numerical model optimization of surface-water...
TRANSCRIPT
NUMERICAL MODEL OPTIMIZATION OF NUMERICAL MODEL OPTIMIZATION OF SURFACESURFACE--WATER INFLOWS TO ACHIEVE WATER INFLOWS TO ACHIEVE RESTORATION SALINITY PERFORMANCE RESTORATION SALINITY PERFORMANCE
MEASURESMEASURES
GEER 2008 ConferenceGEER 2008 ConferenceJuly 29, 2008July 29, 2008
Dawn JamesDawn James11 and Eric Swainand Eric Swain22
11MWH Americas, IncMWH Americas, Inc
22U.S. Geological Survey, Florida Integrated Science U.S. Geological Survey, Florida Integrated Science CenterCenter
PROJECT STRATEGYPROJECT STRATEGY
•• To use inverse techniques on a numerical To use inverse techniques on a numerical model to determine methods of achieving model to determine methods of achieving desired hydrologic objectivesdesired hydrologic objectives
•• Using model simulations to analyze how Using model simulations to analyze how specific waterspecific water--management practices management practices affect factors such as water levels, flows, and affect factors such as water levels, flows, and salinities. salinities.
•• In lieu of trial and error methods, In lieu of trial and error methods, optimization techniques were employed to optimization techniques were employed to more precisely and directly define good water more precisely and directly define good water management alternatives.management alternatives.
Project LocationProject Location
Everglades NationalProject
Florida Bay
SICS Model Area
Inland Flows
Inland Flows
Project Location
Tools used in experimentTools used in experiment
•• SICS Model:SICS Model: Coupled surfaceCoupled surface--water/ water/ groundground--water representation of southern water representation of southern EvergladesEverglades
•• UCODE:UCODE: Parameter estimation code used Parameter estimation code used to optimize target performance measures to optimize target performance measures for physical characteristics in SICS modelfor physical characteristics in SICS model
•• Performance Measures:Performance Measures: The desired The desired objectives for ecological restoration, i.e. objectives for ecological restoration, i.e. salinity rangessalinity ranges
SICS Model PrinciplesSICS Model Principles
•• Uses FTLOADDS coupled hydrodynamic Uses FTLOADDS coupled hydrodynamic surface water and ground water flow surface water and ground water flow model.model.
•• Salinity transport represented in both Salinity transport represented in both ground water, surface water and leakage.ground water, surface water and leakage.
•• Applied in several locations in south Applied in several locations in south Florida.Florida.
UCODE Parameter EstimationUCODE Parameter Estimation
•• Developed to perform inverse modeling Developed to perform inverse modeling for any numerical modelfor any numerical model
•• User defined input and output formatUser defined input and output format•• Varies parameters within allowable rangesVaries parameters within allowable ranges•• Background of successful applicationsBackground of successful applications
Performance Measure StrategyPerformance Measure Strategy
•• Multi agency effort to restore the Florida Everglades Multi agency effort to restore the Florida Everglades and offshore areas to more natural conditionsand offshore areas to more natural conditions
•• Through evaluation of Florida Bay within the context of Through evaluation of Florida Bay within the context of numerous regional water resources issuesnumerous regional water resources issues
•• Evaluate the connection between Florida Bay and Evaluate the connection between Florida Bay and 1.1. EvergladesEverglades2.2. Gulf of MexicoGulf of Mexico3.3. Florida Keys marine ecosystemsFlorida Keys marine ecosystems
•• Identify modification needed to restore the waterIdentify modification needed to restore the water--quality and ecological conditions of Florida Bay.quality and ecological conditions of Florida Bay.
METHODS METHODS
•• SICS represents the coastal salinity SICS represents the coastal salinity response to changing inland flowsresponse to changing inland flows
•• UCODE adjusts the boundary condition UCODE adjusts the boundary condition inland flows in the SICS model based on inland flows in the SICS model based on SICS output salinitiesSICS output salinities
•• UCODE and SICS interact iteratively to UCODE and SICS interact iteratively to determine the inland flows that induce the determine the inland flows that induce the desired salinity value. desired salinity value.
Optimization ProcedureOptimization Procedure
UCODEEstimation of Parameters
Parameters defining
Inland inflows
SICS Preprocessor SICS
model input
SICS Model
SICS model output
SICS Model Postprocessor
Salinities at target locationUCODE determines
statistical relationship between inland inflows
and salinities
SICS Model Area and WaterDelivery Inflows
METHODSMETHODS-- continuedcontinued•• Six Parameter Method: Two multipliers for each Six Parameter Method: Two multipliers for each
discharge inflow locations; one for higher than average discharge inflow locations; one for higher than average flows and one for lower than average flows.flows and one for lower than average flows.
