A conceptual data modeling framework with four levels of abstraction for environmental information

Con en s lis s a ailable a ScienceDi ec
En i onmen al Modelling and So wa e
jou nal homepage: www.else ie .com/loca e/en so
A concep ual da a modeling amewo k wi h ou le els o abs ac ion o
en i onmen al in o ma ion
Da id Ma ínez a, Lau a Po b, Raquel T illo-Lado c, José R.R. Viquei a a,∗
aDepa amen o de Elec ónica e Compu ación, Uni e sidade de San iago de Compos ela, San iago de Compos ela, A Co uña, Spain
bDipa imen o di Ingegne ia "Enzo Fe a i", Uni e si à degli s udi di Modena e Reggio Emilia, Modena, Reggio Emilia, I aly
cDepa amen o de In o má ica e Ingenie ía de Sis emas, Ins i u o de In es igación en Ingenie ía de A agón, Uni e sidad de Za agoza, Za agoza, Za agoza, Spain
ARTICLE INFO
Da ase link: h ps://da a.eu opa.eu/,h ps://
www.me eogalicia.gal,h ps://www.in ecma .
gal
Keywo ds:
Concep ual da a modeling
En i onmen al da a
Da a managemen
Me eo ological da a
Oceanog aphic da a
Ai quali y da a
Da a in eg a ion
ABSTRACT
En i onmen al da a gene a ed by obse a ion in as uc u es and models is widely he e ogeneous in bo h
s uc u e and seman ics. The design and implemen a ion o an ad hoc da a model o each new da ase
is cos ly and c ea es ba ie s o da a in eg a ion. On he o he hand, designing a single da a model ha
suppo s any kind o en i onmen al da a has shown o be a complex ask, and he esul ing ools do no
p o ide he equi ed e iciency. In his pape , a new da a modeling amewo k is p oposed ha enables he
euse o gene ic s uc u es among di e en applica ion domains and speci ic applica ions. The amewo k
conside s ou le els o abs ac ion o he da a models. Le els 1 and 2 p o ide gene al da a model s uc u es
o en i onmen al da a, based on hose de ined by he Obse a ions and Measu emen s (O&M) s anda d o he
Open Geospa ial Conso ium (OGC). Le el 3 inco po a es gene ic da a models o di e en applica ion a eas,
whe eas speci ic applica ion models a e designed a Le el 4, eusing s uc u es o he p e ious le els. Va ious
use cases we e implemen ed o illus a e he capabili ies o he amewo k. A pe o mance e alua ion using
six da ase s o h ee di e en use cases has shown ha he que y esponse imes achie ed o e he s uc u es
o Le el 4 a e e y good compa ed o bo h ad hoc models and o a di ec implemen a ion o O&M in a
Senso Obse a ion Se ice (SOS) ool. A quali a i e e alua ion shows ha he amewo k ul ills a collec ion
o gene al equi emen s no suppo ed by any o he exis ing solu ion.
So wa e a ailabili y
So wa e All he SQL code used o he implemen a ion and e alu-
a ion o he cu en p o o ype o he p esen amewo k is a ailable in
he ollowing u l: h ps://gi hub.com/cog adeusc/en damo .
1. In oduc ion
The obse a ion and p edic ion o he condi ions o ou en i on-
men is a keys one o he p og ess o many scien i ic disciplines.
Scien is s ha e been de eloping in as uc u es ha gene a e and s o e
huge amoun s o en i onmen al da a. The da a s o age and access
subsys ems ange om e y simple ools de eloped in he scope o
speci ic esea ch p ojec s o e y complex da a hubs ha in eg a e he
deluge o in o ma ion gene a ed by sophis ica ed pubic da a gene a ion
∗Co esponding au ho .
E-mail add esses: [email p o ec ed] (D. Ma ínez), [email p o ec ed] (L. Po), [email p o ec ed] (R. T illo-Lado), [email p o ec ed]
(J.R.R. Viquei a).
1h ps://da aspace.cope nicus.eu/
2h ps:// da.uca .edu/
3h ps://da a.noaa.go /ones op/
4h ps://www.hyd osha e.o g/
in as uc u es and la ge scien i ic communi ies. Examples o such com-
plex da a hubs a e he Cope nicus Da a Space Ecosys em,1 he Global
Ea h Obse a ion Sys em o Sys ems (GEOSS) (Na i i e al.,2015),
he NCAR Resea ch Da a A chi e,2 he NOAA OneS op po al3and he
Hyd osha e online collabo a ion en i onmen .4En i onmen al da ase s
ha e c i ical impo ance o use s wi h e y specialized and high skills
o science and enginee ing, howe e , hey a e also he basis o e which
gene al se ices o ci izens may be de eloped (Viquei a e al.,2020).
En i onmen al da a in as uc u es mus p o ide da a s o age s uc-
u es and da a disco e y and access mechanisms. The design and
implemen a ion o he da a model has much impac in he da a s o age
and access e iciency and also in he da a disco e y e icacy. A da a
model mus con ain all he equi ed da a and me ada a o implemen
e ec i e da a disco e y. A he same ime, he s uc u es should be
designed bea ing in mind he ele an que ies ha will be implemen ed
h ps://doi.o g/10.1016/j.en so .2024.106248
Recei ed 16 May 2024; Recei ed in e ised o m 10 Oc obe 2024; Accep ed 11 Oc obe 2024
En i onmen al Modelling and So wa e 183 (2025) 106248
A ailable online 18 Oc obe 2024
1364-8152/© 2024 The Au ho s. Published by Else ie L d. This is an open access a icle unde he CC BY-NC-ND license (
h p://c ea i ecommons.o g/licenses/by-
nc-nd/4.0/ ).
D. Ma ínez e al.
in he da a access mechanisms, o enable an adequa e e iciency, e en
when he size o he da ase s inc ease wi h ime.
Concep ual modeling amewo ks such as he Uni ied Modeling
Language (UML) enable he de ini ion o da a s uc u es using classes,
p ope ies o classes and di e en ypes o associa ions be ween classes.
Classes a e used o model se s o en i ies o he applica ion domain
(ins ances o he class). Thus, he ci y o ‘‘San iago de Compos ela’’ may
be an ins ance o a class Ci y. The model ha de ines he mechanisms
a ailable in a modeling amewo k is called a me amodel (model o he
model) (Gonzalez-Pe ez and Hende son-Selle s,2008). Model elemen s
a e ins ances o he me amodel. Thus, o example, class Ci y will be
an ins ance o me aclass Class and a p ope y ‘‘name’’ o class Ci y will
be an ins ance o me aclass P ope y. Ins ances o he model ( he da a)
a e usually eco ded using da a s o age echnologies wi h speci ic da a
s o age models ( o example da abases wi h he ela ional model). On
he o he hand, ins ances o he me amodel ( he me ada a) a e usually
eco ded in ca alogs (implemen ed also wi h da a s o age echnologies),
and a e o key impo ance o da a disco e ing asks. Gene aliza-
ion/specializa ion associa ions be ween classes enable he de ini ion
o hie a chies whe e subclasses inhe i he p ope ies and associa ions
o supe classes. As an example, a supe class Popula ionCen e may be
specialized in wo subclasses, Ci y and Village. Supe classes a e abs ac
when hey may no ha e ins ances. Models con aining only supe classes
a e called abs ac models and may be specialized o a ious speci ic
pu poses, enabling he euse o da a model s uc u es.
Designing a good ad hoc model o each speci ic da ase is cos ly,
howe e i is clea ly he bes app oach o achie e e icien solu ions
in e ms o da a s o age space and que y esponse ime. Cu en ly,
wo b oad da a s o age models a e used o he de elopmen o such
c a model based solu ions o en i onmen al in o ma ion: he Unida a
Common Da a Model (CDM) and da abase models. Unida a CDM and
i s co esponding Ne CDF (Ne wo k Common Da a Fo m) ile o ma 5
a e b oadly used o ep esen emo e sensing obse a ions and en i on-
men al model ou pu s, whose shape has he o m o some g id o spa ial
and empo al dimensions. On he o he hand, da abase models, mos
commonly ei he he ela ional model (Codd,1970) o some ex ension
o i , a e b oadly used o da ase s gene a ed by in-si u obse a ion
in as uc u es. Such a c a model app oach also has some impo an
d awbacks. Fi s , he cos o ha ing o de elop a new model o each
speci ic da ase is high in e ms o human esou ces. Besides, he quali y
o each model depends comple ely on he skills o he designe s. Finally,
designing independen models o di e en applica ions c ea es a le el
o he e ogenei y ha hinde s he in eg a ion o in o ma ion ob ained
om di e en da ase s.
To add ess he limi a ions o ad hoc solu ions, se e al gene ic
models o en i onmen al da a ha e been p oposed in he li e a u e.
De ining a single, uni e sal da a model o all applica ions is imp ac-
ical due o he di e se na u e o en i onmen al da a needs. Howe e ,
models o speci ic applica ion domains ha e been de eloped and ap-
plied success ully (Ho sbu gh e al.,2008;Mason e al.,2014;Abdallah
and Rosenbe g,2019). The Open Geospa ial Conso ium (OGC) has
es ablished he Obse a ions and Measu emen s (O&M) da a model,
which p o ides a amewo k o en i onmen al obse a ions (Cox,
2013). A key s eng h o O&M is i s ex ensibili y, allowing i o be
ailo ed o he speci ic equi emen s o a ious applica ions. Nume ous
p o iles specialized o di e en con ex s ha e been de eloped based
on O&M (Taylo e al.,2013;Wojda and B ouyè e,2013;Ho sbu gh
e al.,2016;Blodge e al.,2021).
O&M has been ex ensi ely used alongside web se ices such as he
OGC Senso Obse a ion Se ice (SOS) o suppo he dissemina ion o
en i onmen al da a (B ö ing e al.,2012). Solu ions like 52◦No h,6
5h ps://doi.o g/10.5065/D6H70CW6
652◦No h Ini ia i e o Geospa ial Open Sou ce So wa e GmbH h ps:
//52no h.o g/so wa e/so wa e-p ojec s/sos/
is SOS,7and PySOS8p o ide implemen a ions o he SOS s anda d;
howe e , hese echnologies o en encoun e pe o mance challenges
when wo king wi h la ge da ase s, pa icula ly ega ding da a e ie al
and que y e iciency (see Sec ion 7.2).
An e o o speci y s anda d ways o s o e en i onmen al da a
in he Unida a CDM is done wi h he speci ica ion o he Clima e
and Fo ecas (CF) con en ions (Ea on e al.,2023). CF also de ines
a s anda d ocabula y o a ious da a model elemen s (p ope ies,
uni s o measu e, e c.). A gene ic da a modeling app oach based on
seman ic web s anda ds, called NGSI-LD, has been p oposed o con ex
senso da a on he in e ne o hings and sma ci y a eas (ETSI,2023).
Nei he CF no NGSI-LD ep esen all he basic concep s ela ed o
en i onmen al obse a ion, which a e de ined in he O&M model. In
gene al, a e iew o exis ing en i onmen al da a modeling solu ions
shows a big challenge in achie ing bo h comple eness in co e ing all
he equi emen s o all applica ions and e iciency in da a s o age and
que y (see Sec ion 7.3).
In his pape , a amewo k o he concep ual modeling o en i-
onmen al da a is p oposed. The amewo k de ines da a s uc u es
o he ep esen a ion o he main concep s ha a ise in he scope o
bo h obse a ion and modeling in as uc u es. Those s uc u es may be
specialized o inco po a e he equi emen s o di e en domains and
speci ic applica ions. Fou le els o abs ac ion a e conside ed in he
amewo k. The i s wo le els p o ide, espec i ely, wi h a gene ic
da a modeling app oach and wi h a gene ic solu ion o en i onmen al
in o ma ion. In he hi d le el o abs ac ion, gene al da a models o
speci ic applica ion domains may be de ined specializing he s uc u es
o he p e ious le el. Models o speci ic applica ions a e de ined a
he ou h le el o abs ac ion, once again eusing he s uc u es om
he p e ious le el. The ex ensibili y o he amewo k enables i s use
in any en i onmen al applica ion domain. In spi e o i s b oad scope,
he e alua ion o he amewo k has shown in gene al a e y good
pe o mance in e ms o que y esponse ime. The p oposed amewo k
e ol es om he TAQE da a modeling amewo k (Ma ínez e al.,
2022) de ined o a ic and ai quali y moni o ing.
Based on he abo e, he main con ibu ions o he pape can be
summa ized as ollows.
•A comple e se o equi emen s o he concep ual modeling
amewo k a e p oposed, ha we e ex ac ed om an ex en-
si e e iew o exis ing solu ions and om he expe ience o he
au ho s in p ojec s.
•An abs ac da a model specialized on en i onmen al applica ions
la gely based on O&M and a me amodel wi h suppo o mul iple
ocabula ies.
•An illus a ion o he use o he amewo k o de ine specialized
abs ac da a models in wo applica ion a eas: (i) clima e science
and (ii) a ic and ai quali y moni o ing in sma ci ies (al eady
conside ed in TAQE). Eigh speci ic use case da ase s o he abo e
a eas we e modeled and implemen ed wi h he amewo k.
