Directoraat-Generaal Rijkswaterstaat
Ministerie van Verkeer en Waterstaat
Storten van baggerspecie in open putdepots (fase 2) Deelrapport 3: Verspreiding van stikstof tijdens storten van baggerspecie in open putdepots AKWA-rapport 00.002 RIZA werkdocument 2000.42X
AK\AM Advies- en Kenniscentrum Waterbodems
Ministerie van verkeer en waterstaat
~"*»5^**^"'
D i r e c t o r a a t - G e n e r a a l Rijkswaterstaat Rijksinstituut voor Integraal Zoetwaterbeheer en Afvalwaterbehandeling/RIZA
Storten van baggerspecie in open putdepots (fase 2) Deelrapport 3: Verspreiding van stikstof tijdens storten van baggerspecie in open putdepots 8 maart 2000
RIZA werkdocument 2000.042X AKWA-rapport 00.002
Auteur: G.A. van den Berg (RIZA, afdeling WST)
Inhoudsopgave
Voorwoord 5 Samenvatting 7
1 Inleiding 9 2 Stikstofhuishouding in natuurlijke watersystemen 11 2.1 Inleiding 77 2.2 Oppervlaktewater 77 2.3 Liggende waterbodems 14 2.4 Gestratificeerde watersystemen 76 3 Vrijkomen van stikstof tijdens storten 77 3.1 Inleiding 77 3.2 Vrijkomen van stikstof tijdens de stortfase 77 3.2.1 Sedimentatiewater 17 3.3 Consolidatiewater 18 3.4 Berekening emissie van stikstof tijdens de stortfase 18 4 Scenarioberekeningen voor inschatting stikstofbelasting ten gevolge van storten 27 4.1 Inleiding 27 4.2 Niet-gestrabficeerde open putdepots 22
4.2.1 Modelbeschrijving 22 4.2.2 Voorbeeldberekeningen 22 4.2.3 Discussie 23 4.3 Gestratificeerde open putdepots 24 4.3.1 Inleiding 24 4.3.2 Modelbeschrijving 24 4.3.3 Voorbeeldberekeningen 25 4.3.4 Discussie 26 5 Conclusies 29 Referenties 37
Storten van baggerspecie in open putdepots (fase 2)
Storten van baggerspecie in open putdepots (fase 2)
Voorwoord
Het project "storten van baggerspecie in open putdepots" heeft tot doel het ontwikkelen van een beleidslijn voor het storten van verontreinigde specie in open putdepots. Het onderzoek voor dit project wordt gefaseerd uitgevoerd. De resultaten van de eerste fase zijn door het Advies en Kenniscentrum Waterbodems (AKWA) beschreven in een covernotitie (Best, J. de etal., 1999). Op basis van deze covernotitie is vervolgonderzoek gedefinieerd. In de tweede fase van het onderzoek wordt aandacht geschonken aan het ontwerp van een open putdepot, de verspreiding van verontreinigingen ten gevolge van storten van baggerspecie in open putdepots, evenals de consequenties voor normstelling en beheer. Deelproject 3 richt zich op de verspreiding van stikstof als gevolg van storten van baggerspecie in open en halfopen putdepots. Een eerste inventarisatie van bestaande onderzoeksresultaten op dit terrein is uitgevoerd door Jeroen Bakker in het kader van een afstudeerproject van de vakgroep Fysische Geografie (Universiteit Utrecht). In het voorliggende rapport wordt met behulp van een aantal scenarioberekeningen een beeld gegeven van het vrijkomen van stikstof bij het storten van baggerspecie in open putdepots. Tevens wordt aangegeven door welke locatie- en stofspecifieke parameters de gevoeligheid voor verspreiding van stikstof het meest wordt bei'nvloed. De auteur bedankt dr. F. van Luijn (RWS-RDIJ) en dr. P.C.M. Boers (RIZA-WSE) voor het kritisch doorlezen van dit rapport. Gerard van den Berg (maart 2000)
Storten van baggerspecie in open putdepots (fase 2)
Storten van baggerspecie in open putdepots (fase 2)
Samenvatting
Het storten van baggerspecie in open putdepots in uiterwaarden, rivierbeddingen, meren, havens en kanalen wordt sinds enkele jaren beleidsmatig gezien als een mogelijkheid voor het milieuvriendelijk opbergen van verontreinigd baggerspecie. In dit rapport is de verspreiding van stikstof tijdens storten van baggerspecie in open putdepots beschreven. Hierbij is gekozen voor een concentrateen effectbenadering. De belangrijkste conclusie is dat de samenstelling en textuur van de baggerspecie slechts in beperkte mate verantwoordelijk is voor een eventuele extra belasting van het oppervlaktewater met stikstof. Daarnaast beinvloeden voornamelijk locatiespecifieke omstandigheden de verspreiding van stikstof en de ecotoxicologische effecten gerelateerd aan het vrijkomen van stikstof in het oppervlaktewater. In stromende systemen zijn door het hoge uitwisselingsdebiet effecten van het storten van baggerspecie op de stikstofhuishouding in het oppervlaktewater in het algemeen verwaarloosbaar. Hoewel de totale vracht aan stikstof naar het oppervlaktewater gelijk blijft, lijkt storten van baggerspecie in een tijdelijk gestratificeerde put met betrekking tot het vrijkomen van stikstof voordelen te hebben ten opzichte van storten in een niet-gestratificeerde put met vergelijkbare dimensies. Indien de put wordt gebouwd in een stagnant, en mogelijk eutrofieringsgevoelig, niet-gestratificeerd watersysteem of wanneer rond de put een damwand is gebouwd, waardoor een klein uitwisselingsdebiet wordt gecreeerd tussen de put en het omliggende stromende oppervlaktewater (vergelijkbaar met een halfopen putdepot), zijn in en nabij de put gedurende de zomermaanden zowel acute toxiciteit als gevolg van het vrijkomen van ammoniak als effecten op eutrofiering mogelijk. Bij een dergelijke inrichting is er een permanente belasting van het oppervlaktewater met stikstof (dergelijke risico's worden echter niet verwacht wanneer wordt gestort in een tijdelijk gestratificeerde put, omdat omslag van een gestratificeerde situatie naar een gemengde put plaatsvindt in het najaar). Naast de intrinsieke eigenschappen van de baggerspecie is dus ook de keuze en inrichting van de locatie, evenals de keuze van de storttechniek, zeer belangrijk bij de beoordeling of storten van baggerspecie in een open putdepot kan resulteren in risico's met betrekking tot het vrijkomen van stikstof.
Storten van baggerspecie in open putdepots (fase 2)
Storten van baggerspecie in open putdepots (fase 2)
1 Inleiding
Het storten van baggerspecie in open putdepots in uiterwaarden, rivierbeddingen, meren, kanalen en havens wordt sinds enkele jaren beleidsmatig gezien als een mogelijkheid voor het milieuvriendelijk opbergen van verontreinigd baggerspecie. Dit wordt bijvoorbeeld beschreven in de Vierde Nota waterhuishouding (1998). Voor de bouw van dergelijke depots wordt gedacht aan verschillende boven regionale locaties in de Rijkswateren (bijv. Hollandsch Diep, Umeer, Kaliwaal en Molengreend). De randvoorwaarden voor de bouw van dergelijke open putdepots zijn afhankelijk van locatiespecifieke omstandigheden en mogelijkheden ter beperking van emissie richting oppervlaktewater (Absil & Bakker, 1999). In de meeste rapportages over berging van baggerspecie in depots wordt aandacht besteed aan de ecotoxicologische en verspreidingsrisico's met betrekking tot zwevend stof, zware metalen en organische microverontreinigingen. Op dit moment heeft het bestaande modelinstrumentarium voor de inschatting van emissies bij het storten van baggerspecie in open putdepots (WESTSIDE) de mogelijkheid de verspreiding van zware metalen en organische microverontreinigingen te kwantificeren. In een aantal studies wordt echter ook de potentiele bei'nvloeding van stikstofconcentraties in het oppervlaktewater door storten van baggerspecie benadrukt (o.a. Hartnack & Wesseling, 1990; Hartnack, 1994; Hartnack etal., 1996). De ecotoxicologische risico's die samenhangen met verhoogde concentraties aan stikstof in oppervlaktewater zijn voornamelijk gerelateerd aan eutrofiering en acute toxiciteit ten gevolge van het vrijkomen van ammoniak. In de kennisinventarisatie die is uitgevoerd in het kader van de eerste fase van het project storten van baggerspecie in open putdepots (Heijdt, van der etal., 1999), is benadrukt dat aanzienlijke verschillen worden gemeten in concentraties aan ammonium in waterbodems (poriewater) en concentraties in oppervlaktewater. Dit verschil wordt toegeschreven aan de vorming van ammonium bij anaerobe afbraak van organisch stof in de waterbodem (zie Berg, van den & Loch, 1993). Daarom wordt verwacht dat het uittreden van ammonium tijdens storten van baggerspecie tot aanzienlijke verhoging van de stikstof concentratie in het oppervlaktewater kan leiden. Mogelijk kan het vrijkomen van stikstof onder bepaalde condities van storten van baggerspecie zelfs een belangrijke rol spelen in de beoordeling, zoals reeds eerder door Hartnack (1994) is opgemerkt. Slechts in een beperkt aantal studies is het vrijkomen en de verspreiding van stikstof tijdens storten van baggerspecie ook daadwerkelijk in beeld gebracht. Zoals is aangetoond in de studie in de Averijhaven (zie Gerrits, 1999) kan vrijkomen van stikstof in gesloten depots inderdaad een probleem vormen voor de kwaliteit van het retourwater. In de Averijhaven was echter geen sprake van een bovenstaande waterkolom, maar was alle water in het depot afkomstig uit de gestorte specie en regenwater, waardoor verdunning minimaal is. Voor open putdepots is het effect van vrijkomen van stikstof nog niet duidelijk in beeld gebracht.
Storten van baggerspecie in open putdepots (fase 2)
Het doel van dit rapport is het inzichtelijk maken van de verhoogde stikstofbelasting ten gevolge van het storten van baggerspecie in open putdepots in de Rijkswateren. Hierbij spelen de volgende vragen een belangrijke rol: welke hoeveelheid stikstof zal vrijkomen tijdens storten van verontreinigde specie in open putdepots; welke relatie is er met het gehalte aan stikstof in de baggerspecie; welke processen in baggerspecie en oppervlaktewater spelen een belangrijke rol; welke effecten zijn eventueel in het watersysteem te verwachten; en welke mogelijkheden zijn er om deze effecten te beperken. In dit rapport zal worden besproken in hoeverre het mogelijk is op basis van de samenstelling en textuur van de baggerspecie (stofspecifieke risicobenadering) en locale factoren te beoordelen of storten van baggerspecie in open putdepots een mogelijke milieuvriendelijke optie is. Hiertoe bevat dit rapport een beschrijving van de stikstofhuishouding in natuurlijke watersystemen en de beinvloeding hiervan tijdens het storten van baggerspecie in open putdepots. Met een aantal scenarioberekeningen wordt vervolgens aangegeven welke extra belasting van stikstof in het oppervlaktewater kan worden verwacht bij het storten van baggerspecie in open en halfopen putdepots en in welke gevallen een locatiespecifieke studie noodzakelijk kan zijn om tot een goede inschatting te komen van eventuele risico's. Concluderend zal worden aangegeven welke parameters de gevoeligheid voor verspreiding van stikstof het meest beinvloeden.
Storten van baggerspecie in open putdepots (fase 2)
10
2 Stikstofhuishouding in natuurlijke watersystemen
2.1 Inleiding Het in beeld brengen van het vrijkomen van stikstof tijdens storten is van belang, omdat de ecologische samenstelling van oppervlaktewater sterk kan worden bei'nvloed door stikstofgerelateerde processen, zoals eutrofiering (algenbloei) en vrijkomen van ammoniak. Om tot een goede inschatting te komen van de stikstofgerelateerde potentiele en actuele risico's bij het storten van baggerspecie in open putdepots is het in eerste instantie noodzakelijk een beeld te hebben van het gedrag van stikstof onder natuurlijke omstandigheden. In dit hoofdstuk worden achtereenvolgens de stikstofhuishouding in oppervlaktewater, de relatie met processen in de waterbodem en de specifieke processen die plaatsvinden in gestratificeerde watersystemen, beschreven (voor een uitgebreide beschrijving van het gedrag van stikstof in natuurlijke watersystemen wordt verwezen naar bijvoorbeeld Wetzel (1975)). 2.2 Oppervlaktewater De stikstofcyclus in natuurlijke watersystemen is zeer complex. Door de grote variatie in dynamiek kunnen ruimtelijk grote verschillen optreden, waardoor inschatting van locale condities en intensiteiten van processen noodzakelijk is. Het gedrag van stikstof in oppervlaktewater kan worden gemodelleerd door rekening te houden met de processen die omzetting tussen de verschillende compartimenten en stikstofvormen beschnjven. De belangrijkste opgeloste stikstofspecies zijn nitraat (NO s ) en ammonium (NH«*). Ammonium komt voornamelijk in het oppervlaktewater terecht via diffuse bronnen door uitspoeling uit landbouwgronden en nalevering uit de liggende waterbodem. In zuurstofhoudend oppervlaktewater vindt vervolgens omzetting van ammonium naar nitraat (nitrificatie) plaats via redox-reacties 1 en 2, waarvan de eerste de snelheidsbepalende is: 2NH,* + 3 0 2 -> 2N0 2 + 4H* + 2H 2 0 en 2 N 0 2 + 0 2 -> 2N0 3
(1) (2)
Het zuurstofverbruik voor nitrificatie is relatief hoog: per g NH4-N wordt 4,57 g 0 2 verbruikt. Bij lage zuurstofconcentraties treedt remming op, waardoor ophoping van nitriet ( N 0 2 ) kan plaatsvinden in het oppervlaktewater. Door het optreden van nitrificatie zijn ammonium concentraties in zuurstofhoudend oppervlaktewater in het algemeen zeer laag. Snelheden voor nitrificatie liggen voor de Rijn in de orde van 0,1-1,5 mg N H / - N / I per dag (Admiraal & Botermans, 1989).
Storten van baggerspecie in open putdepots (fase 2)
11
Door verschil in aanvoer en temperatuursafhankelijkheid van de meeste stikstof-gerelateerde processen in het oppervlaktewater (bij verhoging van de temperatuur met 10 °C verdubbelt bijvoorbeeld de nitrificatiesnelheid) zijn opgeloste nitraat- en ammoniumconcentraties in oppervlaktewater niet constant. De seizoensvariatie in nitraat en ammonium in het oppervlaktewater is weergegeven in figuur 2.1 voor een stromend systeem (de Rijn) en een stagnant (stilstaand) systeem (Usselmeer). Als gevolg van beperkte nitrificatie (bij een temperatuur lager dan 5 °C zijn nitrificatie- en denitrificatiesnelheden verwaarloosbaar) worden de hoogste concentraties aan ammonium gemeten in het winterhalfjaar.
Figuur 2.1 Concentraties nitraat + nitriet (a) en ammonium (b) in de Rijn (bij Lobith; dichte rode symbolen) en in het Usselmeer (locatie Vrouwezand; open biauwe symbolen) gedurende de periode 1995-1997 (data afkomstig uit MWTL).
••-VV' : ' ; W ' 010195
010197
110196 Datum
Storten van baggerspecie in open putdepots (fase 2)
'2
01019f
(b)
In eutrofieringsgevoelige watersystemen kunnen grote verschillen in concentraties aan stikstof tussen zomer en winter worden aangetroffen (zie Zwolsman, 1999'). Dergelijke verschillen (zie bijv. figuur 2.1) kunnen worden verklaard door gedeeltelijke of volledige assimilatie van opgelost stikstof door algen. Volledige assimilatie leidt derhalve tot stikstof-limitatie. Algen kunnen zowel ammonium als nitraat assimileren. Bij hoge ammoniumconcentraties wordt ammonium preferent opgenomen. Een sterke toename in algengroei wordt verwacht bij een stikstofconcentratie groter dan 2,2 mg N/l (MTR voor oppervlaktewater gedurende de zomermaanden) in eutrofieringsgevoelige (stagnante) wateren (voor een definitie van eutrofieringsgevoelige wateren wordt verwezen naar Huisman etal. (1999)). Omdat een dergelijke opgeloste stikstofconcentratie in veel watersystemen wordt overschreden, is een terugdringing van onder andere de stikstofbelasting een belangrijk doel bij de bestrijding van de gevolgen van eutrofiering (zie Van der Molen, 1996). Boers ef al. (1993) stellen zelfs dat het mogelijk noodzakelijk is om een lagere ecologische streefwaarde (1,0 mg N/l) vast te stellen om een gezond en soortenrijk watersysteem te bereiken. Er moet bovendien rekening mee worden gehouden dat als gevolg van een algengroei de pH van het oppervlaktewater gedurende de zomermaanden kan toenemen. In het Usselmeer worden gedurende de zomer pH-waarden hoger dan 9 gemeten, terwijl gedurende de winter de pH kan dalen tot pH 8 (zie Zwolsman, 1996). De pH voor stromende watersystemen ligt veelal lager (de gemiddelde pH voor de Rijn is 7,75). Omdat zich een evenwicht instelt tussen ammonium en ammoniak leidt een toename in ammonium op korte termijn tot een verhoogde concentratie aan opgelost ammoniak (NHj). Ammoniak is acuut toxisch voor o.a. vissen (de MTR voor ammoniak in oppervlaktewater ligt op 20 u.g NH 3 -N/I, maar acute risico's treden waarschijnlijk pas op bij concentraties groter dan 3 a 4 mg NH,-N/I). De verhouding NH 4 */NH 3 in oplossing neemt af bij stijgende temperatuur en pH (zie figuur 2.2; de relatieve hoeveelheid ammoniak in de waterfase neemt dan dus toe). In tabel 2.1 is de verhouding NH„7NH 3 in oplossing weergegeven voor de winterperiode (10 °C) en de zomerperiode (18 °C) en een drietal pHwaarden. Bovendien is de ammoniumconcentratie weergegeven waarbij de MTR voor ammoniak wordt overschreden onder deze omstandigheden. Zowel uit de figuren als uit tabel 2.1 blijkt dat zowel de verhouding N H / / N H 3 in oplossing als de kritieke ammoniumconcentratie sterk afhankelijk zijn van de pH en in mindere mate van de temperatuur.
1000
Figuur 2.2 Relatie tussen de verhouding NH 4 7NH 3 en pH (a) en temperatuur (b). De figuren zijn gebaseerd op thermodynamische constanten afkomstig uit Smith & Martell (1976).
I
SCO
2»0
!
1500
i
Storten van baggerspecie in open putdepots (fase 2)
PSd
i-itrc
....
