HAL Id: cea-02435102
https://hal-cea.archives-ouvertes.fr/cea-02435102 Submitted on 10 Jan 2020
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
Simultaneous estimation of groundwater recharge and
hydrodynamic parameters for groundwater flow
modeling
F. Hassane Maina, P. Ackerer, Olivier Bildstein
To cite this version:
F. Hassane Maina, P. Ackerer, Olivier Bildstein. Simultaneous estimation of groundwater recharge and hydrodynamic parameters for groundwater flow modeling. SAMO 2016 - International Conference on Sensitivity Analysis of Model Output, Nov 2016, Le Tampon, France. �cea-02435102�
SIMULTANEOUS ESTIMATION OF
GROUNDWATER RECHARGE AND
HYDRODYNAMIC PARAMETERS
FOR GROUNDWATER FLOW
MODELING
29/09/2016 1
Fadji HASSANE MAINA (Université de Strasbourg)
Philippe ACKERER (Université de Strasbourg)
INTRODUCTION
INTRODUCTION
29/09/2016
Groundwater (aquifers): 3rd freshwater reservoir on the planet
Replenished essentially by precipitation through groundwater recharge
.
sH
S
K H
q
t
q
K H
.
sH
S
q
q
t
2INTRODUCTION
29/09/2016
Groundwater (aquifers): 3rd freshwater reservoir on the planet
Replenished essentially by precipitation through groundwater recharge The main available water resources for many countries
Threatened by pollution and over-exploited
Usually described by models Simulations of piezometric levels (Hydraulic heads)
Physical Models are widely used
1. Continuity equation (mass conservation):
2. Darcy’s law (computation of fluxes):
Diffusivity equation
(flow equation):Values of H knowing S, K and q
s S and K are constant, q
sincludes
groundwater recharge
.
sH
S
K H
q
t
q
K H
.
sH
S
q
q
t
229/09/2016
Groundwater flow models: Sophisticated and
accurate
Recharge modeling:
Challenge
o Complex hydrological component
Groundwater recharge depends on:
Climatic conditions Vegetation
Soil and root zone Unsaturated zone
GROUNDWATER RECHARGE
Recharge : water transfer from precipitation to groundwater
BRGM
29/09/2016
Groundwater flow models: Sophisticated and
accurate
Recharge modeling:
Challenge
o Complex hydrological component
Groundwater recharge depends on:
Climatic conditions Vegetation
Soil and root zone Unsaturated zone
Values can not be directly measured
Estimation
Direct methods: lysimeters, TDR Empirical methods
Tracers
Mathematical models
GROUNDWATER RECHARGE
Recharge : water transfer from precipitation to groundwater
BRGM
RECHARGE ESTIMATION:
COUPLED MODEL
29/09/2016
Simulation of flow in both unsaturated and saturated zones
4
Runoff
Precipitations, Evapotranspiration
Infiltration
Flow in the Unsaturated Zone
Interactions between
Atmosphere and Soil
and Hydrodynamic in
the Unsaturated Zone :
Nash
Groundwater Flow:
Diffusivity equation
Recharge
Flow in the Saturated Zone
Coupling
NASH MODEL
Hydrodynamic in the unsaturated zone
t Rn Taue t Tau t Tau Rn 1 1Root
Zone
1 1 1
1
0 Recharge k k n t n k n t k Pe t d
Nash parameters: RUMAX, RN, TAU
Water transfer in the unsaturated zone: delay and spreading
Modeling : propagation of a unit pulse signal
Threshold : RUMAX
ET Rainfall
if RUMAX is reached
the excess water (Pe) goes into the unsaturated zone
5 Infiltration Recharge Time Water Level ET depends on soil saturation
Model uses constants called parameters
Aquifers are highly heterogeneous
Parameters are not known accurately or evenunknown in natural environments
PARAMETERS ESTIMATION
Model uses constants called parameters