•An e icien implemen a ion o he da a s uc u es gene a ed by
he amewo k based on Pos g eSQL, Pos GIS and schema-less ag-
g ega es encoded in JSON da a ypes. A pe o mance e alua ion
shows ha he implemen a ion is in gene al as e icien as good
ad hoc models designed in exis ing o ganiza ions and ou pe o ms
in mos cases a e e ence di ec implemen a ion o he OGC O&M
model used by a SOS ool. Di e en ypes o da ase s we e used
in his e alua ion, including simple da a alues ob ained in s a ic
obse a ion s a ions, ajec o ies gene a ed by mobile pla o ms
and e ical p o iles ob ained a speci ic loca ions in he sea.
7Senso Obse a ion Se ice o Wa e In o ma ion Sys ems h ps://is sos.
o g/
8h ps://gi hub.com/manuGil/py4sos
En i onmen al Modelling and So wa e 183 (2025) 106248
2
D. Ma ínez e al.
•A quali a i e e alua ion o he ul illmen o he equi emen s
posed o he p esen amewo k. O he wel e da a modeling
solu ions we e also e alua ed wi h espec o he same equi e-
men s, showing ha none o hem achie es ully all o hem.
The emainde o his pape is o ganized as ollows. In Sec ion 2,
ela ed wo k and exis ing app oaches a e e iewed in de ail, pay-
ing special a en ion o he OGC O&M da a model and da a s o age
models based on i . The gene al amewo k s uc u e and he se o
equi emen s assumed o i s design and implemen a ion a e desc ibed
in Sec ion 3. The da a models and me amodels de ined o he wo
i s le els o abs ac ion a e desc ibed in Sec ion 4. In Sec ion 5, a
gene ic da a model o clima e science applica ions is p oposed. Two
o he use cases implemen ed wi h he amewo k a e desc ibed in
Sec ion 6. The esul s o he e alua ion o he amewo k a e shown in
Sec ion 7, including an illus a ion o da a in eg a ion among di e en
da ase s, a quan i a i e e alua ion o he que y pe o mance and a
quali a i e e alua ion o he ul illmen o he p oposed equi emen s.
Finally, some conclusions and lines o u he wo k a e depic ed in
Sec ion 8. The pape is comple ed wi h an appendix (Appendix) ha
p o ides addi ional models and use cases ela ed o a ic and ai
quali y moni o ing in sma ci ies.
2. Rela ed wo k
Models a e abs ac ions o eal sys ems ha a e key a i ac s du ing
enginee ing o so wa e p oduc s (B ambilla e al.,2017). Va ious da a
modeling pa adigms a e used a di e en le els o abs ac ion in he
a ea o Da a Managemen . Two main pa adigms a e used a he con-
cep ual le el: (i) Models based on en i ies (objec s) and ela ionships
(associa ions) among hem (Chen,1976;Blaha and Rumbaugh,2005)
and (ii) models based on dimensions and measu able ac s (dimensional
modeling) (Kimball and Ross,2002). A a lowe le el o abs ac ion,
cu en ly, mos applica ions s ill ely on implemen a ions based on
he ela ional model (Codd,1970). Howe e , in some speci ic cases,
non- ela ional (Sadalage and Fowle ,2013) pa adigms p o ide good
pe o mance, by suppo ing complex nes ed da a ypes (agg ega es) and
la ge scale dis ibu ed a chi ec u es. All hose non- ela ional models
ely also in he lack o p ede ined schema, which b ings ad an ages
a da a inse ion and disad an ages a da a que ying. In pa icula , as
applica ions canno que y he da abase ca alog o ge he schema, he
schema mus be encoded in he applica ions code, which is no he
bes place o be. Fu he mo e, he da abase canno use he in o ma ion
o he schema o pe o m que y op imiza ion (Sadalage and Fowle ,
2013).
Ex ensions o suppo bo h complex da a ypes and dis ibu ed
a chi ec u es a e cu en ly a ailable o ela ional DBMSs, enabling
hei use wi h di e en da a models and con igu a ions. An example
is he Pos g eSQL DBMS9and i s Ci us Da a ex ension10 o dis ibu ed
da abases. Howe e , he managemen o scien i ic a ays is s ill nowa-
days no e icien in DBMSs, hus, ele an applica ions ha e o use
ei he speci ic ile o ma s (De ys e al.,2019;The HDF G oup,2024)
o speci ic a ay DBMSs (Baumann,1994;B own,2010). A inal gene al
pu pose da a model pa adigm is he Resou ce Desc ip ion F amewo k
(RDF) (Cyganiak e al.,2014) used in he seman ic web and linked da a
scope.
Concep ual objec -based and dimensional models a e also used in
he a ea o geospa ial da a managemen (Rigaux e al.,2001;Viquei a
e al.,2005). En i ies ypes (objec s classes) o objec -based models a e
called Fea u e Types (Ko man and Reed,2009), and hey may include
geome ic p ope ies (He ing,2020) o ep esen hei loca ion and
shape on he Ea h su ace. Measu es ha change o e geospa ial and
empo al dimensions a e modeled using collec ions o mappings called
9h ps://www.pos g esql.o g/
10 h ps://www.ci usda a.com/
Co e ages (OGC,2007). Fea u e Types and Co e ages wi h spa se
spa ial domains a e e icien ly managed wi h ei he ela ional o non-
ela ional app oaches. On he o he hand, dense co e ages a e usually
ep esen ed wi h a ays o spa io- empo al dimensions, whose size
migh be e y la ge. Thei e icien managemen equi es he e o e he
a ay speci ic solu ions men ioned abo e. Few wo ks ha e a emp ed
he uni o m managemen o Fea u es and Co e ages (Villa oya e al.,
2016;de Bakke e al.,2017), and hus, ele an ma u e and e icien
implemen a ions ha e no been eached ye .
An impo an miles one owa ds he de ini ion o a gene al concep-
ual da a model o en i onmen al obse a ion da a was he p oposal
o he Obse a ions and Measu emen s (O&M) s anda d da a model
by he OGC (Cox,2013). This concep ual model de ines he main
concep s in ol ed in he ep esen a ion o en i onmen al obse a ions
(see Sec ion 2.1 o mo e de ails). The O&M da a model is a key pa
o s anda d in e aces de ined by he OGC o access obse a ion da a
h ough he web, such as he Senso Obse a ion Se ice (SOS) (B ö ing
e al.,2012) and he sensing pa o he Senso Things API (Liang
e al.,2021). Va ious ools o en i onmen al da a wa ehouses ha
implemen he OGC SOS da a access in e ace a e cu en ly a ailable:
52◦No h SOS,11 is SOS12 and PySOS13 a e ep esen a i e examples (see
Sec ion 2.1 o mo e de ails). O&M is also a he co e o he Seman ic
Senso Ne wo k (SSN) on ology (Halle e al.,2017;Comp on e al.,
2012) p oposed by bo h OGC and he Wo ld Wide Web Conso ium
(W3C). SSN was used a he co e o a gene ic model o he medi-
a o componen o a seman ic in eg a ion ede a ed a chi ec u e o
en i onmen al obse a ion da ase s (Reguei o e al.,2017).
O e he pas decade, he hyd ological esea ch communi y has
made signi ican p og ess in da a modeling. Rep esen a i e ou pu s o
such e o a e he Hyd osha e in as uc u e (Ta bo on e al.,2024)
and he Wa e ML p o ile (Taylo e al.,2013) o he O&M da a model.
One o he i s solu ions p oposed was he Obse a ions Da a Model
(ODM) (Ho sbu gh e al.,2008), designed o he s o age o he da a
and me ada a o in-si u obse a ions o moni o ing si es in a ela-
ional da abase. The objec -o ien ed H𝑔
2Omodel (Wojda and B ouyè e,
2013) o g ound wa e da a specializes he O&M model o include
speci ic ea u e ypes o ea u es o in e es and sampling ea u es.
I includes also new s uc u es o simple and complex obse a ion
ypes. The VOEIS Da a Model (VODM) (Mason e al.,2014) ex ends
ODM wi h da a s eams, da ase s, use s, oles, membe ships, me a ags
and si e da a ca alogs. The ela ional model p oposed in Kim e al.
(2015) is based on a geoda abase and enables he eco ding o bo h
i e obse a ions and simula ions. ODM2 (Ho sbu gh e al.,2016)
may be conside ed a p o ile o O&M and i enables he modeling o
disc e e Ea h obse a ions, i.e., hose ha eco d a single alue o
he whole FOI (i does no suppo co e ages as obse a ion esul s).
The ela ional Wa e Managemen Da a Model (WaMDaM) (Abdallah
and Rosenbe g,2019) es ic s also o disc e e obse a ions and i
was designed bea ing in mind he ollowing p inciples: modula i y
and ex ensibili y, inco po a ion o ne wo ks o nodes and links as
ea u es, suppo o scena ios and e sion con ol, eusable con ex
me ada a, suppo o mul iple da a ypes, seman ics speci ied wi h
con olled ex ensible ocabula ies, di ec access suppo o subse s o
da a and me ada a and open-sou ce so wa e en i onmen . The main-
s ems (Blodge e al.,2021) da a model desc ibe hyd ologic ne wo ks
using ea u e ypes p oposed in pa 3 o Wa e ML. Finally, he ame-
wo k p oposed in Salas e al. (2020) o open da a and open modeling
uses agen s o in eg a e da a ob ained om di e en models.
Rega ding oceanog aphic da a, in e na ional hubs such as he U.S.
In eg a ed Ocean Obse ing Sys em (IOOS),14 he Eu opean Ma ine Ob-
se a ion and Da a Ne wo k (EDMODne )15 and he Cope nicus Ma ine
11 h ps://52no h.o g/so wa e/so wa e-p ojec s/sos/
12 h ps://is sos.o g/
13 h ps://sou ce o ge.ne /p/pysos/
14 h ps://ioos.noaa.go /
15 h ps://emodne .ec.eu opa.eu/
En i onmen al Modelling and So wa e 183 (2025) 106248
3
D. Ma ínez e al.
Se ice (CMEMS),16 ely mainly in he Unida a Common Da a Model
(CDM), implemen ed wi h Ne CDF iles o ep esen and s o e bo h in-
si u and emo e-sensed obse a ions and model esul s. The Clima e
and Fo ecas (CF) con en ion (Ea on e al.,2023) is used o speci y how
Ne CDF is used o ep esen speci ic da ase ypes and also o p o ide
a s anda d ocabula y o obse ed and modeled p ope ies.
Unida a CDM and he Ne CDF s anda d is also he keys one o
he ep esen a ion o mos da ase ypes o obse a ion and modeling
in clima e science and me eo ology. An in e es ing ecen con ibu-
ion in his a ea is he axonomy o ea u es o inpu s and ou pu s
o nume ical models p oposed in Ha pham (2020). The axonomy is
based on s anda d spa ial geome y ypes (Poin , Mul ipoin , Polyline,
e c.), whe e spa ial g ids and meshes a e ea ed as specializa ions o
Mul ipolygons. Tempo al a ia ions enable he modeling o da a whose
spa ial inge p in is de ined by a single geome y, bu i a ies o e
ime. Spa io- empo al a ia ions suppo he changing in bo h alue
and spa ial inge p in a each ime ins an , i.e., i enables he modeling
o acks o di e en ypes (poin ack, polyline ack, e c.).
En i onmen al da a is e y impo an in he scope o sma ci ies
and in e ne o hings. In hese a eas, he Eu opean Telecommunica-
ions S anda diza ion Ins i u e (ETSI) has eleased a sui e o speci ica-
ions called NGSI (Nex Gene a ion Se ice In e aces), which includes
he NGSI-LD con ex in o ma ion model (ETSI,2023). The model is
based on he W3C seman ic web s anda ds and i includes h ee laye s:
(i) a Me a Model de ined on op o RDF/RDFS concep s, (ii) a C oss
Domain On ology ha inco po a es empo al and geog aphical p op-
e ies and (iii) Domain Speci ic On ologies ha es ic s he p e ious
laye o a speci ic applica ion domain and de ines his way he speci ic
s uc u e and ocabula y. W3C SSN (Halle e al.,2017;Comp on e al.,
2012) could be inco po a ed as a new laye be ween he C oss Domain
On ology and Domain Speci ic On ologies ela ed o senso obse a ion.
Few e o s may be iden i ied o achie e a uni o m da a model o
ai quali y and a ic da a in he con ex o sma ci y in as uc u e
de elopmen . The on ology p oposed in Op ea (2009) was designed o
model ai pollu an concen a ions a moni o ing s a ions, oge he wi h
hei po en ial pollu ion sou ces. QBOAi base (Galá aga e al.,2017)
was buil as a educed e sion o he Ai base da ase p o ided by he
Eu opean En i onmen al Agency (EEA), which gi es access o pollu an
concen a ions egis e ed a en i onmen al s a ions h ough Eu ope.
QBOAi base has a dimensional model consis ing o h ee dimensions
(yea , s a ion and senso ) and measu emen s o di e en pollu an s.
The da abase is linked o o he seman ic web RDF da a sou ces and i
inco po a es da a p o enance in o ma ion by le e aging he use o he
P o -O on ology (Lebo e al.,2013). In Fe nandez e al. (2016), he au-
ho s de ine an on ology o suppo in elligen anspo a ion sys ems,
which inco po a es mechanisms o model ehicles, oad in as uc u e
and senso s o a ic moni o ing. Ai quali y p edic ion is an ac i e
esea ch opic in he en i onmen al modeling a ea (Johansson e al.,
2022;Pisoni e al.,2024). Howe e , speci ic da a modeling solu ions
ha e no been p oposed, o he bes o hese au ho s knowledge.
