys) and infiltration mass (Mout) are processed using the improved Adaptive Window
and Adaptive Threshold (AWAT) filter routine (Peters et al., 2016) resulting in hourly
values of precipitation, evaporation and infiltration. In order to use the AWAT routine,
which does not include runoff in its balance, for paved surfaces, hourly runoff sums are
added to the precipitation values produced by AWAT.
The AWAT filter routine
The AWAT filter routine was developed to process raw lysimeter data and receive im-
proved separation of precipitation and evaporation from noise (Peters et al., 2014). It
tackles the problem of compromising between too strong and too weak filtering caused by
varying atmospheric conditions (e.g. strong wind) producing varying levels of noise and
requiring different filter set-ups (Peters et al., 2014). Three benchmark events (smooth
evaporation, heavy precipitation, and strong wind) that were used to evaluate the AWAT
filter routine are shown in figure 3.9. The AWAT filter routine combines a smoothing
routine over a certain averaging window with a threshold value to separate noise from
significant weight changes (Peters et al., 2014). The innovation of this routine is the
variation of both, the averaging window width and the threshold value, depending on
the characteristics of the measured data (Peters et al., 2016). First results indicated
that the routine was able to increase overall accuracy and fulfil the requirements set
by the three differing benchmark events (Peters et al., 2014). Since then, it has been
successfully utilised in other studies (Peters et al., 2016). In 2016, an improved version
32
has been published (Peters et al., 2016). While the original AWAT filter routine (Peters
et al., 2014) used stepwise interpolation, the improved version uses linear and spline
interpolations in order to better reflect the smooth and continuous evapotranspiration
processes and enable a higher temporal resolution of the output data (Peters et al.,
2016). Figure 3.9 illustrates the changes between the original filter (denoted as ‘Steps’)
and the improved version (denoted as ‘Linear’ and ‘Spline’). In this study, the improved
routine has been utilised to process raw lysimeter data.
Figure 3.9: Principle of AWAT filter routine. Top: different types of events to be eval-
uated (Peters et al., 2014); middle and bottom: difference between original
and improved filter routine (Peters et al., 2016)
Data preparation In order to use this routine, the automatic drainage of the infiltration
water container has to be corrected beforehand to obtain the required input format.
Beneath each lysimeter, a container collects infiltration water leaving the lysimeter at
33
the bottom. The weight of this container is continuously measured at 1-minute intervals
using scales with a resolution of 0.1 g. If infiltration takes place, the weight will increase
as the container fills up. If no infiltration takes place, the weight may decrease slowly
as small amounts of water evaporate from the container. The sum of weight positive
changes in weight is the amount of water passing the lysimeter body as infiltration.
A level sensor monitors the water level within the container and triggers automatic
draining to prevent overflowing and reduce manual maintenance input. Depending on
the infiltration rate, this process takes 3−4minutes. If one container triggers the
drainage, both containers are drained simultaneously.
During processing, drainage of the container is detected automatically by monitoring
the change of weight. A threshold difference of −500 g min−1, which is high enough to
not be triggered by evaporation processes, initiates a function to replace values recorded
during the drainage process.
When a drainage process is detected, the original value that triggered the routine,
as well as the next 4 values (enabling a 5 minute time window for drainage) are first
removed from the data and then replaced by the value immediately following the drainage
process, preserving the overall increase in weight occurring within the 5-minute interval
of drainage.
The drainage event function used to adjust the 1-minute-resolution infiltration water
data is:
if ∆Mouti<−500g:
∆Mouti,∆Mouti+1 , ..., ∆Mouti+4 = ∆Mouti+5 (3.7)
where
∆Mout Recorded change of weight (infiltration) at 1-minute-resolution [g].
After these corrections, the lysimeter weight and infiltration water weight data have
been converted to the required input data file formats and run through the AWAT filter
which provides hourly data of precipitation, infiltration and evaporation.
Control As a control, precipitation values produced by both lysimeters have been com-
pared, which resulted in a very good fit of 0.98 and 0.99 for hourly and daily values,
respectively (figure 3.10).
34
Figure 3.10: Comparison between measured precipitation (hourly and daily) from
lysimeters, climate station (1.5 m height) and rain gauge at ground level
35
3.4.3 Sensor data
Volumetric water content (θ) and soil temperature (Tsoil) are measured simultaneously
by the built-in sensors (Decagon 5TM). To measure θ, the sensors make use of time
domain reflectometry, which has to be calibrated using the soil temperature (Ledieu
et al., 1986). The following equation by Stoffregen (1998) (based on Ledieu et al., 1986)
was used to calibrate measured θ:
θcorrected =θmeasured ×(︃1 + Tsoil −20
400 )︃(3.8)
Both, V WC and Tsoil are measured at 5-minute intervals. For each depth (5, 15 & 25
cm below paver underside), two sensors were used for measurement in order to generate
a mean value for each depth. While the continuous measurement worked well in general,
individual sensors were prone to drop out for single values or longer periods. One sensor
(lysimeter with concrete slabs at 25 cm depth) proved unreliable and data produced by
this sensor has been removed completely. This results in only one measurement being
available for this particular depth. The sensors used to measure Tsurf and Tpav proved
to be very reliable with mostly continuous measurement. Some additional data gaps
affecting all sensor measurements resulted from power shortage of the logger or other
logging problems.
Data processing of the available sensor data consisted of:
1. Noise removal (for θ)
2. Mean value calculation for sensor pairs (for θand Tsoil)
3. Hourly mean value calculation
4. Daily mean value calculation
First, rare cases in which individual data points of θare wrong had to be filtered out
as noise. To achieve this, a threshold of 1 % change in θwithin the 5-minute resolution
was set. If the change of θexceeded this threshold, the data point was removed. After
this exclusion of noise, mean values of θand Tsoil were produced for each sensor pair. If
one sensor of the pair did not provide data, the value of the corresponding other sensor
was used. Finally, these mean values and the values of Tsurf and Tpav, for which only
one data point per lysimeter was available, were used to produce hourly mean values.
To ensure that enough data points were recorded to accurately assign an hourly mean
value, only hours in which at least 6out of the maximum 12 data points were available
had mean values assigned to them. If less than 6values were available, the hourly mean
value was set to NA. The resulting hourly data was then aggregated to mean daily
values.
36
4 Results
4.1 Lysimeter study
4.1.1 Climatological conditions
Table 4.1 compares the climatological conditions recorded at the lysimeter site during the
measurement period to values of Berlin during the World Meteorological Organization
(WMO) reference period spanning from 1981 to 2010 (DWD, 2018a,d). Within the
measurement period of one year, a total of about 397 mm precipitation was measured
by the lysimeters. This does not include 17 days in October and individual other days
for which lysimeter data is unavailable. As shown in figure 3.10, daily precipitation
data of the lysimeters and the climate station installed in close vicinity to the lysimeters
(see figure 3.2) are not identical but in general fit well (R2= 0.95). Replacing missing
precipitation data from the lysimeters with values recorded by the climate station yields
a total of 476 mm a−1. Based on reference period, the annual mean precipitation of
Berlin is 590 mm a−1(DWD, 2018d). The annual precipitation during the measurement
period was 79.3 % of the reference mean value, indicating a rather dry period.
Apart from the total amount, the intensity of rainfall events plays an important role
for hydrological processes, especially for paved surfaces. Figure 4.1 illustrates hourly
intensities of precipitation events as measured by the two lysimeters. The highest inten-
sity recorded was 10.06 mm h−1, but only 19 out of 2196 hours with rainfall exceeded
2.5 mm h−1. As can be seen in figure 4.1, nearly all hours with rainfall had light inten-
sities below 0.5 mm h−1. The classification of rainfall intensities into light, moderate, or
heavy can differ significantly. Comparing two European classifications, light rainfall can
range from under 0.5(MetOffice, 2012) to under 2.5 mm h−1(DWD, 2018e). Depending
1981 – 2010 Measurement period
Annual mean air temperature (Tair) [◦C] 9.7 10.1
Coldest month: Mean Tair [◦C] JAN: 0.6 JAN: -1.2
Warmest month: Mean Tair [◦C] JUL: 19.5 JUL: 19.8
Precipitation (P) [mm a−1] 590 476
Sunshine hours (n) [h a−1] 1706 2426
Annual global radiation (Rs) [kW h m−2] 1021–1040 1088
Table 4.1: Climatological conditions at the site during the measurement period
(2016-06-03 to 2017-06-02) compared to the long-term climatological mean
values of the WMO reference period 1981 −2010 for Berlin (DWD, 2018a,d)
37
on which classification is used, light rainfall intensities can account for 30 −80 % of the
annual precipitation amount in the measurement period. Heavy rainfall accounted for
3−11 %.
Figure 4.1: Rainfall intensities measured by lysimeters (mean value) from June 2016 to
June 2017. Only events with precipitation >0 mm h−1considered. Top:
frequency of rainfall intensities; middle: classification of rainfall intensities
after 1) DWD (2018e) and 2) MetOffice (2012); bottom: contribution of
rainfall intensities to annual sum of rainfall
4.1.2 Hydrological balance & water transport
4.1.2.1 Hydrological balance
Annual and seasonal (summer and winter) hydrological balances of the two paved lysime-
ters are illustrated in figure 4.2, with detailed values in table 4.2. Infiltration is equal
for both surfaces with 61.9−62.5 % of the sum of infiltration, evaporation and runoff.
Both have higher infiltration in winter, with 67 −70 % of the total amount of infiltration
being recorded between November and April. In contrast, evaporation and runoff differ
significantly for the two surfaces. Runoff amounts to 2.6−16.0 % annually, with summer
runoff being twice as high as in winter. Concrete slabs produced more than five times
more runoff than the cobblestones. In winter, both surfaces evaporate the same amount
of water. In summer, cobblestones evaporate twice as much as concrete slabs, so that
38
annually cobblestones evaporate 65 % more.
Figure 4.2: Hydrological balance of paved surfaces (annual, winter and summer) as mea-
sured June 2016 to June 2017 in Berlin. Hydrological summer May to Octo-
ber, hydrological winter November to April
Runoff generation depends on on the rainfall intensity and duration. Figure 4.3
shows the cumulative hydrological balance based on daily values. For cobblestones,
44 % of all runoff was generated on just two days with high intensity precipitation events.
On June 17th, 2016, precipitation of 18.70 mm d−1or 0.4−3.9 mm h−1led to runoff of
1.07 and 5.76 mm d−1for cobblestones and concrete slabs, respectively. The highest
hourly precipitation and runoff rates were recorded on September 17th, 2016. While
daily precipitation rates were lower than in the previous example (12.36 mm d−1), a
high hourly rate of 10.44 mm h−1results in high runoff of 3.46 and 8.37 mm d−1and 3.46
and 8.25 mm h−1for cobblestones and concrete slabs, respectively. Both of these days lie
within the example periods for hourly hydrological balances illustrated in figure 4.6. The
cumulative hydrological balances further show seasonal trends. While for cobblestones,
evaporation increases in warm months and comes to a halt in cold periods, concrete slabs
have a slow but steady increase of evaporation throughout the year. For both surfaces,
infiltration increases sharply between November and April, when precipitation is high
whereas temperature and evaporation are relatively low.
39
Cobblestones Concrete Slabs
P I RO E P I RO E
mm mm mm mm mm mm mm mm
(%) (%) (%) (%) (%) (%)
Annual 389.16 251.89 10.30 140.55 405.65 257.77 66.71 92.24
(62.5) (2.6) (34.9) (61.9) (16.0) (22.1)
Summer 166.94 74.71 7.14 95.80 175.94 86.32 44.12 49.50
(42.1) (4.0) (53.9) (48.0) (24.5) (27.5)
Winter 222.22 177.18 3.16 44.75 229.71 171.45 22.59 42.74
(78.7) (1.4) (19.9) (72.4) (9.5) (18.1)
June∗48.18 28.57 1.29 25.89 50.97 27.93 12.26 15.51
(51.3) (2.3) (46.4) (50.1) (22.0) (27.9)
July 39.56 13.43 1.51 20.77 40.49 16.23 12.79 10.46
(37.6) (4.2) (58.2) (41.1) (32.4) (26.5)
August 23.94 15.59 0.67 13.81 25.28 13.66 6.74 7.71
(51.9) (2.2) (45.9) (48.6) (24.0) (27.4)
September 16.97 3.74 3.48 11.42 19.16 4.22 9.66 4.98
(20.0) (18.7) (61.3) (22.4) (51.2) (26.4)
October†19.41 10.97 0.01 2.41 18.51 10.35 1.86 2.15
(81.9) (0.1) (18.0) (72.1) (12.9) (15.0)
November 37.51 30.96 0.47 2.65 37.76 24.43 5.19 6.67
(90.8) (1.4) (7.8) (67.3) (14.3) (18.4)
December 37.39 34.32 0.84 4.48 38.45 30.12 4.83 9.98
(86.6) (2.1) (11.3) (67.0) (10.8) (22.2)
January 47.63 40.66 0.99 3.97 49.95 37.28 4.59 4.74
(89.1) (2.2) (8.7) (80.00) (9.8) (10.2)
February 32.41 28.37 0.24 5.84 34.81 28.71 2.70 6.64
(82.3) (0.7) (17.0) (75.5) (7.1) (17.4)
March 46.10 32.66 0.34 13.41 46.83 34.39 3.71 8.78
(70.4) (0.7) (28.9) (73.4) (7.9) (18.7)
April 21.18 10.21 0.28 14.40 21.91 16.52 1.57 5.93
(41.0) (1.1) (57.9) (68.8) (6.5) (24.7)
May 18.88 2.41 0.18 21.50 21.53 13.93 0.81 8.69
(10.0) (0.8) (89.2) (59.4) (3.5) (37.1)
∗Sum of values from June 3rd to June 30th 2016 and June 1st to June 2nd 2017
†No data available for 17 out of 31 days in October 2016
Table 4.2: Results from the lysimeter study: hydrological balance (annual, winter, sum-
mer, monthly) with absolute values [mm] and share of components [%] related
to sum of infiltration, evaporation and runoff. Hydrological summer May to
October, hydrological winter November to April
40
Figure 4.3: Cumulative daily hydrological balance as measured June 2016 to June 2017 in
Berlin. Hydrological summer May to October, hydrological winter November
to April
41
Figure 4.4: Hourly and daily intensities of evaporation, runoff and infiltration events
(only values >0 mm d−1or mm h−1considered), axes in ln(x)
Figure 4.4 offers are more detailed view at intensities of evaporation, infiltration and
runoff events. It considers only values >0, to show the possible extent of intensities.