1
NO
i
XD p. W O
'
a pH
13
5
10
(a)
10
19 Temperatuuf C O
20
n
(b)
Tabel 2.1
Overzicht van de ammoniumconcentr atie waarbij de MTR voor ammoniak wordt overschreden gedurende de winter- en zomerperiode bij verschillende pHwaarden (de verhoudingen NH 4 7NH 3 in oplossing zijn afgeleid uit figuur 2.1).
temperatuur
pH
NH//NH,
CO 10 10 10 18 18 18
7 8 9 7 8 9
500 50 5 300 30 3
NH4* (mg NH..-N/I) 10 1 0,1 6 0,6 0,06
2.3 Liggende waterbodems De liggende waterbodem speelt een belangrijke rol bij de vastlegging en het vrijkomen van stikstof in natuurlijke watersystemen. Opgeloste concentraties aan nitraat en ammonium in waterbodems (in feite het poriewater) worden bepaald door een combinatie van voornamelijk microbiologisch gereguleerde (bio)geochemische processen (Berner, 1980). Omdat de bulk aan gebaggerd sediment in de Nederlandse binnenwateren wordt gekarakteriseerd door anaerobe afbraak van organisch stof (Berg, van den & Loch, 1993), is ammonium verreweg de belangrijkste opgeloste stikstofspecies in baggerspecie: Door het hoge gehalte aan afbreekbaar organisch stof in zwevend stof en recent afgezet sediment, zijn oxidatieve processen beperkt tot de bovenste millimeters van het sediment, waardoor nitraat slechts wordt aangetroffen in de bovenste laag van de waterbodem.
Figuur 2.3 Karakteristieke ammonium- en nitraatprofielen in poriewater, zoals die zijn gemeten in de waterbodem in de Brabantse Biesbosch (Berg, van den, 1998). De profielen zijn gemodelleerd met het multicomponent reactietransport model STEADYSED (Cappellen, van & Wang, 1996).
Storten van baggerspecie in open putdepots (fase 2)
M
In figuur 2.3 zijn karakteristieke nitraat- en ammoniumprofielen in poriewater in een zoete Nederlandse waterbodem weergegeven. Het concentratieprofiel voor nitraat in het poriewater wordt onder andere bepaald door diffusie vanuit het bovenstaande water, denitrificatie (zie vergelijking 4) en nitrificatie. Zoals weergegeven in de vergelijkingen 1 en 2 is nitrificatie alleen mogelijk onder aerobe omstandigheden (aanwezigheid van zuurstof). Denitrificatie vindt alleen plaats onder anaerobe omstandigheden (0 2 < 0,2 mg/l). In de meeste waterbodems is nitrificatie een belangrijke bron voor nitraat dat wordt gebruikt voor denitrificatie (Seitzinger, 1988; Van Luijn, 1997). Denitrificatie kan worden beschouwd als een belangrijke "verliesroute" voor stikstof in watersystemen (Nowicki, 1994), omdat hierbij gasvormige stikstofverbindingen (o.a. N2 en N 2 0) worden gevormd. Het concentratieprofiel van ammonium in het poriewater wordt bepaald door productie van ammonium bij anaerobe afbraak van organisch stof (vergelijkingen 5 t / m 8), uitwisseling (voornamelijk met Ca2*) en verticale diffusie in de waterbodem, gecombineerd met nitrificatie in de toplaag. Zoals wordt weergegeven in onderstaande vergelijkingen voor afbraak van organisch stof, speelt de samenstelling (o.a. de C/N-verhouding) en afbreekbaarheid van het aanwezige organisch stof een belangrijke rol bij de intensiteit van ammoniumvorming. Door het verschil in ammoniumconcentratie tussen waterbodem en oppervlaktewater kan verticale diffusie plaatsvinden naar het bovenstaande water (nalevering van ammonium).
Vergelijkingen voor afbraak van organische stof Zuurstofreductie: (CH20),(NH3),(H3P04)2 + (x+2y)Q1 + (y+2z)HC03 -> (x+y+2z)C02 + y N 0 3 + zHPO„2 + (x+2y+2z)H20
(3)
Denitrificatie: ( C H J O W N H ^ H J P O J , +
-> ((2x+4y)/5)H2 +
CO2 +
(4)
Mn(IV) - reductie: (CH20),(NH3),,(H3P04), + 2xMn0 2 + (3x+y-2z)C02 + (x+y-2z)H 2 0 -> 2xMn 2 * + f4x+y-2z;HC0 3 + y N H / + zHP0 4 2
(5)
Fe(lll) - reductie: (CH 2 0),(NH 3 ),,(H 3 P0 4 )I + 4xFe(OH)3 + (7x+y-2x)C03 -> 4x?f?* + (8x+y-2z)HC03 + yNH 4 * + zHP0 4 2 + (3x-y+2z)H20
(6)
Sulfaatreductie: (CH20),(NH3),(H3PO«), + fx/2;S0 4 2 + (y-2z)CO} + (y-2z)H 2 0 -»fx/2;H 2 S + (x+y-2z)HCO} + yNH 4 * +zHP0 4 2
(7)
Methanogenese: ( C H J O M N H ^ H J P O J , + (y-2z;H 2 0
- • (x/2)CH, + ('Cx-2y+4z;/2;C02 + fy-2z;HC0 3 + yNH 4 * + zHP0 4 2 ' Organisch stof is weergegeven als
Storten van baggerspecie in open putdepots (fase 2)
15
(CH^/NHJ/HfOJ,
(8)
2.4 Gestratificeerde watersystemen In Nederland kan zowel thermische stratificatie als zoet-zoutstratificatie optreden. Thermische stratificatie kan optreden in diepe meren en putten gedurende de zomer- (warme stratificatie) en wintermaanden (koude stratificatie) als gevolg van het relatief grote temperatuurverschil tussen de bovenste en de onderste waterlaag. Door afkoeling van de bovenlaag en menging als gevolg van verhoogde stroming door de invloed van wind kan in het najaar de thermische stratificatie worden opgeheven; dit proces kan afhankelijk van de geometrie van het watersysteem (o.a. diepte van de put, stroming) en locale weersomstandigheden instantaan plaatsvinden (bij snelle afname in temperatuur), maar ook een langere periode (bijv. enkele weken) in beslag nemen. In Nederland vindt door verhoogde turbulentie volledige menging op van het najaar tot het voorjaar. In figuur 2.4 is zichtbaar gemaakt dat thermische stratificatie leidt tot de vorming van een relatief warme toplaag (epilimnion) en een koude onderiaag (hypolimnion). Deze worden gescheiden door een spronglaag (het metalimnion), die in Nederland op ongeveer 10 m waterdiepte ligt. Derhalve kan alleen in putten dieper dan 10 m thermische stratificatie optreden. In diepe watersystemen die worden gekarakteriseerd door een zoet-zoutstratificatie kan stratificatie mogelijk veel moeilijker worden opgeheven. Immers, het zwaardere zoute water zal ook bij hogere stroomsnelheden permanent nabij de bodem blijven. Het optreden van stratificatie wordt uitgebreid beschreven (en gemodelleerd) door bijvoorbeeld Imboden & Wuest (1995) en Martin & McCutcheon (1999). Stratificatie leidt tot een beperkte uitwisseling van zwevend materiaal en opgeloste bestanddelen tussen het epilimnion en het hypolimnion. In gestratificeerde watersystemen (meren en putten) die worden gekarakteriseerd door een relatief hoge toevoer van organisch stof (eutrofe systemen), vindt daardoor een hoog verbruik van zuurstof plaats. Wanneer door het hoge zuurstofverbruik het hypolimnion zuurstofloos is geworden, kan accumulatie van ammonium plaatsvinden in het hypolimnion (zie bijv. Balistrieri etal., 1992). In figuur 2.4 is de verticale distributie van ammonium en nitraat in de waterkolom weergegeven na instellen van stratificatie. Duidelijk is te zien dat in de onderste (zuurstofloze) waterlaag de concentratie nitraat sterk afneemt (gerelateerd aan denitrificatie), terwijl gelijktijdig de concentratie ammonium toeneemt. Vorming van ammonium wordt in natuurlijke watersystemen (bijvoorbeeld meren) voornamelijk gekoppeld aan nalevering uit de liggende waterbodem. Immers, door het verdwijnen van de geoxideerde toplaag van het sediment door zuurstofloosheid in het hypolimnion neemt de snelheid van nitrificatie in de toplaag sterk af. In tijdelijk gestratificeerde watersystemen wordt deze zuurstofloosheid van de diepere laag bij menging weer opgeheven.
Figuur 2.4 Verticale distributie van ammonium (NH4*), nitraat ( N O , ) , zuurstof (0 2 ) en temperatuur (9) in gestratificeerde watersystemen (relatieve schaal; uit Wetzel, 1975).
n
: ! i
U
I
\
0UJ
I I I
•/' !°2
NOy
4
I K 2
i ^ > ' '
Q
I
"/' /
/
\ \
T
\
M °r r,^
Storten van baggerspecie in open putdepots (fase 2)
16
,
3 Vrijkomen van stikstof tijdens storten
3.1 Inleiding De vracht aan stikstof die vrijkomt in het oppervlaktewater tijdens de stortfase (de emissie) kan worden berekend uitgaande van de hoeveelheid baggerspecie die wordt gestort en de hoeveelheid retourwater (voornamelijk sedimentatiewater en consolidatiewater) die daarbij uittreedt. De concentraties aan ammonium in het sedimentatie- en het consolidatiewater zijn afhankelijk van de concentratie in het poriewater van de in situ baggerspecie (bepaald door de samenstelling en afbreekbaarheid van organisch stof) en de hoeveelheid water die wordt bijgemengd tijdens baggeren (afhankelijk van de eigenschappen van de baggerspecie en de verwerkingsmethode).
3.2 Vrijkomen van stikstof tijdens de stortfase 3.2.1 Sedimentatiewater De hoeveelheid sedimentatiewater die vrijkomt tijdens het storten van baggerspecie wordt bepaald door de hoeveelheid baggerspecie die per tijdseenheid wordt gestort, de uitleveringsfactor en de sedimentatiefactor. De uitleveringsfactor geeft het verschil aan tussen het volume van de specie voor en na baggeren en is een functie van de fysische eigenschappen van het materiaal (o.a. korrelgrootte en mate van binding) en de wijze van baggeren. De uitleveringsfactor is een in het geval geen water wordt bijgemengd tijdens baggeren (de specie heeft een massa die gelijk is aan de in situ specie). Dit zal vaak het geval zijn wanneer zand of ongeconsolideerd slib met grijpers wordt gebaggerd. Wanneer wel water wordt bijgemengd tijdens baggeren (bijvoorbeeld bij baggeren van geconsolideerd slib met een zuiger), kan de uitleveringsfactor sterk toenemen. In een studie naar storten van specie in diepe putten in Noord-Holland (o.a. Umeer) is aangenomen dat slibrijk materiaal een uitleveringsfactor van twee heeft. Bij een uitleveringsfactor van twee daalt de soortelijke massa van de specie van 1.400 kg m 3 (bij een porositeit van 0,7 en een droge stofdichtheid van 2.400 kg m 3 ) naar 1.200 kg m'3.
Storten van baggerspecie in open putdepots (fase 2)
17
De sedimentatiefactor is een maat voor het verschil tussen het totale gestorte volume en het volume na het vrijkomen van het sedimentatiewater (er komt alleen sedimentatiewater vrij indien specie wordt gestort met een lagere dichtheid dan de sedimentatiedichtheid). Derhalve speelt het cohesieve karakter van de gestorte specie (o.a de aanwezigheid van aggregaten) een belangrijke rol bij de afleiding van de sedimentatiefactor (Heijdt, van der et al., 1999). De wijze van storten speelt een belangrijke rol bij de afleiding van de in situ sedimentatiefactor. Om ongewenst grote horizontale verspreiding van stortverlies te voorkomen wordt bij storten van baggerspecie in stromende systemen vaak gekozen voor hydraulisch storten met een stortpijp en diffusor.
3.3 Consolidatiewater Tijdens en na de stortfase (nadat de specie is gesedimenteerd) vindt consolidatie plaats. Onder invloed van verhoogde waterspanning komt consolidatiewater vrij en vindt verspreiding van poriewaterbestanddelen plaats via advectief transport. Door toevoer van baggerspecie zal de consolidatiesnelheid tijdens de stortperiode licht toenemen. Na het beeindigen van de stort neemt de mate van consolidatie snel af, maar zal nog langere tijd (zeker 30 jaar) plaatsvinden (dit wordt niet meegenomen in de berekeningen). Indien de specie met grijpers is gebaggerd en vervolgens met onderlossers wordt gestort is de massa van de gestorte specie vrijwel gelijk aan die van de in situ specie. De hoeveelheid consolidatie is dan minimaal.
3.4 Berekening emissie van stikstof tijdens de stortfase Voor het schatten van de totale vracht aan stikstof die vrijkomt tijdens storten wordt uitgegaan van een stikstof- (eigenlijk ammonium-)concentratie in gebaggerd materiaal die vergelijkbaar is met die in anoxisch sediment. Immers de bulk aan gebaggerd materiaal is anoxisch, waardoor de gemiddelde ammoniumconcentratie in het poriewater relatief hoog is ten opzichte van die in zuurstofhoudend oppervlaktewater/sediment. In waterbodems in sedimentatiegebieden van de Rijn (Ketelmeer) en Maas (Brabantse Biesbosch) worden ammoniumconcentraties gemeten in de orde-grootte 10-20 mg NH 4 -N/I (Paalman, 1997; Berg, van den, 1998). De hoogst gemeten concentraties lagen in deze studies op ongeveer 30 mg NH 4 -N/I, terwijl concentraties in zuurstofhoudend oppervlaktewater/sediment zelden hoger zijn dan 1 mg NH 4 -N/I. In organisch-rijk havenslib kunnen hogere concentraties ammonium voorkomen (Zwolsman, 1999b). De gemeten nitraatconcentraties in (anoxisch) poriewater zijn in al deze studies verwaarloosbaar klein. De samenstelling van het sedimentatie- en consolidatiewater wordt bepaald door die van het poriewater en die van het bijgemengde (oppervlakte-)water. Bijmenging met water tijdens baggeren, bepaald door de uitleveringfactor, leidt per definitie tot een verlaging van de stikstof- (ammonium-Jconcentratie in de specie, en derhalve van die in het sedimentatiewater en consolidatiewater (er vindt verdunning plaats van het poriewater). Met een eenvoudige berekening kan een schatting worden gemaakt van concentraties in het sedimentatiewater. Dit is een bovengrens, omdat er ook rekening mee zou kunnen worden gehouden dat geen volledige menging van uitleveringswater en poriewater optreedt en tijdens het storten de kwaliteit van het sedimentatiewater voornamelijk wordt bepaald door die van het bijgemengde oppervlaktewater. Een dergelijk gedrag zou kunnen worden verklaard door de aanwezigheid van aggregaten (vlokken) in het slib die tijdens het storten snel naar de bodem zakken.
Storten van baggerspecie in open putdepots (fase 2)
18
Voor een eerste schatting van de hoeveelheid stikstof die vrijkomt tijdens storten van baggerspecie in open putten wordt aangenomen dat de concentraties aan ammonium in het consolidatiewater vergelijkbaar zijn aan die in het sedimentatiewater (volledige menging van poriewater en toegevoegd water tijdens baggeren). Wanneer wordt uitgegaan van volledige menging wordt bijvoorbeeld bij een in situ gebaggerde specie met een porositeit van 0,7 en een stikstofconcentratie in het poriewater van 30 mg N/l (maximale concentratie in saneringsspecie uit de grote rivieren) bij een uitleveringsfactor van 2 de concentratie stikstof in het sedimentatie- en consolidatiewater ongeveer 12 mg N/l. Uitgaande van de hoeveelheid te storten specie kan dan de maximale vracht aan vrijgekomen ammonium in het oppervlaktewater worden berekend indien het sedimentatie- en consolidatiedebiet bekend is. Verhoogde nalevering vanuit het slib vindt voornamelijk plaats door opwerveling van materiaal tijdens storten (dit valt onder de sedimentatieflux). Diffusieve nalevering wordt bij storten van baggerspecie verwaarloosbaar geacht ten opzichte van de hoeveelheid stikstof die vrijkomt bij consolidatie (dit in tegenstelling tot natuurlijke watersystemen, waarin de consolidatieflux veel lager is door lagere sedimentatiesnelheden).
Storten van baggerspecie in open putdepots (fase 2)
19
Storten van baggerspecie in open putdepots (fase 2)
20
4 Scenarioberekeningen voor inschatting stikstof belasting ten gevolge van storten
4.1 Inleiding Voor een schatting van de vracht aan stikstof die vrijkomt bij storten van baggerspecie in open putten en de effecten hiervan op concentraties aan stikstof in het oppervlaktewater is gekozen voor een tweetal voorbeeldsituaties, P1 en P2 (de dimensies hiervan zijn vergelijkbaar met die in deelproject 2 "verspreiding van verontreinigingen"). P1 betreft een put in ruim open water. Te denken valt hierbij bijvoorbeeld aan een depot in het Hollandsch Diep (hoog effectief uitwisselingsdebiet) of het Ijsselmeer (laag effectief uitwisselingsdebiet). Gedurende een deel van de vulfase kan in deze put (tijdelijk) stratificatie van de bovenstaande waterschijf optreden. Voor de dimensies van P1 worden de volgende waarden aangehouden: 10*10* m3 45 m 10 m 10 jaar
Bergingsvolume Begindiepte Einddiepte Vulperiode
P2 is een halfopen put, bijv. een zand- of grindwinput in een uiterwaard, die door een kleine opening in verbinding staat met de rivier. Denk hierbij bijvoorbeeld aan de Kaliwaal (langs de Waal) of de Molengreend (langs de Maas). Voor de dimensies van P2 worden de volgende waarden aangehouden: 5 * 7 06 m3 25 m 5m 10 jaar
Bergingsvolume Begindiepte Einddiepte Vulperiode
Voor de twee voorbeeldputten wordt ervan uitgegaan dat de baggerspecie hydraulisch wordt gestort en het volume water dat vrijkomt tijdens de stortfase ongeveer 50% is van de totale hoeveelheid gestorte specie (vergelijkbare getallen zijn gebruikt in een studie naar open putten in Noord-Holland). Hiervan bedraagt de hoeveelheid sedimentatiewater 65-70% en de hoeveelheid consolidatiewater 30-35%. De totale bergingscapaciteit van P1 is dan 20"10 6 m 3 gestorte specie (komt overeen met 10*10 6 m3 in situ gebaggerde specie en 10*10 6 m3 sedimentatie en consolidatiewater) en 10*10 6 m 3 gestorte specie voor P2. Hieruit kan eenvoudig een schatting worden gemaakt van de maximale vracht (opgelost) stikstof naar het oppervlaktewater. Voor P1 is de emissievracht bijgevolg 12.000 kg N per jaar en voor P2 6.000 kg N per jaar (in het geval de stortfase 10 jaar duurt) uitgaande van een stikstofconcentratie in het sedimentatie- en consolidatiewater van 12 mg N/l. Deze vracht wordt hoger naarmate de stikstofconcentratie in het poriewater hoger is. In het algemeen kan worden gesteld dat in riviersystemen de totale vracht aan stikstof-emissie verwaarloosbaar klein zal zijn ten opzichte van de totale vracht aan stikstof in het oppervlaktewater.
