Documente Academic
Documente Profesional
Documente Cultură
Albert Einstein observed that our approach to problem solving should be ‘‘as simple as possible
but not simpler’. This philosophy appears to be a firm tenet for modeling environmental systems.
În cursul sec. XIX au fost derivate diverse legi care descriu mişcarea/transferul căldurii, electricităţii
şi a apei subterane printr-un mediu continuu. Aplicând principiul conservării masei şi
încorporând legea Darcy ca relaţie constitutivă se obţine o ecuaţie cu derivate parţiale care
descrie sarcina piezometrică în cadrul mediului poros. Odată determinată sarcina piezometrică
putem aplica legea Darcy pentru a obţine vectorii viteza de curgere în tot sistemul. Aceste principii
constituie baza pentru modelele de curgere şi transport.
Pentru multe probleme de ape subterane cu geometrie simplă, distribuţie simplă a
parametrilor, condiţii la limită simple se pot determina soluţii analitice ale problemei matematice.
Deoarece există o analogie între curgerea apei subterane şi transferul de electricitate sau căldură
adesea se poate găsi o soluţie matematică în literatura de specialitate.
Analogia între transferul termic şi curgerea apei subterane a fost marcată în anii `30 de
modelul Theis (1935) – sugerează că curgerea apei subterane în regim nestaţionar este analogă
transferul termic tranzitoriu şi în anii `40 de modelul Jacob (1940).
În anii 1940-1950 numeroase probleme au fost rezolvate pentru forajele de pompare.
Numeroase proceduri de interpretare a testelor de pompare au fost dezvoltate. Multe din soluţiile
clasice presupun integrarea numerică a funcţiilor matematice.
Acviferul este un sistem care îndeplineşte diverse funcţii şi care trebuie gestionat pentru a
atinge diverse obiective.
Managementul unui sistem presupune luarea unor decizii (=atribuirea unor valori numerice
variabilelor de decizie) cu scopul de a modifica starea sistemului, pentru a atinge anumite
obiective.
Un grup de la USGS, condus de Herb Skibitski, a pus la punct o reţea electrică rezistor-capacitor
ca un model analogic pentru curgerea apei subterane. Astfel s-a introdus în hidrogeologie ideea
de model a unui sistem acvifer.
Eforturi au fost făcute în paralel şi în industria de petrol de către inginerii de zăcământ care
simulează curgerea în rezervoare de hidrocarburi reale, cu ajutorul computerelor. Unii dintre cei
mai buni matematicieni din anii `50-60 au lucrat la dezvoltarea de metode numerice pentru
rezolvarea ecuaţiilor de curgere în medii poroase. Terminologia folosită era simularea rezervorului
în loc de modelare.
Pe măsură ce creşte puterea şi ponderea computerelor digitale, metodele analogice din anii
`50-60 folosite în hidrogeologie au fost înlocuite cu metode numerice (pentru calculator) în cursul
anilor `70. Modelarea devine din ce în ce o practică curentă în hidrogeologie. Programele numerice
pe calculator au avantajul că pot modela şi transportul masic (în soluţie). Setul de ecuaţii
diferenţiale cu derivate parţiale poate fi rezolvat simultan. Modelele analogice se ocupă doar de
soluţia ecuaţiei de curgere a apei subterane. Computerele aduc o nouă dimensiune modelării. O
întreagă industrie a apărut şi s-a dezvoltat pentru a produce modele de curgere şi transport. Există
o serie de pre- şi post- procesoare pentru Modflow şi MT3D – cele mai comune programe de
curgere şi transport, care fac modelarea mai uşoară şi permit rezolvarea de probleme complexe
(reţele mari, număr mare de noduri).
Odată cu dezvoltarea modelului/ecuației Theis (1935), hidrogeologii au fost preocupați de
dinamica răspunsului unui acvifer față de un stress (pompare sau realimentare). După ce Theis
(1935) și Jacob (1940) au arătat analogia dintre curgerea apei subterane și curgerea căldurii,
comunitatea hidrogeologică s-a implicat puternic în găsirea valorilor la limită corespunzătoare
diverselor scheme de dezvoltare.
Anii 1940-1950 au fost perioada în care preocuparea principală a comunității
hidrogeologice a fost curgerea către un foraj singular, fiind elaborate multe soluții. Aceste soluții au
fost folosite pentru a prognoza răspunsul sistemului acvifer și pentru a estima proprietățile
acviferului (T, K, a și S).
Hidrogeologii acelei perioade au sesizat limitările analizei forajelor și au căutat o metodă
mai robustă prin care să analizeze întregul acvifer, incluzând condițiile la limită și heterogeneitatea
acviferului. Astfel, în anii `50, un grup de la USGS condus de Herb Skibitski a elaborat un model
analogic electric pentru un acvifer, constituit dintr-o rețea cu diferențe finite de rezistori și
condensatori: transmisivitatea T era reprezentată prin rețeaua de rezistori, coeficientul de
înmagazinare elastică S prin rețeaua de condensatori. Această rețea era excitată prin generatori
de curent care simulau pomparea (sau alte stresuri). Voltajul este echivalent cu sarcina hidraulică,
curentul electric este echivalentul curgerii apei subterane.
De fapt acestea erau modele elegante cu diferente finite ale sistemelor acvifere. Au fost
dezvoltate (în anii 1960, USGS – Phoenix, Arizona) modele analogice pentru acvifere multistrat,
unele având până la 250000 noduri. La acel moment nu era fezabilă rezolvarea cu ajutorul
computerelor: acestea erau prea mici și prea lente. Odată cu anii 1970 puterea computerelor
digitale a crescut și modelele digitale încep să devină competitive cu modelele analogice. În anii
1980 modelele pe computere digitale înlocuiesc modelele analogice, chiar și la USGS. Ele includ
transportul în soluție, pre- și postprocesare și estimare automată a parametrilor. Majoritatea
problemelor de curgere a apelor subterane erau simulate folosind codul MODFLOW dezvoltat de
USGS.
2.Obiectivele modelării
Utilizarea modelelor matematice în hidrogeologie este un domeniu specializat care cere experienţă
şi pregătire adecvată pentru a fi abordat. Sute de coduri au fost realizate pentru simularea curgerii
şi transportului poluanţilor în sistemele acvifere.
Prin model se poate înţelege orice structură, entitate, mijloc de aproximare a unei situaţii din teren.
Conceptul de "model" are conotaţii diferite, în funcţie de contextul în care este utilizat. Metodele
analitice sunt bazate pe modele matematice simplificate, care nu necesita întotdeauna prezenţa
unui computer. Termenul de "program" (sau "cod") are un înţeles precis, referindu-se direct la
modele destinate a fi utilizate prin intermediul unui computer.
Tabel - Semnificaţii ale principalilor termeni utilizaţi în modelarea curgerii apei subterane
Termen Definiţie
Model a. reprezentare a unui sistem real sau proces;
b. ansamblu de concepte sub forma ecuaţiilor matematice care permit caracterizarea unui fenomen
natural.
Model Interpretare – descriere sintetizată - a caracteristicilor şi comportamentului dinamic ale unui sistem
conceptual fizic.
Model cadru de ecuaţii matematice ce exprimă funcţionarea unui sistem fizic în limitele unor ipoteze
matematic simplificatoare.
Condiţii de Set de expresii matematice ce exprimă starea unui sistem fizic ce condiţionează modelul matematic.
margine
Cod – Ansamblu de tehnici numerice ce, prin intermediul unor limbaje de programare, materializează un
program model de la introducerea datelor până la obţinerea rezultatelor.
Calibrare Proces de redefinire a cadrului hidrogeologic, proprietăţilor hidraulice şi condiţiilor de margine având
scopul de a ajunge la gradul dorit de apropiere între rezultatele simulărilor şi observaţiile de teren.
Senzitivitate Gradul în care rezultatele modelului reflectă schimbările în datele de intrare .
Verificare Utilizarea unui set al parametrilor hidrogeologici şi al condiţiilor de margine – set rezultat în urma unui
proces de calibrare – în scopul unei corecte aproximări a unui al doilea set de date, măsurate în condiţii
similare.
“Complexitatea este în ochii observatorului”. Este introdus astfel observatorul (uman). Dimensiuni
ale complexităţii nu sunt proprietăţi obiective: ele depind de ceea ce distinge observatorul. Aceasta
l-a făcut pe Simon (1969) să argumenteze că: “The central task of natural science is to make the
wonderful commonplace: to show that complexity, correctly viewed, is only a mask for simplicity; to
find pattern hidden in apparent chaos”.
În construirea unui aparat sau set de proceduri este apreciată complexitatea redusă care,
dacă satisface obiectivele pentru care a fost conceput, devine asociat cu eleganţa şi eficienţa.
Model – versiune simplificată a sistemului real, care reproduce aproximativ corect relaţia excitaţie-
răspuns a acestuia. Simplificarea este realizată printr-un/sub forma unui set de ipoteze care
exprimă înţelegerea noastră despre natura şi comportarea sistemului. Fiind o versiune simplificată
Elemente de management
Acviferul este un sistem care îndeplineşte diverse funcţii şi care trebuie gestionat pentru a atinge
diverse obiective.
Managementul unui sistem presupune luarea unor decizii (=atribuirea unor valori numerice
variabilelor de decizie) cu scopul de a modifica starea sistemului, pentru a atinge anumite
obiective.
În general un acelaşi obiectiv sau set de obiective poate fi atins luând diverse seturi de decizii
(=politici). De aceea managementul presupune alegerea celei mai bune politici care să conducă
la atingerea respectivului scop/obiectiv. Pentru aceasta avem nevoie să măsurăm
eficacitatea/eficienţa cu care diverse politici alternative îndeplinesc/se apropie de obiectivele
propuse. Funcţia scalară a variabilelor de decizie care măsoară această eficienţă – funcţie
obiectiv. Nu toate politicile sunt fezabile; unele nu pot fi considerate din cauza unor constrângeri
sociale, economice sau tehnice.