The occurrence of days without any evaporation, infiltration or runoff is shown in table
4.3. On an hourly basis, evaporation events tend to have a very low rate with a median
value of 0.01 mm h−1. The highest hourly evaporation rates are 2.28 and 1.59 mm h−1for
cobblestones and concrete slabs, respectively. On a daily basis, the median rate for evap-
oration events are 0.33 mm d−1(cobblestones) and 0.24 mm d−1(concrete slabs). Again,
cobblestones reach a higher maximus of 7.64 compared to 4.48 mm d−1. Additionally,
the minimal evaporation rate for ET > 0 mm d−1is higher for cobblestones. Includ-
ing days without evaporation, the daily median values are 0.30 mm d−1(cobblestones)
and 0.22 mm d−1(concrete slabs). Infiltration events have higher rates than evapora-
tion, with a median of 0.03 mm h−1or 0.63 mm d−1(cobblestones) and 0.02 mm h−1or
0.44 mm d−1(concrete slabs). The maximum values were 0.90 mm h−1or 9.85 mm d−1
(cobblestones) and 0.73 mm h−1or 8.66 mm d−1(concrete slabs). While cobblestones
tend to higher infiltration rates, nearly 50 % of the days registered no infiltration for this
surface (see table 4.3). Contrary to that, concrete slabs had more consistent infiltration
with only 9 % of all days without infiltration. Including days without any infiltration, the
median rates are 0.02 mm d−1(cobblestones) and 0.38 mm d−1(concrete slabs). Runoff
42
occurred very rarely for both surfaces, with 76 −80 % of all days without runoff. If
runoff takes place, concrete slabs exhibit significantly higher rates with a median of
0.05 mm h−1or 0.26 mm d−1compared to 0.02 mm h−1or 0.04 mm d−1for cobblestones.
The maximum values were 3.46 mm h−1or 3.46 mm d−1(cobblestones) and 8.25 mm h−1
or 8.37 mm d−1(concrete slabs). In both cases, the highest daily runoff rate can be
attributed to a high-intensity rainfall event taking place within a single hour.
Days without process Cobblestones Concrete slabs
E I RO E I RO
number of days 31 165 267 42 29 252
%of total days 9.3 49.6 80.2 12.6 8.8 75.7
Table 4.3: Number and percentage of days for which no evaporation (E), infiltration (I),
or runoff (RO) was recorded. Total number of days = 333
Apart from intensities, a closer and simultaneous look at the processes over time
is necessary to assess interactions. Figure 4.5 and 4.6 depict the hourly hydrological
balance and soil water movement for two periods of four days each.
The first period (Fig. 4.5) spans from June 15th to June 19th, 2016. It contains
numerous precipitation events with daily intensities of 9.08,0.08,18.70 and 0.97 mm d−1.
The four days were warm with mean daily air temperatures (Tair) of 14.96 −17.43 ◦C,
with higher values on the two drier days. On the two days with high precipitation,
daily solar radiation (Rs) was relatively low with about 104 W m−2compared to about
264 W m−2on the 16th and 18th of June, when there was no to very little precipitation.
Consequently, surface temperature (Tsurf ) decreased, with a daily mean Tsurf drop of
2−4 K on rainy days.
These climatological conditions resulted in the highest hourly and daily evaporation
rates recorded during the measurement period. On June 17th, 2016, surfaces evapo-
rated 7.64 mm d−1(cobblestones) and 4.48 mm d−1(concrete slabs). Hourly, evapora-
tion rates reached up to 2.28 mm h−1(cobblestones) and 1.59 mm h−1(concrete slabs).
As the hourly hydrological balance shows, evaporation takes place on all four days. On
June 16th, 2016, when there was very little precipitation (0.08 mm d−1), both surfaces
evaporated more than on the previous day which had higher rainfall intensities, as well
as 6to 10 times as much as the water available from rainfall on that day. Soil water
content (θ) measurements show that this additionally evaporated water originated from
the underlying soil column. Both lysimeters show increasing θin 5and 15 cm depth
for June 16th, 2016. Since evaporation exceeded the income of rain water on that day,
this increase could not be caused by incoming water. With 0.08 mm provided by rain
water, the remaining 0.72 (cobblestones) and 0.41 (concrete slabs) mm of evaporation
occurring on that day has to have been provided by the soil layers. The same effect was
observed in the second period depicted in figure 4.6, which spans from September 16th
to September 20th, 2016. This period features a high-intensity rainfall event on Septem-
per 17th, 2016, with 12.65 mm d−1, of which 10.44 mm occured in a single hour. The
43
previous and the two following days had no precipitation. Again, daily mean Tair was
relatively high with 15.88 −20.82 ◦Cand reduced Rson the rainy day (80 W m−2com-
pared to about 165 W m−2). In this case, the precipitation event increased the overall
soil water content in all layers. In the two following days, no rainfall provided wa-
ter but upward water transport resulted in 0.98 mm d−1(cobblestones) and 0.3 mm d−1
(concrete slabs) on September 18th, 2016, followed by 0.52 mm d−1(cobblestones) and
0.12 mm d−1(concrete slabs) on September 19th, 2016. This pattern of slightly decreas-
ing daily evaporation rates on the following days shows that upward water transport
is especially effective on days following and between precipitation events. It is highly
dependent on precipitation filling up the soil layers. Both periods show that upward
water transport is responsible for evaporation processes.
Both surfaces led to infiltration on all four days with daily rates of 3.21−6.06 mm d−1
(cobblestones) and 1.08 −5.72 mm d−1(concrete slabs).
4.1.2.2 Water transport processes
The combination of lysimeter data revealing water transport processes leaving the system
and sensor data recording the water movement within the system offered new insights
into processes and their interactions. Tables 4.4 and A.1∗summarise measured soil water
content (θ) at a monthly, seasonal and annual scale. On all of these scales, water tended
to accumulate towards the bottom of the lysimeter, with highest hourly θat 25 cm
depth and lowest at 5 cm. Throughout the year, cobblestone paving resulted in slightly
higher θcompared to concrete slabs. In winter, the difference between the two surfaces
is larger than in summer. While mean hourly θat 25 cm is mostly stable throughout
the year, larger seasonal fluctuation can be observed at 5 cm depth. Over the whole
year, hourly θin all depths reached their minimum in winter and their maximum in
summer (θ25COB has its maximum in winter but nearly the same value in summer with a
difference of 0.08 Vol.-%). A closer look at monthly mean values of θ(excluding October
2016 because of a lack of data in that period) shows that the two surfaces have different
patterns. At the top (θ5), both lysimeters had the highest value in June 2016, the
month with the highest Pin the period, which then gradually decreased towards winter
and then increased again starting February/March. The difference between highest and
lowest θ5is 1.71 Vol.-%(cobblestones, June–November) and 3.44 Vol.-%(concrete slabs,
June–January). In the month with lowest θ5, the lysimeters recorded high infiltration,
high precipitation, and low evaporation.
For this upper layer, the months June, January and March are most interesting.
These are the three months with the highest and similar P. For cobblestones, the net
precipitation (P∗=P−RO) is very similar for these months, ranging from 45.76 to
46.89 mm. Hence, they received roughly the same water input. However, evaporation
amounts varied significantly. June recorded the highest ECOB (25.89 mm), January
∗Figures and tables with indices A are found in the appendices
44
Figure 4.5: Hourly hydrological balances and soil water movement for four days in June
2016, period with highest hourly and daily evaporation event
45
Figure 4.6: Hourly hydrological balances and soil water movement for four days in
September 2016, period with highest rainfall intensity event
46
was among the lowest three months (3.97 mm), and March showed a medium value
(13.41 mm). At the same time, all three months had high infiltration amounts. In
January, the highest monthly ICOB was recorded (40.66 mm), and March and June had
similar infiltration (28.57 −32.66 mm). Combining these observations, it becomes again
clear that large amounts of water are transported upwards in the cobblestone lysimeter.
In June, water infiltrating through the pavement-interface to the upper soil layer will
mostly evaporate, which decreases θ5. Yet, June has the highest mean θ5, which shows
that water is transported upward through capillary rise, increasing the amount of water
available for evaporation. This is further supported by θ15 being high and θ25 being
comparatively low. June and March have very similar P∗COB and ICOB, but ECOB
in March is just half of that in June. The distribution of θCOB shows a shift towards
the lower layers, with the overall highest θ25COB being recorded in March, indicating
less upward water transport which fits the lower evaporation rate. With the same P∗,
January has the highest ICOB and very low ECOB. In this month, most of the water
passing the pavement-interface will pass through the lysimeter and result in infiltration
without or with very little upward water transport. Hence, θ5COB and θ15COB reach their
minimum values. Another example for upward water transport can be found in May
2017. With very low PCOB and P∗COB,ECOB was very high and exceeded incoming
precipitation. Again, θ5COB was relatively high. At the same time, θ25COB and ICOB
reached their minimum in May. Assuming that incoming precipitation is reduced by
runoff and infiltration leaving the system, 5.21 mm were supplied for evaporation by
upward water transport, corresponding to 24 % of the overall evaporation of the month.
In June, January and March, concrete slabs, for which runoff has a higher influence,
resulted in P∗CON = 38.71 −45.36 mm. In June, about a fifth of monthly rainfall left
the system as runoff. The remaining water evaporated from the surface storage or passed
the paved layer, increasing θCON throughout the soil column, leading to the maximum
monthly values for all layers. These high values of θCON occur simultaneously with
the highest monthly amount of ECON , which indicates soil layers and upward water
transport playing an active role in evaporation processes. In January, less of the similar
amount of rainfall leaves the concrete slab surface as runoff and evaporation reaches
its lowest monthly value. Hence, water passes into the soil column. This time, low
temperatures lead to less upward water transport and the water passes the soil column
more quickly. This fast downward water transport is also reflected by the lowest monthly
values of θCON for all layers, despite high precipitation and the highest value of ICON . In
March, the processes of the concrete slab surface are very similar to those of cobblestones.
The main difference is the amount of RO affecting how much water is available in the
system, leading to less E. Again, May is another interesting month worth a closer look.
Compared to the three previously described months, Pwas low (yet three times as
much as the lowest value), with the lowest monthly amount and percentage of RO for
this surface (ROCON = 0.81 mm,3.5 %). With less of the rainfall leaving the system as
runoff, the soil water content increased, with all three depths reaching the second highest
47
Cobblestones Concrete slabs
Depth below paver [cm] 5 15 25 5 15 25
Annual
Min θ[Vol.-%] 4.05 11.36 16.00 3.20 7.69 14.00
Mean θ[Vol.-%] 10.70 14.02 18.18 8.79 13.12 16.64
Median θ[Vol.-%] 10.84 14.00 18.20 9.03 13.15 16.70
Max θ[Vol.-%] 15.83 16.75 19.60 13.95 14.65 18.89
Summer
Min θ[Vol.-%] 10.09 13.05 16.62 8.03 12.70 16.00
Mean θ[Vol.-%] 11.19 14.53 18.00 9.68 13.48 16.96
Median θ[Vol.-%] 11.13 14.57 18.00 9.64 13.44 16.92
Max θ[Vol.-%] 15.83 16.75 19.52 13.95 14.65 18.89
Winter
Min θ[Vol.-%] 4.05 11.36 16.00 3.20 7.69 14.00
Mean θ[Vol.-%] 10.22 13.53 18.36 7.94 12.78 16.33
Median θ[Vol.-%] 10.55 13.65 18.50 8.33 12.85 16.38
Max θ[Vol.-%] 13.33 14.70 19.60 10.97 14.00 18.02
Table 4.4: Hourly soil water content θmeasured within the lysimeter soil column.
Hydrological summer May to October, hydrological winter November
to April
monthly values of θCON recorded. This differs from θCOB, which increased for upper
layers but decreased for a depth of 25 cm. For cobblestones, high amounts of water
were transported upwards and evaporated, leading to high Eand very low I. Under
concrete slabs, upward water transport resulted in a medium amount Ebut the highest
monthly percentage of Erelative to P(ECON = 37.1 %). Still, most water that reached
the soil layers left the system as infiltration rather than evaporation. Given that for this
month, the temperature difference between the two surfaces reached its maximum for all
layers with concrete slabs leading to higher values (figure 4.10), the difference in Ecan
not be solely assigned to energy available for evaporating processes. Instead, the effect
of pavers acting as evaporation barrier with upward transported water condensing on
the underside of the paver and travelling back downward (Flöter, 2006; Wessolek, 2001)
is likely to occur for concrete slabs. However, the extent of this effect is lower than
previously assumed, as it does not prevent any kind of upward water transport resulting
in evaporation. The high percentage of evaporation in May is more likely to result
mainly from water evaporating from the surface storage, as low relative RO indicates
low precipitation rates, which gradually and often fill the surface storage.
48
Figure 4.7: Cumulative evaporation during dry periods following days with rainfall. Only
periods of at least four days considered
The upward water transport from underlying soil layers resulting in evaporating can
be further analysed by studying evaporation on consecutive dry days following days
with rainfall. Figure 4.7 illustrates cumulative evaporation during dry periods (of at
least four days length) following precipitation. Within the measurement period, there
were 19 such periods, of which 13 lay in summer. The longest dry period lasted 13 days
in September 2016. Concurrent to previous observations, cobblestones tend to higher
evaporation rates. Generally, the daily evaporation is highest on the first dry and grad-
ually decreases with each further day. For both surfaces, there are periods for which the
daily evaporation rate is nearly constant or may even increase slightly with progressing
time. For cobblestones, there is a distinct differentiation between summer and winter
periods, which can not be observed for concrete slabs. Additionally, the cumulative
evaporation rates are mostly constant for concrete slabs but vary for cobblestones. This
again reflects the differences between the two surfaces functioning as active interfaces
between the soil and the atmosphere.
These examples show the interaction between the overall hydrological balance and
the water flow direction, as well as the importance of upward water transport.
In order to quantify how relevant upward water transport resulting in evaporation
is, table 4.5 summarises how much of the measured evaporation takes place on days with
and without precipitation. Annually, 13.23 % (concrete slabs) to 47.23 % (cobblestones)
of evaporation is recorded on days without precipitation. In winter, more evaporation is
attributed to days without precipitation than in summer. The table only considers days
without any rainfall. When including cases as described for June 16th, 2016, 62.72 %
(cobblestones) and 46.65 % (concrete slabs) of the overall evaporation can be attributed
to days on which rainfall did not provide all of the water evaporating (E[mm d−1]>
P[mm d−1]).