Storten van baggerspecie in open putdepots (fase 2)
21
4.2 Niet-gestratificeerde open putdepots 4.2.1 Modelbeschrijving Voor de inschatting van het temporele verloop van de stikstofconcentraties en -vrachten in een open put is een model opgesteld dat rekening houdt met het debiet van het inkomende water (Q(in)), het stortvolume (Q(stort)), de hoeveelheid sedimentatie- en consolidatiewater en het resulterende uitgaande debiet naar het oppervlaktewater (Q(uit)), evenals met de hoeveelheid stikstof die bij de verschillende uitwisselingsprocessen vrijkomt (zie figuur 4.1). Bij een constant volume van de put wordt het uitgaande debiet mede bepaald door de hoeveelheid sedimentatie- en consolidatiewater die per tijdseenheid vrijkomt tijdens het storten (het retourwater); tijdens de stortfase is het uitgaande debiet dus per definitie groter dan het inkomende debiet. Uitgaande van de vracht aan stikstof die vrijkomt met het sedimentatie- en consolidatiewater kan voor een gegeven debiet een schatting worden gemaakt van de extra belasting van het watersysteem met stikstof.
Q(stort)
Figuur 4.1 Model voor volledig gemengde put. Q(in)
Q(uit)
>
4.2.2 Voorbeeldberekeningen De emissie van de berekende hoeveelheid ammonium in P1 (ongeveer 12.000 kg N per jaar bij een ammoniumconcentratie van 30 mg N/l in in situ poriewater en 12 mg N/l in sedimentatie- en consolidatiewater) levert door de beperkte verblijftijd (ongeveer 1 dag) bij een gemiddeld uitwisselingsdebiet van enkele honderden m3/s (typisch voor bijvoorbeeld het Hollandsch Diep) slechts een verhoging van de totale stikstofconcentratie op van hooguit enkele ug N/l indien instantane menging wordt aangenomen. Een dergelijke verhoging is verwaarloosbaar (factor 1.000 lager) ten opzichte van de aanwezige stikstofconcentratie in oppervlaktewater. Bovendien is de invloed van vrijkomen van stikstof bij afbraak van organisch stof in het zwevende materiaal in de waterkolom door de beperkte verblijftijd in de waterkolom gering ten opzichte van de invloed van de fysische processen tijdens storten.
Storten van baggerspecie in open putdepots (fase 2)
22
Wanneer het uitwisselingsdebiet sterk wordt verlaagd, kan de extra belasting door storten van baggerspecie mogelijk wel resulteren in risico's met betrekking tot stikstof in het oppervlaktewater. In een min of meer stagnant systeem met een debiet van bijvoorbeeld 1 m3/s kan een relatief grote toename van concentraties aan stikstof in en nabij de put worden verwacht (orde-grootte 0,4 mg N/l voor P1; zie figuur 4.3). Hierbij kan worden gedacht aan bijvoorbeeld een put in het Usselmeer (stagnant systeem met een lange verblijftijd) of een depot in een stromend systeem omgeven door een (tijdelijke) damwand met een klein uitwisselingsdebiet. Gedurende de zomermaanden (hoge pH, hoge temperatuur) kan een dergelijke toename in totaal-stikstof in eutrofieringsgevoelige wateren (gekarakteriseerd door een lage stikstofconcentratie) een significante invloed hebben op de algengroei. Daarnaast kan in eutrofieringsgevoelige wateren gedurende de zomermaanden een dergelijke extra belasting met stikstof (ammonium) tijdelijk leiden tot een lichte overschrijding van de MTR voor ammoniak (bij een pH 9 en 18 °C is de berekende maximale ammoniakconcentratie in oplossing 30 pg N/l). De concentratie ammonium in het oppervlaktewater zal overigens slechts in een zeer beperkte mate worden bei'nvloed: Door de aanwezigheid van nitrificerende bacterien vindt een snelle omzetting van ammonium in nitraat plaats na uittreding. In het geval van een halfopen put kan worden gediscussieerd over de vraag of het uitwisselingsdebiet (onder invloed van bijvoorbeeld dynamische peilverschillen tussen put en rivier) direct vertaald kan worden in een effectief debiet die doorspoeling van het in de put aanwezige water en volledige menging tot gevolg heeft. Het is namelijk in de praktijk onduidelijk in hoeverre het instromende water zich mengt met water in de put. Dit hangt samen met de korte periode waarop water in- en uitstroomt. Indien geen menging optreedt, kan de hoeveelheid opgelost stikstof in de put sterk oplopen. Wanneer voor P2 een ondergrens voor het effectieve debiet 1 m3/s wordt aangehouden, is de maximale verhoging van de stikstofconcentratie in het bovenstaande water ongeveer 0,2 mg N/l (zie figuur 4.4). De bijdrage van een dergelijke stikstofvracht aan de stromende rivier (Maas, Waal) door uitstromend water kan echter als verwaarloosbaar worden beschouwd. 4.2.3 Discussie Bij een open putdepot in een (snel) stromende rivier zal de concentratie stikstof in het oppervlaktewater niet significant toenemen indien geen stratificatie optreedt in de put (aangenomen dat menging snel plaatsvindt). Er is alleen een (locale) verhoging van de concentratie stikstof in het oppervlaktewater mogelijk in het geval een relatief groot depot wordt gebouwd in een stilstaand of licht stromend watersysteem. Beinvloeding van de concentraties stikstof in de put kan ook worden verwacht tijdens de stortfase, in het geval rond het depot een tijdelijke damwand (atol) wordt gebouwd, waardoor een beperkte doorstroming met oppervlaktewater plaatsvindt (deze situatie is vergelijkbaar met een halfopen put, die via een nauwe opening in verbinding staat met de rivier). Bij de beoordeling of baggerspecie gestort mag worden in open putdepots zou in eutrofieringsgevoelige wateren overigens rekening moeten worden gehouden met (seizoens-)variatie in pH, temperatuur en afvoer. Door de continue belasting van het oppervlaktewater (in en nabij de put) met stikstof kan in dergelijke stagnante watersystemen mogelijk lokaal de MTR-waarde voor ammoniak worden overschreden gedurende de zomermaanden en zijn mogelijk toxische effecten op bijvoorbeeld vissen te verwachten.
Storten van baggerspecie in open putdepots (fase 2)
23
4.3 Gestratificeerde open putdepots 4.3.1 Inleiding Een inschatting van het optreden van thermische stratificatie in open putdepots kan worden gemaakt op basis van het Orlob criterium (Hurley Octavio etal., 1977). Volgens dit criterium hangt het optreden van stratificatie onder andere af van het effectieve uitwisselingsdebiet en de dimensies (diepte en volume) van de put. In diepe stagnante systemen kan (gedurende een deel van het jaar) thermische stratificatie van de waterkolom worden verwacht, terwijl in open putten in stromende systemen mogelijk geen stratificatie optreedt, omdat door de hogere stroomsnelheden het water in de put gemengd blijft. Afhankelijk van de stortmethode kan een deel van de gestorte specie (verlies) en het sedimentatiewater terechtkomen in het hypolimnion. Bovendien komt na sedimentatie consolidatiewater vrij in deze laag. In het geval hydraulisch in het hypolimnion wordt gestort met een stortpijp en diffusor, fungeert de spronglaag als barriere voor verticale verspreiding van zwevend stof en opgeloste bestanddelen. Omdat vrijwel geen zuurstof wordt bijgemengd (alleen door diffusie aan het grensvlak), wordt bij storten van (anoxische) baggerspecie het in het hypolimnion aanwezige zuurstof snel verbruikt voor oxidatie van gereduceerde bestanddelen (bijv. metaalsulfiden), opgeloste gereduceerde bestanddelen (bijv. sulfide, ammonium, Fe2* en Mn2*) en afbraak van organisch stof. Na het instellen van de zuurstofloosheid neemt de concentratie nitraat af door denitrificatie, waarbij gasvormige stikstofverbindingen worden geproduceerd, terwijl door uittreding van sedimentatie- en consolidatiewater ophoping van ammonium plaatsvindt in het hypolimnion. De duur van stratificatie bepaalt derhalve voor een belangrijk deel hoeveel stikstof (ammonium) zich kan ophopen. Indien permanente stratificatie en zuurstofloosheid optreden, zoals te verwachten is in bijv. diepe putten in zoet-zout-gestratificeerde watersystemen, kan het opgebouwde stikstof (ammonium) alleen worden gemengd met de bovenstaande zuurstofhoudende waterkolom door verdringing van water ten gevolge van het storten van baggerspecie in de put. 4.3.2 Modelbeschrijving Voor de berekening van de hoeveelheid stikstof die vrijkomt tijdens storten van baggerspecie in (tijdelijk) gestratificeerde putten is het model voor volledig gemengde putten (figuur 4.1) uitgebreid met een module voor uitwisseling tijdens stratificatie (figuur 4.2). In het model wordt verondersteld dat zowel het volume van de put als het volume van het hypolimnion constant zijn. Daardoor vindt tijdens stratificatie door storten van specie verdringing van water uit het hypolimnion plaats (Q(stort,uit)) en wordt geleidelijk stikstof gemengd met het bovenstaande water (advectief transport). Er geldt dan dat de belasting van het bovenstaande water met stikstof zal toenemen in de tijd. Indien permanente stratificatie optreedt, hangt de totale vracht aan stikstof die uiteindelijk in de bovenste waterlaag terechtkomt en de daarbijbehorende concentraties sterk af van de diepte van stratificatie en de storthoogte.
Storten van baggerspecie in open putdepots (fase 2)
24
Figuur 4.2 Model voor een tijdelijk gestratificeerde put.
Q(stort) Qdn)
Q(stort,in) Q(uit) .
Q(in)
Epilimnion 1
Q(ujt) ' P
omslag Q(stort.uit)
Hypolimnior
4.3.3 Voorbeeldberekeningen In het geval baggerspecie wordt gestort met een stortpijp en diffusor komt het materiaal voor een belangrijk deel terecht in de onderste laag van de waterkolom (hypolimnion). Wanneer wordt uitgegaan van een stratificatiediepte van 10 m en stratificatie plaatsvindt gedurende een periode van 6 maanden kan zich in P1 maximaal 6.000 kg stikstof (nitraat en ammonium) ophopen in het hypolimnion. De hoeveelheid stikstof die zich verzamelt in het hypolimnion kan door de omslag in korte tijd worden gemengd met het oppervlaktewater, resulterend in een piekafvoer naar het oppervlaktewater. Bij een debiet van 100 m3/s kan dit gedurende de stortfase bij instantane menging (binnen een dag) tijdens de piekafvoer een extra bijdrage leveren van maximaal 0,4 mg N/l in de put (P1) en het uitstromende water. Wanneer rond de put bijvoorbeeld een vrijwel gesloten damwand wordt geplaatst of wanneer de open put wordt gebouwd in een stagnant watersysteem, kan de bijdrage sterk toenemen (bij een laag effectief uitwisselingsdebiet van bijvoorbeeld 1 m3/s en instantane menging is gedurende de stortfase de toename in stikstofconcentratie in P1 maximaal 1,2 mg N/l; zie figuur 4.3). De bijdrage van een dergelijke tijdelijke vracht aan die in het oppervlaktewater blijft echter verwaarloosbaar. Onder bepaalde omstandigheden (hoge pH en temperatuur) kan een dergelijke piekafvoer leiden tot een tijdelijke overschrijding van het MTR voor ammoniak. Echter, hierbij moet worden bedacht dat de omslag in het algemeen plaatsvindt in het najaar, wanneer de temperatuur en pH reeds gezakt zijn. Wanneer de opheffing van de stratificatie niet instantaan plaatsvindt, wordt de piekafvoer lager, en wordt de kans op acute toxiciteit voor vissen lager. Nadat de stratificatie opgeheven is, neemt de ammoniumconcentratie (snel) af door nitrificatie en derhalve ook de concentratie ammoniak in oplossing.
Storten van baggerspecie in open putdepots (fase 2)
29
Door de beperkte stroming wordt vooral in diepe halfopen putdepots thermische stratificatie verwacht. In P2 kan zich tijdens de eerste jaren van de stortfase gedurende een periode van 6 maanden maximaal 3.000 kg N ophopen in het hypolimnion (de laatste jaren van de stortperiode vindt geen stratificatie meer plaats). Na opheffen van de stratificatie wordt deze hoeveelheid gemengd met doorstromend water met een laag effectief uitwisselingsdebiet (1 m 3 /s). Dit leidt gedurende de stortfase tot een maximale toename in stikstofconcentratie in het uitstromende water van 0,8 mg N/l (zie figuur 4.4). Bij gemiddelde pH en temperatuur gedurende het najaar zal een dergelijke verhoging niet leiden tot een overschrijding van de MTR voor ammoniak. Overigens zal door het lage uitwisselingsdebiet bij het storten van specie in halfopen putten vaak gekozen worden om een deel van de vulfase te storten met onderlossers, waardoor de hoeveelheid sedimentatiewater, dat vrijkomt in het hypolimnion, beperkt wordt. Bij de scenarioberekeningen voor halfopen putten is geen rekening gehouden met hoge afvoeren, waarbij de uiterwaarden " meestromen" met de rivier. Onder dergelijke omstandigheden zullen halfopen putten onderdeel gaan vormen van de rivier en wordt de waterschijf in de put grotendeels vervangen door rivierwater. De duur van een dergelijke hoge afvoer is in de orde van enkele dagen. Omdat de waterschijf gedurende de vulfase steeds ondieper wordt, treedt stratificatie in veel putdepots niet op gedurende de gehele stortfase. Duur en dimensie van de stratificatie spelen dus een belangrijke rol bij de beantwoording van de vraag in hoeverre (tijdelijk) een stikstof (ammonium) piek kan worden gemeten in het oppervlaktewater na opheffen van de stratificatie. Er wordt dan ook op aangedrongen beter zicht te krijgen op de duur van de stratificatie en de duur van de omslag in specifieke open putten evenals de effecten van storten op instellen van stratificatie.
4.3.4 Discussie Bij het storten van baggerspecie in tijdelijk gestratificeerde putdepots is gedurende de zomermaanden het effect van vrijkomen van stikstof op concentraties in het oppervlaktewater minimaal. Door de omslag van een gestratificeerde naar een niet-gestratificeerde put worden piekafvoeren aan stikstof gemeten. De omslag vindt echter plaats in het najaar, waardoor eutrofieringseffecten minimaal zijn. Door de lagere temperatuur en de lagere pH is ook de hoeveelheid ammoniak die hierbij vrijkomt, beperkt.
Storten van baggerspecie in open putdepots (fase 2)
26
Figuur 4.3 Verloop van de toename in stikstofconcentratie in het uitstromende water in P1 tijdens storten van baggerspecie bij een uitwisselingsdebiet van 1 m3/s (onderbroken rode lijn indien geen stratificatie optreedt in de put; dichte biauwe lijn indien wel stratificatie optreedt in de put).
0.8
Figuur 4.4 Verloop van de toename in stikstofconcentratie in het uitstromende water in P2 tijdens storten van baggerspecie bij een uitwisselingsdebiet van 1 m3/s (onderbroken rode lijn indien geen stratificatie optreedt in de put; dichte biauwe lijn indien wel stratificatie optreedt in de put).
5 z
1
l
0.4
[Vu y.
-
Storten van baggerspecie in open putdepots (fase 2)
V
VJ
0
2
4
...
i 8
6 ii|d Qaren
27
^ vA )
10
'.
Storten van baggerspecie in open putdepots (fase 2)
28
5 Conclusies
De belangrijkste conclusie van dit onderzoek is dat de samenstelling en textuur van de baggerspecie slechts in beperkte mate verantwoordelijk is voor een eventuele extra belasting van het oppervlaktewater met stikstof. Naast de intrinsieke eigenschappen van de gestorte baggerspecie zijn het voornamelijk locatiespecifieke omstandigheden die verspreiding van stikstof en ecotoxicologische effecten gerelateerd aan het vrijkomen van stikstof in het oppervlaktewater beinvloeden. De belangrijkste locatiespecifieke parameters zijn het optreden en duur van stratificatie in het open putdepot, het uitwisselingsdebiet met het omliggende water, de natuurlijke stikstofvracht en -concentratie in het oppervlaktewater en het optreden van eutrofiering in het oppervlaktewater (pH-effect). Daarnaast is de storttechniek en het daaraan gekoppelde verlies van materiaal van belang voor het vrijkomen van stikstof. Er wordt dan ook aanbevolen om voor een realistische inschatting van de effecten van storten van baggerspecie in open putdepots op de stikstofhuishouding in het oppervlaktewater deze locatiespecifieke parameters dan ook mede te bestuderen. Modelmatig kan een dergelijke inschatting worden gemaakt door een module voor het vrijkomen van stikstof bij storten van baggerspecie in te bouwen in het bestaande instrumentarium voor open putdepots (WESTSIDE). De keuze en inrichting van de locatie is dus belangrijk bij de beoordeling of storten van baggerspecie in een open putdepot kan resulteren in risico's met betrekking tot stikstof. In stromende systemen zijn door het hoge uitwisselingsdebiet van de put effecten van het storten van baggerspecie op de stikstofhuishouding in het oppervlaktewater verwaarloosbaar. Storten van baggerspecie in een tijdelijk gestratificeerde put lijkt met betrekking tot het vrijkomen van stikstof voordelen te hebben ten opzichte van storten in niet-gestratificeerde systemen met dezelfde karakteristieken. Indien de put wordt gebouwd in een stagnant (eutrofieringsgevoelig) niet-gestratificeerd watersysteem of wanneer rond de put een damwand is gebouwd, waardoor een klein uitwisselingsdebiet wordt gecreeerd tussen de put en het omliggende stromende oppervlaktewater (vergelijkbaar met een halfopen putdepot), zijn in en nabij de put gedurende de zomermaanden zowel acute toxiciteit als gevolg van het vrijkomen van ammoniak als effecten op eutrofiering mogelijk (er is een continue belasting van het oppervlaktewater met stikstof). Dergelijke risico's worden echter niet verwacht wanneer onder deze omstandigheden wordt gestort in een tijdelijk gestratificeerde put, omdat omslag van een gestratificeerde situatie naar een gemengde put plaatsvindt in het najaar.