Managementul unui acvifer presupune, în general, determinarea valorilor numerice pentru
variabilele de decizie, cu scopul de a minimiza/maximiza o anumită funcţie obiectiv, în condiţiile
unor constrângeri specificate.
Variabile de stare:
o nivelul apei,
o concentraţia,
o subsidenţa terenului,
o intruziunea de apă marină.
Variabile de decizie:
o distribuţia spaţială şi temporală a pompării;
o distribuţia spaţială şi temporală a alimentării artificiale;
o nivelul apei în corpuri de apă de suprafaţă (râuri, lacuri) aflate în contact cu
acviferul;
o calitatea apei pentru alimentare artificială;
o calitatea apei pompate;
Funcţii obiectiv:
o beneficiu net total din operarea sistemului pentru o perioadă de timp; maximizarea
valorii funcţiei !
o costurile operaţiunilor de remediere pentru eliminarea poluanţilor din acvifer;
minimizarea costurilor !
o consum total de energie; minimizare !
Constrângeri hidrologice:
- nivelul apei, oriunde sau în anumite locaţii, nu trebuie să depăşească o anumită cotă
maximă specificată;
- nivelul apei, oriunde sau în anumite locaţii, nu trebuie să scadă sub o anumită cotă minimă
specificată;
- debitul unui izvor să nu scadă sub o valoare minimă;
- curgerea de bază în reţeaua hidrografică dependentă de acvifer să nu scadă sub un minim
specificat;
- concentraţia anumitor compuşi solubili din apa pompată, în anumite locaţii să nu
depăşească valori prag;
- pomparea totală ar trebui să satisfacă necesarul de apă dintr-o anumită regiune;
- timpul de rezidenţă pentru apa infiltrată (alimentare artificială) ar trebui să depăşească o
valoare minimă;
- lungimea penei de intruziune marină să nu depăşească o valoare impusă.
În acest context este clar că prognoza răspunsului sistemului acvifer este parte intrinsecă a
procedurii de determinare a politicii optime de management.
Trebuie să cunoaştem valorile viitoare ale variabilelor de stare relevante ce vor apare ca urmare a
implementării unui set de decizii:
- pentru a examina dacă respectă constrângerile,
- pentru a examina gradul de îndeplinire a funcţiei obiectiv.
3. Tipuri de modele
Groundwater models are used to solve problems related to groundwater flow, solute transport,
heat flow and land subsidence (Fetter, 2001). Depending on the type of the media through which
the water movement takes place, porous-media models or fractured rock models are used. For
instance, in the case of a sand and gravel aquifer a porous-media model is the only option, while
for a fractured crystalline rock with spaced fractures a fractured rock model is recommended.
Depending on the scale under which the system is studied, even in the case of fractured aquifers a
porous-media model can still be used.
Certain parameters (like conductivity or storage coefficient) can be considered constant all over the
aquifer or variable. In the first case, the aquifer is assumed to be homogeneous and the model
operates with global (lumped) parameters; in the latter case, the aquifer is heterogeneous and the
model is characterized by distributed parameters. Rarely, in the hydrogeological modeling one
operates with global parameters or uses black-box models. Usually, the mathematical models in
this field are physically based with spatially distributed parameters. The reasons are multiple: one
knows the constitutive laws of the porous media which intervene in the basic equations
characterizing the main physical processes, the underground environment is heterogeneous and a
lot of information concerning the structure or the behavior of the aquifers were collected in time.
Mathematical models in hydrogeology are based on the continuum principle, the real system
being considered continuous in space and in time. Using continuous variables the mathematical
model is expressed analytically; in particular cases (homogeneous and isotropic medium), the
solution of the mathematical model can be obtained analytically, being an analytical expression
also continuous in space and in time.
Analytical models and corresponding analytical solutions characterize usually small domains,
allowing important simplifications of the real systems; some groundwater problems can thus be
solved analytically without substantial loss of accuracy. Analytical solutions are obtained for
example in the analyses of pumping tests or in the case of a homogeneous aquifer for two-
dimensional steady-state flow. Even for flow modeling, analytical models cannot be used when the
value of the parameters varies within the studied area or when complex boundary conditions
characterize the aquifer. In transport problems, because of their increasing complexity with the
dimensions of the problem, analytical solutions are limited to only one-dimension. If the
mathematical model cannot be solved analytically it has to be transformed into a numerical model
and then solved numerically.
The direct problem, when the model parameters are known; providing adequate boundary
and initial conditions the system's answer can be obtained for different scenarios (alteration
of internal or external stresses or boundary conditions).
The inverse problem, when the model parameters are not known and have to be estimated
based on observations and measurements. Due to the fact that the model parameters have
a physical meaning, their first approximation is based on the field measurements. Still,
using these data with the same fundamental equations as in the direct problem, one
obtains important differences between the computed and the measured values of the state
variables. As a result, a calibration process is necessary. Using registered state variables
and the fundamental equations of the modeled process (processes), by mathematical
optimization or by trial-and-error procedures one obtains an estimation of the model
parameters.
The most used groundwater models are deterministic mathematical models. Deterministic
models are based on the principle of causality, assuming a unique cause-effect relation. As a
consequence, understanding the behavior of the system based on past or present relations
between input and output makes possible to pre-determine the system's response to any set of
stresses. In reality this is not always true, especially if the new excitation falls outside the range of
historically registered stresses.
The subsurface environment is very complex, presenting a strong variability. Even if difficult, a
stochastic approach is sometimes necessary to evaluate the uncertainties related to mathematical
modeling. Stochastic models are based on the assumption that input values like physical
parameters of the aquifer (porosity, hydraulic conductivity etc) or boundary conditions (water level
fluctuations in a river hydraulically connected with the aquifer) are characterized by variability,
respectively by stochasticity. The hydraulic conductivity or the transmissivity are log-normal
distributed; at the same time, for boundary conditions time series analyses can be applied. No
matter how much data are collected, uncertainties about the hydrogeological properties and
boundary conditions remain. The stochastic approach can offer significant improvements in
characterization of the subsurface environment and its behavior. Still, most mathematical models in
hydrogeology are based on a deterministic approach.
A generic model is the computer code used to solve one or more differential equations.
A site-specific model is a numerical model (the discretized form of the differential equation)
whose parameters (hydraulic conductivities, storage coefficients, dispersivities etc), boundary
conditions (Dirichlet, Neumann, Cauchy) and grid dimensions are chosen to best represent the
studied aquifer. This basic information serves as input data into the computer code (the generic
model) leading after solving the algebraic system of equations to the numerical solution. The most
important steps in preparing and running a site-specific model are:
Some of these steps can be absent; for instance demonstrating the model sensitivity or auditing
the model, although necessary stages in a professional analysis, are often ignored.
Un model determinist presupune că, un sistem sau un proces, prin funcţionarea lor, pentru un
anumit set de date, produc un răspuns definit în mod unic. Ecuaţiile definitorii ale modelului
determinist stabilesc în mod precis o relaţie tip cauză - efect sau solicitare - răspuns.
Depending on the number of space dimensions, one speaks of 0D, 1D, 2D or 3D models.
0D models have no space dependency, only a time dependency. Ecological models concerning populations
of biological species in an environmental compartment are in their majority of that type. As t is the only
independent variable, the analytical formulation leads to ordinary differential equations. These are
differential equations, which depend on one variable only; in contrast to partial differential equations, where
there are at least two independent variables.
Models with no time dependency are denoted as steady, steady state or stationary. The corresponding terms
for time dependent simulations are: unsteady or transient. A steady state is approached in real systems, if the
internal processes have time enough to adjust to constant outer conditions. It is a necessary condition for
steady state that exterior processes or parameters do not change in time. Otherwise steady conditions cannot
be reached.
1D models include one space dimension only. Models for the soil compartment are mostly 1D, as the
changes in vertical direction are of concern: seepage to the groundwater table or evaporation to the ground
surface. Processes in rivers (image a water level peak or a pollutant plume moving downstream) can be
regarded in 1D under certain conditions. Water from surface water bodies infiltrating into aquifers may be
described by a 1D approach, if relevant conditions do not change substantially in the vertical direction and
along the shoreline.
1D steady state models lead to ordinary differential equations. Transient models, including at least one space
direction, lead to partial differential equations. It is important to know about these differences, as
mathematical solution techniques for both types of equations are different.
2D models include two space variables. One may distinguish between 2D horizontal and 2D vertical models.
Terrestrial ecology is a typical field, where this type of model is suitable, describing the distribution or
population of species on the land surface. In streams or estuaries or in shallow water models are often set up
for vertically averaged variables, for which a 2D horizontal description results. Models for 2D vertical cross-
sections are obtained,
• in groundwater flow, where several geological formations are to be included, but no variations of
hydraulic conditions in one horizontal direction
• in cross-sections of streams
• in air pollution modeling, if no space direction is preferential around a source; in that case the
single radial coordinate r replaces two horizontal space variables x and y.
3D models are quite complex in most cases. Numerical algorithms using the methods of Finite Differences,
Finite Volumes or Finite Elements are the methods of choice for modeling in higher space dimensions,
steady and unsteady.
Ecuaţia ce guvernează un model matematic poate fi rezolvată fie analitic, fie numeric.
Modele analitice -folosesc soluţii exacte ale ecuaţiilor diferenţiale, soluţii continue în spaţiu şi timp.
Modelele semianalitice folosesc tehnici şi metode numerice pentru a aproxima soluţii analitice
complicate, generând soluţii discrete în timp şi spaţiu. Modelele ce utilizează pentru unul din
domeniile spaţiu sau timp soluţii analitice, iar pentru celălalt soluţii numerice, sunt de asemenea
considerate semianalitice.