The TAQE da a modeling amewo k (Ma ínez e al.,2022) consid-
e s ou le els o abs ac ion o de ine da a models o a ic and ai
quali y a bo h local and egional scales. Da a abs ac ion le els ange
om he comple ely gene al pu pose le el 1 o he comple ely speci ic
le el 4 o applica ions. Le el 2 es ic s o en i onmen al applica ions
and i is based on OGC O&M. Le el 3 p o ides gene ic models o some
applica ion domains (ai quali y and a ic in his case). The solu ion
p oposed in he p esen pape e ol es om ha o TAQE, he e o e,
bo h p oposals a e comple ely compa ible. The main di e ence is ha
TAQE was designed o a ic and ai quali y whe eas he p esen
model is mo e gene al. The le el 3 model o TAQE includes emo e
sensing da a sou ces, no conside ed in he ele an le el 3 model
shown in his pape , due o he ocus in sma ci ies, i.e., only in he
16 h ps://ma ine.cope nicus.eu/
local scale o which emo e sensing is no p o ided je su icien eso-
lu ion. Main con ibu ions o he p esen wo k wi h espec o TAQE a e
he ollowing: (i) The O&M model conside ed in TAQE is now ex ended
o p o ide uni o m ep esen a ions o obse a ion subsamples and ime
e ol ing p ocess p ope ies. (ii) A me amodel based on he p e ious
model is de ined ha suppo s he de ini ion o da a ypes and he
inco po a ion o mul iple ocabula ies in he ca alog, (iii) a da a model
o le el 3 o clima e science was p o ided o illus a ion pu poses,
(iii) use cases wi h me eo ological and oceanog aphic da a o common
use we e also es ed, (i ) an e icien implemen a ion wi h agg ega e
s uc u es based on Pos g eSQL, Pos GIS and JSON was p o ided and
( ) a de ailed e alua ion, which is bo h quali a i e compa ing o o he
solu ions and o que y pe o mance was unde aken.
2.1. En i onmen al da a s o age implemen a ions based on O&M
The main concep s conside ed by he OGC Obse a ions and Mea-
su emen s (O&M) concep ual model (Cox,2013) a e shown in he UML
class diag am o Fig. 1(a). Ins ances o class GFI_Fea u e a e used o
model obse ed ea u es, whose Fea u e Types a e ins ances o me aclass
GF-Fea u eType. An example o an obse ed ea u e is he Spanish
egion o Galicia, and i s ea u e ype migh be ‘‘Region’’. P ope ies
o obse ed ea u e ypes, such as empe a u e and ain all a e ep e-
sen ed by me aclass GF_P ope yType. In many cases, ea u es may no
be di ec ly obse ed, a leas in hei whole ex en , and some kind
o sampling has o be pe o med. Class SF_SamplingFea u e p o ides
suppo o he ep esen a ion o he di e en ypes o ea u es ele-
an o hose samplings. In pa icula , class SF_Spa ialSamplingFea u e
suppo s he ep esen a ion o samplings o he spa ial ex ension o
he obse ed ea u e. An example is he collec ion o loca ions o
a ne wo k o me eo ological s a ions ha sample a speci ic egion.
Class SF_Specimen enables he ep esen a ion o specimens cap u ed o
obse e h ough hem he ul ima e obse ed ea u e. An example o a
specimen is a bo le o wa e ob ained om a speci ic loca ion in a i e ,
ha is analyzed in a labo a o y o ge alues o obse ed p ope ies
o he i e . Class OM_P ocess is included o inco po a e me ada a o
he p ocesses used o gene a e obse a ions. Finally, class SF_P ocess
enables he desc ip ion o he me hod used o cap u e he ins ances o
SF_Specimen.
An obse a ion (OM_Obse a ion) eco ds he alue gene a ed o
he obse ed p ope y ( esul ) and i e e ences i s obse ed ea u e
(Fea u eO In e es ), i s obse ed p ope y and he p ocess used o gene -
a e i (p ocedu e). Manda o y empo al da a o he obse a ion consis s
o wo elemen s: (i) he phenomenonTime, which eco ds he ime ins an
o ime pe iod du ing which he esul applies o he p ope y o he
obse ed ea u e and (ii) he esul Time ha eco ds he ins an when
he p ocedu e gene a ed he obse a ion. Op ionally, an obse a ion
may eco d addi ional pa ame e s, esul quali y me ada a, and o he
ypes o gene al me ada a. The esul o an obse a ion may be e y
simple, such as an in ege alue ha ep esen s some coun . Howe e ,
complex esul s a e also possible, such as ime se ies o eco ds and
di e en ypes o geospa ial co e ages.
An unusual cha ac e is ic o he O&M concep ual model is ha
i combines elemen s a bo h model and me amodel le els. A me a-
model is a model whose ins ances a e speci ic models (Gonzalez-Pe ez
and Hende son-Selle s,2008). Ins ances o he me amodel a e usually
eco ded in me ada a ca alogs. Thus, classes GFI_Fea u e,OM_P ocess
and OM_Obse a ion a e ins ances o me aclass GF_Fea u eType. In
Fig. 1(a) i is shown how he O&M model de ines an associa ion
be ween class OM_Obse a ion and me aclass GF_P ope yType. This un-
usual cha ac e is ic will lead o po en ial pe o mance p oblems when
speci ic models based on O&M a e implemen ed.
Fig. 1(b) shows he main da a s o age s uc u es o a ep esen a i e
implemen a ion o he O&M da a model de eloped o suppo a gene al
pu pose senso obse a ion se e wi h a Senso Obse a ion Se ice
En i onmen al Modelling and So wa e 183 (2025) 106248
4
D. Ma ínez e al.
Fig. 1. Illus a ion o a gene al pu pose da a s o age model based on OGC O&M.
(SOS) (B ö ing e al.,2012) da a access in e ace. Such an implemen-
a ion model enables he s o age o en i onmen al obse a ion da a
in any applica ion a ea wi h he same se o da a s o age s uc u es,
which eases da a in eg a ion. Classes ea u e,p ocedu e and phenomenon
a e used o ep esen , espec i ely, obse ed ea u es, da a gene a ion
p ocesses and obse ed p ope ies. Pa ame e enables he eco ding o
ea u e p ope ies ha a e no obse ed. Class da ase is inco po a ed
in he model o p o ide suppo o complex obse a ions, such as ime
se ies, e ical p o iles and ajec o ies, using la (non-nes ed) da a
s uc u es. Obse a ions a e eco ded in models like he one o Fig. 1(b)
in SOS ools like 52◦No h SOS, is SOS and PySOS.
A main p oblem o his ype o gene ic da a s o age solu ions
is ha da a que y o e hose s uc u es does no o e he equi ed
pe o mance in many cases, as i will be shown in Sec ion 7.2. Two
main cha ac e is ics o he model a e behind hose pe o mance issues:
(i) The eco ding in he model o me amodel elemen s. In pa icula ,
obse ed and non obse ed p ope ies and hei ela ionships wi h
obse ed ea u es and obse a ions a e eco ded as da a i ems and no
a pa o he schema in he sys em ca alog and (ii) he use o la
non-nes ed s uc u es o ep esen componen s o complex obse a ion
esul s ha a e usually e ie ed as a whole. The amewo k p oposed
in his pape a oids he wo abo e cha ac e is ics while p o iding a
gene al solu ion o en i onmen al da a s o age.
3. Requi emen s and gene al amewo k s uc u e
The amewo k p oposed in his pape inco po a es da a models a
ou le els o abs ac ion.
Le el 1: The UML objec -o ien ed me amodel.
Le e 2: A me amodel and an ele an abs ac da a model o geospa-
ial and en i onmen al applica ions.
Table 1
Gene ic Requi emen s.
Requi emen Desc ip ion
GR01 The amewo k mus be lexible o be easily adap able o
di e en applica ions and applica ion domains, enabling he
de ini ion o applica ion speci ic s uc u es and he euse o
common da a modeling s uc u es among hem.
GR02 The amewo k mus suppo he inco po a ion o con olled
ocabula ies, such as he S anda d Names o he CF
con en ion (Ea on e al.,2023) and he compliance wi h
ele an con en speci ic s anda ds such as O&M and Wa e ML
(Taylo e al.,2013).
GR03 The amewo k mus p o ide models o di e en le els o
abs ac ion o suppo he in eg a ion o da a sou ces
gene a ed in di e en applica ions. Each model combines
gene al elemen s inhe i ed om he model o he p e ious
le el wi h specialized s uc u es designed o he applica ion
a ea objec i e o he cu en le el.
GR04 The amewo k mus suppo he c ea ion o da a models ha
may be e icien ly implemen ed, bo h in e ms o da a s o age
size and que y esponse ime, allowing he use o di e en
scalable da a managemen echnologies in such
implemen a ion.
Le el 3: Abs ac da a models specialized in di e en en i onmen al
applica ion a eas.
Le el 4: Da a models o speci ic applica ions inside each applica ion
a ea.
The me hodology ollowed o design he abo e le els o he ame-
wo k is desc ibed now. Fi s , a wide collec ion o equi emen s was
ex ac ed om he e iew o ela ed app oaches in he li e a u e and
om he expe ience o he au ho s in a ious p ojec s. A discussion
on he e alua ion o he ul illmen o such equi emen s by p oposed
amewo k and by o he app oaches is p o ided in Sec ion 7.3.
En i onmen al Modelling and So wa e 183 (2025) 106248
5

D. Ma ínez e al.
Table 2
Le el 1 equi emen s: Gene al pu pose da a modeling unc ionali y.
Requi emen Desc ip ion
L1R01 The model mus suppo he ep esen a ion o objec s (en i ies)
and hei p ope ies.
L1R02 The model mus suppo an ex ensible da a ype sys em o
he alues o he objec p ope ies. P imi i e da a ypes
(in ege , s ing, e c.) mus be di ec ly suppo ed and hey
migh be ex ended by use de ined da a ypes, including
enume a ion da a ypes and da a ypes wi h complex s uc u e.
L1R03 The model mus suppo he ep esen a ion o associa ions
( ela ionships) be ween objec s.
L1R04 The model mus suppo he classi ica ion o objec s wi h he
same p ope ies and ela ionships in o classes (en i y ypes).
Table 3
Le el 2 equi emen s: Geospa ial and En i onmen al da a modeling unc ionali y.
Requi emen Desc ip ion
L2R01 The model mus p o ide da a ypes o geome ic objec s,
in e als o eal alues and empo al pe iods.
L2R02 The model mus p o ide s uc u es o he ep esen a ion o
Co e ages wi h geospa ial, empo al and e ical dimensions.
Poin s, lines and su aces mus be suppo ed in he geospa ial
dimensions, enabling mul ipoin , mul icu e and mul isu ace
co e ages.
L2R03 The model mus p o ide da a ypes o ep esen measu es and
ca ego ies. A measu e combines a eal alue wi h a uni o
measu e. A ca ego y combines a keywo d wi h a e e ence o
a ocabula y o possible alues.
L2R04 The model mus suppo he ep esen a ion o da a gene a ion
p ocesses (including obse a ion p ocesses and modeling
p ocesses). The ep esen a ion o he e olu ion wi h espec o
ime o he p ocess p ope ies mus be suppo ed, o be able o
iden i y he condi ions ha applied du ing he gene a ion o a
speci ic esul .
L2R05 The model mus suppo he ep esen a ion o objec s
gene a ed du ing he spa ial sampling o he ea u es o
in e es . Examples o spa ial samples a e poin s, p o iles,
ajec o ies, scenes, swa hs, e c.
L2R06 The model mus suppo he ep esen a ion o specimens
collec ed o sample he ea u es o in e es . The p ocesses used
o cap u e hose specimens should also be ep esen ed by he
model.
L2R07 The model mus ep esen he empo al con ex o he
gene a ed da a, including he esul and phenomenon ime
conside ed in he OGC O&M da a model. Resul ime
ep esen s he ins an when he p ocess gene a ed he da a.
Phenomenon ime is ela ed o he ime when he gene a ed
da a applies o he ea u e o in e es .
L2R08 The model mus suppo simple da a alues and also complex
alues esul ing om spa ial, e ical and empo al
subsampling. Complex alues include ime se ies, e ical
p o iles, ajec o ies and co e ages.
The gene ic equi emen s ha guide he o e all design o he ame-
wo k a e shown in Table 1. B oadly, acco ding o hose equi emen s
he sys em mus suppo he in eg a ion o di e en applica ions, bu
adap ing o hei singula i ies, achie ing a he same ime good pe o -
mance in da a s o age and que ying. Da a in eg a ion is achie ed by he
euse o common abs ac s uc u es ha a e specialized o di e en
applica ions and by sha ing common and s anda d ocabula ies o
p ope y names. Da a in eg a ion is illus a ed in Sec ion 7.1. Que y
pe o mance is e alua ed in Sec ion 7.2 wi h espec o ad-hoc models
and o he gene ic da a s o age solu ions desc ibed in Sec ion 2.1.
Requi emen s ha a e independen o any applica ion (see Table 2)
we e conside ed o design Le el 1 o he amewo k. Thus, he UML
me amodel was chosen a his le el, inco po a ing objec s, classes,
p ope ies and associa ions, including gene aliza ion/specializa ion as-
socia ions and composi ion. UML composi ion, ep esen ed wi h a black
diamond, is used o ep esen a s ong o m o ‘‘has a’’ ela ionship,
which is ma e ialized in he implemen a ion wi h dependencies o
exis ence and iden i ica ion om he pa s o hei whole (as in weak
en i y ypes in he En i y-Rela ionships model). The e o e, ca dinali y
in he side o class wi h he ole o ‘‘whole’’ is always 1..1 and i is no
ep esen ed in he diag ams.