49
Evaporation cobblestones Evaporation concrete slabs
P > 0 mm d−1P= 0 mm d−1P > 0 mm d−1P= 0 mm d−1
Annual 52.75 %47.23 %86.77 %13.23 %
Summer 54.30 %45.70 %89.25 %10.75 %
Winter 49.43 %50.57 %83.90 %16.10 %
Table 4.5: Percentage of total amount of evaporation (annual, summer, or winter) mea-
sured on days with (P > 0 mm d−1) and without (P= 0 mm d−1) precipita-
tion
4.1.3 Heat balance & transport
Temperatures were measured in both lysimeters and at their surfaces, with sensors on
the surface (Tsurf ), directly below the pavers (Tpav) and at 5, 15, and 25 cm below
the paver underside (T5, T15, T25). For both surfaces, mean temperatures are highest
within the soil column and lowest at the underside of pavers, with surface temperatures
in between (figure 4.8). With otherwise small differences, there is a large temperature
gradient between Tpav and T5. Concrete slabs lead to higher mean temperatures in all
depths, with the highest difference for surface temperatures. In general, the upper layers,
especially surface and paver, are most exposed and hence influenced by atmospheric
conditions. Temperature fluctuations are highest at the surface and decrease with depth.
Independent of surface, Tsurf reached the highest and lowest hourly temperatures and
T25 had the smallest maximum and highest minimum values of hourly temperatures (see
table A.2).
Figure 4.8: Comparison of annual mean temperature profiles
50
Figure 4.9: Daily solar radiation & mean temperatures (air, surface, below paver, at 5,
15 & 25 cm) during measurement period
Monthly mean temperatures for all depths and Tair can be found in table A.3. For
all layers, highest mean temperatures were recorded in June 2016, which resulted in
high evaporation. Lowest overall temperatures in all depths were recorded in January
2017, a month with very little water evaporating and very high infiltration. During warm
51
months (June to September 2016), mean temperatures were highest at 5 cm depth. With
less incoming energy from above, the highest temperatures are found in the lowest layer
at 25 cm in the cold months (October 2016 to March 2017). Starting April 2017, the
upper layers start heating up again and by May 2017 T5has again the highest mean
temperatures. At a higher temporal resolution, figure 4.9 depicts daily mean values for
all layers, as well as Tair and Rs. It shows how closely daily mean surface and soil
temperatures follow daily Tair and Rs. In summer months, all layers have very similar
mean T, with larger temperature gradients in colder months.
Largest differences between daily mean values of T25 and Tsurf were in January to
mid-February 2017. The figure shows that the overall response to Tair and Rschanges
is very similar between the two lysimeters. As has been shown, concrete slabs generally
lead to higher temperatures in all layers. The difference between the two surfaces differs
depending on layer and time (figure 4.10). Most of the time, concrete slabs result in
higher temperatures, with the highest difference to cobblestones in May 2017. Tsurf
generally differs most between the two surfaces. From October to March, T5is slightly
higher under cobblestones and in January, cobblestones reach higher temperatures in
all layers, with the largest difference for T5. Overall, the temperature differences range
from 1 to 3 K.
Figure 4.10: Differences in monthly mean temperatures between the two surfaces for all
measurement depths
52
Figure 4.11: Hourly air, surface & soil temperatures – period with highest evaporation
from June 15th to June 19th, 2016
53
Corresponding to figures 4.5 and 4.6, hourly temperature data is illustrated for two
example periods of four days each in figures 4.11 and A.1. The hourly data supports
previous observations. For both periods, the two surfaces responded very similarly but
concrete slabs surface leads to higher temperatures and fluctuations than cobblestones.
At this scale, diurnal cycles become apparent. During night, temperature drops in all
layers. With a sharp decrease at the upper layers, which are more exposed to atmospheric
conditions, and smaller changes at lower levels, highest Tis recorded at the lowest level
and increases towards the surface. Towards noon and early afternoon, Tincreases again
and the higher levels heat up rapidly. During most of the cycle, Tpav is smaller than
Tsurf . When Tsurf starts cooling down in the evening, Tsurf responds more rapidly,
which can result in times for which Tsurf =Tpav. In general, a lag of response time
can be observed, with a shift of diurnal cycles between the layers. For example, on June
16th, 2016, Tsurf reaches its daily maximum at 11:00, followed one hour later by Tpav and
nine hours later by T25. This lag is present throughout the diurnal cycle. In this study,
no temperature measurements for other surface covers are available. However, data is
available for bare soils (DWD, 2018c) and forests (Trinks, 2010) from studies carried
out in Berlin with measurements at a common depth of 5 cm (T5). Since one of those
studies provided data on a monthly basis, figure 4.12 compares the relationship between
monhtly Tair and T5. For the two paved surfaces, the relationship is nearly identical,
with slightly higher values of T5CON except for very cold months, as previously described.
Both pavements led to T5exceeding Tair, with less difference in colder months. This is
also the case for bare soil, which still exceeds Tair but led to lower T5than the two paved
surfaces. While these three are very similar, forest cover creates a different pattern.
Here, T5exceeds Tair in cold months and decreases in warmer months. This is likely to
be caused by forest trees reducing solar radiation which contributes to heating up the
bare and paved surfaces and hence, the layers beneath, as well as evapotranspiration
processes leading to cooling. A similar comparison can be found in Wu et al. (2014), in
which daily soil temperature measurements up to 300 cm below the surface were carried
out for bare soil, grass and concrete cover in both urban and rural settings. They too
observed increased soil temperature beneath concrete compared to bare soil or grass in
all depths, with larger differences in warmer months. In their study, bare soil and grass
led to very similar soil temperatures.
54
Figure 4.12: Monthly air & soil temperatures for different surface covers. Bare soil mea-
surements in Berlin for the years 2015–2017 (DWD, 2018c), forest measure-
ments in Berlin between June 2008 and January 2010 (Trinks, 2010). All
soil temperature measurements taken at 5 cm below surface cover
4.2 Process interactions
Hydrological processes are closely linked between individual processes, as well as to atmo-
spheric conditions. For paved surfaces, these interactions are focussed at the pavement
layer as the soil-atmosphere interface. Depending on the properties of the pavement,
their response to hydrological and atmospheric conditions differs greatly. In the next
section, a closer look at the coupled heat and water transport offers insight into how
different the two studied surfaces react to these conditions. After that, all measured
hourly and daily data is paired up to analyse process interactions based on their cor-
relations. Both sections illustrate how complex and varied paved surfaces react to and
impact their surroundings.
4.2.1 Coupled heat and water transport
Water and heat transport within soils are closely coupled and influence the water and
energy exchange at the soil-atmosphere interface (Bachmann and van der Ploeg, 2002;
Heitman et al., 2008; Lakshmi et al., 2003; Menziani et al., 2003; Schrödter, 1985). Soil
moisture (VWC) plays a key role in how much of the incoming energy will be turned
into latent or sensible heat flux (Lakshmi et al., 2003). Latent heat means energy is
utilised to transition between phases (e.g. from liquid to gas) and does not lead to a
change in temperature. In contrast, sensible heat leads to a change in temperature.
55
Evaporation, the amount of water transferred from the soil into the atmosphere in the
form of vapour, is the latent heat flux (Lakshmi et al., 2003; Schrödter, 1985). Energy
for evaporation processes is provided by solar radiation and the energy stored at the soil
surface and in the soil column (Schrödter, 1985). By affecting evaporation processes from
the surface and the soil, changes in Tsurf impact the soil moisture, which in turn affect the
temperature (Lakshmi et al., 2003). As Heitman et al. (2008) point out, evaporation of
soil water can be seen as a large heat sink. When energy is being utilised for evaporation
processes, it is not available to maintain the sensible heat of the surrounding, which leads
to a cooling effect. Sensible heat fluxes in the soil column are driven by temperature
gradients (Heitman et al., 2008), which are influenced by evaporation from the surface
and underlying soil.
Figure 4.13 illustrates hourly observations of volumetric water content (θ) and tem-
perature at 5, 15 and 25 cm depth for the example period in September 2016 (see also
figures 4.6 & A.1). On September 17th, a high precipitation event occurred at noon,
Figure 4.13: Example of hourly soil water content and temperatures at 5, 15, and 25 cm
below the paver underside
56
followed by an increase of θfor all three depths as water infiltrates into the soil. At
the same time, Tsoil decreases in all depths with a slight lag with increasing depth. The
supplied water starts to evaporate, for which energy is consumed as latent heat. As
a result, this energy can not be used as sensible heat and it takes a couple of days
for Tsoil to gradually increase again as the soil dries. This heating up process takes
longer for cobblestones, which tend to higher evaporation rates and hence, to more en-
ergy being used as latent heat. The figure shows upward water transport in form of
evaporation: T15 is decreasing at the same time as θ15 decreases, followed shortly after
by θ5increasing. The same can be observed for depth 25 and depth 15 cm. Water
at the lower depths uses the energy stored in the soil as heat to evaporate and travels
upward, leading to a decrease of θat the lower layer and an increase above. From the
5cm, some of the water will pass the paved soil-atmosphere interface and leave the
soil, using the energy stored in the upper soil layer and the new incoming energy from
the atmosphere. The less water is transported to the upper 5 cm, the less incoming
energy is used as latent heat, leading to slowly increasing T5in the following days and
decreasing evaporation rates (see also figure 4.7). It should be noted that capillary rise
of liquid water is another source of upward water transport, which plays an important
role at the beginning with water vapour being the dominant source of water as the soil
becomes drier. Heitman et al. (2008) carried out spatially high-resolution measurement
of soil temperatures at the first 6.6 cm below the surface and showed that soil thermal
properties near the surface are affected by wetting and drying processes. In order to
analyse this wetting–drying process, an additional experiment was carried out using a
thermal camera (see chapter 3.3). On this day, mean air temperature was 23.87 ◦Cand
daily Rsreached 270 W m−2. The results for one wetter/drying cycle for each surface are
illustrated in figures 4.14 (mean temperature time series) and 4.15 (thermal pictures).
In the time series, the first value (t= 0 s) is the initial dry surface. Next, the surface was
watered, which is excluded in this figure, followed by the first picture after the surface
was completely covered in water (t= 4 s). For concrete slabs, joints and pavers follow
the same pattern, with joints maintaining higher temperatures. Water application leads
to a sharp decline in temperature by 1−2 K, as the applied water was cooler than the
surface temperature in dry conditions. Immediately after, the applied water is heating
up, with sensible heat fluxes dominating the process. Tsurf remains more or less stable
with a slight tendency to decrease until t= 200 s. The fluctuations indicate that both
processes, sensible and latent heat flux, take place. At this point, the water film adhering
to the surface has been completely evaporated but the concrete slabs were visibly still
wet, with the pores storing water. For the next 340 seconds, the surface is first cooling
down and then remaining at this low temperature, with the water evaporating from the
surface storage and the upper soil layers transforming energy to latent heat. Figure 4.15
shows the surface at t= 420 s, when it reached its lowest Tsurf . Overall, it takes 540
seconds until most of the available water has been evaporated and sensible heat fluxes
start to dominate, gradually increasing Tsurf again. Yet, even during this increase, small
57
Figure 4.14: Timeseries of Tsurf during surface wetting and drying process. Based on
thermal pictures with 4-second resolution. Wetting process was cut, so
that t= 0 is initial dry surface and t= 4 is surface immediately after
water application. Drying process complete after 322 (cobblestones) and
908 (concrete slabs) seconds
fluctuations with decreasing Tsurf show that latent heat fluxes still take place. After a
total of 908 seconds, the surface is mostly dry again with only small amount of pores still
visibly containing water and Tsurf again reaching its initial temperature. Throughout
the process, joints has lower temperatures, which can be attributed to higher evaporation
rates compared to the pavers due to their utilisation of soil water.
The wetting–drying process differs significantly between the two surfaces. For cob-
blestones the whole cycle only takes 322 seconds until initial Tsurf is reached again.
Furthermore, the overall decrease in temperature is very small compared to concrete
slabs. Both can be attributed to the lower porosity of the pavers, which leads to no or
very little water being stored on the pavers. The little amount of water adhering to the
paver surfaces follows the same pattern as described for the concrete slabs, with sensible
heat dominating as the surface water heats up, followed by latent heat flux as the water
evaporated and finally an increase of Tsurf after most of the water has been evaporated.
Figure 4.15 shows that individual pavers may react very differently, depending on ma-
terial and micro-topography of the surface. Some pavers included small indentations in
which water accumulated. Different materials (e.g. granite and sand stone) have varying
thermal properties and heat up at different rates. For the cobblestone surface, joints
have a high share of the surface area (20 %) and are influences less heavily by the pavers.
In the first minute after watering, latent heat fluxes dominate and the joints cool down.
After that, joints generally heat up again, but fluctuations again show that both sensible
and latent heat fluxes are taking place. For both surfaces, it has to be taken into ac-
count that evaporation processes increase the relative humidity of its surrounding, which
decreases the evaporation potential. However, since only the two surfaces were watered,
the surrounding area was mostly unaffected and supplied dry air to the surfaces through
advection. This also means that this experiment represents idealised conditions rather
than natural ones. Still, it is useful to observe these underlying processes.
58
Figure 4.15: Thermal pictures of surface wetting and drying process. Timestamp (t) in
seconds
59
4.2.2 Correlation of processes
Correlation is a measure of relationship and interdependence between processes. In
this thesis, the coefficient of determination based on the Pearson correlation coefficient
(R2) is used, to describe if there is a linear fit between to variables. In this section, a
more general relationship is analysed, which does not have to be linear. The Spearman
correlation (ρ) is another method to determine this relationship. The main difference
is, that ρcan also be used for non-linear correlation. Since the analysed data very
seldom follows a linear pattern, ρis used to gain insight into relationships and hence,
possible parametrisation of models. The results are summarised in figures 4.16 and
4.17. A value of 0indicates no correlation and −1or 1indicating highly dependent
variables. Negative values point to one variable decreasing while another increases. For
example, grass-reference evapotranspiration (ET0, see chapter 5) is highly dependent
on relative humidity (RH) with dry air (low RH) being able to absorb more water
vapour, increasing evaporation rates (ρ(RH, ET0) = −0.9). The other way around, high
Tair provides energy to heat up paved surfaces, so that ρ(Tair, Tsurf ) = 1.0. For other
pairs, correlation can exist but be weakened by other factors or temporal resolution.
For example, runoff formation logically is highly dependent on precipitation. Without
precipitation, there can be no runoff. The amount of runoff depends on the intensity of
the rainfall event, as well as the moisture of surface and soil at the start of the event. For
hourly data, ρof Pand RO is 0.2 (cobblestones) or 0.4 (concrete slabs), compared to 0.5
(cobblestones) or 0.7 (concrete slabs) when considering daily data. The relationship is
stronger for concrete slabs which tend to form runoff more often and in higher absolute
amounts. For both surfaces, ρis relatively low because hourly or daily precipitation
intensities do not reflect the behaviour of the event. The hourly amount could be evenly
distributed over one hour, or take place in just a couple of minutes. To accurately
describe precipitation and runoff relationships, precipitation events would have to be
recorded at a higher temporal resolution (Rim, 2011). The relationship may be stronger
for daily data due to possible lags between precipitation and runoff collection. These
examples illustrate the interpretation of ρ. While the amount of information contained
in the two correlation matrices is too high to examine in detail here, it is included in full
to provide data and insights that may be interesting for others. It should be taken into
account that a small ρdoes not necessarily mean that there is complete independence
between two variables. For example, it is possible for a specific combination of variables
to correlate with another variable, without all of the individual ones having high values
of ρwith the aimed for variable.