1. Eficiente când datele sunt dispersate în spaţiu 1. Limitate la cateva condiţii idealizate, cu
Modele şi afectate de incertitudine în determinare; geometrie simplă; de multe ori inaplicabile la
analitice 2. Economice; condiţiile concrete ale modelării unor cazuri
3. Bune pentru determinarea în prima fază a concrete, cu condiţii de margine complicate;
gradului de contaminare; 2. Majoritatea nu pot lua în calcul variaţiile
4. Estimări preliminare, adesea posibile din spaţiale şi temporale ale sistemului.
datele disponibile;
5. Datele de intrare pentru codurile de calcul
sunt în general simple.
În cazul modelelor numerice, alegerea codului de calcul, verificarea lui, realizarea modelului,
calibrarea, analiza de senzitivitate şi, ca ultimă etapă, predicţia, pot fi duse la bun sfârşit prin
intermediul unei metodologii de modelare. Aceasta cuprinde, în mod obligatoriu, următorii paşi:
stabilirea scopului: sunt evidenţiate ecuaţiile matematice ce urmează a fi rezolvate;
dezvoltarea unui model conceptual: se identifică unităţile hidrostratigrafice, condiţiile de
margine; datele referitoare la parametrii hidrogeologici şi la condiţiile hidraulice de margine
impuse sistemului sunt trecute în revistă;
selectarea codului de calcul: este în mod obligatoriu inclusă şi verificarea lui, prin compararea
rezultatelor obţinute pe un set de date, valorizate în paralel cu ajutorul unui alt mijloc de
prelucrare (de exemplu, alt cod numeric de calcul, unanim acceptat);
proiectarea modelului: include alegerea reţelei de discretizare a domeniului, selectarea paşilor
de timp, alegerea condiţiilor de margine şi a condiţiilor iniţiale, precum şi stabilirea , într-o primă
fază, a valorilor parametrilor hidrogeologici;
calibrarea: scopul este de a stabili dacă modelul poate reproduce sarcinile şi debitele măsurate
în teren; prin metode tip “trial and error” sau prin calibrare automată, se obţine o distribuţie
spaţială a parametrilor hidrogeologici şi a conditionărilor hidraulice ce generează, prin
modelare, valori ale sarcinilor şi ale debitelor asemănătoare celor măsurate;
analiza de senzitivitate a calibrării: studiază incertitudinea ce afectează valorile şi distribuţia
spaţială şi temporală a parametrilor hidrogeologici şi a condiţiilor de margine; analiza de
senzitivitate măsoară efectul incertitudinii asupra rezultatelor modelului;
verificarea modelului: creşte gradul de încredere în modelul realizat prin utilizarea unui set
calibrat al parametrilor hidrogeologici şi al condiţiilor de margine pentru a reproduce un al
doilea set de date de teren;
predicţia: cuantifică răspunsul sistemului în diferite variante de funcţionare viitoare;
analiza de senzitivitate a predicţiei: evidenţiază incertitudinea existentă în rezultatele simulărilor
predictive, ca efect al incertitudinii în determinarea parametrilor hidrogeologici şi a condiţiilor de
margine viitoare;
prezentarea rezultatelor;
revizuirea modelului, pe baza unor alte seturi de date, colectate în anii următori;
reproiectarea modelului.
dimensiunea domeniului considerat.Pe baza acestui concept (EVR) domeniul mediului poros poate
fi definit ca o parte a spaţiului ocupată de mai multe faze: faza (matricea) solidă+faze fluide, pentru
care un EVR poate fi găsit.
Astfel, o valoare macroscopică într-un punct al domeniului poros este interpretată ca o
medie a acelei variabile pe EVR centrat pe acel punct. Mediind o variabilă pe toate punctele din
cadrul domeniului se obţine un camp de valori continuu al acelei variabile. Reprezentând mediul
poros actual ca un continuum, este evitată nevoia de a cunoaşte detaliile la scară microscopică ale
spaţiului poros. Totuşi, la nivel macroscopic, geometria complexă a interfeţei solid-spaţiu poros
este înlocuită prin diverşi coeficienţi ai matricii solide (porozitate, permeabilitate, dispersivitate).
Astfel, un coeficient care apare într-un model macroscopic reflectă efectul configuraţiei
microscopice a spaţiului poros.
În practică toate modelele ce descriu curgerea şi transportul în subteran sunt scrise la nivel
de continuum sau macroscopic. Ele sunt obţinute prin medierea pe EVR a modelelor microscopice
corespunzătoare. Aceasta înseamnă că trebuie să începem prin a înţelege fenomenele care apar
la nivel microscopic (ex. limita dintre faze) înainte de a deriva/dezvolta modelul macroscopic.
Acviferul este un domeniu de mediu poros. De aceea curgerea prin spaţiul poros apare în toate
modelele hidrogeologice (fizice şi matematice). Deoarece nu este nici fezabil şi nici necesar să
modelăm în detaliu curgerea, la nivelul porilor vom neglija/face abstracţie de detalii ale geometriei
spaţiului poros prin introducerea conceptului de mediu continuu.
Mediu poros = matrice solidă + spaţiu poros
AEV – element de volum, suficient de mare, plasat oriunde în cadrul domeniului, care
conţine fază solidă.
Curgerea apei prin mediul poros poate fi tratată la scară micro – ce se întâmplă într-un punct în
cadrul fluidului privit ca un mediu continuu (neglijând structura moleculară). De exemplu, pentru un
fluid care ocupă întreg spaţiul poros, putem aplica ec. Navier-Stokes pe domeniul fluidului, cu
condiţii de margine la interfaţa solid-fluid. Totuşi, această abordare de regulă nu este practică
datorită dificultăţii de a descrie configuraţia complexă a acestei limite (interfeţe). Mai mult, chiar
dacă am reuşi să rezolvăm pentru valori ale variabilelor de stare la scară micro (presiune, … ), nu
am putea verifica aceste soluţii prin măsurători la acest nivel.
Pentru a depăşi aceste dificultăţi, este necesară abordarea la o altă scară – macroscopic.
Aici se pot face măsurători iar problema valorilor la limită poate fi rezolvată.
Pentru a obţine descrierea curgerii la acest nivel adoptăm modelul continuu – sistemul
poros real, în care fiecare fază ocupă o parte din AEV este înlocuit cu un sistem fictiv, în care
fiecare fază este privită ca un continuu care umple întregul AEV. Astfel, pentru un AEV vom avea
suprapunere şi posibil interacţiune între 2 sau mai multe medii continue. Pentru fiecare se pot
considera valori medii (macroscopice) valabile pe întregul AEV şi care se atribuie centroidului
acestuia. Traversând întreg sistemul poros cu un AEV mobil vom obţine un câmp de variabile
macroscopice (funcţii diferenţiabile de coordonatele spaţiale).
Configuraţia limitei gol-solid şi a interfeţei fluid-solid este introdusă în descrierea
macroscopică a curgerii (=model macroscopic) sub forma coeficienţilor. Valorile lor numerice
trebuie determinate experimental pentru fiecare mediu poros.
Trebuie selectată dimensiunea corespunzătoare pentru volumul mediat, funcţie de scopul
modelului macroscopic.
Principalul dezavantaj al folosirii AEV este că pentru fiecare valoare medie trebuie
specificat volumul pe care s-a făcut medierea. Pentru a elimina acest inconvenient este nevoie de
o procedură universală care:
- se aplică tuturor mediilor poroase,
- asigură o valoare mai mult sau mai puţin constantă pentru valorile mediate, care să
corespundă domeniului de variaţie al dimensiunilor instrumentului.
Acest volum de mediere universal este cunoscut ca EVR.
Dimensiunea EVR se stabileşte astfel încât valorile mediate ale caracteristicilor geometrice
(=coeficienţilor) microstructurii spaţiului poros sau interfeţei gol-solid, pentru orice punct din
domeniu, să fie o funcţie cu valoare unică de locaţia punctului, independent de dimensiunea EVR.
Bachmat & Bear (1986) au stabilit că volumul EVR ar trebui să fie ales astfel încât medierile
volumetrice să fie considerate estimări satisfăcătoare ale tuturor parametrilor statistici relevanţi
pentru configuraţia spaţiului poros; estimări care nu sunt influenţate de dimensiunea probei.
Dacă notăm cu l – dimensiunea caracteristică a EVR (ex. diametrul sferei) şi cu d –
dimensiunea structurii microscopice a spaţiului poros (ex. raza hidraulică), ei arată că o condiţie
necesară pentru a obţine estimări nealeatoare pentru caracteristicile geometrice ale spaţiului poros
în orice punct al domeniului este
l d
Altă condiţie este: l lmax, cu lmax = distanţa de la care distribuţia spaţială a coeficienţilor
macroscopici relevanţi pentru spaţiul poros (ex. porozitate, permeabilitate), deviază de la cea
liniară cu mai mult de o valoare acceptată.
De asemenea, l L, cu L = lungime caracteristică a domeniului mediului poros peste care apar
schimbări (variaţii) semnificative ale mărimilor (macroscopice) mediate.
Conservation of Mass
The most nearby and most common application of the continuity equation is that for mass. There
are two types of mass conservation. One type is the mass conservation of the medium, which can
be solid, aqueous or gaseous. The mass is expressed in terms of a density ρ with a physical unit
[M/L3]. ‘M’ represents a mass unit and ‘L’ a length unit.
For example the density of fresh water at a temperature of 4◦C and the pressure of
101325 Pa (1 atm) is 1000 kg/m3.
The second formulation of mass conservation in a fluid concerns biogeochemical species
within a fluid. In that case the mass is expressed in terms of the concentration c. The continuity
equation then is formulated in terms of the concentration c of the species. The concentration also
has the unit [M/L3] and measures the mass within a volume of fluid. The content of chloride Cl− in
seawater amounts to 19 g/l.
One has to extend the mass definition in situations, where the fluid does not fill the entire
volume.Then the total mass per space volume is expressed by θc with porosity θ as additional
factor. Porosity is dimensionless and measures the volumetric share of the fluid phase on the total
volume. The physical dimension of the product is thus further [M/L3]. In aquifers groundwater
usually fills approximately 25% of the volume; thus holds: θ=0.25.