Nex , he equi emen s ha a e speci ic o geospa ial and en i-
onmen al applica ions, bu a e a he same ime independen o any
o hose applica ions we e conside ed o design a second le el o
abs ac ion. Le el 2 equi emen s a e shown in Table 3. To design his
le el, OGC and ISO s anda ds we e used as a baseline. In pa icula ,
his le el is la gely based on OGC O&M (Cox,2013). Le el 2 consis o
a me amodel ha de ines me ada a ca alog s uc u es and an abs ac
model ha will be specialized in subsequen le els.
A le el 3, a ious abs ac da a models may be de ined o inco po-
a e da a s uc u es ha a e common in di e en applica ion a eas. This
le el is illus a ed in he p esen pape by p o iding a model ha migh
be eused by may di e en clima e science applica ions (See Sec ion 5).
O he examples o le el 3 abs ac models o a ic and ai quali y
moni o ing a local scale in he con ex o sma ci ies a e shown in
Appendix. In gene al, communi ies o expe s o di e en applica ion
a eas may each ag eemen s in he o m o models o le el 3, which
will be eused in many applica ions o hose a eas. The exis ence o
hose models will ease he in eg a ion o he da a gene a ed by di e en
applica ions in hose a eas.
Finally, da a models ha de ine he da a s uc u es used o eco d
da ase s in speci ic applica ions a e inco po a ed a le el 4. These
speci ic models specialize and euse abs ac da a s uc u es om le el
3.
De ining da a models ha a e specialized a a ious le els is an
app oach al eady ollowed by OGC. Thus o example, he Wa e ML
model specializes he mo e gene al O&M da a model o applica ions
in he a ea o hyd ology. Use s may use di ec ly da a encodings based
on O&M and Wa e ML o ep esen hei da a. Howe e , use s migh
also specialize u he Wa e ML o speci ic applica ions. I is a gued
by he au ho s o his pape ha once he use s a e amilia wi h he
concep s o models om uppe le els, he o mula ion o new models
o hei applica ions is simpli ied by eusing elemen s o hose models.
Addi ionally, in gene al, he quali y o he designs ge s imp o ed and
he in eg a ion o da a among applica ions is also acili a ed.
4. Le el2: Abs ac model and me amodel o en i onmen al da a
The abs ac da a model ha con ains he gene ic da a s uc u es o
en i onmen al da a ep esen a ion is desc ibed below in Sec ion 4.1.
Sec ion 4.2 ou lines he main cha ac e is ics o he cu en implemen-
a ion o he da a s uc u es. Finally, he s uc u es o he me ada a
ca alog p o ided a le el2 o he amewo k a e shown in Sec ion 4.3.
4.1. Abs ac da a model
The abs ac da a model o le el 2 is shown in Fig. 2(a). I is no iced
ha his model is e y simila o he O&M model al eady desc ibed
in Sec ion 2.1 (see Fig. 1(a)). Classes Fea u e,P ocess,Obse a ion,
SamplingFea u e,Spa ialSamplingFea u e,Specimen and SamplingP ocess
a e inco po a ed o suppo ele an concep s o O&M. The da a ypes
p o ided a he op o he igu e a e also suppo ed by O&M and enable
he ep esen a ion o di e en ypes o obse a ion esul s, including
g id and disc e e co e ages, ca ego ies ( e ms a ailable in ocabula ies)
and measu es ( eal alues wi h a uni o measu e). The main di e ences
be ween his abs ac da a model and O&M a e esumed as ollows.
•Obse ed p ope ies a e ep esen ed in O&M wi h associa ions
be ween he obse a ion and a me aclass o he me amodel. In
he p esen amewo k, p ope ies (ei he obse ed o no ) a e
ep esen ed a le el 2 only in he me amodel and hus a ailable
only in he me ada a ca alog (see Sec ion 4.3). Values o obse ed
p ope ies will be eco ded a le el 4 ei he as alues o p ope ies
o speci ic subclasses o Obse a ion, o as alues o p ope ies
o speci ic subclasses o Obse a ionSubsample, o as bands o
co e ages e e enced in ins ances o some subclass o Obse a ion
.
En i onmen al Modelling and So wa e 183 (2025) 106248
6
D. Ma ínez e al.
Fig. 2. Le el 2 abs ac en i onmen al da a model and me ada a ca alog me amodel.
•The cu en e sion o he amewo k does no inco po a e quali y
me a ada a, which is op ional in O&M.
•O&M enables he inco po a ion o a lis pa ame e alues in
obse a ions. In mos cases, hose pa ame e s a e used o eco d
e ical coo dina es and spa ial and empo al p ope ies o com-
ponen s o complex obse a ions. The p esen amewo k does no
p o ide suppo o gene ic pa ame e s, bu speci ic s uc u es
o empo al and geospa ial subsamples o he obse a ion a e
suppo ed by class Obse a ionSubsample, which is no pa o
O&M.
•Ve ical coo dina es a e ep esen ed in he p esen amewo k
sepa a ely om geome ic objec s. Suppo ing heigh coo dina es
sepa a ed om geome ic objec s eases he in eg a ion wi h ools
o he a ea o Geog aphic In o ma ion Sys ems (GIS). In some
cases, i is needed o p o ide speci ic p ope ies o di e en
heigh s o he same geome ic objec (one example is a e i-
cal p o ile). In hese cases, suppo ing he heigh as a sepa a e
dimension helps in achie ing mo e e icien solu ions.
•Da a ypes Ca ego y and Measu e a e pa ame ic in he p esen
amewo k. Thus, i he ocabula y o he uni o measu e a e
speci ied du ing he decla a ion o he da a ype, hen hei alues
a e eco ded only in he me ada a ca alog. O he wise, hey ha e
o be eco ded wi h each obse a ion esul , as i is he case in
O&M.
•Classes P ocess and SamplingP ocess inco po a e in he p esen
model p ope ies o ep esen pe iods o alid ime ( alidTimeS a
and alidTimeEnd). The eason is ha hose classes use o ha e a
highly dynamic na u e, equi ing he eco ding o e olu ion wi h
espec o ime o de ice con igu a ions and/o model hype pa-
ame e s.
En i onmen al Modelling and So wa e 183 (2025) 106248
7
D. Ma ínez e al.
4.2. Da a model implemen a ion
The implemen a ion o he cu en p o o ype o he amewo k is
based on he Pos g eSQL17 da abase managemen sys em. Pos g eSQL
implemen s a ela ional model ex ensible wi h complex s uc u es (ag-
g ega es). Those agg ega es may be inco po a ed in he sys em in
a ious di e en ways, including a combina ion o a ays and use
de ined ypes, XML and JSON. The suppo o XML and JSON s uc u es
ans o ms he unde lying ela ional model o a hyb id SQL and NoSQL
documen -based model.
Geospa ial da a ype implemen a ion elies on he OGC and ISO
s anda d da a ypes (He ing,2020) p o ided by he Pos GIS ex en-
sion.18 Apa om he geome ic objec s, o he complex alues mus
be ep esen ed, including heigh in e als, ime pe iods, co e ages,
ca ego ies, measu es and subsamples. JSON s uc u es a e used o
ep esen all hose nes ed complex s uc u es.
Co e age da a ypes con ain e e ences o ou -o -band ep esen a-
ions o hose co e ages. In he cu en p o o ype o he amewo k
implemen a ion, g id co e ages a e s o ed in GeoTIFF iles (De ys e al.,
2019). O he ypes o geospa ial co e ages (mul ipoin , mul icu e and
mul isu ace) a e eco ded wi h ables in GeoPackage o ma (Yu zle ,
2024). Tempo al and heigh dimensions o co e ages a e ep esen ed
in nes ed JSON s uc u es unde obse a ion subsamples.
4.3. Me ada a ca alog s uc u es
Ca alog s uc u es suppo he eco ding o me ada a o each o he
subclasses o he models de ined a le el 4. The schema o hose s uc-
u es is shown in he UML class diag am o Fig. 2(b). Class Fea u eType
eco ds me ada a o any ea u e ype, including ea u es o in e es , p o-
cesses, sampling p ocesses and sampling ea u es. Fo each ea u e ype,
he ca alog eco ds i s name (which iden i ies he ea u e ype), he se
o p ope ies, he se o e e ences o o he ea u e ypes, an op ional
se o names ha deno e he ea u e ype in di e en ocabula ies,
and he se o supe ype names. Supe ypes a e classes o some da a
model o Le el 3. Each ea u e ype o each speci ic applica ion (Le el
4), may be a subclass o one o mo e classes o Le el 3. Tagging he
ea u e ype o le el 4 in he ca alog wi h he names o he supe classes
p o ides gene al seman ics ha help in in e p e ing he seman ics o i s
ins ances, easing he implemen a ion o da a in eg a ion applica ions,
as i is illus a ed in Sec ion 7.1. Examples o hese me ada a a e gi en
in Sec ion 6 o speci ic use cases.
Each P ope y o each Fea u e Type is iden i ied by a name. Besides,
he ca alog eco ds he name o i s da a ype (da a_ ype), an op ional
se o names ha deno e he p ope y in di e en ocabula ies and a
boolean lag, called epea ed, ha poin s ou whe he he p ope y has
ei he jus one alue o a ious alues.
The da a ype may be ei he a p imi i e one di ec ly suppo ed by
he sys em o a use de ined da a ype. Two classes o da a ypes may
be de ined by use s. An Enume a ionType de ines a se o possible ex
alues o a p ope y. On he o he hand, aComplexType de ines an
s uc u e con aining ields, each o hem again wi h a name and a da a
ype, which migh be p imi i e o use de ined. The names o he da a
ypes and ields ob ained om ocabula ies may also be eco ded in
he ca alog.
ARe e ence ep esen s a link be ween a sou ce ea u e o a speci ic
ea u e ype o one o mo e des ina ion ea u es o ano he ea u e ype.
I enables he implemen a ion o bina y associa ions be ween ea u e
ypes.
The me ada a o each p ocess ype is eco ded in subclass P o-
cessType o class Fea u eType. In addi ion o he me ada a al eady
desc ibed o ea u e ypes, o each p ocess ype, he ca alog eco ds:
17 h ps://www.pos g esql.o g/
18 h ps://pos gis.ne /
Table 4
Le el 3 equi emen : Clima e obse a ion and modeling.
Requi emen Desc ip ion
L3CS01 The da a gene a ed by obse a ion in as uc u es ins alled a
speci ic loca ions, using ei he s a ic o emo able de ices, and
ep esen ing he heigh o he obse ed loca ion whene e
necessa y.
L3CS02 The da a gene a ed along e ical p o iles, e ical sec ions,
linea ansec s and 3D ajec o ies, ei he a he ocean o a
he a mosphe e.
L3CS03 The da a gene a ed by emo e sensing de ices, including
obse a ions esul s o each poin o a ei he egula o
i egula g id. The da a gene a ion p ocess migh gene a e
snapsho s a speci ic o p ede ined scenes o da a gus s
ollowing speci ic o p ede ined swa hs.
L3CS04 The da a gene a ed by oceanog aphic and me eo ological
models, including nowcas , o ecas and eanalysis sys ems.
L3CS05 The da a gene a ed by he indi ec obse a ion o
en i onmen al p ope ies h ough he di ec obse a ion o
cap u ed specimens.
(i) he name o he da a s uc u e o he model ha eco ds he
obse a ions gene a ed by he p ocess ype (obse a ionType), (ii) he
name o he ea u e ype ha eco ds he pla o m whe e he p ocess
is ins alled, (iii) he name o he ea u e ype ha eco ds he FOIs
o he gene a ed obse a ions and (i ) he se o p ope ies ha a e
obse ed by he p ocess. I is eminded ha , in spi e o he use o he
e m ‘‘obse a ion’’, p ocess ypes may be used o ep esen obse a ion
and modeling p ocesses.
Class SamplingFea u eType eco ds me ada a o each sampling ea-
u e ype, including he name o he sampled ea u e ype ( he one
eco ding he inal FOI ha is being sampled). Two ypes o sampling
ea u e ypes a e suppo ed, aSpa ialSamplingFea u eType o suppo he
spa ial sampling o he inal FOI and aSpecimenFea u eType o suppo
he sampling hough specimen cap u e. The names o he coo dina e
e e ence sys ems used o geospa ial and e ical coo dina es a e
eco ded as me ada a o he spa ial sampling ea u e ype. Rega ding
specimen ea u e ypes, he ca alog eco ds he name o he spa ial
sampling ea u e ype ha eco ds he loca ion whe e he specimen
was cap u ed. Besides, class SamplingP ocessType eco ds me ada a o
he p ocesses used o cap u e he specimens.
5. Le el 3 da a model o clima e obse a ion and modeling
The clima e science applica ion domain was chosen o illus a e and
e alua e he amewo k due o he wide a ie y o di e en obse a ion
and modeling in as uc u es a ailable in his a ea. The equi emen s
conside ed o ou pu poses a e shown in Table 4.