Some further observations are noteworthy for this work. As described in the previ-
ous section, soil temperature and water content are governed by coupled processes. This
is reflected by high values of ρ.ECOB is linked to Tsoil and θ, as well as atmospheric
conditions (Tair, RS, RH), which is also reflected by its relationship to ET0. In contrast,
ECON has no or very small correlation to the same variables, but a higher correlation
60
to precipitation compared to cobblestones. For the next chapter, the relationship be-
tween ET0, which represents the evapotranspiration of a grass surface with optimum
water supply, and ECOB and ECON . This relationship is represented by κ, the ratio
between Eand ET0. Figure 4.17 contains daily values of κ, which is the ratio for each
day individually. While ECON had low correlation with atmospheric conditions deter-
mining ET0,κCON has a relatively high correlation with these variables. In most cases,
ρ(κCON , ET0)is equal or even slightly higher than ρ(κCOB, ET0). This indicates that
estimating ECON by first calculating ET0and then adjusting it could be more promising
than attempting to directly deduce ECON from climatological data. Daily values of κ
also show that for concrete slabs, the relationship is weakly correlated to daily precipi-
tation, which does not affect κCOB. This reflects previous observations which showed
that 47 % of ECOB takes place on days without precipitation, compared to 13 % for
concrete slabs. Therefore, κCOB does not depend on precipitation events as strongly.
61
Figure 4.16: Correlation matrix for hourly data (Spearman correlation). All data mea-
sured at lysimeter site, Pas mean of the two lysimeters. Non-significant
values are crossed out
62
Figure 4.17: Correlation matrix for daily data (Spearman correlation). All data mea-
sured at lysimeter site, Pas mean of the two lysimeters. Additional data:
ET0= grass-reference evapotranspiration after Penman-Monteith (Allen
et al., 1998; ASCE-EWRI, 2005), κ= relationship between daily measured
Eand calculated daily ET0with κ=E
ET0. Non-significant values are crossed
out
63
5 Estimating evaporation of paved surfaces
As shown in the previous chapter, evaporation (E) processes have a cooling effect. Paved
surfaces react differently, depending on paver material and sealing degree, leading to
varying heat storage capacities and evaporation potentials. Evaporation takes place on
wet and dry days. For dry days, evaporation originates mostly from underlying soil
layers. On wet days, the surface storage of the paving material and the intensity of
precipitation play key roles. Given the function of evapo(transpi)ration (E(T)) as heat
sink and its potential for countering the Urban Heat Island (UHI) effect, it is essential to
asses how different vegetated and paved surfaces impact urban areas. Because of its ap-
plication in agriculture, evapotranspiration (ET) has been studied for a several decades
with numerous estimation models and measurements. As Schrödter (1985) points out,
estimating actual evaporation is a significant challenge, as it is the result of numerous
interdependent physical processes. For vegetated surfaces, there are many models to es-
timate ET. Combined with a crop coefficient, grass-reference evapotranspiration (ET0)
is used as a standardised model. It has also been used together with a reduction coeffi-
cient to estimate evaporation from paved surfaces (Mansell and Wang, 2010; Wessolek
et al., 2008). In this chapter, the usage of ET0for estimating Efrom paved surfaces is
discussed.
5.1 Grass-reference evapotranspiration ET0
A commonly used method for evapotranspiration estimations is the physically based FAO
Penman-Monteith grass-reference evapotranspiration, which offers high accuracy for var-
ious locations (ASCE-EWRI, 2005; Azhar and Perera, 2011; Nandagiri and Kovoor, 2006;
Raziei and Pereira, 2013; Yoder et al., 2005). It uses the grass reference crop, a hypo-
thetical crop with an uniform height of 0.12 m, an albedo of 0.23 and a surface resistance
of 70 s m−1, which is based on the characteristics of extensive grass surface (Allen et al.,
1998). It assumes vegetation cover throughout the year with an optimal water supply
through capillary rise (ASCE-EWRI, 2005).
64
The general formula for calculating ET0is (ASCE-EWRI, 2005):
ET0=
0.408 ∆ (Rn−G) + γCn
Tair + 273 u2(es−ea)
∆ + γ(1 + Cdu2)(5.1)
where
ET0Standardized reference evapotranspiration [mm d−1]
∆Slope of saturation vapour pressure-temperature curve [kPa ◦C−1]
RnNet radiation at crop surface [MJ m−2d−1]
GSoil heat flux density at soil surface [MJ m−2d−1]
γPsychrometric constant [kPa ◦C−1]
CnConstant changing with calculation time step and crop [K mm s3Mg−1d−1]
Tair Mean daily or hourly air temperature at 1.5to 2.5 m height [◦C]
u2Mean daily or hourly wind speed at 2 m height [m s−1]
esSaturation vapour pressure at 1.5to 2.5 m height [kPa]
eaMean actual vapour pressure at 1.5to 2.5 m height [kPa]
CdConstant changing with calculation time step and reference crop [s m−1].
The 0.408 coefficient is a conversion factor for radiation to gain equivalent evaporation
[mm d−1] from [MJ m−2d−1] and its units are [m2mm MJ−1].
Detailed description of computation can be found in the ASCE-EWRI Task Commi-
tee Report and its appendices (ASCE-EWRI, 2005).
The climate station at the lysimeter site Berlin-Marienfelde offers all required input
data to compute ET0. For three days for which lysimeter measurements were obtained,
the climate station did not provide data. For these days, data from the nearby DWD
climate station at Berlin-Tempelhof (DWD, 2018b) has been used to compute ET0. For
the 333 days with lysimeter results, the FAO Penman-Monteith grass-reference evapo-
transpiration results in ET0= 643 mm. This value is used as reference in the following
considerations.
5.2 Relationship between Eand ET0for paved surfaces
ET0is based on a hypothetical surface with hypothetical conditions. Hence, it is a
representation of the atmospheric potential for evaporation for a specific surface cover.
Consequently, it differs from actual evaporation, even for surfaces that closely resemble
the hypothetical grass vegetation. For vegetated surfaces, crop coefficients are used to
derive ET of different surfaces from ET0. This crop coefficient differs throughout the
year as vegetation develops (Allen et al., 1998). Additionally, water supply is seldom
as optimal as assumed for ET0. Hence, water availability as to be considered when
estimating E(T)(Allen et al., 1998; Schrödter, 1985). For pavements, surface properties
remain the same throughout the year. However, water availability remains a key factor
for estimating E. In the following sections, one model for estimating annual Eand
65
possibilities for a higher temporal resolution are introduced.
5.2.1 TUBGR model
For urban areas, the TUBGR model (Wessolek et al., 2008) can be used to estimate the
annual hydrological balance based on precipitation and ET0. Depending on the sealing
degree (SD), different empirical coefficients (table 5.1) have been determined using data
from other lysimeter studies (Wessolek and Facklam, 1997; Flöter, 2006). The focus of
this chapter is evaporation. For calculation of infiltration and runoff according to this
model, the reader is advised to refer to publications of this model (Wessolek et al., 2008;
Wessolek, Kaupenjohann and Renger, 2009).
The TUBGR formula for estimating annual evaporation is:
E=κ×ET0(5.2)
with
κ=(︃log(0.6×βs×Ps)
log(ET0))︃4
(5.3)
where
EEvaporation from paved surface [mm a−1]
κReduction coefficient [-]
ET0Grass-reference evapotranspiration [mm a−1]
βsSummer infiltration coefficient [-]
PsPrecipitation in summer (April to September) [mm a−1].
It uses ET0and introduces a term reflecting water availability based on net precipita-
tion (precipitation minus runoff and infiltration losses). The model has been tested with
the data from the two paved lysimeters. For this, mean values of both lysimeters have
been used for precipitation. ET0was calculated for days with lysimeter data as described
in the previous section. Results for all four sealing degrees can be found in table 5.2.
Comparing these results to the measured hydrological balance (table 4.2) shows that the
annual actual evaporation has a very good fit between model and measurement for the
two paved surfaces studied. According to the TUBGR model, cobblestones should have
evaporated 141 mm, which corresponds to the measured 140.55 mm. For concrete slabs,
there was a small overestimation by about 7 % with 99 mm (model) compared to 92.24
mm (lysimeter). While annual evaporation was estimated accurately, the distribution
of the remaining water between runoff and infiltration is off. Runoff was significantly
overestimated for both surfaces, being 2.5 (concrete slabs) to 6.6 (cobblestones) as high
as measured runoff. Contrary to that, infiltration is underestimated. Cobblestones were
estimated to infiltrate 188 mm and actually infiltrated 252 mm. The discrepancy is
higher for concrete slabs, an estimation of 131 mm compared to measured 258 mm.
66
Sealing degree Examples βs†βw†
Class I (low: <10%) Grass pavers 0.90 0.95
Class II (medium: 10 −50%) Cobblestones 0.80 0.85
Class III (high: 50 −90%) Concrete slabs 0.55 0.60
Class IV (severe: >90%) Asphalt 0.20 0.25
†infiltration coefficient for summer (βs) and winter (βw)
Table 5.1: TUBGR model coefficients for paved surfaces (Wessolek et al., 2008)
Sealing degree Examples κE RO I
[-] [mm a−1] [mm] [mm a−1]
Class I (low: <10%) Grass pavers 0.24 156 29 213
Class II (medium: 10 −50%) Cobblestones 0.22 141 68 188
Class III (high: 50 −90%) Concrete slabs 0.15 99 168 131
Class IV (severe: >90%) Asphalt 0.05 31 307 59
Table 5.2: Annual hydrological balance estimated for paved surfaces by TUBGR model
In reality both paved surfaces yielded very similar infiltration amounts, yet the model
resulted in a rather large difference between the two. It should be noted however, that
this comparison between measured and estimated values is not sufficient to evaluate the
model. Given annual and seasonal changes in precipitation patterns and other climato-
logical factors, measurement of one year resulting in one data point per paved surface
type can be an indicator but not enough to confirm or calibrate a model. However,
estimation of annual Eis independent of the calculation of RO and I, an therefore, can
be considered separately.
Running the TUBGR model for the same time period, the same climate and the
same soil type results in evapotranspiration sums of 351 mm (arable land), 321 mm
(grassland) and 444 mm (deciduous forest).
5.2.2 Determining the ratio between Eand ET0
In the TUBGR model, the reduction coefficient κis used to derive annual Efrom annual
ET0and summer P. Based on equation 5.2, with measurements of ECOB and ECON
combined with estimated ET0,κcan be calculated from the available data with:
κ=E
ET0
(5.4)
where
κReduction coefficient [-]
EEvaporation from paved surface [mm a−1]
ET0Grass-reference evapotranspiration [mm a−1].
67
Figure 5.1: Relationship between daily evaporation (E) from paved surfaces and grass-
reference evapotranspiration (ET0)
Calculating κon a daily basis (ratio daily Eand daily ET0) results in varying values.
Figure 5.1 illustrates daily values of κfor both paved surfaces, as well as absolute amounts
of daily measured Eand calculated ET0. Since it is the ratio, values of κ < 1indicate a
reduced Ecompared to ET0and κ > 1means that the paved surfaces evaporated more
than a grass surface with optimal water supply. As can be seen, there are some days for
which E > ET0. One case is in mid June 2016, which is the day with highest recorded
68
evaporation from the surfaces, corresponding to figure 4.5. Otherwise, κ > 1occurs in
winter, mostly between November and February. As the bottom plot illustrates, this
is the time for which ET0and ECOB are low in general, so that even a high κresults
in small differences when calculating amounts of E. For concrete slabs, Eis more or
less stable throughout the year and monthly ECON between November and February
can even exceed that of months with high ET0(see also table 4.2). For example, ECON
reaches its third highest value in December. As shown in figure 4.17, daily κis weakly to
moderately correlated to climatological conditions (RH, Rs, Tair) and only κCON shows
a relationship to daily precipitation. However, it does not correlate strongly with a single
easily measured variable. A look at median values of daily κover annual, seasonal, and
monthly periods (table A.4) yields insights into general trends. Annual mean values of
κbased on measurements are very similar to the annual κestimated by the TUBGR
model. For cobblestones, both methods yield an annual κCOB of 0.22 and concrete slabs
differed slightly with κCON = 0.12 (measured) compared to 0.15 (TUBGR). However,
this annual κCON as the median of daily κwould lead to an underestimation of ECON .
As previously shown, ECOB can be attributed to nearly 50 % to dry days, with upward
water transport providing water to evaporate. Concrete slabs are more dependent on
rainfall events and their intensities, as their main evaporative potential stems from water
stored in the porous pavers. This is reflected when differentiating between wet and dry
days. Table A.4 also contains the median values of daily κvalues separated into dry
(κd) and wet (κw) days. While there is very little difference for cobblestones, κwis
significantly larger than κdfor concrete slabs. This effect is especially noticeable in
summer months. It should be noted that some of the monthly values of κ, especially
when separating κdand κware based on few data points. Applying these monthly values
to obtain daily ECOB and ECON did not yield satisfactory results.
Figure 5.2: Monthly estimation of Ewith annual constant κof TUBGR model compared
to measured E
69
So far, the TUBGR model offers good estimation of annual ECOB and ECON . Figure
5.2 shows monthly estimations of Ewhen using the constant annual κestimated by the
TUBGR model (equation 5.3) compared to measured values of monthly E. As can be
seen, the fit is better for ECOB, where the monthly κdo not differ as much and for
which wet and dry conditions lead to similar amounts of evaporation. Still, months with
small amounts of Etend to be underestimated while those with high Eare slightly
overestimated. This pattern is the same for concrete slabs, combined with an overall
poor fit. Monthly values of measured and estimated Eas well as ET0can be found in
table 5.3. This result is to be expected, as the κfrom the TUBGR model is intended
for annual estimations, based on ET0and Pover longer periods.
On a monthly scale, κ(here: ratio monthly Eand monthly ET0) for both surfaces
showed moderate to strong correlation with monthly mean Tair. Other correlations were
tested for deriving κ, including single or combined variables of Tair, P, RH, Rs, ET0. In
the end, the best fit was achieved using monthly mean Tair. This relationship is not
linear but can be approximated using an exponential function, with higher values of κ
for lower temperatures (figure 5.3). The relationship is stronger for cobblestones with
rather high deviations for concrete slabs.
Using the results of the non-linear regression, monthly values of κcan be estimated with:
κ=a×0.9Tair (5.5)
κMonthly reduction coefficient[-]
aSurface type specific constant: 0.6 (COB) or 0.9 (CON)
Tair Monthly mean air temperature at 2 m height [°C].