The described concept can be extended to situations with several phases, where each
phase has its own θ-value. In the unsaturated soil zone below the ground surface, above the
groundwater table there are three phases present: the soil as solid phase, seepage water as liquid
phase and soil air as gaseous phase. The mass of a gas component is expressed in terms of
partial pressure.
According to the ideal gas law the product of pressure and volume is a constant, which
changes only with temperature. Thus mass conservation can be formulated in terms of pressures
instead of volumes. According to Dalton’s Law10 the pressures of a gas mixture have to be
summed up to yield the total pressure.
Conservation of Momentum
The momentum of a fluid is expressed as the product ρv, where v denotes the velocity. The
physical unit of momentum is [M/(L2T)], where the letter ‘T’ represents a time unit. As velocity is a
vector, the momentum also is a vector, with one vector component for each space dimension of
the model.
Conservation of Energy
The kinetic energy of a fluid is expressed as 1/2ρv2 with the physical unit [M/(LT2)]. If energy is
measured in Joule, the given expression measures Joule per volume.
In problems, which include heat transfer, thermal energy is expressed in terms of
temperature T . The energy content per volume is given by the product ρCT , where the new factor
C is the specific heat capacity. The physical unit of C is [L2/(T2K)], where ‘K’ represents the
temperature measure unit, mostly ◦Celsius or ◦Kelvin. In many tables values for the product ρC can
be found, which is addressed simply as heat capacity. Heat capacity has the unit [M/(LT2K)]. If
energy is measured in Joule ρC has the physical unit of J per volume and ◦Kelvin, while C is
measured in the unit Joule per mass and ◦Kelvin.
To solve numerically the groundwater flow and transport equations the following data are
necessary:
A general map of the area containing all relevant information for groundwater modeling:
towns' location, land use (agricultural and industrial areas, forests, ways of communication,
irrigation canals, drainage systems), wells position (both observation and exploitation
wells), and hydrologic information (springs, streams, rivers, swamps and wetlands, lakes,
reservoirs). Based on this data, the recharge and the discharge areas, as well as the
pollution sources can be approximately delineated. The general map can be used as base
map on which other data are overlaid.
A topographic map, to define the ground surface as a boundary limit for the water table in
the lower areas where possible discharges could occur; the topographic map is used also
in connection with the hydrogeological cross-sections to define the areal extent of the
aquifer.
Geological and hydrogeological maps with cross-sectional plots or fence diagrams to show
the stratigraphy. The cross-sections indicate the major changes in geology such as faults,
the aquifer geometry (base, thickness, lateral extent), the lithology and thickness of strata,
the depth of stream channels and other drainage areas, as well as wells and boreholes
characteristics like position, depth, screened interval, intercepted strata etc.
Aquifer characteristics like: hydraulic conductivity or transmissivity, anisotropy, specific yield
or specific storage (for flow modeling), effective porosity, aquifer dispersivity factors, soil
bulk density (for transport modeling).
System's stresses such as history of the pumping rates or natural recharge. It has to be
mentioned that while the withdrawal rates are measurable, the natural recharge is obtained
indirectly, either by a global evaluation or by mathematical modeling of the infiltration
process. In the first case the repartition of the precipitations between surface run-off and
evapotranspiration is necessary; in the latter, the knowledge of temperature and
precipitation regime is compulsory. When using reservoir type models for the description of
the percolation process through the unsaturated zone the soil map and general soil indexes
(total porosity, field capacity and hygroscopicity coefficient) are the necessary data, while in
physically based models even hydraulic characteristics of the unsaturated soils are
requested.
Water levels history in rivers, lakes or channels used for external or internal boundary
conditions.
History of the groundwater levels in the observation wells.
Groundwater chemistry and background values of solutes or contaminants.
Data about the pollution sources: nature (industrial or agricultural), location (point or non-
point sources), source concentration and volume.
Data about the contaminant: solubility, density, viscosity, Henry's law constant, diffusion
coefficient, adsorbtion isotherm, decay rate, biodegradability etc.
A database fed with all relevant data is a necessary step in data retrieval.
Some data are used to create maps for a better understanding of the aquifer and its behavior. For
instance, based on the measured groundwater heads water table maps for unconfined aquifers or
potentiometric surface maps for confined aquifers are realized. These maps represent in two-
dimensions with contour lines of equal elevation three-dimensional surfaces. The maps are
constructed using water levels measured in a number of wells adequately distributed on the
hydrogeological basin of the aquifer.
During water withdrawal dynamic levels characterize the aquifer. Wells that normally provide water
can be used for level measurement on the condition that the pumping has been shut off long
enough for the water level to stabilize at the static level; the non-pumping or the static level is
reached when the hydraulic head stops rising.
In drawing water-table maps a special attention must be paid to the local topography and surface-
water features: the contour lines cannot be above the ground surface or above the level of the
surface waters. Potentiometric surface of the confined aquifer is not subjected to such constraints,
the potentiometric contours having sometimes values above the ground surface; it is the case of
artesian levels.
There is a close relation between the hydraulic gradients and the hydraulic conductivities; thus, a
steep gradient indicates a high hydraulic resistance, meaning small values of hydraulic
conductivities compared with upstream and downstream values on the flow direction of the same
parameter.
Mention must be made that water-table or potentiometric surface maps represent an interpretation
of the real surface, introducing different types of errors (linear interpolation between too spaced
wells, wells measuring a combined level by putting in contact more layers or ignoring the influence
of the surface topography). That is why in many cases for calibration purposes it is recommended
to take into consideration only the point measured levels; if due to insufficient number of monitoring
wells the maps obtained using these levels are used, a different weight should be assigned to the
residues.
In a similar way, on the basis of the measured concentration in different points at different depths
one can prepare a concentration distribution map in horizontal or vertical planes for a given
moment or for several times, together with their source locations.
During the mathematical modeling process some additional data can be found necessary and new
field campaigns have to be organized.
In fact, the conceptual model represents decomposition of a real system into interconnected
subsystems. The output(s) of one subsystem represent input(s) in other subsystem(s); some
subsystems can be regarded as block-box systems characterized by a transfer function that relates
inputs and outputs or as white box systems whose behavior is physically described by differential
equations.
For instance, to model the groundwater flow two options are possible: a) all processes taking place
in the unsaturated and saturated zone can be modeled simultaneously; b) the unsaturated zone
processes are independently studied in a first phase providing the natural recharge of the aquifers;
subsequently the numerical model of the saturated zone uses this value as an input data.
The conceptual model can have an evolution during the modeling phase or when collecting new
data; sometimes the first conceptual model must be improved to better fit the behavior of the
modeled system. The development of an appropriate conceptual model is the key for the success
of modeling process.
During the conceptualization phase, based on the existing information, important decisions have to
be taken:
Areal extent of the model (the whole hydrogeologic basin or only a part of it). This choice
strongly depends on the purpose of the mathematical modeling. If the purpose is the
groundwater resource evaluation, the best is to consider the full areal extent of the aquifer.
If one wishes to evaluate the influence of a water work (reservoir, pond, diversion, drain,
navigable canal, etc) in order to prescribe technical measures for controlling the ground
level increase, only the neighboring areas may be examined. At the same time, for
transport problems, after realizing a general flow model a local model has to be set up.
Vertical extent of the model, examining saturated and/or unsaturated flow. For saturated
flow, depending on the water quality and the exploited strata only medium or deep aquifers
are of interest. In other cases, the whole complex of aquifers has to be studied, the
delineation of the system being given by an aquiclude generally formed by thick clay layers
or solid rocks. At this point, one should also answer to the question if the modeled aquifer is
confined or unconfined.
Number of model dimensions (1-D, 2-D or 3-D) and if the model is not 3-D its orientation
(horizontal or vertical model). Normally, when having enough data, a 3-D model describes
the aquifer and its behavior better than any other model. Generally, one faces the data
scarcity and other schematization should be considered. A most often used approach,
when the third dimension of the aquifer (the thickness) is very small as compared with the
dimensions in plane, is to consider the aquifer as a 2-D structure, formed by one horizontal
layer. Still, it has to be checked if Dupuit's assumption of small hydraulic gradients is valid;
this will not be the case in the vicinity of drainage borders or near the pumping wells and for
a detailed study of this area a local 3-D model should be necessary. A 3-D model is almost
compulsory when the dimensions of the aquifer are comparable or when the examined
process is taking place locally, like in the case of plume evolution near a pollution source. A
2-D vertical model can be used if the perpendicular fluxes (along the 3rd direction) are
negligible, as it is the case for water or pollutant fluxes from a river or a pipeline to an
aquifer. Sometimes, permeable layers more or less in communication through quasi-
permeable formations (named aquitards) form the system. In this case, a multi-layer
schematization can be used: water is moving horizontally through the permeable
formations, while in the aquitards the water fluxes are vertical. Finally, in a relatively flat
zone the infiltration is predominately vertical (unless the existence of less pervious lens
leading to small perched water tables) and a 1-D model can be used for water movement in
the unsaturated zone. On the contrary, when studying infiltration on a slope, a 2-D vertical
model should be used to take into account the two components of the infiltration.
Degree of heterogeneity to be considered. Some parameters like hydraulic conductivity,
transmissivity, specific yield, specific storage or aquifer dispersivity are known on the basis
of field measurements. They present a certain variability that should be introduced in the
numerical model; for example, the hydraulic conductivity is usually log-normally distributed.
The number of homogeneous zones depends on the density of field data, but in some
cases it is also based on the interpretation of the hydrogeologic structure. During the
calibration process, the number of homogeneous areas could be increased from one to
several to assure the best fitting of the computed values with measured state variables
(hydraulic heads, concentration, temperatures etc). It is still not recommended to
excessively increase the number of homogeneous areas for a better fitting of the measured
state variables. It must not be forgotten that the model itself is a simplification of the reality
that involves approximations; at the same time, even the measured data are affected by
errors. The over-parameterization can lead to a better response of the model during the
calibration phase, but does not necessarily mean that the model will provide more accurate
results during the prediction phase. The principle of parsimony, meaning the minimum
number values, respectively of homogeneous areas for a given parameter should be kept in
mind. Similar considerations can be made regarding the number of natural recharge areas.