The le el 3 da a model o clima e obse a ion and modeling appli-
ca ions is shown in he UML class diag am o Fig. 3. P ocesses o in-si u
obse a ion a speci ic loca ions a e modeled as ins ances o subclasses
o CSInsi uS a icP ocess. The obse ed loca ions (usually loca ions o en-
i onmen al s a ions) a e modeled wi h ei he class CSSamplingLoca ion
o class CSSamplingLoca ionHeigh , depending on whe he he e ical
o se has o be eco ded o no . Requi emen L3CS01 is hus suppo ed
by he p e ious classes.
Mo e complex ypes o in-si u obse a ions a e also suppo ed
by he model, o ul ill equi emen L3CS02. In pa icula , classes
a e included o model p ocess ha gene a e e ical p o iles, e i-
cal sec ions, linea ansec s and ajec o ies. The spa ial sampling
ea u e ype o a e ical p o ile CSP o ile eco ds he geosap ial lo-
ca ion and he e ical in e al. The obse a ion esul is complex
(CSP o ileObse a ion), and i has alues o he obse ed p ope ies
a each e ical o se (CSP o ileObse a ionSubsample). The shape o
a e ical sec ion is de ined by a lines ing and a e ical in e al
(CSSec ion). Each sample o i s complex obse a ion eco ds he ob-
se ed p ope ies, a poin inside he lines ing and a e ical o se
En i onmen al Modelling and So wa e 183 (2025) 106248
8
D. Ma ínez e al.
Fig. 3. Le el 3 da a model o clima e science.
(CSSec ionObse a ionSubsample). The shape o a ansec is de ined by
a 3D s aigh line, s o ed wi h a line geome y and a heigh in e -
al (CST ansec ). The combined coo dina es o he poin and heigh
eco ded in each sample (CST ansec Obse a ionSample) mus lie inside
he 3D s aigh line o he ansec . P o iles, e ical sec ions and
ansec s a e associa ed o a single phenomenon ime ins an . Con a y
o his, a ajec o y is gene a ed du ing a ime pe iod. I s shape is also
a 3D line as in he case o a ansec , bu in his case, he line does no
ha e o be s aigh . Thus, each sample (CST ajec o yObse a ionSample)
mus eco d he poin , he e ical o se and he speci ic phenomenon
ime ins an , in addi ion o he obse ed p ope ies.
Two ypes o emo e sensing p ocesses a e suppo ed by he model
(L3CS03), namely CSRemo eSensingSceneP ocess and
CSRemo eSensingSwa hP ocess. The o me gene a es obse a ions a
each loca ion o a p ede ined o spo adic egion, whe eas he la e
gene a es gus s o obse a ions along he pa h de ined by a spa ial
swa h. Two ypes o spa ial samplings a e suppo ed, egula g ids
(CSRSG idScene) and i egula samplings (CSRSI egula Scene). ACSRe-
mo eSensingSceneP ocess gene a es o each phenomenon ime ins an a
complex esul ha is encoded as a spa ial co e age. G id co e ages
o mul ipoin co e ages a e used depending on he ype o egula
o i egula scene (see classes CSRSG idSceneObse a ion and CSI egu-
la SceneObse a ion). On he o he hand, he obse a ions gene a ed by
aCSRemo eSensingSwa hP ocess a e e en mo e complex. In ac , each
obse a ion, which applies o a pe iod o phenomenon ime, con ains a
ime se ies o spa ial co e ages, whose combined geospa ial inge p in
de ine a spa ial swa h. The e o e, each sample o each obse a ion,
eco ds a co e age o a speci ic phenomenon ime ins an o he
obse ed pe iod (see he con en s o classes CSRSG idSwa hSubsample
and CSRSI egula Swa hSubsample).
The model suppo s also he eco ding o esul s gene a ed by
en i onmen al models (L3CS04). A spa ial model (CSSpa ialModel) gen-
e a es an es ima ion o he alues o he p ope ies o in e es o
each cell o a spa ial g id. The g id is always egula in he wo
geospa ial dimensions (CSModelG id). I he model gene a es es ima-
ions a di e en e ical o se s, hen a heigh dimension mus be
added o he g id. Two ypes o e ical dimensions a e suppo ed.
In aCSHeigh Regula ModelG id, e ical o se s a e placed a egula
dis ances. On he o he hand, he e ical o se s a e no placed egu-
la ly in aCSHeigh I egula ModelG id. Spa ial models a e no mally used
o pe o m nowcas s, i.e., eal ime es ima ions o he p ope ies a
e e y disc e e loca ion o he egion o in e es . Fo ecas and eanalysis
a e suppo ed by ins ances o some subclass o CSSpa ioTempo alModel.
Now, he ou pu co e age mus ha e also a empo al dimension, which
is de ined wi h a egula sampling a he obse a ion phenomenon ime
pe iod. Again, he spa ial dimensions o he g id may include a e ical
dimension, wi h a egula o i egula subsampling.
Class CSOceanSampleAnalysisP ocess is used o suppo he obse a-
ion hough he cap u e o specimens in he ocean (L3CS05). Examples
o such specimens a e samples o sea wa e ob ained a speci ic loca-
ions and dep hs and samples o ma ine o ganisms ished a speci ic sea
swa hs. Collec ed specimens a e modeled as ins ances o CSOceanSam-
ple. The p ocess used o collec he specimens is modeled wi h class
CSOceanSamplingP ocess.
6. Use cases
In he ollowing subsec ions, a ious da a models o Le el 4 a e
desc ibed o illus a e and e alua e he use o he amewo k in a
a ie y o di e en applica ions.
6.1. Me eo ological da a in METEOGALICIA
Fou speci ic Le el 4 da a models a e desc ibed in his subsec ion
o illus a e he use o he amewo k wi h me eo ological da a. All
En i onmen al Modelling and So wa e 183 (2025) 106248
9
D. Ma ínez e al.
Fig. 11. Que y pe o mance: CTD p o iles (X axis: Response ime in seconds. Y axis: empo al que y size in minu es).
a e eco ded in he SOS obse a ion able. Response imes o que ies
e u ning a single p ope y a e shown in Fig. 11(a). In gene al, he
pe o mance o he cu en p o o ype o he p esen amewo k is ei he
he bes o nea he bes . The ad hoc model has he wo s pe o mance
in spa ial que ies. Wi h e ical ange que ies, he SOS has a good
pe o mance, since e ical coo dina es a e di ec ly used. In his case,
he JSON agg ega e used by he cu en p o o ype implemen a ion
has o be unnes ed and il e ed, which degene a es i s pe o mance
wi h espec o he SOS model. Howe e , when he opmos e ical
le el is equi ed (obse a ion a sea su ace), he SOS model su e s
a pe o mance p oblem, since i eco ds e ical coo dina es and no
a e ical le el. This is no he case o he ad hoc model, which
eco ds bo h e ical le el in one ield and dep h as an obse ed
p ope y. The use o a JSON a ay s uc u e in he cu en p o o ype
o he p esen amewo k enables also he di ec access o he i s
elemen , wi hou ha ing o compa e coo dina es o ind he opmos .
When wo p ope ies a e equi ed, he cu en implemen a ion o he
p esen amewo k ou pe o ms he o he wo models, as i is shown
in Fig. 11(b). Fig. 11(c) shows how he cu en implemen a ion o
he p esen amewo k main ains a e y good pe o mance when i e
p ope ies a e e ie ed. The SOS model may be used only o e ie e
ew ime ins an s, and e en in ha case wi h a oo poo pe o mance.
The ad hoc model is e en wo se, and i is no shown in he igu e due
o oo high esponse imes.
As a syn hesis o he abo e expe imen s and esul s, i has been
shown ha he SOS model analyzed su e s om impo an pe o -
mance issues when i has o be used o e ie e a ious obse ed
p ope ies o when he size o he da ase is la ge (millions o obse a-
ions). This is so despi e no suppo ing a ious empo al e sions o he
same da a gene a ion p ocess. The pe o mance o he ad hoc models
depends on he objec i es conside ed du ing hei design. O e all, i
he equi ed que y ypes a e conside ed du ing he design phase, hen
i will each a pe o mance ha may be conside ed as a e e ence o
be achie ed by mo e gene ic models. In p ac ice, ad hoc models a e
no always he bes possible models. Finally, he models designed, and
he da a s o age sys ems cons uc ed wi h he cu en p o o ype o he
p esen amewo k o e ed e y good que y pe o mance igu es in all
he analyzed cases. As a inal ema k, i has o be no ed ha as e
da ase s we e no conside ed du ing e alua ion since hei pe o mance
is mainly de e mined by he ex e nal as e da a s o age app oach,
which is ou o he scope o his wo k.
7.3. Quali a i e e alua ion
The model o Le el 1 o he p esen amewo k and some o he
ele an app oaches a e e alua ed in his subsec ion wi h espec o
he equi emen s speci ied in Sec ion 3. A syn hesis o he esul s o
his e alua ion is shown in Table 6. The jus i ica ion and discussion
co esponding o each equi emen is p o ided below.
GR01 Designing models a Le el 4 o he p esen amewo k enables
he euse o s uc u es o he p e ious le els among di e en
applica ion a eas and speci ic applica ions. Thus, he s uc u es
o Le el 4 combine applica ion speci ic s uc u es designed o
speci ic equi emen s wi h gene al pu pose ones. This equi e-
men is also ul illed by speci ic models and p o iles de ined
unde he OGC O&M model, such as Wa e ML, H𝑔
2O, ODM2
and Mains ems. Models ODM, VODM and WaMDaM p o ide
wi h gene al s uc u es ha may be used in many applica ion
domains and hey ha e conside ed he linking wi h applica ion
speci ic s uc u es, bu only o he ep esen a ion o spa ial
sampling ea u es o in e es . The NGSI-LD on ology is o gene al
pu pose and i may be eused in any applica ion. Tha is no he
case o he AIR_POLLUTION_On o and QBOAi base on ologies,
which ha e been designed o ai quali y obse a ions a s a ic
s a ions.
GR02 Va ious ocabula ies may be used in each model o Le el 4
in he p esen amewo k. Va ious names may be p o ided
o ea u e ypes and p ope ies in he ca alog, speci ying he
ocabula y o each o hem. Vocabula ies in he da a alues o
he ins ances may be speci ied using he Ca ego y da a ype.
S anda ds ela ed o he da a s uc u es may be inco po a ed
by de ining hose s uc u es in he Le el 4 da a model. As
an example, i is possible o inco po a e ISO 19115 o Dublin
Co e me ada a elemen s in he de ini ion o p ocess ypes o
inco po a e hose me ada a a he le el o da ase , howe e ,
he amewo k does no o ce his as manda o y. O he models
En i onmen al Modelling and So wa e 183 (2025) 106248
16

D. Ma ínez e al.
Table 6
E alua ion o ela ed app oaches wi h espec o he speci ied equi emen s suppo ed by he p esen amewo k (Y =Yes, N=No, P=Pa ially).
Requi e-
men
O&M
(Cox,
2013)
Wa e ML
(Taylo
e al.,
2013)
ODM
(Ho sbu gh
e al.,
2008)
H𝑔
2O
(Wojda
and
B ouyè e,
2013)
VODM
(Mason
e al.,
2014)
ODM2
(Ho sbu gh
e al.,
2016)
WaMDaM
(Abdallah
and
Rosenbe g,
2019)
Mains ems
(Blodge
e al.,2021)
CDM-CF
(Ea on
e al.,
2023)
NGSI-LD
(ETSI,
2023)
AIR-
_POLLUTION-
_On o Op ea
(2009)
QBOAi -
base
(Galá aga
e al.,
2017)
GR01 Y Y P Y P Y P Y Y Y N N
GR02 P P P P P P Y P Y Y Y Y
GR03 Y Y N Y N Y N Y N P N N
GR04 N N N N N N N N Y – – –
L1R01-
L1R04
Y Y P Y P Y P Y Y Y N N
L2R01 Y Y P Y P Y P Y N Y N N
L2R02 Y Y N Y N N N Y Y N N N
L2R03 Y Y Y Y Y Y Y Y P N N N
L2R04 P P P P P P P P N N N N
L2R05 Y Y P Y Y Y P Y N N N N
L2R06 Y Y Y Y Y Y N Y N N N N
L2R07 Y Y P Y P Y P Y N P P P
L2R08 P P P P P P P P P N N N
enable he use o a ious ocabula ies o name da a elemen s.
Thus, Ne CDF a ibu es may be used in CDM-CF o inco po-
a e s anda d CF names. Seman ic web based solu ions such
as NGSI-LD, AIR_POLLUTION_On o and QBOAi base may also
inco po a e di e en e minologies in hei on ologies. On he
o he hand, all he o he e alua ed app oaches enable he use o
only one ocabula y o he speci ica ion o he ea u e ype and
p ope y names. Rega ding he possibili y o inco po a ing o he
s anda d da a s uc u es, his is possible in hose app oaches
based on he OGC O&M model (O&M, Wa e ML, H𝑔
2O, ODM2,
Mains ems) and also in CDM-CF and NGSI-LD, bu i is no
possible in all he o he .