Storten van baggerspecie in open putdepots (fase 2)
29
Storten van baggerspecie in open putdepots (fase 2)
30
Referenties
• Absil, L.L.M. & T. Bakker, 1999. Inventarisatie waterbodems: Aanbod en bestemming van baggerspecie 1999-2010. AKWA rapport. • Admiraal, W. & Y.J.H. Botermans, 1989. Comparison of nitrification rates in three branches of the lower river Rhine. Biogeochemistry 8, 135-151. • Balistrieri, L.S., J.W. Murray & B. Paul, 1992. The cycling of iron and manganese in the water column of Lake Sammamish, Washington. Limnology and Oceanography 37, 510-528. • Berg, G.A. van den, 1998. Geochemical behaviour of heavy metals in a sedimentation area of the rivers Rhine and Meuse. Proefschrift Universiteit
Utrecht. • Berg, G.A. van den & J.P.G. Loch, 1993. De chemie van verontreinigd baggerslib in depot: Een kennisinventarisatie. Flevobericht 342. RWS-RDIJ. • Bemer, R.A. 1980. Early diagenesis: A theoretical approach. Princeton Univ. Press. • Best, J. de, M. Beek. K. Groen, K. van der Guchte, K. Hartnack, B. van der Heijdt, A. van der Laan, D. van Pijkeren & J. Tuinstra, 1999. Storten van baggerspecie in open putdepots: Covernotitie eerste fase. RIZA rapport 99.053, AKWA rapport 99.013. • Boers, P., W. Laane, L. van Liere, C. Peeters, S. Parma & J. van der Does, 1993. Eutrofiering in Nederland, hoe verder ? RIZA notitie 93.056x. • Cappellen, P. van & Y. Wang, 1996. Cycling of iron and manganese in surface sediments: A general theory for the coupled transport and reaction of carbon, oxygen, nitrogen, sulfur, iron, and manganese. American Journal of Science 296,197-243. • Gerrits, G.W.R., 1999. Onderzoek naar stikstofomzetting en waterkwaliteit in baggerspeciedepot Averijhaven: Uitvoering WVO-vergunning. Haskoning rapport, i.o.v. RWS, directie Noord-Holland. • Hartnack, J., 1994. Beoordeling van de lozing van retourwater uit een baggerspeciedepot. RIZA werkdocument 94.120X. • Hartnack, J. & B. Wesseling, 1990. Gevolgen van de stort van baggerspecie in de ontgrondingsput Kaliwaal te Druten. RIZA werkdocument 90.187X. • Hartnack, J., B.P.C. Steenkamp . & J.M. van Steenwijk, 1996. Overal zindert bagger. Storten van klasse lll/IV specie in diepe putten: Aanbevelingen voor Milieu-Effect Rapportages. RIZA nota 96.073. • Heijdt, L.M. van der, EJ. Houwing, A.K.J, van der Laan & B.P.C. Steenkamp, 1999. Het storten van baggerspecie in open putdepots: Een kennisinventarisatie. AKWA rapport 99.011. • Huisman, J., P. van Oostveen & F.J. Weissing, 1999. Critical depth and critical turbulence: two different mechanisms for the development of phytoplankton blooms. Limnology and Oceanography 44, 1781-1787. • Hurley Octavio, K.A., G.H. Jirka & D.R.F. Harleman, 1977. Vertical heat transport mechanisms in lakes and reservoirs. Report 227, Dept. Civil Engineering, Massachusetts Institute of Technology, Boston, USA. • Imboden, D.M. & A. Wiiest, 1995. Mixing mechanisms in lakes. In: Lerman, A., D.M. Imboden & J.R. Gat (eds), Physics and chemistry of lakes, p. 83-138. Springer-Verlag, Berlin.
Storten van baggerspecie in open putdepots (fase 2)
31
• Martin, J.L. & S.C. McCutcheon, 1999. Hydrodynamics and transport for water quality modeling, chapter 8: Stratification and heat transfer in lakes and reservoirs, p. 335-384. Lewis Publishers, Boca Raton. • Molen, D.T. van der, 1996. Beleidsanalyse WSV: eutrofiering van het zoete oppervlaktewater. RIZA rapport 97.016. • Nowicki, B.L., 1994. The effect of temperature, oxygen, salinity, and nutrient enrichment on estuarine denttrification rates measured with a modified nitrogen gas flux technique. Estuarine Coastal and Shelf Science 38, 137-156. • Paalman, M.A.A., 1997. Processes affecting the distribution and speciation of heavy metals in the Rhine/Meuse estuary. Proefschrift Universiteit Utrecht. • Seitzinger, S.P., 1988. Denitrification in freshwater and coastal marine ecosystems: Ecological and geochemical significance. Limnology and Oceanography 33, 702-724. • Smith, R.M. & A.E. Martell, 1976. Critical stability constants, volume 4, inorganic complexes. Plenum Press, New York. • Steenkamp, B.P.C, L.M. van der Heijdt & D. Ludikhuize, 1996. Prognose waterkwaliteit tijdens storten van specie in de Molengreend. RIZA werkdocument 96.137X. • Vierde Nota waterhuishouding Regeringsbeslissing. 1998. Ministerie van Verkeer en Waterstaat, Den Haag. • Wetzel, R.G. 1975. Limnology, chapter 11: The nitrogen cycle, p. 186-214. W.B. Sanders Company, Philadelphia. • Zwolsman, J.J.G., 1996. Chemische kwaliteit van de Rijkswateren: Ontwikkeling van de waterkwaliteit van Rijn, Maas en Usselmeer (1971-1993). Landschap 13, 133-144. • Zwolsman, J.J.G., 1999". Die Wasserqualitat des Rheins in den Niederianden (1985-1997). In: Hydrologische Dynamik im Rheingebiet. IHP/OHP-Berichte 13, Proceedings Internationale Rhein-Konferenz 27.-28. April 1999, Koblenz. Deutsches IHP/OHP-Nationalkomitee, Koblenz. • Zwolsman, J.J.G., 1999b. Geochemistry of trace metals in the Scheldt estuary. Proefschrift Universiteit Utrecht.
Storten van baggerspecie in open putdepots (fase 2)
32
AKWA het Advies- en Kenniscentrum Waterbodems is een samenwerkingsverband van Rijkswaterstaat op het gebied van vervuilde waterbodems. Hierin zijn DWW, Bouwdienst, RIZA, RIKZ en Directie Noordzee vertegenwoordigd.
Voor meer informatie kan contact worden opgenomen met AKWA via het bureau WAU "Waterbodems Advies en Uitvoering" Adriaen van Ostadelaan 140, 3583 AM Utrecht, telefoon 030-2192020, of via RIZA afdeling Beleidsvoorbereiding Onderzoek en Advisering (BOA), postbus 17, 8200 AA Lelystad, telefoon 0320-298761
Generic Framework Water Program & related projects in the Netherlands
Papers presented at the Hydrolnformatics 2000 conference
July 22-27, Cedar Rapids, IA, USA.
Generic Framework Water Program & related projects in the Netherlands
Papers presented at the Hydrolnformatics 2000 conference
July 22-27, Cedar Rapids, IA, USA.
Generic Framework Water Program & related projects in the Netherlands
Papers presented at the Hydrolnformatics 2000 conference
July 22-27, Cedar Rapids, IA, USA.
Generic Framework Water Program & related projects in the Netherlands
Papers presented at the Hydrolnformatics 2000 conference
July 22-27, Cedar Rapids, IA, USA.
Generic Framework Water Program & related projects in the Netherlands Papers presented at the Hydrolnformatics 2000 conference July 22-27, Cedar Rapids, IA, USA.
RIZA working paper STOWA working paper RIKZ working paper
2000.039X 2000-W-01 RIKZ/OS/2000.110X
may 2000
Contents: Contributing authors and organizations (including e-mail addresses) Preface Blind, M.W.. A. Ubbels, L.R. Wentholt, Th.L. van Stijn, A.H. Bakema, J.D. Bulens, J. J. Noort, B. van Adrichem. J. Stout, F.C. van Geer, "Towards a welloiled model infrastructure for water management: the generic framework water program", Proceedings Hydrolnformatics 2000, 23-27 July 2000, Cedar Rapids, IA, USA. van der Wal, T, and M.J.B. van Elswijk, "Generic Framework for hydroenvironmental modeling". Proceedings Hydrolnformatics 2000, 23-27 July 2000, Cedar Rapids, IA, USA. ten Cate, H.H., S. Hummel, M.R.T. Roest, "An Open Model System for 2D/3D hydrodynamic simulations" , Proceedings Hydrolnformatics 2000, 23-27 July 2000. Cedar Rapids, IA, USA. Scholten, H., R.H. van Waveren, S. Groot, F.C. van Geer, J.H.M. Wosten, R.D. Kroeze, and J.J. Noort, "Good modelling practice in water management". Proceedings Hydrolnformatics 2000. 23-27 July 2000, Cedar Rapids, IA, USA. Wentholt, L.R., J.J. Noort, B. van Adrichem, J.H.P.M. Tacke, "Watermark: a general quality mark for software in the Netherlands, from idea to reality". Proceedings Hydrolnformatics 2000, 23-27 July 2000, Cedar Rapids, IA, USA.
1
9
14
22
29
•
c o
^ *
•*->
03
N
c
ed IM
o -o c Cd
co 3 00 •3
< -' r-
o U
11IIII
CO
=
EC
-a = co ca U
q
s.
~ 1 CO "o' f
«.
—
«^ o o o c2 •S Z CO
111 a a
C CJ
< .y —•
. 3 2 LO
-J
— — r",
I
u tu
fe" a»
O
CO
>
£ £ 3 y.
>
Preface Application of computer models has become increasingly important in water management over the past decades. Many organizations dealing with water management problems have developed Simulation models in order to convert their expertise intot operational knowledge for supporting policy developement and decision making. As an integrated approach in water management, covering various disciplines and organizations, is getting customary, various problems in utilizing models became more evident. These problems relate to the aspects shown in the table below: technical
organizational
utilization quality assurance
technical quality assurance financial
=> the current models and databases are insufficient flexible and interconnec table see e.g. van der Wal et al, p. 9 and ten Cate et al.. p. 13 _z> the current models and databases are often poorly accessible due to ownership rights. see e.g. Blind et al.. p. I => as more complex model systems are utilized errors are easily overlooked and. due to propagation may have large effect on simulation outcomes see e.g. Scholten et al. p 21 => the current models vary in technical quality (software and maintenance) which is not desirable in complex model systems. see e.g. Wentholt. et al. p. 29 => in developing and maintaining individual simulation models and complex model chains many activities, such as developing and upgrading user interfaces, are performed repeatedly. Resources for developing simulation models are more and more limited due to increasing cost of maintenance. see e.g. Blind et al. p. I
In the Netherlands many projects are currently carried out which aim at solving one or more of these problems. The paper by Blind et al presents an overview of the ongoing projects, with emphasis on the projects which are carried out within the Generic Framework Water Program. The subsequent papers present several of the projects in more detail, one on the architecture of the "Generic Framework for hydro-environmental modeling', one on the Open Model System for 2D/3D hydrodynamic simulations, one on 'Good modelling practice in water management' and one on 'Watermark: a general quality mark for software in the Netherlands'. This working paper aims at giving insight in these developments in an attempt to communicate developments here and to provoke reactions from other countries, either on similar projects in terms of comments ande questions. All papers have been accepted for publication at the Hydrolnformatics 2000 conference, 23-27 July 2000, Cedar Rapids, IA, USA. Michiel Blind RIZA. Institute for Inland Water Management & Waste Water Treatment. Lelystad. The Netherlands may2000
1
Towards a well-oiled model infrastructure for water management: the generic framework water program
M.W. Blind RIZA. Institute for Inland Water Management & Waste Water Treatment. Lelystad. The Netherlands A. Ubbels RIZA. Institute for Inland Water Management & Waste Water Treatment. Lelystad. The Netherlands L.R. Wentholt STOWA, Foundation for Applied Water Research. Utrecht. The Netherlands Th.L. van Stijn RIKZ. National Institute for Coastal and Marine Management. The Hague. The Netherlands A.H. Bakema RIVM. National Institute of Public Health and Environmental Protection. Bilthoven, The Netherlands J.D. Bulens ALTERRA. Green World Research. Wageningen. The Netherlands J.J. Noort SEPRA. Berkel en Rodenrijs, The Netherlands B. van Adrichem EDS. Leidschendam. The Netherlands J. Stout WL\Delft Hydraulics, Delft, The Netherlands F.C. van Geer 77ie Netherlands Institute of Applied Geoscience TNO. Delft. The Netherlands
ABSTRACT: The past decades have shown an increase in the use of information technology (IT) in water resources management. The development of IT applications at different institutions in The Netherlands and over a long period of time has resulted in major difficulties in technically linking applications and dealing with uncertainties in model chains. However, combining the models and data from different disciplines is considered a prerequisite for effective integrated water resources management. These and other findings are the driving force in the current activities within the Generic Framework Water (GFW) Program. The GFW Program's main aim is to improve the contribu-
Towards a well-oiled model infrastructure for water management: the generic framework water program Paper presented at Hydrolnformatics 2000, 23-27 July. 2000. Cedar Creeks. IA, USA.
tion of models in policymaking, management and research, by facilitating easy access and linking of models and databases, reduce the costs of engineering, maintenance and utilizing models, improve flow of expertise and modeling practice. Within this program three projects are carried out concerning (a) The development of an IT-framework for modeling, (b) an umbrella agreement on the accessibility of models and databases of various institutes, and (c) a good modeling practice handbook. The program is supported by all major players in the water management field in the Netherlands. This paper describes the different projects being carried out within the GFW Program and their coherence.
INTRODUCTION Due to the ever increasing complexity of water management the utilization of models has steadily increased and become more and more complex. In 1997, Dutch regional and national water authorities investigated the bottlenecks and subsequent requirements of model development and utilization (Zanting et al., 1997; STOWA, 1997). The removal of such bottlenecks is believed to be an important prerequisite to enhance the quality of water management. The main conclusion presented in the studies is that the main problem of today's model utilization lies in linking various models and databases covering different disciplines of water resources management or originating from different institutes. Linking generally occurs 'on demand' and is time and money demanding. Combining models and databases from different disciplines, however, is considered a prerequisite for effective integrated water resources management. Various other problems were pointed out: a) management and maintenance of software puts an increasing burden on budgets; b) new scientific insights are difficult to incorporate in existing model structures; c) combination and exchange of expertise in joint projects of different institutes is constrained due to poor connectivity and accessibility of models; d) an increasing dependence on specific software developers exists; e) the multitude of models on the same subject allows many answers: for policy development a single but reliable answer is necessary. Based on these findings it was concluded that an IT framework for model studies in water
resources management was desirable. In 1997, as part of the AQUEST program, a discussion concerning the possible development of such a system, named the Generic Framework Water Program (GFW-program) was launched. This paper describes the general ideas, the setup and progress within the project, and relates various other developments in The Netherlands to this project. GENERIC FRAMEWORK WATER (GFW) PROGRAM The GFW Program's main aim is to improve the contribution of models in policymaking, management and research, by facilitating easy access and linking of models and databases, reduce the costs of engineering, maintenance and utilizing models, improve flow of expertise and modeling practice. Since a large number of model users and developers exists in Dutch water management a steering committee was established in which all major players participate (column A. Figure 1). The players concluded that, in order to achieve these goals, an application named 'Generic Framework for Modeling in Water Resources Management' (GFW) needs to be developed. A project group concerning the design and implementation of this framework was established, which also consisted of many players in the field of water management and research (column B. Figure 1). It was immediately realized that facilitating easy linking of models itself, would not be sufficient to tackle the problems of easy access to models developed by different institutes. Two other problems were identified and subsequently tackled.
Towards a well-oiled model infrastructure for water management: the generic framework water program Paper presented at Hydrolnformatics 2000,23-27 July. 2000, Cedar Creeks. IA. USA.
Federal Institutes National Institute for Inland Water Management and Waste Water Treatment RIZA National Institute for Coastal and Marine Management. RIKZ National Institute of Public Health and Environmental Protection. RIVM (Semi-) Governmental institutions STOWA. Foundation for Applied Water Research Province South-Holland Public Works Directorate, Department South-Holland Water Supply Company Gelderland Municipal Water Supply Company Amsterdam Water Authorities "Friesland"; "Rijnland"; "Uitwatcrende Sluizen"
A X X X
B c D X X X X X X X X X
X X X X
x
X X X X
Great technological institutes WLIDelft Hydraulics The Netherlands Institute of Applied Geoscience TNO(NITG-TNO)
X X X X X X X X
Consultants & IT companies Alterra (formerly Winand Staring Centre for Integrated Land, Soil and Water Studies) Sepra DHV EDS Geodan - IT BV IWACO Resource Analysis Witteveen en Bos
X X X X X X X X X X X X X X
Universities Wageningen University Delft Technical University Free University of Amsterdam
X X X
Figure 1: Participating parties in the various projects (A: Steering committee GFW; B: Design and realization group GFW; C: Umbrella agreement: D: Good modeling practice)
The first problem was the of right of ownership of models and databases, often leading to high costs and redundant developments. A project group was set up to design an 'umbrella agreement' on the accessibility of the members' models and databases (column C, Figure 1). The second problem was today's modeling practice. As problems and model(-chains) get increasingly complex, more and more choices need to be made in the modeling process. Subsequently, the number of possible sources of error increase and interpretation of results gets more complex. A third working party was set up to establish recommendations for good modeling practice (project GMP, members column D, Figure 1). Figure 2 shows the GFW Program structure and past and present projects. In the next sections the main features of the three projects are described.
THE PROJECTS Generic Framework Water The general idea to solve many of today's problems in model utilization by the GFW is far-reaching modularization of software. The models itself, that is die actual computing core of the software, containing the scientific expertise, should be separated from data and pre- and post-processing tools. Though not quite according to the technical IT-implementation of the GFW, Figure 3 illustrates the basic principle of such a modularization. The advantages of such an approach are described in 3.1.1. Within the framework models, data and tools should be lined up to form a working model application. Component-based technology is a necessity for such modularization. This approach heavily relies on unambiguous application protocol interfaces (API's) and clear
Towards a well-oiled model infrastructure for water management: the generic framework water program Paper presented at Hydrolnformatics 2000.23-27 July. 2000, Cedar Creeks, IA. USA.