One can admit a unique value for the recharge by percolation, or a spatial distribution
according to the soil texture. The anisotropy is usually considered in the vertical models
taking into account that the vertical conductivity is generally an order of magnitude smaller
than the horizontal conductivity. In some cases (fractured rocks), even in the case of 2-D
horizontal models, the anisotropy should be considered.
Miscible or immiscible character of the studied pollutants. For example, in the case of the
oil pollution, some components are miscible with water, being more dangerous because
they are transported with the water velocity, thus affecting an important area in a relatively
short time . Other components are immiscible, and due to their viscosity are characterized
Developing the conceptual model is one of the most important phases in mathematical modeling.
An oversimplification of the reality will lead to false results during the prediction phase. Some
important features are sometimes overlooked: the natural recharge is considered as a unique
value for the whole aquifer, too little homogeneous areas for different parameters are proposed,
the transport in bedrock or in low permeability units is neglected due to the fact that the flow is
unimportant, the 3-D character of transport is strongly constraint by the choice of 2-D models or the
density effects are ignored. Still, an excessive sophisticated model will suffer because of lack of
adequate data. The choice of the conceptual model is thus strongly dependent on the quantity and
quality of collected data, and it can be improved on the basis of newly obtained data.
Primul pas în modelare constă în construirea modelului conceptual = set de aproximări care
descriu compoziţia sistemului, procesele de transport ce au loc, mecanismele ce le guvernează şi
proprietăţile relevante ale mediului.
Acesta constă dintr-un set de ipoteze care reduc problema reală şi domeniul real, la o versiune
simplificată, acceptabilă din punct de vedere al obiectivelor şi al problemelor de management
asociate. Modelul conceptual se poate exprima printr-o formă matematică, numită model
matematic.
Aproximarile se refera la:
- geometria limitelor domeniului acvifer studiat;
- tipul de material din cadrul acviferului (inclusiv heterogeneitate, anizotropie);
- tipul de curgere în acvifer (3D sau 2D orizontal);
- regimul de curgere (laminar, nelaminar/turbulent);
- proprietăţile apei (cu referire la omogenitate, compresibilitate);
- efectul substanţelor dizolvate şi/sau al temperaturii asupra densităţii şi vîscozităţii;
- prezenţa unor limite nete fluid-fluid: nivelul hidrostatic, limita apă dulce – apă sărată;
- variabilele de stare relevante, împreună cu suprafaţa/volumul pe care se consideră media
(medierea) acestora;
- surse sau pierderi de apă şi/sau poluanţi, în cadrul domeniului sau pe limitele acestuia (cu
precizarea dacă sunt punctuale sau distribuite).
Rushton (1998):
“Aquifer systems are so complex that it is not possible to study every detail”
“This leads to the question of what needs to be included in an aquifer study and what can be
ignored”
“For most aquifer systems there are a small number of crucial factors which must be examined
in detail; if only one of these is ignored the conclusions may be seriously in error”.
Aproximările care stau la baza unui model conceptual trebuie să se refere la următoarele aspecte:
• the geometry of the boundaries of the investigated aquifer domain;
• the kind of solid matrix comprising the aquifer (with reference to its homogeneity, isotropy, etc.);
• the mode of flow in the aquifer (e.g., 1D, 2D sau 3D);
• the flow regime (laminar or nonlaminar);
• the properties of the water (with reference to its homogeneity, compressibility, effect of dissolved
solids and/or temperature on density and viscosity, etc.);
• the presence of assumed sharp fluid-fluid boundaries, such as a phreatic surface;
• the relevant state variables and the area or volume over which the averages of variables are
taken;
• sources and sinks of water and of relevant contaminants, within the domain and on its boundaries
(with reference to their approximation as point sinks and sources, or distributed sources);
• initial conditions within the considered domain; and
• the conditions on the boundaries of the considered domain that express the interactions with its
surrounding environment.
Selectarea unui model conceptual corespunzător pentru o problemă dată este o etapă esenţială în
procesul de modelare. Simplificarea în exces poate conduce la un model căruia îi lipsesc
informaţiile necesare în vreme ce sub-simplificarea poate rezulta într-un model costisitor, sau în
lipsa datelor necesare pentru calibrarea modelului şi estimarea parametrilor.
Selectarea unui model conceptual corespunzător şi a gradului de simplificare pentru un caz anume
depinde de:
• obiectivele problemei de management;
• resursele disponibile;
• datele de teren disponibile;
• cadrul legal.
Următorul pas în procesul de modelare este exprimarea modelului conceptual sub forma modelului
matematic. Soluţia modelului matematic reprezintă predicţii asupra comportării sistemului real ca
răspuns la diverse surse sau pierderi. Majoritatea modelelor nu exprimă altceva decât un bilanţ al
cantităţii extensive considerate (masa de apă sau de solut).
În abordarea mediu continuu ecuaţiile de bilanţ sunt scrise “într-un punct al domeniului” şi trebuie
interpretate ca medie, pe unitatea de suprafaţă sau volum în vecinătatea punctului. În aceste
condiţii bilanţul ia forma ecuaţiei diferenţiale parţiale.
Modelul conceptual se poate exprima printr-o formă matematică, numită model matematic. Acesta
trebuie să conţină:
Definirea geometriei domeniului considerat şi a frontierelor;
Ecuaţii care exprimă bilanţul mărimilor considerate;
Ecuaţii care descriu fluxurile cantităţilor extensive considerate (variabile de stare relevante);
Ecuaţii constitutive care definesc comportarea materialelor (fluid, solid);
Condiţii iniţiale care descriu starea sistemului la momentul iniţial;
Condiţii pe frontiere, care descriu interacţiunea acviferului cu mediul înconjurător.
Toate ecuaţiile trebuie exprimate funcţie de variabilele dependente selectate pentru problemă.
Selectarea variabilelor adecvate pentru a fi folosite într-un caz particular depinde de datele
disponibile. Numărul de ecuaţii incluse în model trebuie să fie egal cu numărul variabilelor
dependente. Condiţiile la limită trebuie să fie astfel încât permită o soluţie stabilă, unică.
Modelul matematic conţine aceleaşi informaţii ca şi cel conceptual dar exprimate printr-un set de
ecuaţii care poate fi rezolvat analitic şi numeric.
Odată construit modelul matematic, în funcţie de variabilele de stare relevante, acesta trebuie
rezolvat pentru cazuri de interes practic. (Sunt preferate metodele analitice, acestea dând soluţii
general valabile). De cele mai multe ori metodele analitice nu pot fi folosite din cauza dificultăţii
modelului. În acest caz se folosesc metode numerice.
În trecerea de la sistemul real la modelul conceptual şi apoi la modelul matematic sunt introduşi
diverşi coeficienţi de transport şi înmagazinare cum ar fi: permeabilitatea mediului poros,
transmisivitatea, coeficientul de înmagazinare elastică, dispersivitatea mediului poros.
Permeabilitatea şi dispersivitatea sunt exemple de coeficienţi care exprimă efectele
macroscopice ale configuraţiei limitei solid-fluid la scară microscopica. Ele sunt introduse la
trecerea de la scara micro de descriere la cea macro, a mediului continuu. Toţi aceştia sunt
coeficienţi ai modelului şi de aceea, în ciuda denumirii similare în diferite modele, interpretarea lor
precum şi valorile actuale pot fi diferite de la un model la altul.
Evident, nici un model nu poate fi folosit fără o cunoaştere prealabilă a valorilor coeficienţilor
folosiţi. Această activitate de obţinere a parametrilor/coeficienţilor – problemă de identificare
(problema inversă). În principiu singura cale este de a investiga sistemul acvifer real şi de a
căuta în trecut o perioadă pentru care există informaţii disponibile asupra:
a. condiţiilor iniţiale,
b. acţiuni asupra sistemului (pompare, realimentare artificială sau naturală, injectare de
poluanţi/trasori, modificarea condiţiilor de margine);
c. răspunsul sistemului acvifer (distribuţia spaţială/temporală a variabilei de stare).
Principiul metodei MDF constă în a aproxima soluţia ecuaţiei cu derivate parţiale, care descrie
mişcarea fluidelor, cu soluţia unui sistem algebric de ecuaţii (soluţia finală reprezentând valorile
nivelului piezometric într-un număr mare de puncte, dar finit). Aproximarea prin MDF a ecuaţiei de
curgere constă în a înlocui derivatele parţiale spaţiale şi temporale prin diferenţe finite. Mediul
continuu este înlocuit cu un mediu discret constituit dintr-o reţea de celule rectangulare (cazul 2D)
sau paralelipipedice (cazul 3D), conectate prin intermediul nodurilor corespunzătoare.
Ca şi la metoda diferenţelor finite, principiul metodei MEF constă în a aproxima soluţia
ecuaţiei cu derivate parţiale care descrie mişcarea fluidelor, cu soluţia unui sistem algebric de
ecuaţii. Se realizează astfel un model numeric discret al mediului continuu considerat. Ecuaţiile
sunt deduse din necesitatea ca soluţia finală să aproximeze în medie soluţia reală cât mai bine.
Construirea modelului numeric implică realizarea unui program de calcul complex care
conduce în final la determinarea şi rezolvarea sistemului algebric de ecuaţii. Acest program
foloseşte ca date de intrare:
coordonatele fiecărui nod;
structura fiecărui element;
transmisivitatea din cadrul fiecărui element;
o estimare iniţială a nivelului piezometric;
în cazul curgerii nestaţionare: coeficientul de înmagazinare şi valorile paşilor de timp.