Figure 5.3: Deriving measured monthly reduction coefficient κfrom monthly mean Tair
70
Using these monthly values of κderived from Tair, monthly Eis estimated. Figure
5.4 compares the fit of estimated and measured Efor constant κ(TUBGR model)
and monthly κ(derived from Tair). Corresponding monthly values and annual sums
can be found in table 5.3. For cobblestones, estimating Ewith κCOB derived from
Tair generally improves the fit. However, for three months there is a large deviation,
leading to an underestimation of annual ECOB. Two of these month are June and July,
when evaporation potential of the surfaces was high, with the highest daily and hourly
evaporation event recorded in June. For concrete slabs, the relationship between κ
and Tair was weaker, which is conveyed to estimating monthly ECON using these κCON
values. Both, constant and monthly κdid not yield satisfying results on a monthly scale.
Using the derived monthly κimproves the fit for months with low ECON , yet leads to
some significant overestimations for months with high values, leading to a significant
overestimation on an annual scale. For April, this overestimation is especially high with
estimated 21.15 mm compared to measured 5.93 mm, which could be caused by long dry
periods and transitioning climatological conditions, as reflected in figure 5.1. The fit for
concrete slabs could potentially be improved by separating κdand κw. For this, daily
values would be needed as basis. Figure A.2 shows the median values of daily κdand
κwfor each month related to monthly mean values of Tair (only months with at least
10 days classified as dry or wet were considered). For concrete slabs, this results in a
better fit than combined κ, especially for wet days. Applying these monthly values of
κdand κwto daily values of ET0improves the fit for concrete slabs slightly compared
to using the combined monthly κas in figure 5.4. While some months still have high
deviations, the sum of ECON using monthly values of κdand κwderived from Tair is
96.5 mm, which is close to the measured sum of 92.24 mm.
Cobblestones Concrete slabs
ET0Lys TUBGR with κTLys TUBGR with κTwith κw,d
Jan 5.92 3.97 1.30 3.66 4.74 0.89 6.18 4.33
Feb 15.6 5.84 3.43 7.50 6.64 2.34 10.72 7.90
Mar 38.98 13.14 8.58 13.55 8.78 5.85 15.64 12.12
Apr 60.79 14.40 13.37 19.39 5.93 9.12 21.15 16.44
May 106.94 21.50 23.53 21.04 8.69 16.04 16.67 13.05
Jun 127.66 25.89 28.09 19.28 15.51 19.15 12.80 10.99
Jul 107.72 20.77 18.19 15.57 10.46 16.16 10.04 8.75
Aug 82.70 13.81 18.19 13.39 7.71 12.41 9.32 7.60
Sep 73.98 11.42 16.28 12.06 4.98 11.10 8.43 6.29
Oct 7.03 2.41 1.55 2.18 2.51 1.05 2.34 1.86
Nov 8.17 2.65 1.80 3.64 6.67 1.23 4.96 3.69
Dec 7.13 4.48 1.57 3.39 9.98 1.07 4.82 3.48
Sum 642.62 140.55 141.38 134.65 92.24 96.39 123.08 96.50
Table 5.3: Monthly and annual estimated ECOB and ECON based on ET0. Lys = mea-
sured E, TUBGR = Eestimated using constant κfor each month as derived
from annual TUBGR model, κT=Eestimated using monthly κderived from
monthly Tair,κw,d =Eestimated using monthly median κwand κdderived
from monthly Tair applied to daily data depending on wet or dry conditions
71
Figure 5.4: Comparison of monthly estimations of Ecompared to measured E, with
annual constant κof TUBGR model and monthly κderived from Tair
5.2.3 Outlook
For annual estimations, the TUBGR model has yielded very good results in this study
as well as previous ones. Overall, monthly Efrom cobblestones and concrete slabs
proved difficult to estimate using ET0. For practice, it could be useful to make a general
differentiation between summer and winter conditions. In that case, the usage of the
seasonal values of κdetermined from daily measurements (table A.4) would be feasible.
For cobblestones, these are κ= 0.16 (summer) and κ= 0.30 (winter). In case of
concrete slabs, the seasonal values are κ= 0.08 (summer) and κ= 0.34 (winter). If
monthly reduction coefficients are required, they may be derived from mean Tair. This
method seems promising but would require additional data sets of monthly ET0,E,
and Tair, ideally originating from different climate zones. With this additional data,
a more universal equation for deriving monthly κvalues could be developed. This
would be valuable, as the usage of ET0includes atmospheric conditions at different
sites and deriving κfrom Tair rather than assigning fixed monthly values would further
enable its potential application at different sites. For concrete slabs, water storage
on the surface plays an important role, so that a differentiation between wet and dry
days could be promising. This would again require data of ET0,E, and Tair, as well
as P, but on a daily basis. Since the available data from this study is limited, the
analysis, ideas, and methods presented in this chapter should be seen as conceptual
approaches that might contribute to the collective development of an estimation model
for evaporation from paved surfaces on an improved temporal scale. Compared to the
grass-reference evapotranspiration model, which is the result of decades of research and
numerous lysimeter studies providing calibration data, the development of a robust
model for paved surfaces is still at its beginning.
72
6 Conclusion
Paved surfaces are a key element of urban areas. The modified soil-atmosphere interface
alters the urban hydrological balance and can lead to significant challenges. With a focus
on stormwater management, most existing hydrological models treat all paving materials
as the same and assume little or no infiltration and evaporation losses. Few studies
providing measurements of the hydrological balance and water transport processes of
pavements have been published in the past. They show that the impact of different
paving materials varies considerably, and even severe sealing with asphalt exhibited in
infiltration and evaporation processes.
In this study, two common pavement types (cobblestones and concrete slabs) were
researched. The combination of weighable lysimeters and sensors to measure soil water
content and temperature provided insights into their hydrological balance, as well as wa-
ter and heat transport processes, at a high temporal-resolution. Over the measurement
period of one year, the surfaces led to similar infiltration and differed in evaporation and
runoff production. Cobblestones, which are characterised by large joint areas, evapo-
rated 25 % and rarely produced runoff (3 %). Concrete slabs, which represent a higher
degree of sealing with narrow joint areas, tended to produce more runoff (16 %) and less
evaporation (22 %). Both surfaces reacted differently depending on seasonal conditions.
Evaporation processes did not occur exclusively on days for which precipitation provided
water. Upward water transport from underlying soil layers led to evaporation processes
during dry periods. This effect contributed 47 % of evaporation for cobblestones and
13 % for concrete slabs. Previous descriptions of pavements acting as evaporation bar-
rier were not supported by the results. Both surfaces led to increased heat storage in
the soil compared to natural surfaces, with concrete slabs tending to slightly higher
temperatures. Thermal and water transport processes are closely linked. Compared to
cobblestones, evaporation from concrete slabs surfaces relies more on water supply from
precipitation events.
Estimating evaporation of paved surfaces by reducing the common parameter of
grass-reference evapotranspiration produced mixed results. For annual estimations, the
pre-existing TUBGR model provided a good fit. On a higher temporal resolution, the
reduction coefficient changes depending on month or season. For concrete slabs, a dis-
tinction between wet and dry conditions should be made. Monthly coefficients might be
derived from air temperature. The presented modelling ideas and approaches did not
yield satisfying results for monthly evaporation estimations.
Overall, it could be shown that pavements, as the urban soil-atmopshere interface,
73
are highly active systems determining the urban hydrological balance. They are more
than runoff generators. Upward water transport is essential when assessing evaporation
from pavements. Varying types of paving material and degree of surface cover lead
to different impacts on heat and water transport. Understanding and utilising these
differences has the potential to improve the design of urban areas.
74
Bibliography
Allen, R. G., Pereira, L. S., Raes, D. and Smith, M. (1998), ‘Crop evapotranspiration
(guidelines for computing crop water requirements)’, FAO Irrigation and Drainage
Paper No. 56. URL: http://www.kimberly.uidaho.edu/water/fao56/fao56.pdf, last
accessed 2018-06-16.
Allen, R., Smith, M., Perrier, A. and Pereira, L. (1994), ‘An update of reference evapo-
transpiration’, ICID Bulletin, 43(2), 1–34.
Andersen, C. T., Foster, I. D. L. and Pratt, C. J. (1999), ‘The role of urban sur-
faces (permeable pavements) in regulating drainage and evaporation: development
of a laboratory simulation experiment’, Hydrological Processes, 13(4), 597–609. doi:
10.1002/(sici)1099-1085(199903)13:4<597::aid-hyp756>3.0.co;2-q.
Angrill, S., Petit-Boix, A., Morales-Pinzón, T., Josa, A., Rieradevall, J. and Gabar-
rell, X. (2017), ‘Urban rainwater runoff quantity and quality – a potential endoge-
nous resource in cities?’, Journal of Environmental Management, 189, 14–21. doi:
10.1016/j.jenvman.2016.12.027.
ASCE-EWRI (2005), The ASCE standardized reference evapotranspiration equa-
tion, Technical report, American Society of Civil Engineers, Environmental & Wa-
ter Resource Institute. URL: https://www.kimberly.uidaho.edu/water/asceewri/
ascestzdetmain2005.pdf, last accessed 2014-10-10.
Azhar, A. H. and Perera, B. J. C. (2011), ‘Evaluation of Reference Evapotranspiration
Estimation Methods under Southeast Australian Conditions’, Journal of Irrigation
and Drainage Engineering, 137(5), 268–279. doi: 10.1061/(asce)ir.1943-4774.0000297.
Bach, P. M., Rauch, W., Mikkelsen, P. S., Mccarthy, D. T. and Deletic, A. (2014), ‘A
critical review of integrated urban water modelling – Urban drainage and beyond’,
Environmental Modelling & Software, 54, 88–107. doi: 10.1016/j.envsoft.2013.12.018.
Bachmann, J. and van der Ploeg, R. R. (2002), ‘A review on recent develop-
ments in soil water retention theory: interfacial tension and temperature effects’,
Journal of Plant Nutrition and Soil Science, 165(4), 468. doi: 10.1002/1522-
2624(200208)165:4<468::AID-JPLN468>3.0.CO;2-G.
Berlekamp, L. (1987), ‘Bodenversiegelung als Faktor der Grundwasserneubildung (Soil
sealing as factor of groundwater recharge)’, Landschaft + Stadt, 19(3), 129–136.
75
Berthier, E., Dupont, S., Mestayer, P. and Andrieu, H. (2006), ‘Comparison of two
evapotranspiration schemes on a sub-urban site’, Journal of Hydrology, 328(3-4), 635–
646. doi: 10.1016/j.jhydrol.2006.01.007.
Bhaduri, B., Minner, M., Tatalovich, S. and Harbor, J. (2001), ‘Long-Term Hydrologic
Impact of Urbanization: A Tale of Two Models’, Journal of Water Resources Planning
and Management, 127(1), 13–19. doi: 10.1061/(asce)0733-9496(2001)127:1(13).
Bonicelli, A., Giustozzi, F. and Crispino, M. (2015), ‘Experimental study on the effects
of fine sand addition on differentially compacted pervious concrete’, Construction and
Building Materials, 91, 102–110. doi: 10.1016/j.conbuildmat.2015.05.012.
Booth, D. B. and Leavitt, J. (1999), ‘Field Evaluation of Permeable Pavement Systems
for Improved Stormwater Management’, Journal of the American Planning Associa-
tion, 65(3), 314–325. doi: 10.1080/01944369908976060.
Brattebo, B. O. and Booth, D. B. (2003), ‘Long-term stormwater quantity and quality
performance of permeable pavement systems’, Water Research, 37(18), 4369–4376. doi:
10.1016/s0043-1354(03)00410-x.
Brewer, C., Harrower, M., Sheesley, B., Woodruff, A. and Heyman, D. (2013), ‘Color
Brewer 2.0 – color advice for cartography’. URL: http://colorbrewer2.org/, last ac-
cessed 2017-06-01.
Bricker, S., Banks, V., Galik, G., Tapete, D. and Jones, R. (2017), ‘Account-
ing for groundwater in future city visions’, Land Use Policy, 69, 618–630. doi:
10.1016/j.landusepol.2017.09.018.
Carbone, M., Mancuso, A. and Piro, P. (2014), ‘Porous Pavement Quality Modelling’,
Procedia Engineering, 89, 758–766. doi: 10.1016/j.proeng.2014.11.504.
Chatzidimitriou, A. and Yannas, S. (2015), ‘Microclimate development in open urban
spaces: The influence of form and materials’, Energy and Buildings, 108, 156–174. doi:
10.1016/j.enbuild.2015.08.048.
Chen, J., Theller, L., Gitau, M. W., Engel, B. A. and Harbor, J. M. (2017), ‘Urbanization
impacts on surface runoff of the contiguous United States’, Journal of Environmental
Management, 187, 470–481. doi: 10.1016/j.jenvman.2016.11.017.
Chiang, Y., Sullivan, W. and Larsen, L. (2017), ‘Measuring Neighborhood Walkable
Environments: A Comparison of Three Approaches’, International Journal of Envi-
ronmental Research and Public Health, 14(6), 593. doi: 10.3390/ijerph14060593.
Corazza, M. V., Mascio, P. D. and Moretti, L. (2016), ‘Managing sidewalk pave-
ment maintenance: A case study to increase pedestrian safety’, Journal of
Traffic and Transportation Engineering (English Edition), 3(3), 203–214. doi:
10.1016/j.jtte.2016.04.001.
76
Daniel, M., Lemonsu, A. and Viguié, V. (2018), ‘Role of watering practices in large-
scale urban planning strategies to face the heat-wave risk in future climate’, Urban
Climate, 23, 287–308. doi: 10.1016/j.uclim.2016.11.001.
Delleur, J. W. (2003), ‘The Evolution of Urban Hydrology: Past, Present, and Fu-
ture’, Journal of Hydraulic Engineering, 129(8), 563–573. doi: 10.1061/(asce)0733-
9429(2003)129:8(563).
Diestel, H. and Schmidt, M. (2001), Untersuchungen an der Lysimeteranlage zur Ermit-
tlung von Versickerungs- und Oberflächenabfluß für unterschiedliche Gehwegbefesti-
gungen (Investigations using lysimeters to determine infiltration and runoff of different
pavements). Unpublished Research Report fort he water works of Berlin, executed by
Technical University Berlin, Berlin - Germany, Department of Architecture, Environ-
ment and Society, Institute Landscape and Environmental Planning.
Dirckx, G., Daele, S. V. and Hellinck, N. (2016), ‘Groundwater Infiltration Potential
(GWIP) as an aid to determining the cause of dilution of waste water’, Journal of
Hydrology, 542, 474–486. doi: 10.1016/j.jhydrol.2016.09.020.
Dowle, M. and Srinivasan, A. (2017), data.table: Extension of ‘data.frame‘. URL: https:
//CRAN.R-project.org/package=data.table, last accessed 2018-06-10.