S w i n g ("omiice Generic F m n ork Water (GFW) Project Group Development GFW Project Imcmory Generic Tools Invrniory Generic Tools report. 1998 Project Program of Requirements Program of Requirements GFW. 1998 Project Benefits of GFW Benefits ofGFW report. 1999 Project IT - .Architecture GFW Architecture report. 1999 Project Design & Realisation Report & Application, due 2000 Project Group Good Modelling Practise I Good Modelling Practise Handbook. 1999
Project Group Umbrella Agreement
]
1,'mbrclla Agreement, due 2000
Figure 2: GFW project organization structure and projects. specifications of application domains. Models within the same domain should be exchangeable trouble-free, in other words without imperative changes in the models or interfaces. On a small scale this concept has been proven in the national research program on desiccation (van Baalen, 1998) and in several other projects in the LWI-program. Though other frameworks for models in water management exist, none is sufficiently generic to achieve a high degree of trouble-free exchangeability of models and tools. It was decided that the GFW needs to be built from scratch since (a) the transformation of existing systems often raises new problems and redesign and integration of existing systems is very difficult (b) redesigning an existing framework of one institute might reduce the bearing surface other institutes.
plex problems in integrated water management: The once-only investment to adapt to framework specifications is compensated by then redundant time and money demanding development or redesign of models. The time between the rise of a new issue in integrated water management and solutions supported by models should therefore be shortened. Of course, this only holds true if models are also accessible and modeling practice is reliable (see paragraphs 31 and 32). Since these complex problems are rarely covered by a single field of expertise, and therefor by a single institution, increasing connectivity of models also improves the possibilities for cooperation between institutes and thus stimulates the exchange of expertise. For many aspects of modeling in water management various models are available, which try to answer the same question but are based on varying scientific insights, or are only applicable within a specified range of time and scale resolution. Improvements in linking allows electing the 'best' chain of models given the problem at hand and available data. Besides that, the consequences of using different scientific approaches on the outcome of a study may be researched. Of course creating model chains is not the only advantage of GFW. The building and testing of single model components by researchers may speed up and become more effective since little effort is necessary to connect data and tools to models. Scientists therefore may spend more time on their core-business, such as developing new algorithms. user interface proces flow management tool
selection tool M i l ) i l l tool
calibration tool
presentation tool
simulation control program
The advantages of the approach conversion
The modularization together with the domain specifications and API's results in easy linking of models from different application domains, allowing responsible analysis of com-
.!jt.ib.i,c.
djuhase III
database I
ZZ:
Figure 3: Illustration of the GFW principle.
Towards a well-otled model infrastructure for water management: the generic framework water program Paper presented at Hydrolnformatics 2000. 23-27 July. 2 0 0 0 . Cedar Creeks. IA. USA.
Apart from the already mentioned advantages speeding-up model building and model utilization has many financial benefits, since less time is needed and additional software engineering to link models is redundant. Furthermore, modularization results in smaller systems which are easier to manage and maintain. Upgrading algorithms in the various components and updating user interfaces to modern standards gets easier. More important, the possibility of re-using software will increase. While model algorithms are currently repeatedly implemented (amongst others due to poor modularization), a model component may be reused in various kinds of studies in the future. Since scientific principles in modeling and residue analysis are not changing (frequently), model components may not demand changes for long periods of time. User interfaces, visualization tools etc. are frequently newly implemented because of lack of modularization and upgraded due to necessary modernization. Within the framework basically only one tool for each of the generic purposes is necessary. These tools may then be reused in al kinds of studies. This also means that a smaller number of updates is required, additionally reducing costs. The project 'Inventory of generic tools' (Ubbels et al., 1998) investigated the availability of generic tools. 23 out of 59 mainly Dutch companies were questioned on the subject. Tools for the following procedures were distinguished: process control; data editing and management; selection and definition; calibration: analysis; presentation; on-line control; and uniform data access. The results of this study, including a 'generic tools database', are available to the public on http://index. waterland.net/aquest/ zoektool.html It was concluded that especially for presentation, analysis and uniform data access a large number of tools is available, meaning that much resources are spent on very similar software. The project 'Benefits of the Generic Framework Water' (Noort & Ubbels, 2000) researched the claim of reduced costs for model development: though it proved difficult to obtain reliable estimates it appears that approximately 50% savings can be achieved in model development if all necessary tools are already available. Finally, the dependence on few software de-
veloper may reduce since other software bureaus may create new models and tools without knowledge of the implementation of other models. The disadvantages and challenges The most important disadvantage is the risk that the GFW may lead to inflexible, unworkable standards. In such a situation updating the entire framework to reach the then current state of the art will be difficult, thus expensive. Inflexible standards may also result in partners doing developments outside the framework, resulting in a reduction of all benefits. Thus the framework should not be allowed to become a straitjacket but instead should be useful to all the parties involved. It should support current needs and be open towards future developments. The development of the framework requires a change of working habits for software developers. It is necessary that developers comply to the framework. For the developers this means a reduction of freedom in developing applications. To facilitate the transition to the framework-compliant programming it is necessary to invest in courses and workshops for application developers. Only few of the current models and tools in integrated water management will easily be modified to fit in the framework. In many cases major adjustments will be necessary, demanding much of the resources for model development in the next years. The complete rebuilding of models may also turn out to be necessary, indicating a complete loss of investments of the then redundant models. Current status In 1999 the IT-architecture has been designed (van der Wal, 1999). The report comprises a domain analysis of models in integrated water management and interface specifications. The domain analysis is necessary to translate the subject of integrated water management in abstract IT definitions. The paper by van der Wal and Elswijk (2000) deals with the architecture in detail. In spring 2000 the detailed specification and implementation of version 1.0 of GFW has
Towards a well-oiled model infrastructure for water management: the generic framework water program Paper presented at Hydrolnformatics 2000. 23-27 July, 2000. Cedar Creeks. IA. USA.
commenced. This first version will contain some models and tools and have limited functionality. In time more and more models and functionality will be added, based on agreement on how to proceed and the integrated water management projects at hand. Umbrella Agreement Creating the technical possibilities to connect models and databases is insufficient to allow utilization of models and databases originating from different institutes. The use of external software and databases must be authorized. To avoid procedures for each individual model and database an umbrella agreement covering the use of many models and databases is being drawn up. The main aim is to agree on accessibility of models and databases enhancing the efficiency of application, development, management and maintenance of models and databases. The umbrella agreement also states an agreement on the development of quality assurance requirements with respect to the development, management and maintenance of models and database, modeling practice (see 32) and a method to review and maintain these requirements. The parties also agree on stimulating the use of existing preconditions and standards within their organizations and cooperation in new developments. The agreement specifically states that ownership remains with the originating institutes and commercial application is only possible with permission. A draft list of models and databases that will be included in the agreement has been created. Models concerning all aspects of integrated water management are included, reflecting the commitment to the agreement by all parties involved and the eagerness to achieve the greatest possible freedom in the exchange of models. Since the requirements on development, management and maintenance have not yet been defined it is not clear which models will be included in the final agreement. Good Modeling Practice (GMP) Proper, reliable use of models is a necessity for integrated water management. Due to the many choices to be made and the (increasing) complexity of nowadays model studies, errors are
easily made: Examples of such errors are erroneous input data, poor calibration and validation or application of parameters with unrealistic values. Thus, the quality of model studies very much relies on the skills and knowledge of the modeling teams. It is easily realized that increasing accessibility and easy linking of models may increase improper use of models: studies become more complex and models are used by parties less familiar with the peculiarities of the models. As a result recommendations on good modeling practice (GMP) in water management have been developed (STOWA/RIZA, 1999). The recommendations support all aspects of model studies in a seven step procedure: (1) starting a logbook. (2) defining the modeling project, (3) building the model, (4) analyzing the model, (5) using the model, (6) interpretation and (7) reporting and (reproducible) filing. Though the handbook contains more specific information including tests for various procedures it is not designed as a straitjacket: If, for example, steps are not performed (e.g. due to lack of data, time or budget), this should not render a study useless: Applying the recommendations however does result in better insight in the ins and outs of the study and the uncertainties surrounding the results, which is very important for policy making. The recommendations are also meant to be dynamic, that is they will be updated as new insights become available. Scholten et al. (2000) discus good modeling practice in detail. In 1999 a workshop on GMP, visited by both clients and contractors of model studies took place. Many clients consider GMP as a prerequisite for future modeling studies. Contractors subscribed to the value of GMP and intend to incorporate GMP within their organizations. OTHER DEVELOPMENTS Open Modeling System (OMS) The Open Modeling System also concerns creating a software infrastructure for modeling. It is specifically at aimed at two- and threedimensional models for modeling coastal and marine environments and lakes. More specifically the model systems SIMONA (ten Cate el al. 2000) and DELFT-3D (http://www.wl.delft
Towards a well-oiled model infrastructure for water management: the generic framework water program Paper presented at Hydrolnformatics 2000. 23-27 July, 2000. Cedar Creeks, IA. USA.
.nl/d3d/) need to be integrated in a modular, flexible manner. These models are frequently applied in operational management such as storm surge control, but also in policy analysis issues. OMS will adhere to the principles of GFW. In addition to the GFW specifications computing performance and independence of platform and operating system are specific issues in OMS will be tailored to. These issues will therefore determine some of the framework's specifications and the level of modularization. Ten Cate et al. (2000) discusses OMS and its relation to GFW in detail. 'Watermark' The project watermark aims at developing a hallmark for models. Hallmarks are being developed for programming and software and software management and maintenance. The major parties involved in water resources management in The Netherlands are represented in the project. The programming standards will be based on international accepted standards and specific in-house standards of the different participating institutes. Consensus between the parties is a necessity for success. Since the hallmarks are not intended to frustrate development and accessibility of software they should not result in a straitjacket. The standards agreed upon need to be efficient, that is suitable for automation. Furthermore, international software may not be banned - a hallmark for non-Dutch software is currently under consideration. Wentholt et al. (2000) discusses the Watermark project in more detail. Adventus and 'stekkerdoos water' In the Adventus and 'Stekkerdoos water' ('multiple socket water", both at http:// www. adventus.nl/adventus. in Dutch) project the aim is to reach standardization on the level of data. The Adventus data-dictionary contains information on the definitions of all kinds of data relevant to water management. The 'Stekkerdoos water' is library which allows building interfaces for databases and applications designed according to Adventus data-dictionary specifications. The Adventus data-dictionary and the 'Stek-
kerdoos water' play a major role in both the GFW and Watermark. CONCLUDING REMARKS Many projects are on their way to increase the possibilities and the reliability of model studies in integrated water management in The Netherlands. Though there is still a long way to go. the first steps are taken and the results so far are promising. The involvement of so many players in the field poses a tremendous challenge: interests of the various institutes vary due to the nature of the institutes, commercialization, short and long term expectations, etc. As a result much attention needs to be given to meet specific wishes of institutes and reach consensus on the way to proceed. This requires intensive communication, which in itself is challenging: Misunderstandings due to slight differences in interpretation of oral and written communication easily occur if many parties are involved. The parties in the various platforms need to report 'back home", creating the necessary support for ongoing and future developments. This requires much effort of the individual representatives in the project groups: GFW, 'Umbrella Agreement' and GMP. However, realization of the GFW and the Umbrella Agreement will have important advantages for every single modeler: improvement of the utilization of models of various kinds, reduction of costs and improvement of flow of expertise. But never forget models are just tools. Expertise on integrated water management related issues is still the most important part in modeling. Following the recommendations of the GMP handbook will contribute to reliable model studies and thus effective integrated water management. The GFW may be useful in other areas of modeling as well. Though the current emphasis is on water management, the architecture seems to be sufficiently generic to be useful in other (environmental) application domains. Cooperation has given a good basis for improving future modeling. This cooperation now also has an international component. The U.S. Environmental Protection Agency is interested in the results of die GFW Program. A Memorandum of Understanding is signed for future
Towards a well-oiled model infrastructure for water management: the generic framework water program Paper presented ai Hydrolnformatics 2000. 23-27 July. 2000, Cedar Creeks. IA. USA.
X
cooperation. This cooperation may lead to a new international standard on modeling in integrated water resources management. ACKNOWLEDGMENT The authors wish to thank R.H. van Waveren, F.H.M. van de Ven, H.F. Terveer (all RIZA) and H. Scholten (Wageningen University) for their valuable contributions. REFERENCES Ubbels, A., R. Koeze. R. Terveer, C. Sprengers, F. Keppel, and B. van Adrichem, "Inventory of generic tools", CUR/LWI. Gouda, The Netherlands, s.p., in Dutch, 1998. Noort, J.J. and A. Ubbels, "Benfits of the Generic Framework Water", STOWA, Utrecht, The Netherlands, in preparation, in Dutch, 2000. Scholten, H., R.H. van Waveren, S. Groot, F.C. van Geer. J.H.M. Wosten, R.D. Koeze, and J.J. Noort, "Good modelling practice in water management". Proceedings Hydrolnformatics 2000, 23-27 July 2000, Ceder Rapids, IA, USA. STOWA, "Research on the requirements of consensus model instruments", STOWA report 97-01, Utrecht, The Netherlands, 30p. in Dutch, 1997. STOWA/RIZA. "Good Modeling Practice
Handbook", STOWA-report 99-05, Utrecht. The Netherlands, s.p., 1999. Zanting, H.A., G. Baarse, P. Kouwenhoven, M. Taal, J. Tacke. S. Ooms. "Evaluation of the PAWN model instruments". EDS, Resource Analysis report RA-97/281, Rijswijk, The Netherlands, 40 p. in Dutch, 1997. ten Cate H.H.. Hummel, S. & M.R.T. Roest, "An open model system for 2D/3D hydrodynamic simulations". Proceedings Hydrolnformatics 2000, 23-27 July 2000, Ceder Rapids. IA, USA. van Baalen, S.J.A. (ed.), "Process chain management system for desiccation: chain of models in water management", NOV-report 13-2, RIZA-report 98.019, Lelystad, The Netherlands, 44 p., in Dutch, 1998. van der Wal, T. (ed.), "Generic Framework Water Architecture". RIZA report 99.063, Lelystad, The Netherlands 92 p., in Dutch, 1999. van der Wal, T. & M.J.B. van Elswijk, "A generic framework for hydroenvironmental modelling" Proceedings Hydrolnformatics 2000, 23-27 July 2000, Ceder Rapids, IA, USA. Wentholt, L., 2000, "Watermark (a general quality mark for software in the Netherlands, from idea to reality)". Proceedings Hydrolnformatics 2000, 23-27 July 2000. Cedar Rapids, IA, USA.
Towards a well-oiled model infrastructure for water management: the generic framework water program Paper presented at Hydrolnformatics 2000. 23-27 July. 2000. Cedar Creeks. IA. USA.
Generic Framework for hydro-environmental modelling T. van der Wal Alterra - Group Software Engineering. Wageningen University and Research centre. The Netherlands M.J.B. van Elswijk Software Engineering Research Centre. Utrecht. The Netherlands
ABSTRACT: A group of knowledge institutes in the Netherlands initiated the development of a Generic Framework for Modeling in Water Resources Management. The aim of this framework is to improve the coupling of models and enhance the development of new algoritms and models. The architecture of this framework is finished recently and is the subject of this paper. The architecture is based on a domain model as derived by a large group of specialists in different fields. It contains concepts for software development as well as domain concepts. A first version of the framework is under construction.
INTRODUCTION The utilization of models increases steadily and becomes more and more complex. A group of knowledge institutes concerning integral water management in the Netherlands therefore initiated the development of an IT framework for model studies, known as the Generic Framework for Modeling in Water Resources Management (GFW). The GFW is part of a larger program aiming at improvement of the contribution of models in policymaking, management and research, by facilitating easy access and linking of models and databases, reduce the costs of engineering, maintenance and utilizing models, improve flow of expertise and modeling practice (Blind et al., 2000). The GFW development is one action within this program. Though other frameworks for models in water management exist, none is sufficiently generic to achieve a high degree of trouble-free exchangeability of models and tools. It was decided that the GFW needs to be built from scratch since (a) the transformation of existing systems often raises new problems and redesign and integration of existing systems is very difficult (b) redesigning an existing framework of
one institute might reduce the bearing surface other institutes (Blind et al., 2000). The architecture consists of both software engineering aspects and domain specific aspects. The architecture is designed for maximum flexibility in use, adaptability, userfriendliness and robustness. It contains mix & match concepts (Fayad, 1999), OpenGIS and state-of-the-art software engineering standards. The first version of die Standard Framework will be entering the beta-testing fase in the second quarter of 2000. DOMAIN MODEL The process leading to the architecture for the standard framework was highly interactive and participative. A group of fifteen scientists, modelers, information architects and software engineers drew up together a domain model. Based on the abstract version of this domain model, the architecture was designed. For the first time, the design of such a system has been carried out by a very broad group of organisations active in the development of software tools for water management.
Generic Framework for hydro-environmental modelling Paper presented at Hydrolnformatics 2000. 23-27 July, 2000, Cedar Creeks. IA. USA.
10
The GFW architecture is based on a shared perception of the domain 'Simulation modeling of water resources management'. This is captured in a so called domain model. In several workshops the domain model was established in a discussion with hydrologists. First, a conceptual model was derived. All hydrologists agree on the notion that the domain of simulation modeling of water resources management can be divided into subdomains. Coupling of two models is often complicated when both contain algorithms for the same subdomain. This should therefore be avoided. Different subdivisions in subdomains are feasible, but we found enough discrimination in a subdivision in environmental compartments
It becomes clear from figure 1 that any model element is connected to only one single model application. The simulation takes place in the model application. A model application within the GFW has to conform to a standard interface that is defined for all framework components. COMPONENTS Besides model applications the GFW architecture describes 'generic tools' as framework components. Both model applications and generic tools implement the same framework component interface (see figure 2 below). We have identified the in process tool, such as fil-
Framework Component
~A~ Model Application ModelComponent
ModelElement
Connector
Figure 1: Class diagram (UML) of the conceptual model in the architecture of the GFW (van der Wal, 1999). such as atmosphere, soil, saturated soil, surface ters, converters and other tools that influence water etc. The core of the GFW architecture is the output of model applications in a process an abstract description of this conceptual chain. In process tools behave like model applimodel, as illustrated in figure 1. It shows how a cations and therefore inherit both the Genericmodel application can be linked to many model Tool and the ModelApplication interface. components: Model elements and connectors. Linked tools do not have any influence on ModelElements represent physical objects in the simulation results. An example of a linked water resources management such as a ditch, a tool is a presentation tool that for instance pressoil compartment or a part of an aquifer. Conents model results in a graph. nectors link modelelements together.
Generic Framework for hydro-environmental modelling Paper presented at Hydrolnformatics 2000.23-27 July. 2000. Cedar Creeks. IA. USA.
11 «
Model
interface » Framework Component
« interface »> GenericTool
«interface » Modelapplication
Data « interface » InProcessTool
Figure 2: Class diagram (UML) of the framework ModelApplications are linked to a model as can be seen in Figure 2. This delegation allows interfaces to be built around existing model systems (so called wrapping). Also in figure 2 it is illustrated that a Model is linked to data and a Model 'knows' where and how to get its data.