Domeniul considerat este discretizat cu o reţea de elemente finite de forme variabile şi se
presupune că în interiorul fiecărui element finit variaţia nivelului piezometric este polinomială, ceea
ce înseamnă că funcţia = (x,y) este aproximată de o funcţie diferenţiabilă pe fiecare element
finit. Valorile nivelului piezometric - în fiecare nod al reţelei, sunt considerate necunoscute. Astfel,
variabila continuă (x,y) este înlocuită de un număr mare de necunoscute k = 1,2,...,n.
The FD method is based on the principle that any complex function (f(x)) can be approximated with a simple
linear function for a small increment of the independent variable (x), which could for example be space or
time. Mathematically this can be described as:
In the FD method, a system to be modelled is mapped onto some grid or lattice. Values of the state variables
(e.g. temperature) are only recorded at the grid nodes. The values are assumed to vary linearly between
these nodes. It is clear that the spacing between these nodes must be small enough to describe the real
system accurately enough: Eq. 2.8 must be satisfied. Time-dependent changes to the system, like cooling of
a dyke, are calculated in small increments, again assuming that changes are linear for small enough
increments of time. Ideally, a system would be described by an extremely large number of nodes and all time
increments should approach zero. However, limited computer memory and calculation time constraints mean
that the user must find a compromise between accuracy on the one hand, and computing time and memory
use on the other hand.
In Eq. 2.8 we used the derivative of f(x) to approximately predict the value of f(x + Δx) after adding a small
increment of Δx. In many cases, however, both f(x) and f(x + Δx) are known, but not the derivative of the
function. The derivative can easily be obtained from Eq. 2.8:
This is called the “forward” approximation of the derivative (Fig. 2.16). Two other possibilities are the
“backward” and “central” approximations,
respectively (Fig. 2.16):
Often one is also interested in a second derivative of the function of interest (e.g. when Fick’s second law is
applied for diffusion). The second derivative of a function, of which the values are know at discrete
increments x, can be approximated in a similar way as the first derivative:
Fig. - To determine the local gradient at one point (x) on a lattice, one can use the difference with the next point (x+Δx;
forward method), the previous point Δx; backward method), or the points on both sides (x-Δx and x+Δx; central
method). Note that the latter is closest to the true gradient at point x
The Finite Difference method has one major drawback, in that it is not well suited for irregular grids, and that
it heavily relies on the assumption that functions can be linearised for small increments.
The FE method is the standard method for modelling deforming materials, and is widely used in engineering
and studies of rock deformation.
2. We assume that the state variables of interest anywhere in the system can be approximated by
certain functions. Such functions can be simple linear functions, but can also be complex non-linear
functions.
3. We solve all functions for all nodes in the FE-mesh, taking into account (a) all the state- and
kinematic equations, (b) the continuity conditions between the elements, and (c) the boundary
conditions.
Numerical Solutions
For complex layered, heterogeneous aquifers, a numerical approximation is required in non-uniform
flow,
Several types — Finite Difference, Finite Element, and Method of Characteristics,
Finite difference approximations involve applying Taylor’s expansions to the equations (flow and
transport) and approximating the derivatives in the equation,
Other methods involve different approximations, but all are based generally on the Taylor
expansion.
The computer code to be used during the mathematical modeling must be selected according
mainly to the capacity of the code to adequately simulate the site conditions or relevant processes,
like:
The areal and the vertical extent of the aquifer were established during the development of the
conceptual model. As a general rule, for water resources evaluation it is preferable to examine the
whole hydrogeologic basin; for other flow problems, only part of the aquifer can be modeled by
using adequate boundary conditions. Usually the flow models are two-dimensional, either
horizontal-plane models (one layer) or quasi 2-D models (multi-layers). For transport problems, in
most of the cases a 3-D model is recommended. One basic principle when choosing the
boundaries of the model is that the internal stresses should not interfere with the boundaries; that
means that the effect of different solicitations (pumping for example) should not reach the
boundaries.
Generally, the accuracy of predicted results is improved when using a finer calculation mesh, but
the computational time and computer resources increase accordingly. Accuracy requirements and
computational time vary inversely and a compromise must be found; this compromise is to use a
finer mesh in the area of interest. This can be easily obtained both in finite element and in finite
difference method. For example, mesh resolution must be finer in the vicinity of rivers and natural
boundaries, waterworks (pumping or injection wells), pollution sources, or in areas of steep
hydraulic or concentration gradients. The mesh can be coarser in areas with scarce information, in
areas presenting a reduced interest or where the gradients are small.
To avoid the numerical deviation from a correct computation or to minimize it, an adequate cell
size, time step and mesh orientation must be chosen.
In principle, variable cell sizes allow a greater flexibility, needed especially in zones with a high
degree of heterogeneity in hydraulic or transport parameters or in boundary conditions. Still, a
highly variable cell sizes introduces a loss of accuracy. To facilitate the model convergence, to
increase the stability and to reduce inaccuracy, a gradual variation in cell sizes is necessary, a
ratio of maximum 2.0 between adjacent cell sizes being recommended.
The ratio of the sides' length of every cell is defined as a function of the travel time along the cell in
each direction; values close to unity are considered ideal, but larger rates (up to 5) are also
acceptable from the point of the introduced numerical errors. For transport problems, the cell size
should not be larger than twice the dispersivity to ensure numerical stability and minimize the
numerical dispersion; this constraint is derived from Peclet criterion:
(7.1)
where:
dx cell size in the x-direction [L]
αx longitudinal dispersivity in the x-direction [L]
In the unsaturated zone a finer mesh is needed and the Pe number should not exceed 0.5.
The mesh orientation has to be chosen parallel with large-scale geologic features; ranges of
elements or cells parallel to a fault easily represent it and its surroundings. In this way, the water
flow and transport is best modeled. When the mesh is oriented along the direction of predominant
groundwater flow, the numerical dispersion is minimized due to the fact that the groundwater
velocity is almost parallel to the calculation-mesh axis.
(7.2)
where:
S storativity [1/L]
T transmissivity [L2/T]
Δx cell width in the x direction [L]
Δy cell width in the y direction [L]
This constraint generally leads to small time steps and thus to quite long computing time. From the
previous relation, it can be seen that the time step and the cell size are related. As a consequence,
the mesh size is chosen on the basis of dispersivity (relation 7.1), while the time step results from
the relation 7.2. It is indicated that the time steps thus obtained should be reduced for a short
interval immediately after changing boundary conditions (hydraulic head or concentration).
For transport problems, in order to minimize the numerical dispersion and maximize the numerical
stability, the time step has to observe the condition that the cell Courant number for the smallest
cell should be no greater than unity:
(7.3)
The boundary conditions represent the connection of the aquifer with external systems and are
very important to define a well posed problem, leading to a unique solution. In fact, these
connections represent either input (discharge, energy, pollutants) in the aquifer or output
(discharge, energy, pollutants) from the aquifer.
One of the most used conditions is known as first kind or Dirichlet condition and represents
prescribed pressure or prescribed hydraulic head. This condition is typical for the calibration stages
when generally prescribed values are used both at the external and the internal boundaries. The
model provides fluxes that enter or leave the aquifer, used subsequently in the next stages of the
mathematical modeling.
The upstream limit of an aquifer at the contact zone with the hill slope is an external boundary
representing a recharge area. During the rainy episodes, the run-off from the slope infiltrates at the
contact with the piedmont plain; at the same time, the saturated zone on the slope provides a
continuous recharge of the aquifer. This continuous or episodic recharge is modeled using a
prescribed hydraulic head obtained from the annual or multi-annual water table map. The
prescribed hydraulic heads lead to hydraulic gradients in the border region, which multiplied by the
hydraulic conductivities from that region provide input discharges. In a similar way, the downstream
limit of the aquifer (usually a river, but in some cases a swamp, wetland, lake or the sea) is
modeled using a prescribed hydraulic head or a prescribed pressure. As a result of model running,
output discharges are obtained.
The internal boundaries of the aquifer (rivers in good hydraulic connection with the aquifer, springs,
ponds, reservoirs, wetlands, swamps, deep excavations, navigable canals, locks, constant-head
wells, seepage faces etc) are also modeled using the Dirichlet condition. For linear boundary
conditions like rivers, the model calculation cells have to be small enough to define the surface
water body accurately. Generally, finite element discretization is more flexible than finite
differences obtained mesh in reproducing the river shape. The prescribed head can be constant or
variable in time. For example, in case of rivers or reservoirs whose water level fluctuates, in
unsteady-state simulations a variable head is recommended; for other boundary conditions, where
such fluctuations do not occur (drains, ground surface, sea level etc) a constant head will be used.
For a two-dimensional, vertical-plane model with sinks (wells withdrawing water), the prescribed
boundary condition can be used to compute water fluxes abstracted from the aquifer: considering
the atmospheric pressure at the well-aquifer contact and ignoring the head loss through the
screen, the prescribed head will approximately be equal with the elevation head of each point
situated above the prescribed water level in the well. In the case of seepage faces along drains or
tunnels, the water level in the hydraulic structure will be used as a prescribed head. Finally, one
still has to mention that the Dirichlet conditions lead to the easiest way to find the solution of the
numerical model; still, abusing of this type of condition makes the model too constrained.
Second kind or Neumann condition or finally, prescribed flux condition is used when water fluxes,
constant or variable in time, are known. Due to the relationship between flux and hydraulic
gradient, a prescribed flux is equivalent to a prescribed hydraulic gradient, or what is the same
thing the derivative of the hydraulic head. The flux condition is less constraining, and thus the
numeric model is moderately difficult to be solved. When the output discharges can be measured
quite precisely (water production wells, drains, springs) or can be roughly approximated by a global
budget of surface water bodies (swamps, wetlands or lakes), this type of condition is
recommended. The discharges used as Neumann condition can be in some cases obtained during
the calibration phase. For instance, the water fluxes at the upstream limit of the aquifer are
immeasurable. As a consequence, in the prediction phase the discharges obtained during the
calibration phase will be used, despite the degree of uncertainty associated with these values. One
has to mention that using only prescribed fluxes as boundary condition will lead to an ill- posed
problem, because of the non-uniqueness of the mathematical solution; at least one hydraulic head
or pressure head had to be prescribed.