Dreelin, E. A., Fowler, L. and Carroll, C. R. (2006), ‘A test of porous pavement effec-
tiveness on clay soils during natural storm events’, Water Research, 40(4), 799–805.
doi: 10.1016/j.watres.2005.12.002.
Dupont, S., Mestayer, P. G., Guilloteau, E., Berthier, E. and Andrieu, H. (2006), ‘Pa-
rameterization of the Urban Water Budget with the Submesoscale Soil Model’, Journal
of Applied Meteorology and Climatology, 45(4), 624–648. doi: 10.1175/jam2363.1.
DVWK (1996), Ermittlung der Verdunstung von Land- und Wasserflächen (Determi-
nation of evaporation from land and water surfaces), DVWK-Merkblatt (DVWK-
Bulletin) 238/1996, Deutscher Verband für Wasserwirtschaft und Kulturbau e.V.
DWD (2018a), ‘Globalstrahlung in der Bundesrepublik Deutschland. Mittlere Jahres-
summen, Zeitraum: 1981–2010 (Global radiation in the Federal Republic of Germany,
average annual sum, period 1981–2010)’. URL: https://www.dwd.de/DE/leistungen/
solarenergie/lstrahlungskarten_mi.html?nn=16102, last accessed 2018-06-08.
DWD (2018b), ‘Klimadaten Deutschland – Monats- und Tageswerte (Archiv) (Climate
data for Germany – monthly and daily data (archive))’. URL: https://www.dwd.de/
DE/leistungen/klimadatendeutschland/klarchivtagmonat.html?nn=16102, last ac-
cessed 2018-06-08.
77
DWD (2018c), ‘Klimadaten Deutschland - Stundenwerte (Archiv) (Climate data for
Germany – hourly data (archive))’. URL: https://www.dwd.de/DE/leistungen/
klimadatendeutschland/klarchivstunden.html?nn=16102, last accessed 2018-06-08.
DWD (2018d), ‘Vieljährige Mittelwerte (Long-term mean values)’. URL: https://www.
dwd.de/DE/leistungen/klimadatendeutschland/vielj_mittelwerte.html, last accessed
2018-06-08.
DWD (2018e), ‘Wetterlexikon – Niederschlagsintensität (Weather Encyclopedia – Pre-
cipitation intensities)’. URL: https://www.dwd.de/DE/service/lexikon/Functions/
glossar.html;jsessionid=4B114C00C68DCF3517F3329C67D52136.live21074?lv2=
101812&lv3=101906, last accessed 2018-03-28.
EC (2012), ‘Soil sealing’, Science for Environment Policy, DG Environment News Alert
Service. In-depth report, URL: http://ec.europa.eu/environment/soil/pdf/guidelines/
pub/soil_en.pdf, last accessed 2015-11-10.
Eckley, C. S. and Branfireun, B. (2009), ‘Simulated rain events on an urban roadway
to understand the dynamics of mercury mobilization in stormwater runoff’, Water
Research, 43(15), 3635–3646. doi: 10.1016/j.watres.2009.05.022.
Fletcher, T., Andrieu, H. and Hamel, P. (2013), ‘Understanding, management and mod-
elling of urban hydrology and its consequences for receiving waters: A state of the art’,
Advances in Water Resources, 51, 261–279. doi: 10.1016/j.advwatres.2012.09.001.
Flöter, O. (2006), Wasserhaushalt gepflasterter Straßen und Gehwege. Lysimeterver-
suche an drei Aufbauten unter praxisnahen Bedingungen unter Hamburger Klima
(Water balance of roads and sidewalks. Lysimeter experiments for three surfaces un-
der realistic conditions with Hamburg climate)., in ‘Hamburger Bodenkundliche Ar-
beiten’, Vol. 58.
Forghani, M. and Delavar, M. (2014), ‘A Quality Study of the OpenStreetMap Dataset
for Tehran’, ISPRS International Journal of Geo-Information, 3(2), 750–763. doi:
10.3390/ijgi3020750.
Fukahori, K. and Kubota, Y. (2003), ‘The role of design elements on the cost-effectiveness
of streetscape improvement’, Landscape and Urban Planning, 63(2), 75–91. doi:
10.1016/s0169-2046(02)00180-9.
Garcia, A., Hassn, A., Chiarelli, A. and Dawson, A. (2015), ‘Multivariable analysis of
potential evaporation from moist asphalt mixture’, Construction and Building Mate-
rials, 98, 80–88. doi: 10.1016/j.conbuildmat.2015.08.061.
García, L., Barreiro-Gomez, J., Escobar, E., Téllez, D., Quijano, N. and
Ocampo-Martinez, C. (2015), ‘Modeling and real-time control of urban drainage
78
systems: A review’, Advances in Water Resources, 85, 120–132. doi:
10.1016/j.advwatres.2015.08.007.
García, P. and Pérez, E. (2016), ‘Mapping of soil sealing by vegetation indexes and
built-up index: A case study in Madrid (Spain)’, Geoderma, 268, 100–107. doi:
10.1016/j.geoderma.2016.01.012.
Göbel, P., Coldewey, W., Dierkes, C., Kories, H., Meßer, J. and Meißner, E. (2007), ‘Ein-
fluss von Gründächern und Regenwassernutzungen auf Wasserhaushalt und Grund-
wasserstand in Siedlungen (Effect of green roofs and usage of rainwater on the water
balance and groundwater level of residential areas)’, Grundwasser, 12(3), 189–200.
doi: 10.1007/s00767-007-0032-y.
Göbel, P., Dierkes, C. and Coldewey, W. (2007), ‘Storm water runoff concentration
matrix for urban areas’, Journal of Contaminant Hydrology, 91(1-2), 26–42. doi:
10.1016/j.jconhyd.2006.08.008.
Göbel, P., Starke, P. and Coldewey, W. G. (2008), Evaporation measurements on en-
hanced water-permeable paving in urban areas, in ‘11th International Conference on
Urban Drainage, Edinburgh, Scotland, UK, 2008’.
Gessner, M., Hinkelmann, R., Nützmann, G., Jekel, M., Singer, G., Lewandowski, J.,
Nehls, T. and Barjenbruch, M. (2014), ‘Urban water interfaces’, Journal of Hydrol-
ogy, 514, 226–232. doi: 10.1016/j.jhydrol.2014.04.021.
Gilbert, J. K. and Clausen, J. C. (2006), ‘Stormwater runoff quality and quantity from as-
phalt, paver, and crushed stone driveways in Connecticut’, Water Research, 40(4), 826–
832. doi: 10.1016/j.watres.2005.12.006.
Glugla, G., Eyrich, A., König, B. and Fürtig, G. (1987), ‘Wasserhaushaltsuntersuchun-
gen im Raum Berlin (Investigating the water balance of Berlin)’, Wasserwirtschaft
Wassertechnik, 5, 113–116.
Glugla, G., Goedecke, M., Wessolek, G. and Fürtig, G. (1999), ‘Langjährige Abflussbil-
dung und Wasserhaushalt im urbanen Gebiet Berlin (Long-term formation of runoff
and water balance of the urban area Berlin)’, Wasserwirtschaft, 89, 34–42.
Gogate, N. G., Kalbar, P. P. and Raval, P. M. (2017), ‘Assessment of stormwater man-
agement options in urban contexts using Multiple Attribute Decision-Making’, Journal
of Cleaner Production, 142(4), 2046–2059. doi: 10.1016/j.jclepro.2016.11.079.
Google (2017), ‘Google street view’. URL: https://www.google.com/streetview/, last
accessed 2017-11-21.
Grimmond, C. S. B. and Oke, T. R. (1991), ‘An evapotranspiration-interception
model for urban areas’, Water Resources Research, 27(7), 1739–1755. doi:
10.1029/91wr00557.
79
Grodek, T., Lange, J., Lekach, J. and Husary, S. (2011), ‘Urban hydrology in mountain-
ous middle eastern cities’, Hydrology and Earth System Sciences, 15(3), 953–966. doi:
10.5194/hess-15-953-2011.
Grolemund, G. and Wickham, H. (2011), ‘Dates and Times Made Easy with lubridate’,
Journal of Statistical Software, 40(3), 1–25.
Haase, D. (2009), ‘Effects of urbanisation on the water balance – A long-term
trajectory’, Environmental Impact Assessment Review, 29(4), 211–219. doi:
10.1016/j.eiar.2009.01.002.
Hannes, M., Wollschläger, U., Schrader, F., Durner, W., Gebler, S., Pütz, T., Fank,
J., Unold, G. V. and j. Vogel, H. (2015), ‘A comprehensive filtering scheme for high-
resolution estimation of the water balance components from high-precision lysimeters’,
Hydrology and Earth System Sciences, 19(8), 3405–3418. doi: 10.5194/hess-19-3405-
2015.
Haselbach, L. M., Valavala, S. and Montes, F. (2006), ‘Permeability predictions for
sand-clogged Portland cement pervious concrete pavement systems’, Journal of Envi-
ronmental Management, 81(1), 42–49. doi: 10.1016/j.jenvman.2005.09.019.
Hassn, A., Chiarelli, A., Dawson, A. and Garcia, A. (2016), ‘Thermal properties of
asphalt pavements under dry and wet conditions’, Materials & Design, 91, 432–439.
doi: 10.1016/j.matdes.2015.11.116.
Heitman, J. L., Horton, R., Sauer, T. J. and DeSutter, T. M. (2008), ‘Sensible Heat
Observations Reveal Soil-Water Evaporation Dynamics’, Journal of Hydrometeorol-
ogy, 9(1), 165–171. doi: 10.1175/2007JHM963.1.
Hendel, M., Gutierrez, P., Colombert, M., Diab, Y. and Royon, L. (2016), ‘Mea-
suring the effects of urban heat island mitigation techniques in the field: Appli-
cation to the case of pavement-watering in paris’, Urban Climate, 16, 43–58. doi:
10.1016/j.uclim.2016.02.003.
Hibbs, B. J. and Sharp, J. M. (2012), ‘Hydrogeological Impacts of Urbanization’, Envi-
ronmental & Engineering Geoscience, 18(1), 3–24. doi: 10.2113/gseegeosci.18.1.3.
Hirschi, M., Michel, D., Lehner, I. and Seneviratne, S. I. (2017), ‘A site-level comparison
of lysimeter and eddy covariance flux measurements of evapotranspiration’, Hydrology
and Earth System Sciences, 21. doi: 10.5194/hess-21-1809-2017.
Hollis, G. E. and Ovenden, J. C. (1988a), ‘One year irrigation experiment to assess
losses and runoff volume relationships for a residential road in Hertfordshire, England’,
Hydrological Processes, 2(1), 61–74. doi: 10.1002/hyp.3360020106.
80
Hollis, G. E. and Ovenden, J. C. (1988b), ‘The quantity of stormwater runoff from ten
stretches of road, a car park and eight roofs in Hertfordshire, England during 1983’,
Hydrological Processes, 2(3), 227–243. doi: 10.1002/hyp.3360020304.
Howell, T., Schneider, A. and Jensen, M. (1991), History of Lysimeter Design and Use for
Evapotranspiration Measurements, in ‘Proceedings of the International Symposium on
Lysimetry, Honolulu’, pp. 1–9.
InfraTec (2008), IRBIS 3 Infrared Thermography Software – User Manual.
InfraTec (2012), VarioCAM hr head incl. IRBIS remote 3.0 software description – User
Manual.
Jacobson, C. R. (2011), ‘Identification and quantification of the hydrological impacts of
imperviousness in urban catchments: A review’, Journal of Environmental Manage-
ment, 92(6), 1438–1448. doi: 10.1016/j.jenvman.2011.01.018.
Jacqueminet, C., Kermadi, S., Michel, K., Béal, D., Gagnage, M., Branger, F., Jankowf-
sky, S. and Braud, I. (2013), ‘Land cover mapping using aerial and VHR satellite
images for distributed hydrological modelling of periurban catchments: Application
to the Yzeron catchment (Lyon, France)’, Journal of Hydrology, 485, 68–83. doi:
10.1016/j.jhydrol.2013.01.028.
Jung, H., young Lee, S., Kim, H. S. and Lee, J. S. (2017), ‘Does improving the physical
street environment create satisfactory and active streets? evidence from seoul’s design
street project’, Transportation Research Part D: Transport and Environment, 50, 269–
279. doi: 10.1016/j.trd.2016.11.013.
Kakar, M. R., Hamzah, M. O. and Valentin, J. (2015), ‘A review on moisture damages
of hot and warm mix asphalt and related investigations’, Journal of Cleaner Produc-
tion, 99, 39–58. doi: 10.1016/j.jclepro.2015.03.028.
Kaparias, I., Bell, M., Biagioli, T., Bellezza, L. and Mount, B. (2015), ‘Behavioural
analysis of interactions between pedestrians and vehicles in street designs with ele-
ments of shared space’, Transportation Research Part F: Traffic Psychology and Be-
haviour, 30, 115–127. doi: 10.1016/j.trf.2015.02.009.
Kardos, A. J. and Durham, S. A. (2015), ‘Strength, durability, and environmental prop-
erties of concrete utilizing recycled tire particles for pavement applications’, Construc-
tion and Building Materials, 98, 832–845. doi: 10.1016/j.conbuildmat.2015.08.065.
Kassambara, A. (2016), ggcorrplot: Visualization of a Correlation Matrix using ’ggplot2’.
URL: https://CRAN.R-project.org/package=ggcorrplot, last accessed 2018-06-10.
Kayhanian, M., Anderson, D., Harvey, J. T., Jones, D. and Muhunthan, B. (2012),
‘Permeability measurement and scan imaging to assess clogging of pervious concrete
81
pavements in parking lots’, Journal of Environmental Management, 95(1), 114–123.
doi: 10.1016/j.jenvman.2011.09.021.
Kelly, G., Delaney, D., Chai, G. and Mohamed, S. (2016), ‘Optimising local council's
return on investment from annual pavement rehabilitation budgets through target-
ing of the average pavement condition index’, Journal of Traffic and Transportation
Engineering (English Edition), 3(5), 465–474. doi: 10.1016/j.jtte.2016.09.008.
Kodešová, R., Fér, M., Klement, A., Nikodem, A., Teplá, D., Neuberger, P. and Bureš,
P. (2014), ‘Impact of various surface covers on water and thermal regime of Technosol’,
Journal of Hydrology, 519, 2272–2288. doi: 10.1016/j.jhydrol.2014.10.035.
Kottek, M., Grieser, J., Beck, C., Rudolf, B. and Rubel, F. (2006), ‘World Map of the
Köppen-Geiger climate classification updated’, Meteorologische Zeitschrift, 15(3), 259–
263. doi: 10.1127/0941-2948/2006/0130.
Lakshmi, V., Jackson, T. J. and Zehrfuhs, D. (2003), ‘Soil moisture-temperature relation-
ships: results from two field experiments’, Hydrological Processes, 17(15), 3041–3057.
doi: 10.1002/hyp.1275.