GR03 Da a in eg a ion may be done in he p esen amewo k a Le els
1, 2 and 3, by using common da a s uc u es de ined a hose le -
els. Thus, a Le el 1, in o ma ion sys ems may be implemen ed
o p o ide wi h unc ionali y o comple e gene al pu pose o e
ea u es ypes and hei ela ionships (as he one p o ided by
gene ic DBMS clien s). Gene ic en i onmen al da a ools may
be implemen ed assuming he model o Le el 2, p o iding now
speci ic unc ionali ies ela ed o da a gene a ion p ocesses, ea-
u es o in e es , obse a ions and hei geospa ial and empo al
con ex . Sys ems ha may in eg a e di e en da ase s coming
om speci ic applica ion a eas may be implemen ed assuming
he models a Le el 3 and speci ic applica ions wi h hei speci ic
unc ionali ies use he models o le el 4. An example o da a
in eg a ion has been shown in Sec ion 7.1. The abo e op ions
may also be enabled by all he models based on OGC O&M
(O&M, Wa e ML, H𝑔
2O, ODM2, Mains ems), by de ining speci ic
p o iles o he p oposed models. NGSI-LD conside s only wo
le els o abs ac ion, and all he o he de ine only one da a
model.
GR04 Pe o mance e alua ion esul s we e discussed in Sec ion 7.2. I
was shown ha he cu en p o o ype o he p esen amewo k
has in gene al a good pe o mance, only clea ly bea en by some
ad hoc simple models in speci ic cases. On he o he hand, he
di ec implemen a ion o he OGC O&M model in a SOS ool
has shown poo que y esponse imes in many cases, speci ically
when ei he he da ase is e y la ge o a ious p ope ies ha e
o be e ie ed o he obse a ion ype is complex. The key
cha ac e is ic o such a di ec implemen a ion o he OGC O&M
da a model ha causes bad pe o mance is he use o a ious
ows o ep esen a ious obse a ion componen s (p ope ies
and/o subsamples). This cha ac e is ic is sha ed by all he mod-
els based on O&M (O&M, Wa e ML, H𝑔
2O, ODM2, Mains ems)
and by ODM, VODM and WaMDaM. I is ob ious ha speci ic
models implemen ed wi h CDM-CF may yield good pe o mance.
The pe o mance o he senso web based solu ions (NGSI-LD,
AIR_POLLUTION_On o and QBOAi base) was no es ed, due
o he lack o implemen a ions ha suppo he conside ed
da ase s.
L1R01-L1R04 The gene al me amodel o Le el 1 o he p esen ame-
wo k p o ides wi h suppo o ep esen a ion o en i ies wi h
hei ele an p ope ies and ela ionships. The model suppo s
also use de ined da a ypes. This is also he case o all he
models based on he OGC O&M (O&M, Wa e ML, H𝑔
2O, ODM2,
Mains ems), since hey a e de ined as specializa ions o he
gene al OGC model o geosp ial ea u es (Ko man and Reed,
2009). The CDM-CF app oach p o ides also he gene al pu pose
s uc u es (dimensions, a ibu es and a iables) ha enable he
suppo o he abo e ea u es. NGSI-LD is based on he RDF
model (Cyganiak e al.,2014), which p o ides he equi ed
gene al pu pose da a ep esen a ion capabili ies. O he models
a e mo e speci ic. In pa icula , ODM, VODM and WaMDaM
enable he inco po a ion o any s uc u e, bu only as a spa ial
sampling ea u e o in e es ype. Finally, AIR_POLLUTION_On o
and QBOAi base a e speci ic models o ai quali y obse ed in
s a ions.
L2R01 OGC s anda ds a e used in he p esen amewo k o model ge-
ome ic objec s and ime pe iods. Following a simila app oach,
e ical in e als ha e also been added. The same, excep o
he lack o speci ic suppo o e ical in e als, applies also o
all he solu ions based on OGC O&M (O&M, Wa e ML, H𝑔
2O,
ODM2, Mains ems). VODM suppo s geome ies bu does no
de ine pe iods and in e als. Geome ies may also be used in
NGSI-LD by inco po a ing geome ic RDF ep esen a ions (Ca
e al.,2024). Only poin s ep esen ed by geog aphic coo dina es
a e suppo ed in ODM, WaMDaM, AIR_POLLUTION_On o and
QBOAi base.
L2R02 The ep esen a ion o geospa ial co e ages o di e en ypes
is suppo ed by ele an da a ypes o he model o Le el 2
in he p esen amewo k. In he cu en p o o ype, hose ep-
esen a ions e e ence ou -o -band ex e nal da a s uc u es ha
e icien ly s o e he co e ages. Di e en encodings a e consid-
e ed in OGC s anda ds o ep esen co e ages, and hey may be
inco po a ed in all he solu ions ha ex end OGC O&M (O&M,
Wa e ML, H𝑔
2O, ODM2, Mains ems). All he o he app oaches do
no conside he ep esen a ion o co e ages as an objec i e.
En i onmen al Modelling and So wa e 183 (2025) 106248
17
D. Ma ínez e al.
L2R03 Da a alues ha eco d measu es wi h uni s o measu e and
ca ego ies o speci ic ocabula ies a e suppo ed in he p esen
amewo k as alues o da a ypes Measu e and Ca ego y. These
da a ypes a e inhe i ed om OGC s anda ds, hus, all he ap-
p oaches based on OGC O&M ollow he same app oach (O&M,
Wa e ML, H𝑔
2O, ODM2, Mains ems). Measu es and ca ego ies
a e also suppo ed by ODM, VODM and WaMDaM. CDM-CF
use Ne CDF a ibu es wi h CF s anda d names o ep esen
he uni s o measu e o a iables, bu i does no ha e a s an-
da d mechanism o ep esen ca ego ies (only o da a quali y
lags). No ele an speci ic suppo is p o ided in NGSI-LD,
AIR_POLLUTION_On o and QBOAi base.
L2R04 Class P ocess o he Le el 2 da a model o he p esen ame-
wo k, and ele an subclasses in models o lowe le els, a e used
o ep esen con ex da a o da a gene a ion p ocesses and hei
e olu ion wi h espec o ime. No speci ic and s anda d suppo
o e olu ion wi h espec o ime is p o ided in any o he
model, al hough many o hem p o ide some kind o s uc u e o
ep esen p ocesses (O&M, Wa e ML, ODM, H𝑔
2O, VODM, ODM2,
WaMDaM, Mains ems).
L2R05 Class Spa ialSamplingFea u e o Le el 2 p o ides suppo o spa-
ial sampling in he cu en amewo k. Gene al spa ial sampling
con ex da a is also inco po a ed by all he models based on
OGC O&M (O&M, Wa e ML, H𝑔
2O, ODM2, Mains ems). ODM,
AIR_POLLUTION_On o and QBOAi base suppo only sampling
poin s, which a e also suppo ed in WaMDaM in he con ex o
a hyd og aphic ne wo k.
L2R06 Specimens a e ep esen ed in he p esen amewo k wi h in-
s ances o class Specimen. The me hods used o collec hose
specimens a e desc ibed wi h ins ances o class SamplingP ocess.
The eco ding o con ex da a o specimens and ela ed p o-
cesses a e suppo ed in all he models specialized om he OGC
O&M model (O&M, Wa e ML, H𝑔
2O, ODM2, Mains ems). Da a
o specimens may also be eco ded in ODM and VODM. This
ype o sampling is no conside ed explici ly in any o he o he
analyzed models.
L2R07 Bo h esul and phenomenon ime s amping a e suppo ed
in he esul s o he p esen amewo k. This is also he case
o o he models based on OGC O&M (O&M, Wa e ML, H𝑔
2O,
ODM2, Mains ems). Only one ime ins an is conside ed in many
app oaches, which is usually in e p e ed as phenomenon ime
(ODM, VODM, WaMDaM, NGSI-LD, AIR_POLLUTION_On o and
QBOAi base). CDM-CF is o mo e gene al pu pose, hus, i does
no assume any speci ic seman ics o ime s amps.
L2R08 Class Obse a ionSubsample o Le el 2 and ele an subclasses
in lowe Le els o he p esen amewo k a e used o suppo
alues wi h complex s uc u e. Such complex alues may be
inco po a ed also in o he models based on OGC O&M, al hough
hey a e no explici ly pa o he model. Some kinds o subsam-
ples ha e been conside ed in ODM2 ( ime se ies, sec ions and
ansec s) and WaMDaM ( ime se ies and elec onic iles). CDM-
CF de ines s uc u es o g id co e ages, poin s, ime se ies,
ajec o ies, p o iles, ime se ies o p o iles and ajec o ies o
p o iles. Complex alues a e ep esen ed wi h g oups o simple
ones in ODM and wi h g oups and da ase s in VODM. No speci ic
suppo has been de ined in NGSI-LD, AIR_POLLUTION_On o
and QBOAi base.
8. Conclusions and u u e wo k
A concep ual da a modeling amewo k o en i onmen al in o ma-
ion was p oposed. The amewo k p o ides models a ou le els o
abs ac ion. Le els 1 and 2 de ine a gene al en i onmen al da a model
based on he OGC O&M s anda d. A his le el, he amewo k p o ides
also a me ada a ca alog ha suppo s he use o mul iple ocabula ies.
De ining gene ic da a models o Le el 3 enables he adap ion o he
amewo k o he speci ic needs o di e en applica ion a eas. The
gene al pu pose s uc u es o he abo e le els o abs ac ion may be
eused in many applica ions, combining gene al pu pose s uc u es
wi h applica ion speci ic ones. The euse o such common s uc u es
and he po en ial use o s anda d ocabula ies eases he in eg a ion o
di e en and he e ogeneous da ase s a a ious le els o abs ac ion.
The amewo k suppo s he ep esen a ion o con ex da a ela ed o
da a gene a ion p ocesses, ea u es o in e es , sampling ea u es and
gene a ed alues, which include simple alues and complex ones, such
as ime se ies, ajec o ies, ansec s, p o iles and di e en ypes o
spa ial and spa io- empo al co e ages.
Gene al pu pose designs may be used di ec ly by end-use s wi hou
any design e o o be done, howe e hey use o ail in p o iding
he e iciency ha may be achie ed by good ad hoc designs. I was
shown ha he gene al pu pose di ec O&M implemen a ions p o ided
by SOS ools ha e impo an pe o mance p oblems in many cases.
The cu en amewo k p o ides a gene al design a le el 2, bu i
equi es some design e o o p oduce le els 3 and 4. On he o he
hand, he pe o mance achie ed is in gene al close o ha o ad hoc
models (o e en be e when he design o hose ad hoc models is
no good). Applica ion domain expe s wi h some da a modeling skills
should ag ee in he design o le el 3 models. This is no a simple ask
in gene al, bu when i is achie ed, he bene i s a e e y impo an .
Nex , end-use s may bene i om he exis ence o hose le el 3 models
o ease he design o hei le el 4 models. They need o ha e some
da a modeling knowledge, howe e much da a modeling expe ience
would no be equi ed o achie e good solu ions since hey a e al eady
suppo ed by exis ing models o le el 3.
Di ec O&M implemen a ions p o ided by SOS ools and some o he
app oaches do no equi e a p e ious decla a ion o he p ope ies o
ea u es and o he obse ed p ope ies measu ed by p ocesses. This is
close o he idea o no equi ing schema de ini ion o many NoSQL
da abases (Sadalage and Fowle ,2013). On he o he hand, he use
mus decla e he p ope ies o ea u e ypes and he obse ed p ope ies
o p ocesses in he p oposed amewo k du ing he design o le el 4
models. Those decla a ions a e eco ded oge he wi h o he pa s o
he schema in he ca alog de ined a le el 2. Ad an ages and disad an-
ages o wo king wi h and wi hou schema ha e la gely been s udied
by he da abase communi y. In gene al, he lack o schema eases he
inco po a ion o schema changes du ing da a inse ion. Howe e , he
lack o schema b ings impo an p oblems o applica ions ha pe o m
da a que ying, since he changes in he schema a e no documen ed in
he sys em.
Di ec ions o u u e wo k a e mainly o ien ed o he de eloping o
gene al echnologies o he sea ching and in e ac i e explo a ion o
en i onmen al da ase s modeled wi h he p esen amewo k. Besides,
esea ch wo k is s ill needed o adap he amewo k o suppo a ious
e sions o he models o le els 3 and 4, enabling he e olu ion o
he da abase schema. Finally, a s ep o wa d in he da a modeling line
will be he p oposal o a me amodel ha enables he speci ica ion o
applica ion speci ic solu ions wi hou he need o in e media e le els
o abs ac ion.
CRediT au ho ship con ibu ion s a emen
Da id Ma ínez: So wa e, In es iga ion. Lau a Po: W i ing – e-
iew & edi ing, In es iga ion, Concep ualiza ion. Raquel T illo-Lado:
W i ing – e iew & edi ing, In es iga ion, Concep ualiza ion. José R.R.
Viquei a: W i ing – e iew & edi ing, So wa e, In es iga ion, Concep-
ualiza ion.
En i onmen al Modelling and So wa e 183 (2025) 106248
18
D. Ma ínez e al.
Table 7
Le el 3 equi emen s: T a ic in Sma Ci ies.
Requi emen Desc ip ion
L3TF01 The da a gene a ed by he au oma ic moni o ing o a ic a
speci ic loca ions o he oad ne wo k.
L3TF02 The eal ime a ic condi ions es ima ed by a ic
econs uc ion models a each sec ion o he oad ne wo k.
L3TF03 The u u e a ic condi ions p edic ed by a ic models o
each sec ion o he oad ne wo k.
Table 8
Le el 3 equi emen s: Ai Quali y in Sma Ci ies.
Requi emen Desc ip ion
L3AQ01 The da a gene a ed by s a ic ai quali y moni o ing s a ions.