The natural recharge can also be considered a condition of this type, although it refers to an
areal extent and not to a linear boundary. The natural recharge can be obtained in different ways:
a) based on the water balance between precipitation, evapotranspiration and run-off at the level of
the hydrogeologic basin for a long period; b) using reservoir models for the ground surface and the
unsaturated zone; c) based on the Richards equation. Theoretically, the last option is the most
rigorous; still, it needs more data than any other method: initial humidity θ0(z) for the soil profile at
different depths, soil characteristic curve h(θ) and hydraulic conductivity k(θ) or k(h) for unsaturated
medium. Finally, to characterize the recharge at the level of the whole hydrogeologic basin and not
in different locations, the local curves should be averaged in order to obtain synthetic curves.
When the water table is close to the ground surface (less than 2 m usually), water is lost by
evaporation via capillarity, while the transpiration is a function of the depth of the root zone. The
Neumann condition can also be used in this case.
Another case of impervious boundary is the free surface of the water table in a two-
dimensional, vertical-plane model; free surface problems occur when modeling the infiltration
through a dike or the drainage effect of a line of wells. The water table, representing the delineation
between saturated and unsaturated flow, is a flow line for which the hydraulic head is equal with
the elevation head. If only a saturated flow model is used, successive runs will lead to the position
of the free surface water.
Third kind, mixed, Cauchy's or Fourier's condition is indicated when a semi-permeable layer
controls the seepage flow. Usually in the lower part of the rivers the water velocity is small and fine
sediments lay down at the bottom of the river; they are driven with the infiltrated water towards the
aquifer, but because of their dimensions they are blocked quite close to the entrance, clogging the
riverbed. Thus, a hydraulic resistance controls the flux exchange between the river and the aquifer.
If the hydraulic conductivity of the riverbed is approaching the hydraulic conductivity as the aquifer,
the third kind condition reduces to the first boundary condition (head boundary condition). When
the hydraulic conductivity of the riverbed decreases, for the same elevation of the surface water
the flux is diminished more and more; at the limit, the river and the aquifer can be completely
isolated. The flux between the surface water and the aquifer depends on the head difference
between the two water bodies, but the groundwater head is not known in advance. As for the first
two boundary conditions, the water levels can be constant or variable in time. Cauchy boundary is
the least constrained condition, making the solving of the numerical model the most difficult. The
hydraulic conductivity of the riverbed is difficult to be calibrated because the fluxes exchanged
between aquifer and rivers cannot be measured using surface hydrology methods; in fact, aquifer-
river fluxes are of the same order of magnitude or even smaller as the errors made in the surface
discharge evaluation.
By initial conditions one understands the distribution of the hydraulic heads (or pressures) and
concentrations at the start of simulation for the entire flow domain.
For steady flow problems these values represent a numerical facility, in order to help the iterative
solving of the equations system obtained after the discretization of the flow governing differential
equation. In fact, the initial conditions are used as starting point for the iterative computation of the
numerical solution. On the contrary, for transient flow problems the initial state of the system must
be defined, representing a physical requirement. Only knowing the initial state of the system, the
internal structure (in this case the hydrogeological parameters and the governing equation) as well
as the command variables (natural recharge and withdrawals), the trajectory of the system
(evolution of the states variables) and the output variables (output discharges) can be uniquely
determined. The initial state for transient flow can be either measured values, either obtained as a
result of steady flow simulations. In this later case, oscillations of the solution can occur at the first
steps of computation, especially if the transient conditions are far from the equilibrium situation
reached in steady flow simulation. Anyway, a good estimate of the initial conditions, compatible
with previous states and stresses in transient state, accelerate convergence.
For transport problems, the initial conditions are represented by the extent and concentrations of
an existent contaminant plume at the starting time of simulation. The measured concentration
distribution of a plume cannot be used as starting point for a transient transport simulation due to
the scarce information; the monitoring program usually does not foreseen a large number of
measuring points and frequently the highest-occurring concentrations are not even registered. A
solution is to consider estimated source terms as starting point for the transport model; after
transport model calibration against measured concentration in the monitoring network, starting
conditions for transport simulations are obtained.
Data provided by field measurements represent local estimation of different parameters. The
numerical model needs zoned parameters, which are usually different by the local values. In the
frame of each zone, a certain parameter is considered constant. During the calibration process the
uncertain parameters' values will be varied within measured or likely ranges of values; at the same
time, parameters' distributions should be consistent with geological hypotheses. The purpose of
the calibration is the numerical model to provide computed values of the state variables (levels or
concentrations) as close as possible to the measured values of the same variables. If the
calibrated site-specific model is able to reproduce observed responses, it is considered that the
model will provide good results in the prediction phase. This premise is valid only if the future
stresses will be in the same range as during the calibration phase or if the system's behavior will
be linear even outside the registered situations. Sometimes a good agreement of the historical
data can be obtained for unrealistic input data (hydrogeological parameters or natural recharge);
the site-specific model will still not provide accurate results during the prediction phase.
If the volume of available data is small or the model extends significantly beyond the area with a
satisfactory volume of data, more model calibrations can be obtained from the same data set. The
"calibrated" numeric model is not unique, and the obtained parameters of each "successful"
calibration can be considered realization of a statistical process. As a result, even the predictions'
results can be considered as statistical variables. Still, in the current practice, even for an
unsatisfactory volume of data, only one calibration is undertaken; the calibrated model being
considered unique, a deterministic vision is implicitly admitted.
Depending on the available information, the calibration of the flow model parameters can be
achieved either in steady state or in transient state. If a steady state calibration is chosen, the
water-table or the potentiometric map should of course correspond to steady state conditions. The
best calibration in steady state is thus obtained for multi-annual values of natural recharge and
water withdrawals. The water-table or the potentiometric map will be drawn on the basis of multi-
annual levels; the head boundary conditions correspond also to multi-annual levels. It must be
mentioned that the multi-annual value of the natural recharge is relatively well estimated on the
basis of global water balance at the level of the corresponding hydrogeologic basin.
If due to the lack of level registration or water abstraction at the beginning of the water withdrawal
period, or due to the continuous increase of the wells' number and of their influence on the water
levels, a valid multi-annual map cannot be obtained, a shorter period (1 year usually) is chosen for
calibration. Normally, the calibration year should correspond to the beginning of the water
withdrawal, when the abstracted discharges were very reduced. Unfortunately, for that period the
aquifers were less investigated, and the monitoring network was formed by a small number of
wells. Thus, because of few measured levels, the contour lines are drawn less precisely than those
corresponding to recent maps. As a compromise, a recent period is chosen. Ideally, the calibration
year should be a medium year regarding the precipitation regime, with other stress factors as
constant as possible. Still, this year should not follow after an excessively dry or a wet year,
because of its influence on the current levels. Calibration is performed using averaged yearly data
for water levels and stress factors. Still, the natural recharge is less precisely evaluated than in the
case of the multi-annual data.
Manual calibration is strongly connected with the sensitivity analyses, which put into evidence the
variation of the model response as a result of variation of uncertain parameters. From a
mathematical point of view, the sensitivity is given by the partial derivative of the predicted variable
(hydraulic head for instance) to the change in input parameter value:
(8.1)
where:
ΔH difference in predicted value between the base case and the sensitivity case, evaluated
in one ore more locations
ΔP change in input parameter value
s sensitivity index
The parameters that must be firstly modified during the calibration process are those for which the
sensitivity index is greater. For a good calibrated model, the sensitivity index is very reduced being
theoretically at the limit equal to zero. For a model not calibrated enough, the sensitivity index is a
measure of the disagreement degree between computed and measured values, or in other words a
measure of the calibration weakness; in this case, the solution is to continue the calibration or to
organize new data collecting before continuing calibration.
The optimisation method is based on a mathematical model, formed by the discretized
governing equation(s), constraints concerning the ranges of the different parameters and an
objective function to be minimized. The objective function is usually a weighted measure of the
differences between the computed and measured values of a state variable (hydraulic head,
concentration). In some cases, one can add terms reflecting the differences between the measured
and the calibrated parameters (hydraulic conductivities for example) or between the measured and
computed outputs (discharges in drains, wells or sources).
Due to the fact that the computed levels or discharges are numerical values, the objective
function is not analytically derivable to the unknown parameters. In finding its optimal value and
corresponding configuration of the parameters, either numerical derivation is used in steepest
descent methods, either algorithms which do not involve the calculation of the objective function
are employed. In both cases, the optimization process involves large computational effort. If the
mathematical model used for the inverse problem is not complex enough (without limits for the
parameters variation or adequate constraints to avoid steep variation of the parameters etc) or
cannot be adapted to the specificity of the studied aquifer so as to introduce modeler experience,
the results of the inverse method can be used as possible values for the parameters; a trial-and-
error procedure may follow to improve the automated calibration.
After the flow model calibration follows the calibration of the flow and transport model (with
the exception of density-coupled transport, when the calibration process is simultaneous). The
calibration of the transport parameters is much more difficult than the calibration of the flow
parameters. Thus, the source term is often uncertain and can be approximated calculating the
mass of contaminant contained in the volume of the polluted part of the aquifer. Different
processes may produce a decrease in concentration: dispersion, decay or variation of the source
term; the identification of the true process is not an easy thing.
A good understanding of the investigated problem is essential to consider if the calibrated data
represent the natural system or to take the decision to continue improving the quality of these data.
Failure in calibrating the numerical model may have different causes:
The solution to these problems is to check the data and to eliminate those susceptible of being
wrong, to collect new data, or to revise the conceptual model and the governing processes.
Este necesar ca problema calibrării să fie “bine pusă” astfel încât să obţinem un set unic de
parametri care să asigure cea mai bună potrivire (“best fit”) între rezultatele modelului şi
echivalenţii observaţi/măsuraţi.