«interface » LinkedTocJ
components (van der Wal. 1999). solver and the output is not just calculated for the model element that asked for it, but for all model elements in this matrix. This concept allows complex numeric models to be incorporated. DISCUSSION
DYNAMIC BEHAVIOUR The architecture of GFW also describes the dynamic behaviour of the framework components. Framework components (model applications and generic tools) can be used as building blocks to construct a modeling system. The GFW will contain a Process Chain Management Tool to co-ordinate this action. When a modeling system is run this tool will control whether calculations are performed in the right order. Another aspect of dynamic behaviour is the concept of 'pull' calculation rather that the traditional push' or batch calculation. A modeller using a GFW model system can ask the system for output at one of the model elements. The model element 'knows' whether this output is already calculated and otherwise it will start deriving it using the chain of models and tools described in the chain manager. In addition, it can retrieve data from its neighbours with whom it is connected through its connectors. Moreover, when these neighbours have not yet calculated the required output the process will iterate from there, creating a domino effect. Model elements are linked to model applications, where is decided how a certain output is calculated. As an example, it can very well be that the model application contains a matrix
The GFW is part of a larger action within the Dutch water management society (Blind et al., 2000). Where the framework is a technical action leading to improved simulation modeling in water resources management. Good Modelling Practise (Scholten et al., 2000) is a more procedural action leading to a common basis of performing simulation studies. The GFW obviously intends to support the GMP by implementing its concepts and by creating hooks for tools e.g. for sensitivity analysis and uncertainty calculations. A major key to success for the GFW is the amount of effort required to modify existing model systems to adapt to the framework interfaces. Initial effort will be put in wrapping existing models but in the long run this will not be enough. Another issue is the schematisation used in GFW. The architecture is intended to work with one schematisation for all subdomains but it works just as well at one schematisation per model application. The principle of having a uniform schematisation for a large group of models can be found in several model families but they are often dedicated to a small application domain. It would be a major achievement when GFW can introduce a uniform schemati-
Gencric Framework for hydro-environmental modelling Paper presented at Hydrolnformatics 2000.23-27 July, 2000. Cedar Creeks. IA. USA.
12 sation for models of surface water, unsaturated flow and saturated flow (van der Wal & van Waveren, 2000). A first version of the GFW is under construction since spring 2000 and an initial filling of several model applications will demonstrate the abilities of a generic framework for simulation modeling. This first version will contain model applications for two compartments (most likely the compartment "groundwater" and the compartment 'surface water') and models for these two compartments will collaborate together to dynamically simulate both compartments. Interaction between the models will take place on timestep basis. For the compartment surface water a second model application will be available to replace the other. This will show the Mix principle.
Lelystad, The Netherlands 92 p., in Dutch. 1999. van der Wal T. and R.H. van Waveren, "Standaard Raamwerk Water", 17 Matrix jaargang 8 nummer I (53) (in Dutch) february 2000.
The first version of the GFW will be extended and improved in the next years. The modularisation of modeling systems will continue and all possible hydro-environmental models will be made suitable to work with the GFW. ACKNOWLEDGEMENT The Standard Framework is a joint activity of RIZA, RIVM, STOWA, WL | Delft Hydraulics, EDS, Geodan-IT, TNO-NITG and Alterra. REFERENCES Blind M.W., A. Ubbels, L.R. Wentholt, Th. L. van Stijn, A.H. Bakema, J.D. Bulens, J.J. Noort, B. van Adrichem, J. Stout and F.C. van Geer. "Towards a well-oiled model infrastructure for water management: the generic framework water program", proceedings Hydrolnformatics 2000, 23-27 July 2000, Ceder Rapids. IA, USA. Fayad M., D.C. Schmidt and R.E. Johnson (eds.), "Building application frameworks", Wiley Computing Publishing. 1999, Scholten, H.. R.H. van Waveren, S. Groot, F.C. van Geer, J.H.M. Wosten, R.D. Koeze, and J.J. Noort, "Good modelling practice in water management". Proceedings Hydrolnformatics 2000, 23-27 July 2000, Ceder Rapids, IA, USA. van der Wal, T. (ed.), "Generic Framework Water Architecture", RIZA report 99.063,
Generic Framework for hydro-environmental modelling Paper presented at Hydrolnformatics 2000.23-27 July, 2000. Cedar Creeks, IA. USA.
13
An Open Model System for 2D/3D hydrodynamic simulations H.H. ten Cate National Institute for Coastal and Marine Management /RIKZ, Rijkswaterstaat. P.O. Box 20907. 2500 EX, The Hague. The Netherlands, email: [email protected]
S. Hummel WL\Delft Hydraulics. P.O. Box 177. 2600 MH Delft, The Netherlands, email: Slef.Hummel® wldelft. nl
M.R.T. Roest VORtech Computing, P.O. Box 260. 2600 AG Delft. The Netherlands, email: Mark. Roest® vortech. nl
ABSTRACT: In 1998 the Dutch Rijkswaterstaat/RIKZ and Delft Hydraulics started a project called Open Model System (OMS). Such a system facilitates the integration of process models that describe the complex hydrodynamic and ecological phenomena (which may be developed by different teams), by providing a software architecture for communication, and standards for data exchange. The component-based approach also supports domain- and process decomposition and parallel computation. The features offered by the OMS support the trend towards more comprehensive modeling, which is nowadays required to obtain more realistic answers and estimations in case of, for instance, human interception in environmental processes. Usually huge amounts of money are involved, and wrong decisions may lead to financial claims, so only the results of the state-of-the-art simulation models are acceptable. Also, in operational practice like nautical control of harbors, the requested resolution of the flow calculation is increasing. The work towards the OMS is presented in the context of the application domain of the modeling systems (SlMONA and DELFT3D) which are currently used. Characteristics specific for that application domain will be defined, such as the huge amount of data transfer between processes. These characteristics influence the software architecture of the OMS framework.
1 INTRODUCTION Dynamic and flexible modification of geometries and processes is an essential facility for end users in integral water management ( Blind et al.. 2000, Cate et al., 1998). Integral water management requires integral simulations of large scale processes and relatively small or local processes (which have to be calculated with high resolution or even with specialized models). For example, current research on improving cargo exchange between sea and river ships in a harbor studies several solutions, amongst which is re-opening an existing dam. This al-
ternative connects a shallow fresh water channel with a (relatively) deep salt water channel, so effects on local currents have to be studied. That requires a sophisticated turbulence model in the area right behind the dam opening, while elsewhere the default available turbulence model will be sufficient. So its important that one can define and model all relevant physical phenomena, and therefore a user should be able to easily add, modify and/or delete processes and their interactions. Furthermore the user would like to plug in models developed by any expert(s) into the used software package.
An Open Model System for 2D/3D hydrodynamic simulations Paper presented at Hydrolnformatics 2000, 23-27 July. 2000, Cedar Creeks. IA. USA
14
Figure I: A schematisation of the Rhine-Meuse Delta. An important condition is that the performance of model runs is still satisfactory (in Section 2 we will show the importance of this requirement), so an OMS environment for coupling processes should support efficient data exchange. Another important requirement when creating a 'new environment' is the smooth migration from the existing operational systems to this environment. SIMONA and DELFT3D are proven technology, implemented in thousands of lines of Fortran77 code. Despite new programming languages and tools, totally re-engineering the software is more or less impracticable. So data exchange operations should be easily insertable in the existing code. Today the technology to implement the requirements mentioned above is available. The Open Model System project initiated by Delft Hydraulics and RIKZ aims to implement the requirements in their operational systems: DELFT3DandSlMONA. In (Blind et al.. 2000) Blind presenis an overview of related work. The focus of that paper is much wider. The background of that paper is mainly the application domain of ID simulations, where the variety of models is much larger and which have other characteristics than 2D/3D simulations. Section 2 introduces the characteristics of the 2D/3D application domain.
2 THE APPLICATION DOMAIN SlMONA and DELFT3D are information systems for computing physical phenomena in water. By means of numerical-mathematical models water levels, water flow, wave heights and transport of currents can be simulated in rivers, lakes, estuary and shelf seas. Simulation of the phenomena often require a lot of computational time and huge amounts of data. To illustrate the relevance of performance in our application domain we included a figure of the mesh for the Rhine-Meuse Delta (Figure 1). The figure presents the curvilinear grid on which the computations are performed. In the horizontal plane, the grid has a maximum of 320 points in the x-direction and 592 points in the y-direction. Vertically, there are up to 10 layers. All together, there are up to 610.000 'active' grid points (e.g. non-land points). For each grid point at least four unknowns have to be solved, which may result in quite some data to be stored or exchanged. For example, when coupling a hydrodynamics and a morphology module (as presented in Section 3). the depth values are exchanged, which means that at least 61000*8 bytes = 0.5 megabytes are exchanged at every time step in the simulation. One day simulation of the Rhine-Meuse model requires one hour of computation time on 32 processors of a massively parallel computer (CRAY) located at SARA in Amsterdam (the domain has been partitioned in 32 parts, a partitioning in 4 parts is presented in Figure 1). which illustrates the fact that optimal performance is indeed an issue here. 3 COUPLING HYDRODYNAMICS AND MORPHOLOGY At present, the first step towards an OMS software architecture has been set. Before elaboration details, a specific illustrative example will be presented as to clarify the approach. The example concerns the coupling of one of SIMONA's hydrodynamics modules (WAQI M and DELFT3D'S morphology system (MORSYS). They can be used to simulate hydrodynamic and morphology processes in for example a river.
An Open Model System for 2D/3D hydrodynamic simulations Paper presented at Hydrolnformatics 2000, 23-27 July, 2000. Cedar Creeks. IA. USA.
15 model, and indicates the required contents of such a data model. For that a two dimensional depth field is used as an illustration.
Figure 2: The hydrodynamics module (WAQUA) of SIMONA, the morphology modules (MORSYS) of DELFT3D and a conversion module as three interacting components in a computation. In Hummel et al., (1999) the example is implemented as a prototype.
3.2 Example of different data representations A depth field is an object that exists in both applications WAQUA and MORSYS. In the applications the field has different representations. That makes conversion inevitable. Before presenting the representations we will first explain the physical meaning of the depth field: Physical description: 2D horizontal depth field, defined with respect to a reference plane in a specific coordinate system (e.g. Paris coordinate system).
3.1 Exchanged data The data exchange between the coupled modules WAQUA and MORSYS is presented in Figure 2. First WAQUA provides water levels, currents etc. assuming a certain river bed profile. Starting from that (and other) information MORSYS computes a new river bed profile, which is input for the next WAQUA simulation step. The two steps are repeated until the process converges. In general the interactions are more complicated than described here (MORSYS is composed out of two other processes, which also interact with WAQUA). In fact a system of equations (which describe the relation between river bed changes, current changes and changes in water levels) has to be solved, so a 'correct' numerical algorithm has to be used. Therefore, the actual exchange of data can be viewed as a 'coupling algorithm", as will be presented in Section 4.2. As will be described in the next subsection, the data representation (e.g. of the flow field) in WAQUA and in MORSYS are different. Therefore the data that has to be exchanged also has to be converted. This is the reason to insert a conversion module between WAQUA and MORSYS. To be able to exchange and convert data between the WAQUA and MORSYS application, a well designed data model (and data dictionary) is required. A data model defines the data and its implementation in the software system. The next section will motivate the need for a data
Since WAQUA and MORSYS are based on the finite difference/finite volume method, the depth field has a discretised representation: Numerical representation: 2D horizontal depth grid function, where the grid is a Arakawa C grid (Kowalik et al.. 1993). The depth grid function returns for each pair of coordinates of a grid point the depth value. Finally the representations of the depth field in the two software packages WAQUA and MORSYS are presented. The actual computations for the flow simulation and the morphological simulation are carried out on a rectangular grid (the computational grid). The software representation of the depth field is based on that rectangular grid, but the internal data representation is different in both applications. Software representation for MORSYS and WAQUA: In both cases two integer variables MMAX and NMAX denote the dimen-
sions of the rectangular grid. In MORSYS the values of the depth field are stored in a real array DEPTH(MSIZE+4,NMAX). The actual depth values at the grid points are the ones stored in DEPTH(3:MsiZE-2. LNMAX)1, but only for points that are marked as active by means of a boolean array A C T I V E ( M S I Z E + 4 , N M A X ) .
An Open Model System for 2D/3D hydrodynamic simulations Paper presented at Hydrolnformatics 2000, 23-27 July. 2000. Cedar Creeks, IA, USA.
16 In WAQUA the depth values are stored in a real array DEPTH(NACTIVE). NACT1VE represents the number of active grid points. The relation between the grid point numbers and the (m.n)-indices is defined by an integer array ACTIVEGRID(MMAX.NMAX).
It contains the
numbering of a grid point denoted by an (m.n)-index (i.e. coordinates in the rectangular grid). The curvilinear coordinates of the grid points are also represented in the code, but they don't play a role in the current coupling. Since the data-representations in WAQUA and MORSYS differ, data-transfer is not trivial. For conversion first the 'semantics' of the arrays has to be studied, for example which index in the MORSYS array corresponds to which index in the WAQUA array. Therefore a data model would be of great use here. In Section 2 we have shown that the amount of data which is transfered between the two modules is enormous, so its a challenge to implement the OMS in such a way that the requirement on a satisfactory performance can be met. Therefore the data exchange has to be restricted to what is really necessary. The example in the present section (communicating a depth field) illustrates how performance can be improved. Since the curvilinear coordinates are not used in the kernel of the computations in MORSYS and WAQUA it is not necessary to exchange the curvilinear coordinates at each simulation step, so data-transfer is restricted to that of the depth values. On top of that the 'conversion information' (the different active point administrations for) is only provided to the conversion module at start up, so it is only communicated once. 4
DESIGN AND IMPLEMENTATION OF AN OMS-ENVIRONMENT The present section presents the most important demands and requirements for an OMSenvironment. and the resulting functional and 'The elements (1:2. I N M A X ) and (MsiZE-l:MsiZE. I:NMAX) are added for implementation reasons.
detailed design. 4.1 Requirements for an OMS-environmenl In Hummel et al.. (1999) some generic requirements like genericity.flexibility,portability, performance, future developments, etcetera, are stated and elaborated in more detail. In the present section two fundamental functionalities for an environment for a complex modeling system are identified: data exchange (including data conversion) and process-coordination. As can be seen in the example of Section 3 data-exchange has to deal with different datarepresentations for the same object in separate software-packages. Conversions from one datarepresentation to another are generally speaking complex. Often model-specific knowledge is required. The use of a neutral data-format is not always efficient. Especially when the differences between neutral data-format and the dataformat of software-package are large. In that case the programming of the conversion can be complex and the performance of the conversion can be low. Process coordination can also become a complex activity. For example in morphology, if not well defined, the iterative processes may not be stable, which means that no solution is found. In general this is an underestimated problem. In fact the coupled simulation should be considered as a solution of a set of very large system of equations. In that system subsystems are solved by the processes. The overall solution is found by solving the system of equations over the subsystems. In general an iterative scheme will be used, so efficient and flexible process coordination definitely is required. Data-exchange between the processes is performed at certain fixed moments, which are known because of the numerical algorithm which solves the system of equations. Instead of synchronous algorithms, asynchronous algorithms (Bertsekas et al.. 1989) could also be used, but in practice they have not proven to be efficient in the numerical sense. As a consequence no event-based techniques are required for the coupling. This is an advantage since event-based implementations suffer from lack of
An Open Model System for 2D'3D hydrodynamic simulations Paper presented at Hydrolnformatics 2000. 23-27 July, 2000, Cedar Creeks. IA. USA.
17 performance. 4.2
Functional Design of the OMSenvironment Based upon the considerations in the previous section we suggested the following concept for an Open Model System. Processes like WAQUA and MORSYS are considered to be components between which data is exchanged. Below a definition of a component will be given. The coupling (and thus the protocol for data-exchange) between the components is specified in a coupling algorithm. This coupling algorithm solves the overall process. A data-model (or datadictionary) defines the data that is exchanged. The data-model also describes the internal representation in the objects that are involved in the data, and therefore facilitates the development of conversion components. Since the concept is heavily based upon components we first have to define 3 component. The definition we use comes from Krachten. (1998). The programming language in which WAQUA and MORSYS are currently programmed is Fortran 77. So for a component, we need a definition broad enough to address conventional components (such as COM/DCOM, CORBA and JavaBean components) as well as alternative ones (such as executables using proprietary communication), yet not so broad as to encompass every possible artifact of a well structured architecture. A component is a non-trivial, nearly independent, and replaceable part of a system that fulfills a clear function in the context of a well defined architecture. A component conforms to and provides the physical realization of a set of interfaces. Krachten further examines the definition: First, a component is non-trivial: it is functionally and conceptually larger than a single class or a single line of code. This is obviously true for models such as MORSYS and WAQUA. In each model a set of equations is solved and they are build upon objects like a depth field, flow field, etc. Second, a component is nearly independent of
other components. It rarely stands alone. A given component collaborates with other components and in so doing assumes a specific architectural context. The WAQUA and MORSYS components assume that they can send and receive data (e.g. a flow field and a depth field, respectively). Third, a component is a replaceable part of a system. A component is substitutable for any other component which realizes the same interfaces. This aspect is the key for an Open Model System. Parts of a system can be replaced by mature, robust implementations or by research versions. It also supports the evolution of a system, once deployed by making it possible to upgrade and evolve parts of the system independently. Fourth, a component fulfills a clear function. A component is logically and physically cohesive, and thus denotes a meaningful structural and/or behavioral chunk of a larger system, e.g. a solver for the shallow water equations. Fifth, a component exists in the context of a well defined architecture. A component represents a fundamental building block upon which systems can be designed and composed. This definition is recursive: a system at one level of abstraction may simply be a component at a higher level of abstraction. Components never stand alone, however. Every component presupposes an architectural and technological context wherein it is intended to be used. The components MORSYS and WAQUA in the example assume the existence of certain objects, such as WAQUA fields, MORSYS fields, etc. They also assume that these object can be exchanged. Finally, a component conforms to a set of interfaces. A component that conforms to a given interface means that it satisfies the contract specified by that interface and may be substituted in any context wherein that interface applies. Krachten, (1998) defines an interface as follows:
An interface is a collection of operations that are used to specify a service of a component. An interface serves to name a collection of operations and specify their signatures and protocols. An interface focuses upon the behavior, not the
An Open Model System for 2D/3D hydrodynamic simulations Paper presented at Hydrolnformatics 2000. 23-27 July. 2000, Cedar Creeks. IA. USA.