Aquifers are highly heterogeneous
Parameters are not known accurately or evenunknown in natural environments
Available data: meteorological and piezometric levels
PARAMETERS ESTIMATION
6 0.00 50.00 100.00 150.00 200.00 250.00 300.00 294 296 298 300 302 304 306 308 Ra inf a ll ( mm) Hy d raulic Heads ( m) Model uses constants called parameters
Aquifers are highly heterogeneous
Parameters are not known accurately or evenunknown in natural environments
Available data: meteorological and piezometric levels
PARAMETERS ESTIMATION
Simultaneous estimation of groundwater flow and recharge parameters from hydraulic heads
variations
Recharge is estimated by calibrating piezometric levels
6 0.00 50.00 100.00 150.00 200.00 250.00 300.00 294 296 298 300 302 304 306 308 Ra inf a ll ( mm) Hy d raulic Heads ( m)
GLOBAL SENSITIVITY ANALYSIS
Correlation between recharge and storage capacity (S and qs)
Recharge and storage capacity lead to the same effect
29/09/2016
q
K H
Determine q et K knowing H ! Calibration is performed over time and space
But… recharge is variable over time while the storage capacity is constant When Determination of K
When Determination of q knowing K Global sensitivity analysis
Assess the calibration approach and guide the parameter estimation
Recharge Storage capacity
0
K H
0
q
q
0
High recharge and high storage Low recharge and low storage
7
Uncertain parameters: variability ranges known Quantify the effects of parameters uncertainty Global sensitivity analysis: focuses on the output uncertainty over the entire range of values of
the input parameters both single and in combination with one another
GLOBAL SENSITIVITY ANALYSIS
29/09/2016
Variance based approach:
Sobol Indices
Surrogate model: Chaos polynomial expansion
8
q
K H
q
andK
are parameters andH
the state variableq
andK
are unknowns
0.0 0.5
q K105 101n
simulations with different values ofq
andK
n values of HStudy the effects of variations of q and K on H
29/09/2016
SOBOL INDICES :
CHAOS POLYNOMIAL EXPANSION
Let us consider a model with y as output and n input parameters, X the parameters vector ANOVA decomposition :
Orthogonality of f
Sensitivity (Sobol) Indices :
Computation of f Polynomial expansion
Chaos polynomial expansion
0 1,2,..., 1 2 1 1,
...
,
,...,
n n i i ij i j n n i jy
f
f x
f
x x
f
x x
x
0 s K RN RN K,,
S K,,
....
H
H
f
S
f
K
f
RN
f
RN K
f
S K
;
ij i i ijV
V
S
S
V
V
1
0,...,
j j n jy
a
x
x
1,...., 1 1...
n n i ij n i jV
V
V
V
fi : contribution of single parameter i to the output
fij : contribution of the combined effect of parameters i and j to the output
x : variable
𝚿 : orthogonal polynomial (Legendre, Hermite)
GLOBAL SENSITIVITY ANALYSIS
29/09/2016
A subdomain of the Upper Rhine alluvial aquifer widely studied at Université de Strasbourg
Average rainfall per year: 1100 mm
Average of evapotranspiration per year: 900 mm
Simulate variations of piezometric levels during 1500 days Sensitivity analysis performed over the last 450 days
Investigate spatial and temporal variations of sensitivity indices
1500 simulations of Quasi Monte Carlo
Field based sensitivity analysis
GLOBAL SENSITIVITY ANALYSIS:
NASH MODEL
Nash parameters for recharge: RUMAX, RN, TAU Aquifer parameters: K and S
Parameters defined by zones (Zonation) 20 parameters
Mean and Variance of the hydraulic head
RN and TAU Permeability Storage capacity
11
GLOBAL SENSITIVITY ANALYSIS:
NASH MODEL
29/09/2016