Aceasta înseamnă că numărul de parametri estimaţi este mai mic decât numărul de observaţii +
numărul de informaţii independente despre parametri. Mulţi modelatori cred că un model bine
calibrat va produce în mod natural predicţii corecte.
Rezolvarea problemei inverse a potenţialului, calibrarea sau, simplu, problema inversă, poate fi
formulată si în modul următor (Medina, 1999):
- cunoscându-se valori măsurate ale sarcinii piezometrice şi informaţii preliminare asupra
parametrilor sistemului, se urmăreşte obţinerea unor estimări ale parametrilor apropiate de
valorile preliminare, şi care duc la o reproducere “optimă” a sarcinilor măsurate de către
cele calculate.
Calibration is an important step in groundwater modeling. Still, a model may be calibrated using
arbitrary input parameters with the only purpose in mind to fit observed state variables or output
values. The model accuracy is not implicitly guaranteed by a successful calibration; thus, the
model should be tested against other situation(s) before concluding that it adequately represents
the physical system and declaring it reliable for predictions. In fact, the validation's purpose is to
eliminate the fear concerning the parameters' non-uniqueness and thus to consider that the unique
solution (the only valid set of parameters) was obtained.
Model validation consists in applying the model for other data set than that used in calibration.
Normally, the data set is divided into two parts, not necessarily having the same volume: one set is
used for calibration, and the other set for validation. It is preferable to undertake the validation for
situations that differ significantly from those encountered during calibration. If the validation is a
success, one has good reasons to believe that the unique solution or a solution close to it was
obtained; as a consequence, the predictions can be trusted. If the validation fails, both data sets
are used for calibration. The model validation will be achieved by different other tests. For
example, one can try to reproduce the natural water levels before starting the water withdrawals. If
the water levels result in some places above the ground level, the model cannot be validated and
the calibration should start again: either the hydraulic conductivities or storage coefficients are
locally under-evaluated, either the natural recharge is over-evaluated, either both situations can
occur.
To calibrate the model for steady state conditions, while for validation transient conditions
are considered.
To perform both calibration and validation in transient state, of course for different periods
and for different boundary conditions.
Model predictions are in fact one of the main purposes of the mathematical modeling. Provisional
simulations can be conducted for different scenarios of introducing new border conditions or
changing some of the calibration border conditions, while the natural internal structure of the
aquifer is normally not modified. Sometimes, previous boundary conditions used in the calibration
phase are partially altered. Modifying surface water levels due to the realization of a reservoir,
increasing water withdrawals by existing wells or extending water works by adding new wells,
introducing artificial recharge or increasing percolation due to the inauguration of an irrigation
system, or constructing an oil refinery or a chemical factory, which represent important sources of
ground environment pollution, are examples of important alteration of the boundary conditions;
prescribed hydraulic heads, fluxes or concentrations are modified. Other times, engineering works
introduce new internal boundary conditions, which did not exist during the calibration phase:
introducing a new drainage border due to a huge excavation or to a tunnel are examples of such
alterations.
In the prediction phase, based on different scenarios of future water demands and for different
schemes concerning the location of additional wells one obtains the water levels in the proximity of
the water works. If the decrease of water levels is not acceptable, other schemes for the extension
of the water works will be considered.
By using both climatic and unsaturated zone models to obtain modified values of the percolation
for different scenarios of CO2 evolution, the influence of climate changes on the groundwater levels
can be modeled. The results may be used to prepare in due time response actions.
In other cases, like the realization of a reservoir, the efficiency of different technical measures to
control groundwater levels in the surrounding area are tested. Drains placed at different depths or
wells characterized by their number, position or discharge are introduced as new boundary
conditions, resulting their effect on the groundwater level control.
The effect of the rehabilitation measures on the concentration values for different schemes of
pumping (position of the captured zones) can also be simulated, choosing between different
alternatives that lead to the best results.
For the prediction phase, a special care should be paid to the use of the boundary conditions.
Generally, for the prediction phase at the input borders it is recommended to use the prescribed
flux condition. The corresponding values were evaluated during the calibration phase, based on
prescribed hydraulic heads. Actually, if the drawdowns due to water abstraction interact with the
limit of the model, supplementary resources can be mobilized; if it is the case the use of a
prescribed flux will overestimate the decrease of the water level due to the pumping wells. Using a
prescribed head condition when there are no additional resources outside the modeled zone, due
to increased hydraulic gradients the mathematical model leads to an overestimation of the input
discharge at the border; thus the decrease of the water level inside the domain will be
underestimated. In both cases, the water balance will not have real values. This example shows
the importance of correctly choosing the limits of the model; one recalls the recommendation that
the limits of the model should be the hydrogeological limits of the aquifer. In these conditions, no
additional resources can be mobilized from outside and a prescribed discharge is the best option
for the upstream limit of the aquifer.
At the output limits, a prescribed head condition or a mixed condition is more indicated than a
prescribed flux condition. The output discharge (in a river, for example) evaluated during the
calibration cannot be used as a boundary condition during the prediction phase if the groundwater
withdrawal leads to a general decrease of the water table reaching the border; in this case the
effective output discharge will be smaller than the value obtained during calibration conditions. If
the drawdowns are very important and the surface water level becomes higher than the
groundwater level, the water fluxes are reversed.
The consequences of changing boundary conditions must be studied not only in steady state, but
also in transient conditions. This is valid especially for local aquifers, having a small capacity of
water storage. In these aquifers, water level fluctuations can be important and despite the fact that
an average situation is acceptable, very severe conditions can occur especially during low levels.
In fact, small aquifers cannot be used as a significant source of water supply due to the low
capacity of storage and compensation of irregular recharge.
11. Incertitudine
Neconcordanţa dintre valorile măsurate şi cele calculate ale sarcinii piezometrice şi debitelor
vehiculate este una din problemele principale cu care se confruntă utilizatorul. Aceste erori pot
avea mai multe cauze (Carrera şi alţii, 1993; Medina, 1999):
erori numerice:
discretizare insuficient de rafinată a modelului;
când se calculează concentraţii, se adaugă erorile obţinute în calcularea sarcinilor;
probleme neliniare;
erori conceptuale:
ipoteze simplificatoare;
date puţine;
condiţii de margine greşite.
În general, erorile conceptuale au un ordin de mărime uriaş; nu există reguli de evitare, doar o
cunoaştere mai bună a sistemului.
Erorile numerice au un ordin de mărime mic; pentru evitare, trebuie schimbat algoritmul de calcul
numeric sau rafinată discretizarea în spaţiu şi/sau timp.
Erorile în aprecierea valorilor parametrilor au un ordin de mărime mediu/mare, dar pot fi diminuate
prin calibrare/ rezolvarea problemei inverse.
Modelarea problemelor hidrogeologice are asociată o incertitudine datorată unor cauze multiple:
Gradul de incertitudine este crescut în multe cazuri datorită insuficienţei datelor pentru estimarea
parametrilor şi validarea modelului. Erori în datele/măsurătorile folosite la identificarea parametrilor
reprezintă de asemenea o sursă de incertitudine.
Erori la modelare
Aşa cum este unanim recunoscut etapa esenţială în modelarea hidrogeologică este dezvoltarea
modelului conceptual. Dacă acesta este greşit (de ex. nu reprezintă fenomenele relevante de
curgere şi transport), restul modelarii – trecerea modelului conceptual in modele matematice şi
numerice şi rezolvarea problemei, sunt pierdere de timp şi bani.
Greşelile din modelare pot fi grupate în 4 categorii (Mercer and Faust, 1981):
(Rivera A, 2007)
APPROACHES TO GROUNDWATER MODELLING
Hydrogeological modelling requires numerical methods to provide both a suitable representation of the
subsurface and an adequate base for the simulation of flow and transport processes required for
environmental studies (water resources, climate change). As a prerequisite to building a (numerical)
geological model, it is necessary first to generate a conceptual model. This conceptual model combines data
and knowledge from various disciplines concerning geometry, geology, physical parameters and processes
of interest. This process however should be iterative; the conceptual model should be updated as additional
environmental, geologic and hydrogeologic studies, as well as simulations with the original model, provide
new data and new understanding of the groundwater flow systems.
ConnectFlow model with a porous medium representation of the near surface and a fractured basement (Serco, 2000)
A combination of the two approaches described above (EPM and DFN) is possible and has been utilized in
the radioactive waste disposal industry. In this category, the porous media and the fracture media are
integrated into a single package that enables groundwater flow models to be constructed from sub-domains
of porous media and discrete fracture networks (Figure 2). This allows the user greater flexibility in
representing the flow in hydrogeological units. Groundwater flow is coupled across the interface between the
sub-domains by ensuring continuity of water pressures and conservation of mass at the interface. This is
possible because the conceptual model in this case is based on Darcy’s law and finite-element methods are
then used to solve the resulting partial differential equations, which in today’s advanced computer
environment is straightforward. The key of this approach however remains the model grid in the
superposition of two distinct sub-domains.
Two different “schools” of groundwater scientists developing the multidisciplinary science of hydrogeology:
fundamental hydrogeologists and environmental hydrogeologists. Fundamental hydrogeologists study and
develop laws and methods to quantify groundwater flow from a theoretical perspective. Environmental
hydrogeologists use those laws and methods to study real aquifer systems, that is, geological formations
containing and conducting water. The second group is the one that most closely interacts with geologists.
Environmental hydrogeologists develop and apply models where groundwater interacts with rocks, rivers,
lakes, ecosystems, land uses, etc. For instance, this group must monitor that groundwater exploitation is
kept within a sustainable use and must provide a clear path of potential contaminants underground. The
challenges for this group of scientists are too many: how to integrate too much existing data and knowledge?
How to deal with so many expertises involved? The current international practices for water resources
management (IWRM), for instance, represents a huge challenge for this group. The IWRM approach
requires the building of an integrated water-resources model which includes the aquifer in relation with the
geology of a watershed, interactions with surface water (rivers, lakes, wetlands), land use practices
(agriculture, urbanization) and most recently, elements to evaluate climate change impacts.