IS structure, of a given service. An interface offers no implementation for any of its operations. The interfaces of the 'model' components (e.g. WAQUA and MORSYS) consist out of a set
of subroutines with help of which at one moment a certain amount of data can be requested. The origin of the data is not known for the component which requests the data. In that way the component can be used in different environments. It is hidden whether data is transferred by means of memory (pointers), by means of files, or by means of message passing between other processes. In Section 4.3 examples of subroutines in the interface can be found. The OMS-environment consists on the one hand of components for abstract data-types such as fields (e.g. for depth field), providing a set of high level routines, which are used to couple components. On the other hand the OMSenvironment consists of support level components which are a set of generic subroutines for communication and coordination. In first instance the high level components will be developed for specific applications, but as soon as a certain functionality appears very often a generic component can be created for that. In this way we explicitly choose for a practice driven way of working since it is unlikely that a full specification of the required functionality can be given right at the start of the work. The suppon-level components offer a set of elementary and uniform operations for data exchange and coordination. Dependent on the context in which the components are used (e.g. a network of workstations, a shared memory system or a system providing coupling by means of files), they provide different implementations. 4.3 Detailed design of the OMS-environment. The functional design as described in Section 4.2 has further elaborated into a detailed design, which will be presented in the present chapter. The /n',?/i level components will be discussed first, followed by the support-level will be presented. The definition of a component has been presented in Section 4.2. In the OMS-environment we have two kind of high level components.
which only differ in complexity. One kind of component represents physical models and the other kind represents data conversion2. The components are based upon objects which are physical entities, such as WAQUA FIELD. For such objects interfaces are defined, which amongst others contain operations for data exchange. The component WAQUA and MORSYS only use the objects in their own application domain, the WAQUA-objects and the MORSYSobjects respectively. The conversion component uses the objects from both application domains, such that it can do the conversions. The data exchange operations used in the interfaces of the high level components are arranged in two levels. Firstly we have the level 1 operations which are used for exchange of complete data-structures. There exists a set of operations for WAQUA-IN-SIMONA and a set for
the semantics of those is the same, but the implementation differs. Examples are: put operation for grid-sizes, put operation for non-transient grid information, put operation for field-values, etc. MORSYS-IN-DELFT3D;
The implementation of separate operations for the exchange of grid sizes enables a dynamic creation of data-structures in DELFT3D for example. At any time after creation the grid information can be filled in with help of the get operation for non-transient grid information. Secondly we have the level 2 operations in the interface. These are composed of the level I operations. For example there exists an operation for the exchange of all initial data, that is required for a MORSYS simulation. At the suppon-level we identify basic objects for communication and objects for process coordination and control. The communication objects provide two basic communication operations: PUT (which can also be read as: MAKE AVAILABLE) and get (which can also be read BS: OBTAIN). The first operation provides data which can be used by other processes. The second operation handles the delivery and use of the data. Dependent on the context (i.e. the harde n principle Iherc is no difference between It) two kinds, since one could think of the physical model in the conversion component as a very simple one
An Open Model System for 2D/3D hydrodynamic simulations Paper presented at Hydrolnformatics 2000, 23-27 July. 2000, Cedar Creeks. IA. USA.
19
ware platform etc.) communication is realized by means of a file, shared memory or message passing (PVM, MPI). In the prototype OMSenvironment operations for initializing and stopping the communications, and PUT and GET operations for reals and integers are defined. Since, at the level of the operating system, one or more components can be implemented as a process, a separate component for process coordination and control is required in an OMSenvironment. For example in a multiprocessing environment processes must be started and stopped. The component can either exists as a separate executable (e.g. when the platform is a cluster of workstations) or it can be embedded in the first process that is started (e.g. on a SPMD platform). The component will be based upon low-level operations for starting and initializing processes. Just like the put and the get operations, these process control operations will be worked out dependent on the context in which they are used. In the prototype OMSenvironment (used in the case-study) a simple Executive component has been implemented. It starts the required processes and handles error messages. The control of component processing should obey the coupling algorithm, see Section 4.1 and 4.2. The coupling algorithm specifies which put operation is connected to which get operation. So it connects the interfaces of the components. The coupling algorithm is currently specified in a file which is read by the Executive. The prototype OMS-environment uses currently parts of the CouPLE library developed by VORtech Computing for parallel processing of WAQUA and its '3D version' TRIWAQ. 5 COUPLING HYDRODYNAMICS AND MORPHOLOGY REVISITED In Section 3 we showed that WAQUA computes water levels and currents and sends that information to MORSYS. subsequently MORSYS computes the new river bed after which the depth field is returned to WAQUA. In the present section the process coupling is elaborated. Since water levels and currents are computed on a grid, information concerning the grid has
M\Ot \ : PutWAQUAGrid put NFlowTimeSteps do 1: NFlowTimeSteps calculate flow field PutWAQUAFlow(U) PutWAQUAFlow(V) GetWAQUAField (Bottom) done MORSYS . GetDELFT3DGrid get NFlowTimeSteps do 1:NFlowTimeSteps GetDELFT3DField(UI GetDELFT3DPield(V) calculate transport every 1 to M timesteps do calculate bottom done PutDELFT3DPield(Bottom) done
Figure 3: The pseudo codes for WAQUA and MORSYS
to be exchanged. The pseudo code in Figure 3 shows that WAQUA first provides the grid information (PutWAQUAGrid), which MORSYS receives (GetDELFT3DGrid). After that only the values of the water levels and currents need to be transferred. Sending and receiving grid information, currents, etc. and the conversion from WAQUArepresentation to MORSYS-representation and vice versa are collected in separated components. For example in one component interface the operation PutWAQUAGrid is defined. It provides the WAQUA grid. In another component interface the operation GetDELFT3DGrid which receives the data. Currently it includes the conversion to a MORSYS grid. The information concerning the MORSYS grid is also stored in the component, such that on later occasions it is sufficient to exchange the values of water levels, currents, etc. These values are exchanged by the operations Put/GetWAQUAField and Put/GetDELFT3DField. One of the advantages of the developed
An Open Model System for 2D/3D hydrodynamic simulations Paper presented at Hydrolnformatics 2000. 23-27 July. 2000. Cedar Creeks. IA. USA.
20 concept is that MORSYS. by means of GetDELFT3DField. receives the field in its own data-structure. The conversion from the WAQUA data-structures to the MORSYS data-structures is handled by the OMS-environment. Currently data conversion is included in the receiving process (or thread), which is not necessary. It could be included in a sending process (or thread) or in an intermediate process (such as in Figure 2). The reason for the chosen implementation is performance optimization. The WAQUA datastructure is 'smaller' than that of MORSYS, so time for data-exchange is minimized. The operations PUT... and GET... in Figure 3 are part of the interface of the components WAQUA and MORSYS. The coupling of the interfaces is specified by an algorithm, see also Section 4.2. Currently the interface coupling is stored in a file, the coupling file. In the future it can be generated from the coupling algorithm. 6 CONCLUSIONS AND FUTURE WORK We have presented a specification for an Open Model System environment for coupling models. We have shown that the environment allows an efficient implementation of inter-process communication. Existing software packages (e.g. written in Fortran 77) can use the environment by inserting the specific communication subroutines in the source-code. That means that no wrapping techniques are used and that there is no need to fully restructure the sourcecode. The specification follows a component-based approach offering the accompanying advantages, such as flexibility, etc. Since the specification is a generalization of software which supports parallel computing, parallelization is also supported. In the prototype the put and get operations have been implemented in DELFT3D and SIMONA as a replacement for the original communication method (based on storage in a file), which shows that migration for the two software packages is easily possible. The paper motivates why a well described data-model is required. A first guess for structuring such a description is presented. The structuring is based on the experiences presented
in Cate etal., 1996. REFERENCES Bertsekas, D.P. and J.N. Tsitsiklis, "Parallel and Distributed Computation". Prentice-Hall. 1989 Blind. M.W., A. Ubbels, L.R. Wentholt, Th.L. van Stijn. A.H. Bakema. J.D. Bulens, J.J. Noort, B. van Adrichem, J. Stout, F.C. van Geer, "Towards a well-oiled model infrastructure for water management: the generic framework water program", Proc. 5th int. Hydroinformatics Conf., 2000 Cate. H.H. ten, M.R.T. Roest, and E.A.H. Vollebregt, "On the integration of software packages", Proc. 2th int. Hydroinformatics Conf, 9-13 sept 1996 Cate, H.H. ten, H.X. Lin, R. Salden. and A.E. Mynett. "A case study on integrating software packages, Proc. 3th int. Hydroinformatics Conf., 1998 Hummel, S., M.R.T Roest. and E.A.H. Vollebregt, "Technical documentation on coupling Triwaq-Triwaq and Triwaq-Trisula". Report E.0145, Delft/The Hague. WL | Delft Hydraulics / RIKZ, 1998 Hummel. S.. and M.R.T. Roest, "Pilot-studie Proceskoppeling". Report E.0195. Delft/The Hague, WL | Delft Hydraulics / RIKZ (in Dutch). 1999 Kowalik. Z.. and TS. Murty, Numeric Modeling of Ocean Dynamics, World Scientific, 1993 Kruchten. P., "Modeling Component Systems with the Unified Modeling Language, "International Workshop on Component-Based Software Engineering", 1998 Scherjon. F, and A. Staakman. "Demonstrator: Architecture and design of complex modeling systems." The Netherlands: LWI Foundation Report (in Dutch), November 1997
An Open Model System for 2D/3D hydrodynamic simulations Paper presented at Hydrolnformatics 2000. 23-27 July. 2000. Cedar Creeks. IA. USA.
21
Good Modelling Practice in water management H. Scholten Wageningen University. Applied Computer Science Group, Wageningen. The Netherlands R.H. van Waveren RIZA. Institute for Inland Water Management and Waste Water Treatment. Lelystad. The Netherlands S. Groot WL | Delft Hydraulics. Delft. The Netherlands F.C. van Geer The Netherlands Institute of Applied Geoscience TNO, Delft. The Netherlands J.H.M. Wosten Alterra. Green World Research. Wageningen. The Netherlands R.D. Koeze WL | Delft Hydraulics. Delft. The Netherlands J.J. Noort STOWA. Foundation for Applied Water Research. Utrecht. The Netherlands
ABSTRACT: Despite its ever increasing capability to solve complex problems, empowered by exponential growing computer potentials, modelling and simulation is a frequently used, but ill-defined methodology. This ends in irreproducible model based studies of unknown quality. To assess these shortcomings a Good Modelling Practice handbook for water management has been developed, in which the modelling and simulation process is defined in detail and modellers are made familiar with model sensitivities and the pitfalls in which other modellers are trapped. This handbook has been achieved by negotiation with representatives of all stakeholders involved.
INTRODUCTION
Modelling and simulation is nowadays a standard tool in the toolbox of many workers in the different fields of applied sciences. In The Netherlands several hundreds of modellers are involved in modelling for water management. Modelling might be a routinely applied method to investigate problems, it is not straightforward, but rather subjective, depending on the modelling team and its skills. Therefore one often refers to it as to die 'art
of modelling'. This artistic and creative label sounds as a positive designation, but it stresses the unscientific and ambiguous aspects of modelling. The major risks of modelling are related to the many choices that have to be made, the complexity of the problem and the object system at hand, often inadequately supported by an incomplete and controversial theoretical body of knowledge. This results in many uncertainties in the model and in its results. Recently several Dutch newspaper articles discussed the arbitrariness of making policy
An Open Model System for 2D/3D hydrodynamic simulations Paper presented at Hydrolnformatics 2000, 23-27 July, 2000, Cedar Creeks. IA, USA
22 based on models without a proper reference to and quantification of uncertainties involved. Most scientists engaged in modelling and simulation are aware of these problems and many initiatives have been developed to improve the quality of models and modelling. The approach discussed in this paper is one of the most comprehensive examples, initiated within the Generic Framework Water Program (Blind et al., 2000). In this program all main players in Dutch water management co-operate to improve the efficiency and quality of modelling in water management in The Netherlands. The stakeholders decided to tackle this problem by developing a technological standard (an IT framework, discussed in Van der Wal et al., 2000), and a conceptual standard, using a Good Modelling Practice approach, discussed in this paper. METHODS
Capability Maturity Model (CMM), which was developed to improve software engineering (Humphrey, 1989). Just like CMM, SMM distinguishes five stages of maturity: ad hoc, repeatable, defined, managed and optimized. In this study the definition stage is the most essential stage, as defining the modelling and simulation process promotes its repeatability and facilitates an efficient and effective audit. The definition stage of SMM has been used as the backbone of a Good Modelling Practice (GMP) handbook, in which the modelling and simulation process and its products have been defined. This GMP approach has to be accompanied by normal software engineering quality assurance efforts, but at present most of these efforts are based on the ISO-9xxx paradigm, which allows a (too) flexible implementation and does not guarantee anything at the level of GMP and its products.
Quality and simulation models Towards an ontology of the M&S process Simulation models are used habitually in all domains of water management. The easiness to use models enlarges the risk of injudicious use, often caused by the modeller. Careless handling of input data, insufficient calibration and validation, using a model outside its scope are examples of malpractice, which result in unreliable model outcomes. Partly overlapping with this is a modelling approach without predefined quantitative requirements to models used in model based studies. The consequences of this practice are irrepeatable model based studies, with an unknown trustworthiness, not matching predefined prerequisites. The qualiry improvement process Quality management is quite common practice in software engineering, but in the field of modelling and simulation quality management is often restricted to verification and validation issues. Scholten & Udink ten Cate (1999) have proposed a Simulation Maturity Model (SMM), comparable to Humprey's
In the field of knowledge engineering, the word 'ontology' has been defined as 'an explicit specification of a conceptualization' (Gruber, 1993), and more specific, an ontology represents shared knowledge which can be reused. Many authors have proposed schemes to define the modelling and simulation process, of which several divide the process in activities to build or test intermediate products of the modelling and simulation process (Sargent. 1982, 1984a,b, Knepell & Arangno, 1993, Scholten & Udink ten Cate, 1999). Most of these schemes lack approval of a large group of modellers. Our efforts to define the modelling and simulation process focus on general agreement and negotiation, both from a user-participation point of view, as well as from a knowledge-acquisition point of view. In our attempt to improve the overall quality of modelling and simulation in water management, we formulated an ontology
Good Modelling Practice in water management Paper presented at Hydrolnformatics 2000,23-27 July, 2000. Cedar Creeks, IA, USA.
23 »tep1
step 7 start model journal
check model journal and templates
step 2 report and file
describe P'oDiem
I
step 6
define objective
1 analyse consequenc. for problem
describe context
I
I
specify requirements
agree on justification
check objective
I
make a working plan
describe conclusions
step 3 discuss results
describe results
step check run
cnoose a model program
•»-*j
choose numerical approach
puttnlnary inspection nMUtit
choose discretisation (space/time)
implement model
M OK
make plan application run
verify model
not OK
-ii. P-T-
step 4 check mass balances
run with standard input
validation
not OK robuustness test
global behaviour test
carbaHgn
uncertainty analysis
not OK MMTMy analysis
formal ioenUication OK
Figure l. The first two decomposition levels of the Simulation Modelling Process.
Good Modelling Practice in water management Paper presented at Hydrolnformatics 2000. 23-27 July. 2 0 0 0 , Cedar Creeks, IA, USA.
OK
24 on modelling by decomposing the modelling and simulation process in steps. This decomposition was continued to an abstraction level, at which modellers in water management all do more or less the same things. This ontology was subsequently used as the core of a Good Modelling Practice handbook. The seven steps of the first and the steps of the second decomposition level are shown in Figure 1.
Good Modelling Practice handbook The Good Modelling Practice handbook has been written in a stepwise approach, based on the cross-fertilization within the multidisciplinary project team, an active supervising committee, the strong commitment of the 50 participants in a workshop and, finally, on the efforts of two series of experienced and inexperienced 'field testers' (users, not involved in developing the product). The handbook will not be a static document, but it will be updated every 6 to 12 months and perhaps additional tools will be developed to support its use. Users are encouraged to share their experiences with the authors, or with the organization, which is responsible for updates and adaptations. RESULTS
handbook is its ontology of shared and approved 'simulation and modelling concepts'. The consensus on the shared views on modelling and simulation was a surprise for many of the modellers in the field of water management. The core of the handbook The backbone of the handbook is the structured ontology. At die highest level of the decomposition the steps are: (1) starting with a logbook, (2) defining the modelling project, (3) building the model, (4) analyzing the model, (5) using the model, (6) interpreting the results, and (7) reporting and archiving (Figure 1). Some of these steps are further decomposed to a level of a single modelling activity (e.g. determining for which factors the model is most sensitive or checking if all model objectives are met). The core part of the handbook includes also a series of tests widiout prescribing with which methods or widi which algorithms these tests have to be carried out. These tests comprise of conceptual model validation, some aspects of verification (dimension check, mass or energy balance control), a robustness test, a sensitivity analysis, a calibration, a (historical data) validation, and an uncertainty analysis. In this way the handbook supports all activities related to modelling and simulation and passing these tests improves the credibility of the model.
Outline of the handbook Some details The GMP handbook should represent a selfexplaining document to support the entire procedure of the modelling and simulation process. It consists of a clear demarcation of the types and domains of models for which it is intended, a glossary of all concepts, a structured ontology of the modelling and simulation process, a checklist and summary, a tool to document and archive the many steps and tests in the modelling and simulation process, the collective experience of large group of modellers on pitfalls and sensitivities in general and for specific modelling domains and finally references to specific literature and addenda on specific problems. The major payoff of developing such a
Concepts The ontology of the modelling and simulation process is based on an unstructured list of basic terms. A subsequent paragraph on the conceptual framework describes the relations between these terms, spanning up die domain of modelling and simulation. Conceptual modelling A conceptual model describes the functional relationships between those components of the object system (reality) that will be included in the model. The representation of
Good Modelling Practice in water management Paper presented at Hydrolnformatics 2000, 23-27 July, 2000. Cedar Creeks, IA, USA.