Nash parameters for recharge: RUMAX, RN, TAU Aquifer parameters: K and S
Parameters defined by zones (Zonation) 20 parameters
170 165 160 155 150 145 140 10 km 6 8 9 2 1 4 10 5 3 7 11 12 13 14 15
Frame 00131 Mar 2015Zonation
100.1 90.1 80.1 70.1 60.1 50.1 40.1 30.1 20.1 10.1 0.1 10 km 6 8 9 2 1 4 10 5 3 7 11 12 13 14 15 Frame 00108 Jun 2016Zonation
Mean and Variance of the hydraulic head
Mean Variance
Hydraulic heads range between 140 and 165 m
Variance ranges between 0 and 100: the imposed boundaries have the lowest variance
0
100
200
300
400
500
0,0
0,2
0,4
0,6
0,8
1,0
1,2
1,4
Total Sensitivity Indices
Time (days)
S1
K3
RN3
RUMAX1
Point 1
29/09/2016GLOBAL SENSITIVITY ANALYSIS:
NASH MODEL
Influential parameters: S, K, RUMAX, RN
12
29/09/2016
GLOBAL SENSITIVITY ANALYSIS:
NASH MODEL
Influential parameters: S, K, RUMAX, RN
0 100 200 300 400 500 0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4
Recha
rge (mm/day)
Time (days)
Recharge
12 0 100 200 300 400 500 0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4Total Sensitivity Indices
Time (days) S1 K3 RN3 RUMAX1 Point 1 0 100 200 300 400 500 140,6 140,8 141,0 Hydrau lic He ads (m) Time (days)
H
29/09/2016
GLOBAL SENSITIVITY ANALYSIS:
NASH MODEL
Influential parameters: S, K, RUMAX, RN
0 100 200 300 400 500 0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4
Recha
rge (mm/day)
Time (days)
Recharge
12 0 100 200 300 400 500 0,1 0,2 0,3 0,4 0,5 0,6Total Sensitivity Indice of K3
Time (days)
S
K3 0 100 200 300 400 500 0,1 0,2 0,3 0,4 0,5 0,6Total Sensitivity Indice of S1
Time (days)
S
S1Decrease of piezometric level
Decrease of the influence of
permeability 0 100 200 300 400 500 0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4
Total Sensitivity Indices
Time (days) S1 K3 RN3 RUMAX1 Point 1 0 100 200 300 400 500 140,6 140,8 141,0 Hydrau lic He ads (m) Time (days)
H
Temporal variation of sensitivity indices at point 1
Increase of piezometric level
Increase of the influence of
29/09/2016
GLOBAL SENSITIVITY ANALYSIS:
NASH MODEL
0 100 200 300 400 500 0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4Recha
rge (mm/day)
Time (days)
Recharge
0 100 200 300 400 500 0,0 0,1 0,2 0,3 0,4 0,5Total Sensitivity Indice of RN
Time (days)
S
RN13
Temporal variation of sensitivity indices at point 1
RN: water transfer in the unsaturated zone
(delay and spreading)
29/09/2016
If RU>RUMAX
Decrease of the sensitivity
of RUMAX
Increase of the recharge
Increase of the sensitivity of
RN
GLOBAL SENSITIVITY ANALYSIS:
NASH MODEL
0 100 200 300 400 500 0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4Recha
rge (mm/day)
Time (days)
Recharge
0 100 200 300 400 500 0,0 0,1 0,2 0,3 0,4 0,5Total Sensitivity Indice of RN
Time (days)
S
RN 0 100 200 300 400 500 0,00 0,05 0,10 0,15 0,20 0,25 Ru Rumax Water Level (m) Time (days)Temporal Variation of RU and RUMAX
13 0 100 200 300 400 500 0,0 0,1 0,2 0,3 0,4 0,5
Total Sensitivity Indice of RUMAX Time (days)
SRUMAX
Temporal variation of sensitivity indices at point 1
RN: water transfer in the unsaturated zone
(delay and spreading)
29/09/2016
GLOBAL SENSITIVITY ANALYSIS:
NASH MODEL
0 100 200 300 400 500 0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4Recha
rge (mm/day)
Time (days)
Recharge
0 100 200 300 400 500 0,0 0,1 0,2 0,3 0,4 0,5Total Sensitivity Indice of RN
Time (days)
S
RN13
Temporal variation of sensitivity indices at point 1
RN: water transfer in the unsaturated zone