Lanthaler, C. (2004), Lysimeter Stations and Soil Hydrology Measuring Sites in Europe
– Purpose, Equipment, Research Results, Future Developments, Master’s thesis. URL:
http://www.lysimeter.at/HP_EuLP/reports/THESIS_LYSIMETERS.pdf, last ac-
cessed 2018-06-03.
Ledieu, J., Ridder, P. D., Clerck, P. D. and Dautrebande, S. (1986), ‘A method of
measuring soil moisture by time-domain reflectometry’, Journal of Hydrology, 88(3-
4), 319–328. doi: 10.1016/0022-1694(86)90097-1.
Lehmann, A. and Stahr, K. (2007), ‘Nature and significance of anthropogenic urban
soils’, Journal of Soils and Sediments, 7(4), 247–260. doi: 10.1065/jss2007.06.235.
Lerner, D. N. (2002), ‘Identifying and quantifying urban recharge: a review’, Hydroge-
ology Journal, 10(1), 143–152. doi: 10.1007/s10040-001-0177-1.
Liu, Z. and Hansen, W. (2016a), ‘Effect of hydrophobic surface treatment on freeze-
thaw durability of concrete’, Cement and Concrete Composites, 69, 49–60. doi:
10.1016/j.cemconcomp.2016.03.001.
Liu, Z. and Hansen, W. (2016b), ‘Freeze–thaw durability of high strength concrete under
deicer salt exposure’, Construction and Building Materials, 102(1), 478–485. doi:
10.1016/j.conbuildmat.2015.10.194.
López-Montero, T. and Miró, R. (2016), ‘Differences in cracking resistance of asphalt
mixtures due to ageing and moisture damage’, Construction and Building Materi-
als, 112, 299–306. doi: 10.1016/j.conbuildmat.2016.02.199.
82
Mansell, M. and Rollet, F. (2006), ‘Water balance and the behaviour of different
paving surfaces’, Water and Environment Journal, 20(1), 7–10. doi: 10.1111/j.1747-
6593.2005.00015.x.
Mansell, M. and Rollet, F. (2009), ‘The effect of surface texture on evaporation, infiltra-
tion and storage properties of paved surfaces’, Water Science & Technology, 60(1), 71.
doi: 10.2166/wst.2009.323.
Mansell, M. and Wang, S. (2010), ‘Water balance modelling in glasgow and beijing’,
Proceedings of the Institution of Civil Engineers - Water Management, 163(5), 219–
226. doi: 10.1680/wama.2010.163.5.219.
Meissner, R., Prasad, M., Du Laing, G. and Rinklebe, J. (2010), ‘Lysimeter appli-
cation for measuring the water and solute fluxes with high precision’, Current Sci-
ence, 99(5), 601–607.
Meissner, R., Seeger, J., Rupp, H., Seyfarth, M. and Borg, H. (2007), ‘Measurement
of dew, fog, and rime with a high-precision gravitation lysimeter’, Journal of Plant
Nutrition and Soil Science, 170(3), 335–344. doi: 10.1002/jpln.200625002.
Menziani, M., Pugnaghi, S., Vincenzi, S. and Santangelo, R. (2003), ‘Soil moisture
monitoring in the Toce valley (Italy)’, Hydrology and Earth System Sciences, 7(6), 890–
902. doi: 10.5194/hess-7-890-2003.
MetOffice (2012), ‘National meteorological library and archive fact sheet 3 – water in
the atmosphere (version 01)’. URL: https://www.metoffice.gov.uk/binaries/content/
assets/mohippo/pdf/f/c/fact_sheet_no._3.pdf, last accessed 2018-06-04.
Miller, J. D., Kim, H., Kjeldsen, T. R., Packman, J., Grebby, S. and Dearden, R. (2014),
‘Assessing the impact of urbanization on storm runoff in a peri-urban catchment us-
ing historical change in impervious cover’, Journal of Hydrology, 515, 59–70. doi:
10.1016/j.jhydrol.2014.04.011.
Mitchell, V., Mein, R. and McMahon, T. (2001), ‘Modelling the urban water cycle’, Envi-
ronmental Modelling & Software, 16(7), 615–629. doi: 10.1016/s1364-8152(01)00029-9.
Montzka, C., Canty, M., Kunkel, R., Menz, G., Vereecken, H. and Wendland, F.
(2008), ‘Modelling the water balance of a mesoscale catchment basin using re-
motely sensed land cover data’, Journal of Hydrology, 353(3-4), 322–334. doi:
10.1016/j.jhydrol.2008.02.018.
Morabito, M., Crisci, A., Messeri, A., Orlandini, S., Raschi, A., Maracchi, G. and
Munafò, M. (2016), ‘The impact of built-up surfaces on land surface temperatures
in italian urban areas’, Science of The Total Environment, 551-552, 317–326. doi:
10.1016/j.scitotenv.2016.02.029.
83
Morel, J. L., Chenu, C. and Lorenz, K. (2014), ‘Ecosystem services provided by soils
of urban, industrial, traffic, mining, and military areas (SUITMAs)’, Journal of Soils
and Sediments, 15(8), 1659–1666. doi: 10.1007/s11368-014-0926-0.
Nakayama, T. and Fujita, T. (2010), ‘Cooling effect of water-holding pavements made
of new materials on water and heat budgets in urban areas’, Landscape and Urban
Planning, 96(2), 57–67. doi: 10.1016/j.landurbplan.2010.02.003.
Nandagiri, L. and Kovoor, G. M. (2006), ‘Performance Evaluation of Reference
Evapotranspiration Equations across a Range of Indian Climates’, Journal of Ir-
rigation and Drainage Engineering, 132(3), 238–249. doi: 10.1061/(asce)0733-
9437(2006)132:3(238).
Nehls, T., Jozefaciuk, G., Sokołowska, Z., Hajnos, M. and Wessolek, G. (2006), ‘Pore-
system characteristics of pavement seam materials of urban sites’, Journal of Plant
Nutrition and Soil Science, 169(1), 16–24. doi: 10.1002/jpln.200521724.
Nehls, T., Jozefaciuk, G., Sokolowska, Z., Hajnos, M. and Wessolek, G. (2008), ‘Filter
properties of seam material from paved urban soils’, Hydrology and Earth System
Sciences, 12(2), 691–702. doi: 10.5194/hess-12-691-2008.
Nehls, T., Menzel, M. and Wessolek, G. (2015), ‘Depression storage capacities of different
ideal pavements as quantified by a terrestrial laser scanning-based method’, Water
Science and Technology, 71(6), 862–869. doi: 10.2166/wst.2015.025.
Nehls, T., Rim, Y. N. and Wessolek, G. (2011), ‘Technical note on measuring run-off dy-
namics from pavements using a new device: the weighable tipping bucket’, Hydrology
and Earth System Sciences, 15(5), 1379–1386. doi: 10.5194/hess-15-1379-2011.
Niemczynowicz, J. (1999), ‘Urban hydrology and water management – present and future
challenges’, Urban Water, 1(1), 1–14. doi: 10.1016/s1462-0758(99)00009-6.
Oakdale Engineering (2014), ‘DataFit’. URL: http://www.oakdaleengr.com/datafit.htm,
last accessed 2018-06-19.
Open Street Map Contributors (2017), ‘Open Street Map’. URL: https://www.
openstreetmap.org, last accessed 2017-11-17.
Open Street Map Wiki Contributors (2017), ‘Open Street Map Wiki – Key:surface’.
URL: http://wiki.openstreetmap.org/wiki/Key:surface, last accessed 2017-11-17.
Penttala, V. (2006), ‘Surface and internal deterioration of concrete due to saline and
non-saline freeze–thaw loads’, Cement and Concrete Research, 36(5), 921–928. doi:
10.1016/j.cemconres.2005.10.007.
Peters, A., Nehls, T., Schonsky, H. and Wessolek, G. (2014), ‘Separating precipitation
and evapotranspiration from noise – a new filter routine for high-resolution lysimeter
84
data’, Hydrology and Earth System Sciences, 18(3), 1189–1198. doi: 10.5194/hess-18-
1189-2014.
Peters, A., Nehls, T. and Wessolek, G. (2016), ‘Technical note: Improving the AWAT
filter with interpolation schemes for advanced processing of high resolution data’,
Hydrology and Earth System Sciences, 20(6), 2309–2315. doi: 10.5194/hess-20-2309-
2016.
Pistocchi, A., Calzolari, C., Malucelli, F. and Ungaro, F. (2015), ‘Soil sealing and flood
risks in the plains of Emilia-Romagna, Italy’, Journal of Hydrology: Regional Stud-
ies, 4(B), 398–409. doi: 10.1016/j.ejrh.2015.06.021.
Qin, H., Li, Z. and Fu, G. (2013), ‘The effects of low impact development on urban
flooding under different rainfall characteristics’, Journal of Environmental Manage-
ment, 129, 577–585. doi: 10.1016/j.jenvman.2013.08.026.
Qin, Y. (2015), ‘A review on the development of cool pavements to mitigate urban
heat island effect’, Renewable and Sustainable Energy Reviews, 52, 445–459. doi:
10.1016/j.rser.2015.07.177.
R Core Team (2016), R: A Language and Environment for Statistical Computing, R
Foundation for Statistical Computing, Vienna, Austria. URL: https://www.R-project.
org/, last accessed 2018-06-04.
Ragab, R., Rosier, P., Dixon, A., Bromley, J. and Cooper, J. D. (2003), ‘Experimental
study of water fluxes in a residential area: 2. Road infiltration, runoff and evaporation’,
Hydrological Processes, 17(12), 2423–2437. doi: 10.1002/hyp.1251.
Ramier, D., Berthier, E. and Andrieu, H. (2004), ‘An urban lysimeter to assess
runoff losses on asphalt concrete plates’, Physics and Chemistry of the Earth, Parts
A/B/C, 29(11-12), 839–847. doi: 10.1016/j.pce.2004.05.011.
Ramier, D., Berthier, E. and Andrieu, H. (2011), ‘The hydrological behaviour of ur-
ban streets: long-term observations and modelling of runoff losses and rainfall-runoff
transformation’, Hydrological Processes, 25(14), 2161–2178. doi: 10.1002/hyp.7968.
Ramier, D., Berthier, E., Dangla, P. and Andrieu, H. (2006), ‘Study of the water budget
of streets: experimentation and modelling’, Water Science & Technology, 54(6-7), 41.
doi: 10.2166/wst.2006.587.
Rana, G. and Katerji, N. (2000), ‘Measurement and estimation of actual evapotran-
spiration in the field under Mediterranean climate: a review’, European Journal of
Agronomy, 13(2-3), 125–153. doi: 10.1016/s1161-0301(00)00070-8.
Raziei, T. and Pereira, L. S. (2013), ‘Estimation of ET0with Hargreaves–Samani and
FAO-PM temperature methods for a wide range of climates in Iran’, Agricultural
Water Management, 121, 1–18. doi: 10.1016/j.agwat.2012.12.019.
85
Revitt, D. M., Lundy, L., Coulon, F. and Fairley, M. (2014), ‘The sources, impact
and management of car park runoff pollution: A review’, Journal of Environmental
Management, 146, 552–567. doi: 10.1016/j.jenvman.2014.05.041.
Richards, D. R. and Edwards, P. J. (2017), ‘Quantifying street tree regulating ecosys-
tem services using Google Street View’, Ecological Indicators, 77, 31–40. doi:
10.1016/j.ecolind.2017.01.028.
Rim, Y. (2011), Analyzing Runoff Dynamics of paved Soil Surface Using Weighable
Lysimeters, PhD thesis, Technical University Berlin.
Rodriguez, F., Andrieu, H. and Morena, F. (2008), ‘A distributed hydrological model
for urbanized areas – Model development and application to case studies’, Journal of
Hydrology, 351(3-4), 268–287. doi: 10.1016/j.jhydrol.2007.12.007.
Rodríguez, M. C., Dupont-Courtade, L. and Oueslati, W. (2016), ‘Air pollution and
urban structure linkages: Evidence from European cities’, Renewable and Sustainable
Energy Reviews, 53, 1–9. doi: 10.1016/j.rser.2015.07.190.
Salvadore, E., Bronders, J. and Batelaan, O. (2015), ‘Hydrological modelling of urban-
ized catchments: A review and future directions’, Journal of Hydrology, 529(1), 62–81.
doi: 10.1016/j.jhydrol.2015.06.028.
Scalenghe, R. and Marsan, F. A. (2009), ‘The anthropogenic sealing of
soils in urban areas’, Landscape and Urban Planning, 90(1-2), 1–10. doi:
10.1016/j.landurbplan.2008.10.011.
Schrader, F., Durner, W., Fank, J., Gebler, S., Pütz, T., Hannes, M. and Wollschläger,
U. (2013), ‘Estimating Precipitation and Actual Evapotranspiration from Preci-
sion Lysimeter Measurements’, Procedia Environmental Sciences, 19, 543–552. doi:
10.1016/j.proenv.2013.06.061.
Schrödter, H. (1985), Verdunstung – Anwendungsorientierte Meßverfahren und Bestim-
mungsmethoden (Evaporation – Applied methods for measurement and estimation),
Springer Berlin Heidelberg. doi: 10.1007/978-3-642-70434-5.
Seiferling, I., Naik, N., Ratti, C. and Proulx, R. (2017), ‘Green streets - Quantifying and
mapping urban trees with street-level imagery and computer vision’, Landscape and
Urban Planning, 165, 93–101. doi: 10.1016/j.landurbplan.2017.05.010.
Shuster, W. D., Bonta, J., Thurston, H., Warnemuende, E. and Smith, D. R. (2005),
‘Impacts of impervious surface on watershed hydrology: A review’, Urban Water Jour-
nal, 2(4), 263–275. doi: 10.1080/15730620500386529.
Šimůnek, J., Genuchten, M. T. V. and Šejna, M. (2016), ‘Recent Developments and
Applications of the HYDRUS Computer Software Packages’, Vadose Zone Jour-
nal, 15(7), 25. doi: 10.2136/vzj2016.04.0033.
86
Smith, M. L., Zhou, W., Cadenasso, M., Grove, M. and Band, L. E. (2010), ‘Evalua-
tion of the National Land Cover Database for Hydrologic Applications in Urban and
Suburban Baltimore, Maryland’, JAWRA Journal of the American Water Resources
Association, 46(2), 429–442. doi: 10.1111/j.1752-1688.2009.00412.x.
Solpuker, U., Sheets, J., Kim, Y. and Schwartz, F. (2014), ‘Leaching potential of pervious
concrete and immobilization of Cu, Pb and Zn using pervious concrete’, Journal of
Contaminant Hydrology, 161, 35–48. doi: 10.1016/j.jconhyd.2014.03.002.