25 tions behind a model are made fully explicit. A difficult, final step here is the validation of the conceptual model, especially as there are no formal methods for this.
system: input data. analysis data
Calibration
describe structure
define relations
establish assumptions
not OK
verify conceptual model
not OK
OK choose a
model program
Figure 2. Developing a conceptual model, such a conceptual model may be textual or in mathematical equations, supported by drawings, graphs or diagrams. A conceptual model is particularly meant to summarize the ideas behind a model, when discussing these with other stakeholders. The major steps in developing a conceptual model are summarized in Figure 2. An essential step in conceptual modelling consists of the definition of the boundaries: what has to be included and what not (i.e. how far in time and space die model has to be larger than the object of study), which peripheral processes are relevant and what kind of interactions have to be modelled. It is important to find a balance between the degrees of detailing of the various elements of the model. Further, one has to choose the type of model (which independent variable(s), how many dimensions). Essential elements of a conceptual model are both die structure (which (state) variables, which input) and the relations between these. A reconstruction of a modelling and simulation project is only possible, if all assump-
One of the most important activities of the modelling and simulation process is the model analysis. This may range from a set of simple, preliminary tests, via a sensitivity analysis and a calibration, to validation and determining the scope. Of these analyzing activities, calibration often consumes most resources, and therefore it will be discussed in some more detail. Calibration aims at the reduction of differences between field observations and corresponding model outcomes. All methods for calibration change model parameter values and evaluate some objective function to get a better fit and reach the calibration goals. In this way. calibration is transformed to an optimization problem. The handbook distinguishes several steps in calibration (Figure 3) and these will be discussed here briefly. First one has to choose which parameters will be calibrated, based on the results of a sensitivity analysis and on the expertise of the modeller. In many cases the available field observations do not allow to calibrate all parameters. 'Well known' parameter values can be kept constant and sets of parameters maybe treated as a single parameter (zoning) or be replaced by a geostatistical relation. Specification of an objective function will often ends with a sum of least square approach. In the multivariate case (more model variables have data) the situation is more complex, as this belongs to the domain of multi-criteria-analysis. The next step consists of choosing a proper method to perform the parameter optimization. First one has to choose whether to do it 'by hand' or by using an automatic optimization algorithm. A modeller learns more with the former approach, but the latter will generate reproducible results and calibration in this way is in fact a sort of sensitivity analysis (i.e. learning how the model responds to changes in parameters and other calibration factors). Nowadays there are
Good Modelling Practice in water management Paper presented at Hydrolnformatics 2000. 23-27 July. 2000, Cedar Creeks, IA, USA.
lb
many algorithms to do the job, but most software packages provide only a few (or even less). WnWivit, anaiyi*
1
>
Forms
choose ophmieation parameters
1
Mwcrfy o U t d .-.• function
1
choose optimization method
—
define Mop cmena
1 do parameter optimization
1 WOK
A major part of the handbook consists of a series of fill-in forms, which facilitate keeping records of the modelling and simulation process in a logbook. These forms have a 1:1 relationship with the ontology, i.e. the decomposition steps of the modelling and simulation process. In this way the handbook supports a proper management of the modelling and simulation process, both for experienced and inexperienced modellers. Sensitivities and pitfalls
analyse rretrtuat emx
i uncenamy XMlysn
Figure 3.
process. With or without a new sensitivity analysis one can choose new calibration parameters, or change the model (discretization, numerical approach, or the conceptual model) or even collect more field data.
>
r tiec.ilibration step S.
Simultaneously with choosing an optimization method, one has to specify a stop criterion for the optimization (i.e. calibration), if the software does not provide this automatically. The best criterion is: stop when one (or in other cases all) parameter vectors) give model outcomes which deviate less from the field data than the criterion. Other stop criteria use the rate of convergence or similar performance aspects. In the next step the actual optimization is carried out. This is a time consuming step, especially as it often had to be repeated several times. The final step consists of the evaluation of the calibration results. This includes two major aspects: has the stop criterion been met and are the residual errors small and not systematic? If one of these questions has a negative answer, one has to go back to previous steps in the calibration or even to previous steps in the overall modelling and simulation
The handbook contains also descriptions of many domain independent pitfalls and sensitivities of model based studies in water management. Further this part of the handbook consists of a summary of specific problems with models in the 13 domains of water management covered by this handbook (groundwater quality and quantity models, hydrodynamic models, ecological models, etc.). This part on sensitivities and pitfalls has the same structure as the ontology, which enables an efficient lookup when following the network of steps of the modelling and simulation process. The list of items of the general part of 'sensitivities and pitfalls', which have to be avoided, includes: • •
•
not starting a model journal: this makes model based studies hard to reproduce; not defining in an early stage what results are needed, especially in the case of chains of models; not specifying early requirements to the size of the area in view of the boundaries, to a proper choice of spatial and temporal resolutions (scales) in view of the problem at hand and to the choice of the model in view of the required functionality;
Good Modelling Practice in water management Paper presented at Hydrolnformatics 2000. 23-27 July, 2000. Cedar Creeks. IA, USA.
:7 not tuning the manager's model objective to its technical translation; not defining a sound model concept with a proper choice of processes: errors widi this will easily be masked in a calibration; specifying a model with too much details compared to the available data: choosing the wrong software for the problem at hand; allocating too little resources (time) for model analysis; not specifying criteria to the model accuracy at an early stage; improper harmonizing the number and choice of model parameters to calibrate with the available data and its measuring frequency; using the theoretical concepts of observability will avoid this; using a mix of insensitive and sensitive parameters in calibration: calibration of the latter will be hindered by the former; comparing point observations with model outcomes representing an area or a volume in any model evaluation; believing the model is good since the calibration is perfect; using a model outside its scope: this often occurs in prediction and scenario analysis; not using sufficient run-in time before the actual simulation period; not accounting for output uncertainty (ranges, distributions) when comparing different scenarios; improper jargon in communicating the results to clients (decision makers); restricting communication of the results to written reports: verbal communication may avoid incorrect use of the results. Sensitivities and pitfalls of specific domains of water management will not be discussed here, as these will interest specialists only. The final component of this part of the G;MP handbook consists of bloopers, illustrating the effects of the sensitivities and pitfalls.
References Without attempting to provide all relevant literature, all topics discussed in the book refer to some relevant references allowing nonspecialist modellers to gain more in-depth knowledge of the subject at hand. Results of the tests The results of the preliminary tests during the development of the handbook are incorporated in the text in a stepwise approach. The tests by experienced modellers indicated that the handbook is an adequate tool to support the overall modelling and simulation process, forcing modellers to work in a structured manner, without being a straitjacket by strictly prescribing certain methods. How it improves reproducibility of a model based study is still an unanswered question. Further test results are now collected. Especially those of students are easy to get and illustrate the problems encountered by inexperienced modellers. All test results, as well as other comments on the handbook will be used for updates. DISCUSSION There does not exist a simple or universal solution to all modelling and simulation problems, but this handbook is a tool to increase model quality and the repeatability or better reconstructability of the simulation and modelling process. Our approach (van Waveren et al., 1999) has drawn attention of many national and international organizations involved in modelling and simulation. At die moment these include research institutes, governmental organizations and universities, covering many domains in water management, but also landuse planning and nature conservation. The IS09xxx guidelines are the most related activity to this GMP handbook. These define the business process and are adopted nowadays by many companies and organizations. Despite the enormous improvements of the business processes in general, and those of for instance the software engineering process particularly, there are no ISO-guidelines for
Good Modelling Practice in water management Paper presented at Hydrolnformatics 2000. 23-27 July. 2000. Cedar Creeks. IA. USA.
28 modelling and simulation. In some cases ISOguidelines are used in model based studies or in operational water management, but the ISO-guidelines do not refer to any knowledge on modelling and simulation, as does the ontology in the handbook. In ttiis way the handbook may be an extension of the ISOguidelines, at least as the stakeholders agree upon all of its content. In modelling and simulation for water management quite a number of different stakeholders are involved. These include the customer or client, who has to pay for the job and decides on the requirements and acceptance, the model builder, domain experts, the project manager, the model user (often called modeller), the policy/decision maker, the (internal) tester, and the (external) auditor. All these actors can profit of the handbook, whether in formulating simulation projects, recording its progress, assessing its quality, or otherwise. Based on the experiences until now. it can be concluded that using the Good Modelling Practice handbook structures model based studies in water management, improves its quality and enables reconstruction of its results. REFERENCES Blind, M.W., A. Ubbels, L.R. Wentholt, Th.L. van Stijn, A.H. Bakema, J.D. Bulens. J.J. Noort, B. van Adrichem, J. Stout, F.C. van Geer, "Towards a welloiled model infrastructure for water management: the generic framework water program". Proceedings Hydrolnformatics 2000, 23-27 July 2000, Cedar Rapids, IA, USA, 2000. Gruber, T. R., "Towards principles for the design of ontologies used for knowledge sharing," Stanford University, Stanford, CA. USA, Technical Report KSL 93-04, 1993. Humphrey W.S., "Managing the software process", Addison-Wesley, Reading. 1989. Knepell P.L. and D.C. Arangno. "Simulation validation: a confidence assesment methodology," IEEE Com-
puter Society Press, Los Alamitos, CA, USA, 1993. Sargent R.G., "Verification and validation of simulation models," In: Progress in modeling and simulation (F.E. Cellier, Ed.), Academic Press, London, etc, 159-169, 1982. Sargent R.G., "Simulation model validation," In: Simulation and model-based methodologies: an integrative view (T.I. Oren, B.P. Zeigler and M.S. Elzas, Eds.), 10 in the series: NATO ASI Series F: Computer and Systems Sciences, Springer Verlag, Berlin, etc., 537-555, 1984(a). Sargent R.G., "A tutorial on verification and validation of simulation models," In: Proceedings of the 1984 winter simulation conference (S. Sheppard, U. Pooch and D. Pegden, Eds.), 115-121, 1984(b). Scholten H. & A.J. Udink ten Cate, "Quality assessment of the simulation modeling process." Computers and Electronics in Agriculture, 22. pp. 199-208, 1999. Van der Wal, T. & M.J.B. van Elswijk, "A generic framework for hydroenvironmental modelling". Proceedings Hydrolnformatics 2000, 23-27 July 2000. Cedar Rapids, IA. USA, 2000. Van Waveren, R.H.. S. Groot, H. Scholten, F. Van Geer, H. Wosten, R. Koeze, J. Noort. "Good Modelling Practice Handbook," STOWA, Utrecht, RWS-RIZA. Lelystad, The Netherlands, http://www.Waterland .net/'riza/aquest/projecten/modellen/gmph. shtml, 2000.
Good Modelling Practice in water management Paper presented at Hydrolnformatics 2000.23-27 July. 2000, Cedar Creeks. IA. USA.
29
Watermark: a general quality mark for software in the Netherlands, from idea to reality L.R. Wentholt STOWA. Foundation for applied Water Research. Utrecht, the Netherlands
J.J. Noort SEPRA. Berkel en Rodenrijs. the Netherlands B. van Adrichem EDS International BV. Leidschendam. the Netherlands J.H.P.M. Tacke EDS International BV. Leidschendam. the Netherlands
ABSTRACT: In the past decade, national and regional water management in the Netherlands is characterized more and more by an integrated, multidisciplinary approach. Such an approach implies an increasing need for close cooperation between the various water authorities on the one hand and both commercial and non-commercial organizations on the other. It goes without saying that the key to successful cooperation lies in the strength of mutual agreement on various topics. For example: Watermark defines standards certifying, in a broad sense, the quality of software systems as applied in water management. Compliance with standards like Watermark appears to be a prerequisite for successfully linking software systems, originating from different sources. This paper highlights the basic ideas and the successive steps to be taken towards a generally accepted Watermark.
INTRODUCTION Driven by the necessity for an integrated, multidisciplinary approach of water management in the Netherlands, parties involved have joined forces to combine their knowledge. This knowledge is expressed in software such as mathematical models and database systems. In order to combine knowledge successfully it is of the utmost importance to link those software systems. This is discussed in detail elsewhere in these proceedings; see Blind et al. (2000). However, when combining different products, the result is about as strong as the weakest part. Therefore, the parties involved consented to mutual agreements certifying quality of the software systems as an essential prerequisite for
cooperation. These agreements describe for example interfaces between systems, good modelling practices and quality ensuring software standards like 'Watermark'. WHY WATERMARK The rationale behind Watermark is that the only way to get well-behaved and trustworthy software is to have an accepted set of rules by which the software should be developed. This way, a party can ensure that its product is not a once-only development, but a supportable and maintainable system with a reasonable life span, where reported bugs are repaired and new insights are added.
Watermark: a general quality mark for software in ihc Netherlands, from idea to reality Paper presented at Hydrolnformatics 2000.23-27 July, 2000, Cedar Creeks. IA, USA.
M)
As a rule an approach like this has several pitfalls. A standard has to be sound, being neither too strict nor too lenient. Furthermore, a standard has to be accepted by the people using it, and a standard should not exclude third-party software. Last but not least, a standard must be enforced, developed and maintained. In short. to obtain (a) Watermark, an organizational structure has to be arranged. Although Watermark initially focuses on rules, standards and organizational structure, the main objective is to guarantee the end user good support and software expandability. For this purpose the availability of appropriate documentation is of paramount importance. SOFTWARE QUALITY Many organizations throughout the world deal with the subject of defining software quality characteristics. Well-know institutes like ANSI (the American National Standards Institute), ISO (the International Organization for Standardization), CEN (the European Committee for Standardization) and IEC (the International Electrotechnical Commission) promote the development of standardization in many areas, including Information Technology. Of particular interest are standards that relate to the software development process as well as to the product (the software or software system) resulting from it. For example ISO-9001: 'Model for quality assurance in design, development, production, installation and servicing" may apply to software development. However, compliance with this standard is no guarantee for the caliber of the software itself. The quality of software systems is subject of various specific standards, such as ISO/ANSI standards on programming languages like Fortran, C/C++ and so on. Furthermore, ISO-9126: 'Software product evaluation - Quality characteristics and guidelines for their use' deals primarily with the definition of quality characteristics to be used in the evaluation of software products. In total six main quality characteristics are distinguished: functionality, reliability, usability, efficiency, maintainability and portability. In 1996 SERC (the Software Engineering Research Centre, the Netherlands) improved the ISO-9126 model to a so-called Extended ISO-model of Software Quality (see Zeist et al..
1996). In this model the software quality characteristics are expanded with additional indicators and metrics, to quantify the extent to which the specified characteristics are actually met. PLAN OF ACTION When the major parties involved in water management in the Netherlands reached an agreement on defining a set of standards to ensure the quality of their software systems, a plan was made to develop such a set. This plan consists of four stages: 1. Defining the standards for software. 2. Defining the rules of engagement for the support and maintenance of the software. 3. Defining the minimal functional demands for the software. 4. Defining the standards for third-party software. Apart from these stages, it was seen to be vital to ensure acceptance and continuity of the Watermark by defining an organization that bears responsibility for its development and maintenance. Standards for software The standards for software have as a primary goal to define a setting in which support and maintenance can be organized successfully. As numerous standardization efforts have shown, it is far from simple to define a workable standard that satisfies the objectives. A standard must be strict enough to guarantee that the software can be maintained and supported, but flexible enough that people can and will conform to it. For this reason, de facto standards as well as widely accepted industry standards are looked upon as likely candidates. These standards have in common that they are both fitting and usable. Because of the broad field of water management and the numerous development platforms used, the first standards to be developed are standards concerning the use of algorithmic programming languages like C and Fortran (77 and 90), (more or less) object oriented languages such as C++, Java, Delphi and Visual Basic and specific platforms like GIS. For the near future it seems that industrial standards
Watermark: a general quality mark for software in the Netherlands, from idea to reality Paper presented al Hydrolnformatics 2000. 23-27 July. 2000. Cedar Creeks. IA. USA.
31 such as CORBA, DCOM and OpenGIS are becoming more important. International ISO/ANSI-standards for programming languages primarily deal with the syntax and semantics of those languages, i.e. the standards specify the form and establish the interpretation of programs written in a specific language (Fortran, C/C++, etc.). The purpose of such a standard is to promote portability, reliability and efficient execution of programs for use on a variety of computing systems. In addition to these standards it is essential to prescribe extra programming regulations with a view to guarantee maintainability, reusability and expandability of the software. Such regulations reflect, among others, to an orderly, welldocumented and structured way of writing source code. At this time, additional standards are available for Fortran77 and C; a standard for Fortran90 is under development. Support and maintenance The ultimate goal is to ensure well-supported and maintained products. In the end, this is the responsibility of the various parties. It is however desirable that certain minimum requirements will be met. These requirements deal for example with response-time when a question arises, the number of product releases per year and the number of staff capable of handling questions. In case a contractor is responsible for the support the requirements will be made clear in the service level agreement; in those situations the support is provided by the party itself, requirements must be made more specific. In both cases the minimum set of demands agreed upon must be maintained and supported. Minimal functional demands The minimal functional demands can be divided between general demands and specific demands. The most important general demand is that the product is compliant with the functional descriptions (read interfaces) defined by the generic framework water; see Blind et al. (2000). This ensures that the product can cooperate with other products and with generic tools. When this framework will evolve, the products using this framework must evolve as well.
Which specific set of functional demands must be met is open to discussion. Obviously the functional demands for a computationally intensive model will be quite different from demands for a relatively simple spreadsheet model. The minimal demands require that these specifications are made public. It is for example necessary to know the accuracy of the model. Third-party software It would be arrogant to believe that a small local group can enforce a standard in the entire world. It would be just as arrogant to believe that the entire world, overwhelmed by the brilliance of the standard will submit itself voluntarily to such a standard. Nevertheless, some standard must be available to ensure the quality of third-party software, because this kind of software will always be used. The development of such a standard is subject of investigation in the near future. ENSURING SUCCES The greatest challenge in setting a standard is not in defining one but in getting the standard accepted and used by the intended audience. Although several critical success factors have been met, a lot of work still needs to be done. Because many of the parties that should use the standard are to some degree involved with the definition of the standard, acceptance of the standard by organizations will be a minor problem. However, in general these organizations do not develop the software, but use contractors to do the actual development. Most of these contractors are not involved in the definition phase. The contracts, defining the work to be done should state that the work must comply with the Watermark-standard, and it should be a criterion for the acceptance of die products. This criterion must be made clear before starting the work. and. when appropriate, whenever a development is tendered. A second necessity is control. A standard that is not enforced and not controlled will tend to lose its effectiveness. Because not all of the parties involved have the capability to perform the necessary tests, a trusted third party should be engaged to perform assessments.
Watermark: a general quality mark for software in the Netherlands, from idea to reality Paper presented at Hydrolnformatics 2000.23-27 July. 2000. Cedar Creeks. IA, USA.
32 But the most important success factor is, as always, communication. The standard and the rationale behind the standard must be communicated as often as possible. It should be plain to developers that Watermark is not invented nor intended to make their lives miserable, but to make their great achievements even greater. Communication is not the responsibility of a small committee; it is the responsibility of all parties involved, and it should not stop at an obligatory leaflet or a once-only presentation. CONCLUDING REMARKS Watermark is one of the four cornerstones of the generic framework water program in the Netherlands. The major players in Dutch water management participate in die project. Standards to be set will be based both on international standards as well as on specific in-house standards of the participating institutes. At this moment (spring 2000) stage 1 and 2 of the plan of action are being carried out. It is most likely that consensus between the participating members will be reached in the near future, thus accomplishing a first step towards success. REFERENCES Blind, M.W., A. Ubbels, L.R. Wentholt. Th.L. van Stijn, A.H. Bakema, J.D. Bulens, J.J. Noort, B. van Adrichem, J. Stout and F.C. van Geer, "Towards a well-oiled model infrastructure for water management: the generic framework water program". Proceedings Hydrolnformatics 2000, 23-27 July 2000, Cedar Rapids. IA, USA. Zeist, B. van, P. Hendriks, R. Paulussen and J. Trienekens. "Kwaliteit van softwareprodukten; praktijkervaringen met een kwaliteitsmodel". Kluwer - Deventer, the Netherlands, 1996.
Watermark: a general quality mark for software tn the Netherlands, from idea to reality Paper presented at Hydrolnformatics 2000, 23-27 July. 2000, Cedar Creeks, IA. USA.