(delay and spreading)
29/09/2016
If RU>RUMAX
Decrease of the sensitivity
of RUMAX
Increase of the recharge
Increase of the sensitivity of
RN
GLOBAL SENSITIVITY ANALYSIS:
NASH MODEL
0 100 200 300 400 500 0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4Recha
rge (mm/day)
Time (days)
Recharge
13Temporal variation of sensitivity indices at point 1
RN: water transfer in the unsaturated zone
(delay and spreading)
29/09/2016
Sensitivity maps over time
GLOBAL SENSITIVITY ANALYSIS:
NASH MODEL
S1
RUMAX3
14RUMAX3: sensitive from the
1
st to185
th dayS1: influential between the
185
th and the346
th dayWeak or even negligible correlations between the 185th and the 346th day
Identification of S
During the first days
Determination of influential parameters
Nash: RUMAX and RN Groundwater flow: K and S
Temporal variation of parameter sensitivity
Nash: saturation of the 1st reservoir (RUMAX) and recharge (RN)
Permeability: decrease of the hydraulic heads Storage capacity: increase of the hydraulic heads
Weak interactions between aquifer and recharge parameters
Confirmation of the feasibility of simultaneous estimation of recharge and
groundwater flow parameters
29/09/2016
q
K H
K constant, q variable q=0, détermination of KGLOBAL SENSITIVITY ANALYSIS:
CONCLUSION
Thank you for
GLOBAL SENSITIVITY ANALYSIS
29/09/2016
Parameters Interactions
Nash model
Richards model
14 0 100 200 300 400 500 0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4 1,6 Varian ce Temps (jour) RUMAX3 RN1 RN2 RN3 RN4 K3 S1 S2 Variance Point 4 0 100 200 300 400 500 0,00 0,02 0,04 0,06 0,08 0,10 0,12 0,14 Ind ices d e Sen sibilité d 'Or dr e 2 Temps (jour) K3&S1 RN3&S1 RN2&K3 RU3&S1 RUMAX3&K2 RUMAX3&RN2 Point 4 0 100 200 300 400 500 0,00 0,05 0,10 0,15 0,20 0,25 0,30 0,35 0,40 Varian ce Temps (jour) KS1 K3 S1 Variance Point 1 0 100 200 300 400 500 0,00 0,05 0,10 0,15 0,20 Ind ices d e Sen sibilité d 'Or dr e 2 Temps (jour) K3&S1 HE1&S1 HE1&K3 S1 KS1&S1 KS1&K3 Point 1 0 100 200 300 400 500 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 Varian ce Temps (jour) RUMAX3 RN1 TAU1 K2 S2 S3 Variance Point 2 0 100 200 300 400 500 0,00 0,02 0,04 0,06 0,08 0,10 0,12 0,14 Ind ices d e Sen sibilité d 'Or dr e 2 Temps (jour) K2&S3 RN1&S3 RN1&K2 RN1 & TAU1 RUMAX3&RN1 Point 2
PARAMETER ESTIMATION
29/09/2016
Diffusivity equation (Transient state with recharge )
Steady state without recharge
Steady state with recharge
Transient state without recharge
14
.
sH
S
K H
q
t
.
0
K H
.
K H
q
s
.
0
H
S
K H
t
OVERESTIMATION OF RECHARGE
29/09/2016
14
If the hydraulic heads decreases, the imposed pressure decreases, and the flux calculated is positive and strong
If the hydraulic heads increases, the imposed pressure increases, and the flux calculated is negative
Assumption of constant hydraulic head over decade
1 1 2 ( ) 1 n n n i i i i i h h q K h z K h z z
RESOLUTION OF RICHARDS
EQUATION
29/09/2016
14 Pression à l’état initial
M étho de s it ér a ti v es : P ica rd o u N ew to n Ajustement des paramètres Ajustement du pas de temps Calcul du pas de temps Résolution du système : Calcul de la pression Calcul du pas de temps suivant Calcul des paramètres Calcul des paramètres Calcul du pas de temps Résolution du système : Calcul de la pression Vérification des critères de convergence Vérification des critères de convergence M étho de s n o n ité ra tiv es
Pression au pas de temps suivant h=f(K,θ), K= f(h,θ) et θ = f(h)