Starke, P., Göbel, P. and Coldewey, W. G. (2010), ‘Urban evaporation rates for
water-permeable pavements’, Water Science & Technology, 62(5), 1161. doi:
10.2166/wst.2010.390.
Starke, P., Göbel, P. and Coldewey, W. G. (2011), ‘Effects on evaporation rates from dif-
ferent water-permeable pavement designs’, Water Science & Technology, 63(11), 2619.
doi: 10.2166/wst.2011.168.
Stoffregen, H. (1998), ‘Hydraulische Eigenschaften deponiespezifischer Materialien unter
Berücksichtigung von Temperaturveränderungen (Hydraulic properties of landfill ma-
terials related to temperature changes)’. Rote Reihe - Bodenökologie und Bodengenese
Heft 32.
Thoen, E. (2017), padr: Quickly Get Datetime Data Ready for Analysis. URL: https:
//CRAN.R-project.org/package=padr, last accessed 2018-06-10.
Timm, A., Kluge, B. and Wessolek, G. (2018), ‘Hydrological balance of paved surfaces
in moist mid-latitude climate – a review’, Landscape and Urban Planning, 175, 80–91.
doi: 10.1016/j.landurbplan.2018.03.014.
Trinks, S. (2010), Einfluss des Wasser- und Wärmehaushaltes von Böden auf den Be-
trieb erdverlegter Energiekabel (Influence hydrological and heat balance of the soil on
operation of underground power cables), PhD thesis, Technical University Berlin.
UN (2014), World urbanization prospects - 2014 revision, Technical report, United
Nations. URL: https://esa.un.org/unpd/wup/publications/files/wup2014-report.pdf,
last accessed 2018-06-21.
Unold, G. V. and Fank, J. (2007), ‘Modular Design of Field Lysimeters for Spe-
cific Application Needs’, Water, Air, & Soil Pollution: Focus, 8(2), 233–242. doi:
10.1007/s11267-007-9172-4.
Ward, H., Kotthaus, S., Järv, i. L. and Grimmond, C. (2016), ‘Surface Urban Energy
and Water Balance Scheme (SUEWS): Development and evaluation at two UK sites’,
Urban Climate, 18, 1–32. doi: 10.1016/j.uclim.2016.05.001.
87
Wessolek, G. (1993), Erarbeitung eines Schlüssels zur Einschätzung von Versickerung
und Oberflächenabfluß versiegelter Flächen Berlins (Development of a tool to estimate
infiltration and runoff of sealed surfaces in Berlin). Unpublished Research Report
commissioned by the Federal Agency for Hydrology Germany, exectued at Technical
University Berlin, Germany.
Wessolek, G. (1994), Auswertung von Versuchen zur Ermittlung der Abflußverhältnisse
unterschiedlich versiegelter und kanalisierter Flächen Berlins (Evaluating experiments
investigating drainage from different sealed and canalised areas in Berlin). Unpub-
lished Research Report commissioned by the Federal Agency for Hydrology Germany,
exectued at Technical University Berlin, Germany.
Wessolek, G. (2001), ‘Bodenüberformung und -versiegelung (Transformation and sealing
of soil)’, Handbuch der Bodenkunde. doi: 10.1002/9783527678495.hbbk2001002.
Wessolek, G., Duijnisveld, W. and Trinks, S. (2008), ‘Hydro-pedotransfer functions
(HPTFs) for predicting annual percolation rate on a regional scale’, Journal of Hy-
drology, 356(1-2), 17–27. doi: 10.1016/j.jhydrol.2008.03.007.
Wessolek, G. and Facklam, M. (1997), ‘Standorteigenschaften und Wasserhaushalt von
versiegelten Flächen (Characteristics and water balance of sealed surfaces)’, Journal
of Plant Nutrition and Soil Science, 160(1), 41–46. doi: 10.1002/jpln.19971600109.
Wessolek, G., Kaupenjohann, M. and Renger, M. (2009), Bodenphysikalische Ken-
nwerte und Berechnungsverfahren für die Praxis (Soil physical parameters and cal-
culation methods for use in practice), Vol. 40 of Bodenökologie und Bodengenese,
TU Berlin. URL: https://www.boden.tu-berlin.de/fileadmin/fg77/_pdf/Rote_Liste/
Rote_Reihe_Heft_40.pdf, last accessed 2018-06-25.
Wessolek, G., Kluge, B., Nehls, T. and Kocher, B. (2009), ‘Aspekte zum Wasser-
haushalt und Stofftransport urbaner Flächen (Aspects of water balance and solute
transport of urban surfaces)’, Korrespondenz Wasserwirtschaft, 2(4), 205–210. doi:
10.3243/kwe2009.04.001.
Wickham, H. (2009), ggplot2: Elegant Graphics for Data Analysis, Springer-Verlag New
York.
Wickham, H. (2011), ‘The Split-Apply-Combine Strategy for Data Analysis’, Journal of
Statistical Software, 40(1), 1–29.
Wiles, T. J. and Sharp, J. M. (2008), ‘The Secondary Permeability of Impervious Cover’,
Environmental and Engineering Geoscience, 14(4), 251–265. doi: 10.2113/gsee-
geosci.14.4.251.
88
Willuweit, L. and O'Sullivan, J. J. (2013), ‘A decision support tool for sustainable
planning of urban water systems: Presenting the Dynamic Urban Water Simulation
Model’, Water Research, 47(20), 7206–7220. doi: 10.1016/j.watres.2013.09.060.
Wu, J., Tang, C., Shi, B., Gao, L., Jiang, H. and Daniels, J. (2014), ‘Effect of Ground
Covers on Soil Temperature in Urban and Rural Areas’, Environmental & Engineering
Geoscience, 20(3), 225–237. doi: 10.2113/gseegeosci.20.3.225.
Xu, H., Guo, W. and Tan, Y. (2016), ‘Permeability of asphalt mixtures exposed
to freeze–thaw cycles’, Cold Regions Science and Technology, 123, 99–106. doi:
10.1016/j.coldregions.2015.12.001.
Yao, L., Wei, W. and Chen, L. (2016), ‘How does imperviousness impact the urban
rainfall-runoff process under various storm cases?’, Ecological Indicators, 60, 893–905.
doi: 10.1016/j.ecolind.2015.08.041.
Yoder, R. E., Odhiambo, L. O. and Wright, W. C. (2005), ‘Evaluation Of Methods
For Estimating Daily Reference Crop Evapotranspiration At A Site In The Humid
Southeast United States’, Applied Engineering in Agriculture, 21(2), 197–202. doi:
10.13031/2013.18153.
Yong, C., McCarthy, D. and Deletic, A. (2013), ‘Predicting physical clogging
of porous and permeable pavements’, Journal of Hydrology, 481, 48–55. doi:
10.1016/j.jhydrol.2012.12.009.
Zeileis, A. and Grothendieck, G. (2005), ‘zoo: S3 Infrastructure for Regular and Irregular
Time Series’, Journal of Statistical Software, 14(6), 1–27. doi: 10.18637/jss.v014.i06.
Zhang, W., Min, H. and Gu, X. (2016), ‘Temperature response and moisture transport
in damaged concrete under an atmospheric environment’, Construction and Building
Materials, 123, 290–299. doi: 10.1016/j.conbuildmat.2016.07.004.
Zoppou, C. (2001), ‘Review of urban storm water models’, Environmental Modelling &
Software, 16(3), 195–231. doi: 10.1016/s1364-8152(00)00084-0.
89
A Appendices
A.1 Extended tables & figures
Volumetric water content (θ) [Vol.-%]
Cobblestones Concrete slabs
5cm 15 cm 25 cm 5cm 15cm 25cm
2016-06 11.46 14.79 18.09 10.18 13.71 17.23
2016-07 11.32 14.74 17.94 9.90 13.53 17.05
2016-08 11.45 14.91 17.94 9.87 13.63 17.07
2016-09 11.14 14.68 17.92 9.53 13.35 16.67
2016-10†10.91 14.06 18.29 9.32 13.15 16.78
2016-11 9.75 13.48 18.32 8.21 12.60 16.36
2016-12 10.00 13.25 18.06 7.86 12.61 16.16
2017-01 9.88 13.28 18.32 6.74 12.53 16.04
2017-02 10.29 13.44 18.46 7.75 12.71 16.20
2017-03 10.87 13.92 18.74 8.51 13.14 16.80
2017-04 10.69 13.87 18.25 8.66 13.14 16.52
2017-05 11.02 14.27 17.83 9.43 13.60 17.05
†No data available for 17 out of 31 days in October 2016
Table A.1: Monthly mean volumetric water content (θ) [Vol.-%] measured at 5, 15 and
25 cm below paver
I
Cobblestones Concrete slabs
surface paver∗5cm 15 cm 25 cm surface paver∗5cm 15cm 25cm Tair
Annual
Min T[◦C] -8.03 -6.58 -0.80 0.63 0.95 -7.33 -7.17 -1.86 0.49 1.10 -11.30
Mean T[◦C]11.73 10.26 13.28 13.52 13.59 12.81 11.00 13.66 14.01 14.06 10.13
Median T[◦C]9.61 7.85 11.05 11.29 11.50 10.32 8.38 11.20 11.74 11.89 9.20
Max T[◦C]47.23 43.92 39.63 35.57 33.43 51.52 47.12 44.15 38.41 34.62 35.50
Summer
Min T[◦C] 0.75 0.63 5.61 7.08 7.94 1.99 1.23 5.01 7.45 8.68 -0.30
Mean T[◦C]19.62 18.35 21.04 20.96 20.78 21.31 19.46 21.96 21.81 21.47 16.40
Median T[◦C]18.98 18.07 21.90 22.20 22.18 20.09 18.71 22.29 22.83 22.81 16.35
Max T[◦C]47.23 43.92 39.63 35.57 33.43 51.52 47.12 44.15 38.41 34.62 35.50
Winter
Min T[◦C] -8.03 -6.58 -0.80 0.63 0.95 -7.33 -7.17 -1.86 0.49 1.10 -11.30
Mean T[◦C] 4.13 2.47 5.78 6.34 6.66 4.62 2.84 5.64 6.49 6.90 3.76
Median T[◦C] 3.75 1.83 5.55 6.29 6.65 4.01 2.00 4.97 6.20 6.70 3.60
Max T[◦C] 26.72 24.40 21.54 18.89 17.59 30.73 28.68 26.64 21.57 19.08 24.00
∗Measurement underside paver
Table A.2: Hourly surface & soil temperatures Tmeasured on top and within the lysimeter soil column. Hydrological summer May to October,
hydrological winter November to April
II
Monthly mean temperature [◦C]
Cobblestones Concrete slabs
surface paver∗5cm 15 cm 25 cm surface paver∗5cm 15cm 25cm Tair
2016-06 24.18 23.22 25.37 24.96 24.50 26.21 24.63 26.73 26.10 25.38 19.03
2016-07 23.45 21.83 24.63 24.42 24.10 25.21 23.14 25.81 25.44 24.89 19.77
2016-08 23.21 21.85 24.34 24.14 23.87 24.61 22.80 25.15 24.87 24.35 18.28
2016-09 20.43 18.90 21.78 21.84 21.77 21.59 19.72 22.37 22.41 22.23 17.80
2016-10†9.54 7.72 11.02 11.57 11.89 10.12 8.09 10.98 11.78 12.24 8.73
2016-11 3.02 1.21 4.70 5.36 5.76 3.41 1.43 4.45 5.54 6.15 3.56
2016-12 2.18 -0.03 3.17 3.71 4.00 2.25 0.02 2.85 3.74 4.24 2.61
2017-01 0.19 -1.63 2.09 3.04 3.61 0.05 -1.68 1.32 2.75 3.42 -1.16
2017-02 2.81 1.09 4.68 5.45 5.86 2.93 1.19 4.21 5.13 5.856 2.05
2017-03 7.09 5.62 8.83 9.33 9.61 7.80 6.28 8.83 9.64 9.91 7.22
2017-04 9.72 8.85 11.60 11.62 11.59 11.49 10.06 12.39 12.48 12.25 8.29
2017-05 19.88 19.56 21.73 21.28 20.82 23.14 21.46 23.49 22.74 22.03 14.98
∗Measurement underside paver
†No data available for 17 out of 31 days in October 2016, Tair unaffected
Table A.3: Monthly mean air, surface & soil temperatures Tmeasured on top and within the lysimeter soil column
III
Year Season Month
S W 1 2 3 4 5 6 7 8 9 10 11 12
Annual
κcobblestones 0.22 0.16 0.30 0.62 0.41 0.29 0.25 0.20 0.11 0.13 0.13 0.13 0.32 0.35 0.36
κconcrete slabs 0.12 0.08 0.34 0.63 0.44 0.19 0.06 0.08 0.08 0.08 0.08 0.05 0.29 0.59 1.14
Dry days
κdcobblestones 0.21 0.15 0.33 1.31 0.41 0.26 0.26 0.21 0.08 0.10 0.12 0.12 0.47∗0.47 0.83∗
κdconcrete slabs 0.08 0.05 0.31 0.61 0.87 0.12 0.05 0.05 0.06 0.05 0.07 0.04 0.38∗0.74 6.13∗
Wet days
κwcobblestones 0.24 0.18 0.28 0.25 0.40 0.42 0.23 0.18 0.21 0.16 0.13∗0.30∗0.20∗0.18 0.24
κwconcrete slabs 0.23 0.15 0.39 0.69 0.28 0.37 0.15 0.13 0.14 0.17 0.11∗0.38∗0.09∗0.52 0.64
number dry days 183 101 82 14 12 15 20 21 16 15 17 27 5 13 8
number wet days 149 59 90 15 14 15 10 10 14 14 9 3 9 15 21
∗Values based on few data points (<10 values)
Table A.4: Ratio (κ) of evaporation Eof paved surfaces to grass-reference evapotranspiration ET0. Based on daily measurements (E)
and calculations (ET0) for yearly, seasonal and monthly data, and for both wet and dry days. Dry days defined as days with
P= 0 mm d−1, all other days considered wet days. S = Summer, W = Winter
IV
Figure A.1: Hourly air, surface & soil temperatures – period with highest precipitation
event from September 16th to September 20th, 2016
V
Figure A.2: Deriving median measured monthly wet (κw) and dry (κw) reduction coef-
ficients from monthly mean values of Tair
VI
A.2 Lysimeter reconstruction
(a) empty lysimeter (b) mesh protecting infiltration outlet
(c) installation of gradient drainage layer (d) compacting of layers
(e) preparing slope (f) installation of cobblestone surface
(g) fresh surface – concrete slabs (h) fresh surface – cobblestones
Figure A.3: Impressions from lysimeter reconstruction in April 2016
VII
Figure A.4: Installation of sensor for soil water content and temperature measurement,
sensor is pushed into already compacted soil
VIII