L3AQ02 The da a gene a ed by in-si u emo able de ices, ins alled a
g ound s a ic pla o ms, g ound mobile pla o ms and lying
pla o ms.
L3AQ03 The da a gene a ed by nowcas and o ecas models o ai
quali y.
Decla a ion o compe ing in e es
The au ho s decla e ha hey ha e no known compe ing inan-
cial in e es s o pe sonal ela ionships ha could ha e appea ed o
in luence he wo k epo ed in his pape .
Acknowledgmen s
This wo k was pa ially suppo ed by he ollowing p ojec s.
TRAFAIR p ojec (2017-EU-IA-0167), co- inanced by he Connec ing
Eu ope Facili y o he Eu opean Union. Galicia Ma ine Science p o-
g amme, which is pa o he Complemen a y Science Plans o Ma ine
Science o Minis e io de Ciencia, Inno ación 𝑦Uni e sidades included
in he Reco e y, T ans o ma ion and Resilience Plan (PRTR-C17.I1),
unded h ough Xun a de Galicia wi h Nex Gene a ionEU and he
Eu opean Ma i ime Fishe ies and Aquacul u e Funds. Ea hDL-USC
(PID2022-141027NB-C22) and NEAT-AMBIENCE (PID2020-113037RB-
I00) p ojec s, unded by Agencia Es a al de In es igación, Minis e io
de Ciencia e Inno ación, h ough he na ional plan o scien i ic and
echnical esea ch and inno a ion 2021–2023.
Appendix. Da a models o a ic and ai quali y in sma ci ies
A.1. Le el 3 equi emen s and da a models
In he scope o he de elopmen o sma ci y in as uc u es, i is
usual o include se ices ela ed o he moni o ing o a ic condi ions
and ai quali y. In his sec ion, o illus a ion pu poses, equi emen s
and abs ac models a e p oposed o such applica ion a ea. In pa ic-
ula , Tables 7and 8show he equi emen s o a ic and ai quali y
models, espec i ely.
The abs ac da a model o le el 3 o a ic moni o ing in he
con ex o sma ci ies is shown in Fig. 12. B oadly, he model de ines
s uc u es o suppo he eco ding o obse a ions gene a ed by a ic
senso s loca ed a speci ic loca ions o he oad ne wo k (L3TF01). I
inco po a es also s uc u es o eco d ou pu s o wo ypes o models:
(i) models ha es ima e he a ic in each sec ion o he oad ne wo k
in eal ime ((L3TF02)) and (ii) models ha p edic he e olu ion wi h
espec o ime o he a ic in he oad ne wo k ((L3TF02)).
Fig. 13 shows he abs ac da a model o le el 3 ha suppo s he
eco ding o da a ob ained om he moni o ing o ai quali y in sma
ci ies. The model p o ides s uc u es o suppo all he equi emen s
de ined in Table 8.
A.2. T a ic and ai quali y da a in he TRAFAIR p ojec
The main objec i e o he TRAFAIR p ojec (Po e al.,2019) was
he moni o ing and p edic ion o ai quali y a high scale inside ci ies.
T a ic in he ci ies was moni o ed a speci ic loca ions using a ic
senso s (Bachechi e al.,2022b). Using hose a ic obse a ions, a
a ic econs uc ion model was execu ed o p o ide eal ime es ima-
ions o a ic low in each oad sec ion (Bachechi e al.,2022a;Bilo a
and Nesi,2021). Fig. 14(a) shows a Le el 4 da a model ha suppo s
he eco ding o he a ic da a gene a ed in he TRAFAIR p ojec .
Classes oad_sec ion, oad_segmen and oad_node a e used o eco d he
oad ne wo k ob ained om OpenS ee Map.21 T a ic senso s (Class
a ic_senso ) a e loca ed a speci ic segmen s and a e used o gene a e
a ic obse a ions. No ice ha class a ic_senso has he ole o bo h
aT a icS a ion and aT a icPoin Obse a ionP ocess. Each a ic obse -
a ion (Class senso _ a ic_obse a ion) p o ides a alue o he a ic
low, i.e., numbe o ehicles pe hou , and o he a ic occupancy
(pe cen age o ime when he senso was de ec ing a ehicle), e e y
5 min. The oad ne wo k was il e ed and ans o med in he p ojec o
gene a e a g aph o main s ee oad a cs (Class oad_a c). The a ic
econs uc ion model (class a ic_ low_model) p o ides an es ima ion
o he a ic low e e y 15 min o each oad a c, which is eco ded in
class a ic_ low_model_ou pu .
Ai quali y moni o ing was pe o med wi h low cos senso s (Rollo
e al.,2023;Casa i and Po,2024) and calib a ion models buil wi h
machine lea ning (Bachechi e al.,2024). Ai quali y p edic ion was
pe o med wi h he GRAL pollu an dispe sion model,22 using a ic
emission es ima ions as main pollu an sou ces. Fig. 14(b) shows a
Le el 4 da a model ha enables he eco ding o he ai quali y da a
gene a ed in he TRAFAIR p ojec . Ai quali y legal s a ions (class
aq_legal_s a ion) p o ide wi h obse a ions wi h legal co e age inside
he ci y. These ea u es model bo h ai quali y s a ions (AQS a ion) and
ai quali y in-si u s a ic p ocesses (AQInSi uS a icP ocess), since hei
senso s a e no mo ing h ough di e en loca ions du ing hei li e ime.
Each obse a ion (class aq_legal_s a ion_obse a ion) con ains he concen-
a ion o a ious gases eco ded in di e en obse ed p ope ies. Low
cos senso s (class senso _low_cos ) a e used a di e en loca ions du ing
he p ojec . In ac , hey mus be colloca ed wi h a legal s a ion du ing
some pe iods o gene a e aining da ase s ha enable he gene a ion
o he calib a ion models. Each aw obse a ion gene a ed by a low
cos senso (class senso _ aw_obse a ion) e e s o a speci ic sampling
loca ion (class senso _low_cos _ ea u e) and i con ains a ba e y ol age
measu e, a humidi y measu e, a empe a u e measu e, and a couple o
aw ol age measu es, gene a ed by an elec ochemical cell, o each
o he ollowing gases: CO,NO,NO2,O3. Low cos senso aw ol age
obse a ions a e ans o med o gene a e gas concen a ions using a
calib a ion model (class senso _calib a ion). A calib a ion model is gen-
e a ed o a speci ic senso and i has an algo i hm o each gas. The
algo i hm consis s o a machine lea ning model ained wi h a da ase
gene a ed du ing he colloca ion o he senso wi h a legal s a ion. All
he me ada a o each algo i hm is eco ded in a JSON s uc u e, as i is
shown in he diag am, and i includes he e e ence o he legal s a ion,
he aining pe iod, he inpu a iables used by he model (including
a leas senso ol ages), a e e ence o he model implemen a ion
used (sklea n.ensemble.RandomFo es Reg esso was used in he p ojec )
and he alues o he used hype -pa ame e s. Calib a ed obse a ions
(senso _calib a ed_obse a ion) p o ide wi h es ima ions o gas concen-
a ions a he spa ial sampling ea u es. Finally, he da a model enables
also he eco ding o he ou pu s gene a ed by he GRAL pollu an
dispe sion model, which ha e he o m o spa io empo al co e ages o
NO𝑥concen a ion alues. Class ai _quali y_model eco ds he me ada a
o he GRAL con igu a ion used o gene a e he p edic ions and also he
cha ac e is ics o he spa io- empo al g id conside ed. Model ou pu s,
i.e., spa io- empo al co e ages o NOx alues a e eco ded in class
ai _quali y_model_ou pu .
21 h ps://www.opens ee map.o g/
22 h ps://g al. ug az.a /
En i onmen al Modelling and So wa e 183 (2025) 106248
19
D. Ma ínez e al.
Fig. 12. Le el 3 da a model o a ic.
Fig. 13. Le el 3 da a model o ai quali y.
En i onmen al Modelling and So wa e 183 (2025) 106248
20
D. Ma ínez e al.
Fig. 14. Le el 4 da a models o a ic and ai quali y da a in he TRAFAIR p ojec .
Da a a ailabili y
The da a o all he da ase s is publicly a ailable in hei ele an
da a p oduce o ganiza ions. T a ic and ai quali y da a o he ci y o
San iago de Compos ela may be downloaded om he Eu opean Da a
Po al h ps://da a.eu opa.eu/. The me eo ological da a is a ailable
a open da a in as uc u e o Me eoGalicia h ps://www.me eogalicia.
gal. The oceanog ahic da a is a ailable a he open da a in as uc u e
o In ecma h ps://www.in ecma .gal. To ep oduce he expe imen s
unde aken du ing he e alua ion o he amewo k, CSV iles adap ed
o he schema o each o he app oaches may be p o ided upon eques
o he co esponding au ho o his pape .
Re e ences
Abdallah, A.M., Rosenbe g, D.E., 2019. A da a model o manage da a o wa e
esou ces sys ems modeling. En i on. Model. So w. 115, 113–127. h p://dx.doi.
o g/10.1016/j.en so .2019.02.005.
Bachechi, C., Po, L., Rollo, F., 2022a. Big da a analy ics and isualiza ion in a ic mon-
i o ing. Big Da a Res. 27, 100292. h p://dx.doi.o g/10.1016/j.bd .2021.100292,
URL: h ps://www.sciencedi ec .com/science/a icle/pii/S221457962100109X.
Bachechi, C., Rollo, F., Po, L., 2022b. De ec ion and classi ica ion o senso anomalies
o simula ing u ban a ic scena ios. Clus e Compu . 25, 2793–2817. h p://dx.
doi.o g/10.1007/s10586-021-03445-7.
Bachechi, C., Rollo, F., Po, L., 2024. HypeAIR: A no el amewo k o eal- ime low-
cos senso calib a ion o ai quali y moni o ing in sma ci ies. Ecol. In o m.
81, 102568. h p://dx.doi.o g/10.1016/j.ecoin .2024.102568, URL: h ps://www.
sciencedi ec .com/science/a icle/pii/S1574954124001109.
En i onmen al Modelling and So wa e 183 (2025) 106248
21

D. Ma ínez e al.
Baumann, P., 1994. Managemen o mul idimensional disc e e da a. VLDB J. 3 (4),
401–444. h p://dx.doi.o g/10.1007/BF01231603.
Bilo a, S., Nesi, P., 2021. T a ic low econs uc ion by sol ing inde e minacy on
a ic dis ibu ion a junc ions. Fu u e Gene . Compu . Sys . 114, 649–660. h p:
//dx.doi.o g/10.1016/j. u u e.2020.08.017.
Blaha, M., Rumbaugh, J., 2005. Objec -o ien ed Modeling and Design wi h UML.
Pea son Educa ion.
Blodge , D., Johnson, J.M., Sondheim, M., Wieczo ek, M., F azie , N., 2021. Mains ems:
A logical da a model implemen ing mains em and d ainage basin ea u e ypes
based on Wa e ML2 pa 3: HY ea u es concep s. En i on. Model. So w. 135,
104927. h p://dx.doi.o g/10.1016/j.en so .2020.104927.
B ambilla, M., Cabo , J., Wimme , M., 2017. Model-D i en So wa e Enginee ing in
P ac ice: Second Edi ion, second ed. Mo gan & Claypool Publishe s.
B ö ing, A., S asch, C., Ech e ho , J., 2012. OGC Senso Obse a ion Se ice In e ace
S anda d. Open Geospa ial Conso ium Inc., OpenGIS Implemen a ion S anda d.
h ps://www.ogc.o g/s anda d/sos/.
B own, P.G., 2010. O e iew o sciDB: La ge scale a ay s o age, p ocessing and
analysis. In: P oceedings o he 2010 ACM SIGMOD In e na ional Con e ence on
Managemen o Da a. SIGMOD ’10, New Yo k, NY, USA, ACM, pp. 963–968.
h p://dx.doi.o g/10.1145/1807167.1807271.
Ca , N.J., Hombu g, T., Pe y, M., Knibbe, F., Cox, S.J., Abhaya a na, J., Bonduel, M.,
C ipps, P.J., Janowicz, K., 2024. OGC GeoSPARQL - A Geog aphic Que y Language
o RDF Da a. Open Geospa ial Conso ium Inc., OpenGIS Implemen a ion S anda d.
h p://www.opengis.ne /doc/IS/geospa ql/1.1.
Casa i, M., Po, L., 2024. Mi H: A amewo k o mi iga ing hyg oscopici y in low-
cos PM senso s. En i on. Model. So w. 173, 105955. h p://dx.doi.o g/10.
1016/j.en so .2024.105955, URL: h ps://www.sciencedi ec .com/science/a icle/
pii/S1364815224000161.
Chen, P.P.-S., 1976. The en i y- ela ionship model— owa d a uni ied iew o da a. ACM
T ans. Da abase Sys . 1 (1), 9–36. h p://dx.doi.o g/10.1145/320434.320440.
Codd, E.F., 1970. A ela ional model o da a o la ge sha ed da a banks. Commun.
ACM 13 (6), 377–387. h p://dx.doi.o g/10.1145/362384.362685.
Comp on, M., Ba naghi, P., Be mudez, L., Ga cía-Cas o, R., Co cho, O., Cox, S.,
G aybeal, J., Hauswi h, M., Henson, C., He zog, A., Huang, V., Janowicz, K.,
Kelsey, W.D., Phuoc, D.L., Le o , L., Leggie i, M., Neuhaus, H., Nikolo , A.,
Page, K., Passan , A., She h, A., Taylo , K., 2012. The SSN on ology o he
W3C seman ic senso ne wo k incuba o g oup. J. Web Seman . 17, 25–32. h p:
//dx.doi.o g/10.1016/j.websem.2012.05.003.
Cox, S., 2013. Obse a ions and Measu emen s. Ve ion 2.0. Open Geospa ial Conso ium
Inc., The OpenGIS Abs ac Speci ica ion. h p://www.opengis.ne /doc/as/om/2.0.
Cyganiak, R., Wood, D., Lan hale , M., 2014. RDF 1.1 Concep s and Abs ac Syn ax.
Wo ld Wide Web Conso ium, W3C Recommenda ion. h ps://www.w3.o g/TR/
d 11-concep s/.
de Bakke , M.P., de Jong, K., Schmi z, O., Ka ssenbe g, D., 2017. Design and demons a-
ion o a da a model o in eg a e agen -based and ield-based modelling. En i on.
Model. So w. 89, 172–189. h p://dx.doi.o g/10.1016/j.en so .2016.11.016, URL:
h ps://www.sciencedi ec .com/science/a icle/pii/S1364815216305898.
De ys, E., Habe mann, T., Heazel, C., Lo , R., Rouaul , E., 2019. OGC GeoTIFF
S anda d. Open Geospa ial Conso ium Inc., OGC Implemen a ion S anda d. h p:
//www.opengis.ne /doc/IS/GeoTIFF/1.1.
Ea on, B., e al., 2023. Ne CDF clima e and o ecas (CF) me ada a con en ions.
h ps://c con en ions.o g. (Online; Accessed 14 Ap il 2024).
ETSI, 2023. Con ex In o ma ion Managemen (CIM);NGSI-LD In o ma ion Model.
Eu opean Telecommunica ions S anda diza ion Ins i u e (ETSI), G oup Speci ica-
ion (GS). h ps://www.e si.o g/deli e /e si_gs/CIM/001_099/006/01.02.01_60/gs_
CIM006 010201p.pd .
Fe nandez, S., Had i, R., I o, T., Ma sa-Maes e, I., Velasco, J.R., 2016. On ology-based
a chi ec u e o in elligen anspo a ion sys ems using a a ic senso ne wo k.
Senso s 16 (8), h p://dx.doi.o g/10.3390/s16081287.
Galá aga, L., Ma hiassen, K.A.M., Hose, K., 2017. QBOAi base: The Eu opean ai
quali y da abase as an RDF cube. In: Niki ina, N., Song, D., Fokoue, A., Haase, P.
(Eds.), P oceedings o he ISWC 2017 Pos e s & Demons a ions and Indus y T acks
Co-Loca ed wi h 16 h In e na ional Seman ic Web Con e ence. ISWC 2017, Vienna,
Aus ia, Oc obe 23 d - To - 25 h, 2017, In: CEUR Wo kshop P oceedings, ol.
1963, CEUR-WS.o g, URL: h ps://ceu -ws.o g/Vol-1963/pape 507.pd .
Gonzalez-Pe ez, C., Hende son-Selle s, B., 2008. Me amodelling o So wa e
Enginee ing. Wiley Publishing.
Halle , A., Janowicz, K., Cox, S., Phuoc, D.L., Taylo , K., Le ançois, M., 2017. Seman ic
Senso Ne wo k On ology. Open Geospa ial Conso ium and Wo ld Wide Web
Conso ium, W3C Recommenda ion. h ps://www.w3.o g/TR/ ocab-ssn/.
Ha pham, Q., 2020. A simple axonomy o desc ibing he spa io- empo al s uc u e o
en i onmen al modelling da a. En i on. Model. So w. 133, 104810. h p://dx.doi.
o g/10.1016/j.en so .2020.104810.
He ing, J., 2020. Topic 1: Spa ial Schema.. Open Geospa ial Conso ium Inc., The
OpenGIS Abs ac Speci ica ion. h ps://www.ogc.o g/s anda ds/as/.
Ho sbu gh, J.S., Au denkampe, A.K., Mayo ga, E., Lehne , K.A., Hsu, L., Song, L.,
Jones, A.S., Damiano, S.G., Ta bo on, D.G., Valen ine, D., Zasla sky, I., Whi e-
nack, T., 2016. Obse a ions da a model 2: A communi y in o ma ion model
o spa ially disc e e Ea h obse a ions. En i on. Model. So w. 79, 55–74. h p:
//dx.doi.o g/10.1016/j.en so .2016.01.010.
Ho sbu gh, J.S., Ta bo on, D.G., Maidmen , D.R., Zasla sky, I., 2008. A ela ional model
o en i onmen al and wa e esou ces da a. Wa e Resou . Res. 44 (5), h p://dx.
doi.o g/10.1029/2007WR006392, URL: h ps://agupubs.onlinelib a y.wiley.com/
doi/abs/10.1029/2007WR006392.
Johansson, L., Ka ppinen, A., Ku ppa, M., Kousa, A., Niemi, J.V., Kukkonen, J., 2022.
An ope a ional u ban ai quali y model ENFUSER, based on dispe sion modelling
and da a assimila ion. En i on. Model. So w. 156, 105460. h p://dx.doi.o g/10.
1016/j.en so .2022.105460, URL: h ps://www.sciencedi ec .com/science/a icle/
pii/S1364815222001657.
Kim, D., Mus e, M., Me wade, V., 2015. A GIS-based ela ional da a model o mul i-
dimensional ep esen a ion o i e hyd odynamics and mo phodynamics. En i on.
Model. So w. 65, 79–93. h p://dx.doi.o g/10.1016/j.en so .2014.12.002.
Kimball, R., Ross, M., 2002. The Da a Wa ehouse Toolki : The Comple e Guide o
Dimensional Modeling, second ed. John Wiley & Sons, Inc., USA.
Ko man, C., Reed, C., 2009. Topic 5: Fea u es. Open Geospa ial Conso ium Inc., The
OpenGIS Abs ac Speci ica ion. h ps://www.ogc.o g/s anda ds/as/.
Lebo, T., Sahoo, S., McGuinness, D., 2013. PROV-O: The PROV On ology. Wo ld Wide
Web Conso ium, W3C Recommenda ion. h p://www.w3.o g/TR/p o -o/.
Le y, A.Y., 2000. Logic-based echniques in da a in eg a ion. In: Minke , J. (Ed.),
Logic-Based A i icial In elligence. Sp inge US, Bos on, MA, pp. 575–595. h p:
//dx.doi.o g/10.1007/978-1-4615-1567-8_24.
Liang, S., Khala beigi, T., an de Schaa , H., 2021. OGC Senso Things API Pa
1: Sensing Ve sion 1.1. Open Geospa ial Conso ium Inc., OGC Implemen a ion
S anda d. h p://www.opengis.ne /doc/is/senso hings/1.1.
Ma ínez, D., Po, L., T illo-Lado, R., Viquei a, J.R.R., 2022. TAQE: A da a modeling
amewo k o a ic and ai quali y applica ions in sma ci ies. In: B aun, T.,
C is ea, D., Jäschke, R. (Eds.), G aph-Based Rep esen a ion and Reasoning. Sp inge
In e na ional Publishing, Cham, pp. 25–40.
Mason, S.J., Cle eland, S.B., Llo e , P., Izu ie a, C., Poole, G.C., 2014. A cen alized ool
o managing, a chi ing, and se ing poin -in- ime da a in ecological esea ch lab-
o a o ies. En i on. Model. So w. 51, 59–69. h p://dx.doi.o g/10.1016/j.en so .
2013.09.008.
Na i i, S., Mazze i, P., San o o, M., Papeschi, F., C aglia, M., Ochiai, O., 2015. Big
da a challenges in building he global ea h obse a ion sys em o sys ems. En i on.
Model. So w. 68, 1–26. h p://dx.doi.o g/10.1016/j.en so .2015.01.017.
OGC, 2007. Topic 6: Schema o Co e age Geome y and Func ions. Ve sion 7.0.0..
Open Geospa ial Conso ium Inc., The OpenGIS Abs ac Speci ica ion. h ps://
www.ogc.o g/s anda ds/as/.
Op ea, M.M., 2009. AIR_POLLUTION_On o: an on ology o ai pollu ion analysis
and con ol. In: Iliadis, Maglogiann, Tsoumakasis, Vlaha as, B ame (Eds.), A i-
icial In elligence Applica ions and Inno a ions III. Sp inge US, Bos on, MA, pp.
135–143.
Pisoni, E., De Ma chi, D., di Ta an o, A., Bessagne , B., Sajani, S.Z., De Meij, A.,
Thunis, P., 2024. SHERPA-cloud: An open-sou ce online model o simula e ai
quali y managemen policies in Eu ope. En i on. Model. So w. 176, 106031.
h p://dx.doi.o g/10.1016/j.en so .2024.106031, URL: h ps://www.sciencedi ec .
com/science/a icle/pii/S1364815224000926.
Po, L., Rollo, F., Viquei a, J.R.R., Lado, R.T., Bigi, A., López, J.C., Paolucci, M., Nesi, P.,
2019. TRAFAIR: Unde s anding a ic low o imp o e ai quali y. In: 2019 IEEE
In e na ional Sma Ci ies Con e ence. ISC2, pp. 36–43. h p://dx.doi.o g/10.1109/
ISC246665.2019.9071661.
Reguei o, M.A., Viquei a, J.R., S asch, C., Taboada, J.A., 2017. Seman ic media ion o
obse a ion da ase s h ough senso obse a ion se ices. Fu u e Gene . Compu .
Sys . 67, 47–56. h p://dx.doi.o g/10.1016/j. u u e.2016.08.013.
Rigaux, P., Scholl, M., Voisa d, A., 2001. Spa ial Da abases Wi h Applica ion o GIS.
Mo gan Kau mann.
Rollo, F., Bachechi, C., Po, L., 2023. Anomaly de ec ion and epai ing o imp o ing
ai quali y moni o ing. Senso s 23 (2), h p://dx.doi.o g/10.3390/s23020640, URL:
h ps://www.mdpi.com/1424-8220/23/2/640.
Sadalage, P.J., Fowle , M., 2013. NoSQL Dis illed: A B ie Guide o he Eme ging Wo ld
o Polyglo Pe sis ence. Addison-Wesley, Uppe Saddle Ri e , NJ.
Salas, D., Liang, X., Na a o, M., Liang, Y., Luna, D., 2020. An open-da a open-
model amewo k o hyd ological models’ in eg a ion, e alua ion and applica ion.
En i on. Model. So w. 126, 104622. h p://dx.doi.o g/10.1016/j.en so .2020.
104622.
Ta bo on, D.G., Ames, D.P., Ho sbu gh, J.S., Goodall, J.L., Couch, A., Hoope , R.,
Bales, J., Wang, S., Cas ono a, A., Seul, M., Idaszak, R., Li, Z., Dash, P., Black, S.,
Rami ez, M., Yi, H., Calloway, C., Cogswell, C., 2024. Hyd oSha e e ospec i e:
Science and echnology ad ances o a comp ehensi e da a and model publica ion
en i onmen o he wa e science domain. En i on. Model. So w. 172, 105902.
h p://dx.doi.o g/10.1016/j.en so .2023.105902.
Taylo , P., Cox, S., Walke , G., Valen ine, D., Sheahan, P., 2013. Wa e ML2.0:
de elopmen o an open s anda d o hyd ological ime-se ies da a exchange. J.
Hyd oin o m. 16 (2), 425–446. h p://dx.doi.o g/10.2166/hyd o.2013.174.
The HDF G oup, 2024. HDF5 API e e ence. h ps://docs.hd g oup.o g/hd 5/ 1_14/.
(Accessed: 23 Feb ua y 2024).
Villa oya, S., Viquei a, J.R.R., Reguei o, M.A., Taboada, J.A., Co os, J.M., 2016. SODA:
A amewo k o spa ial obse a ion da a analysis. Dis ib. Pa allel Da abases 34
(1), 65–99. h p://dx.doi.o g/10.1007/s10619-014-7165-7.
En i onmen al Modelling and So wa e 183 (2025) 106248
22
D. Ma ínez e al.
Viquei a, J.R.R., Lo en zos, N.A., B isaboa, N.R., 2005. Su ey on spa ial da a modelling
app oaches. In: Manolopoulos, Y., Papadopoulos, A., Vassilakopoulos, M. (Eds.),
Spa ial Da abases: Technologies, Techniques and T ends. Idea G oup, pp. 1–22.
Viquei a, J.R.R., Villa oya, S., Me a, D., Taboada, J.A., 2020. Sma en i onmen al
da a in as uc u es: B idging he gap be ween ea h sciences and ci izens. Appl.
Sci. 10 (3), h p://dx.doi.o g/10.3390/app10030856.
Wojda, P., B ouyè e, S., 2013. An objec -o ien ed hyd ogeological da a model o
g oundwa e p ojec s. En i on. Model. So w. 43, 109–123. h p://dx.doi.o g/10.
1016/j.en so .2013.01.015.
Yu zle , J., 2024. OGC GeoPackage Encoding S anda d. Open Geospa ial Conso ium
Inc., OGC Encoding S anda d.. h p://www.opengis.ne /doc/IS/geopackage/1.4.
En i onmen al Modelling and So wa e 183 (2025) 106248
23