•• Two Parameter Method A: Two multipliers for TSB Two Parameter Method A: Two multipliers for TSB discharge, one for higher than average flows and one for discharge, one for higher than average flows and one for lower than average flows. Discharges at the other two lower than average flows. Discharges at the other two inflow points are set in the proportional to TSB as inflow points are set in the proportional to TSB as existing. More computationally efficient than six existing. More computationally efficient than six parameter method.parameter method.
•• Two Parameter Method B: Same as method A, save Two Parameter Method B: Same as method A, save multipliers are selected for higher than average multipliers are selected for higher than average salinitiessalinities and lower than average and lower than average salinitiessalinities
FBFKFS Zones for Defining Performance Measures
West Lake Joe Bay
TSB
0
5
10
15
20
25
30
ORIGINAL
OPTIMIZED
L-31W
0
5
10
15
20
25
C-111
0
5
10
15
20
25
30
35
DIS
CH
AR
GE
,IN
CU
BIC
ME
TE
RS
PE
RS
EC
ON
D
1/1
/1996
3/1
/1996
5/1
/1996
7/1
/1996
9/1
/1996
11/1
/1996
1/1
/1997
DATE
EXPLANATION
1/1
/1996
3/1
/1996
5/1
/1996
7/1
/1996
9/1
/1996
11/1
/1996
1/1
/1997
DATE
1/1
/1996
3/1
/1996
5/1
/1996
7/1
/1996
9/1
/1996
11/1
/1996
1/1
/1997
DATE
ORIGINAL
OPTIMIZED
EXPLANATION
ORIGINAL
OPTIMIZED
EXPLANATION
DIS
CH
AR
GE
,IN
CU
BIC
ME
TE
RS
PE
RS
EC
ON
DD
ISC
HA
RG
E,
INC
UB
ICM
ET
ER
SP
ER
SE
CO
ND
RESULTS
Original and optimized model inflows for six-parameter method
0
1
2
3
4
5
6
7
8
9
10
ORIGINAL
OPTIMIZED
1/1
/1996
2/1
/1996
3/1
/1996
4/1
/1996
5/1
/1996
6/1
/1996
7/1
/1996
8/1
/1996
9/1
/1996
10/1
/1996
11/1
/1996
12/1
/1996
1/1
/1997
EXPLANATION
SA
LIN
ITY,
INP
RA
CT
ICA
LS
ALIN
ITY
UN
ITS
(PS
U)
DATE
RESULTS- CONTINUED
Zone 1 original and optimized salinities for six parameter method.
RESULTS- Continued
Original and optimized model inflows for two-parameter method, flow criteria.
TSB
0
5
10
15
20
25
30
35
40
45
L-31W
0
2
4
6
8
10
12
14
16
18
C-111
0
5
10
15
20
25
30
ORIGINAL
OPTIMIZED
1/1
/1996
3/1
/1996
5/1
/1996
7/1
/1996
9/1
/1996
11/1
/1996
1/1
/1997
EXPLANATION
ORIGINAL
OPTIMIZED
EXPLANATION
ORIGINAL
OPTIMIZED
EXPLANATION
1/1
/1996
3/1
/1996
5/1
/1996
7/1
/1996
9/1
/1996
11/1
/1996
1/1
/1997
1/1
/1996
3/1
/1996
5/1
/1996
7/1
/1996
9/1
/1996
11/1
/1996
1/1
/1997
DIS
CH
AR
GE
,IN
CU
BIC
ME
TE
RS
PE
RS
EC
ON
D
DATE
DATE
DATE
DIS
CH
AR
GE
,IN
CU
BIC
ME
TE
RS
PE
RS
EC
ON
DD
ISC
HA
RG
E,
INC
UB
ICM
ET
ER
SP
ER
SE
CO
ND
RESULTS- CONTINUED
Zone 1 original and optimized salinities for two parameter method, flow criteria.
0
1
2
3
4
5
6
7
8
9
10
ORIGINAL
OPTIMIZED
SA
LIN
ITY,
INP
RA
CT
ICA
LS
ALIN
ITY
UN
ITS
(PS
U)
1/1
/1996
2/1
/1996
3/1
/1996
4/1
/1996
5/1
/1996
6/1
/1996
7/1
/1996
8/1
/1996
9/1
/1996
10/1
/1996
11/1
/1996
12/1
/1996
1/1
/1997
YEAR
EXPLANATION
1.1. In order to bound the simulated inflows, a In order to bound the simulated inflows, a parameter transformation was necessary.parameter transformation was necessary.
2.2. By matching to a synthetic timeseries, possible By matching to a synthetic timeseries, possible alternate solutions are not considered.alternate solutions are not considered.
3.3. The SICS model inflow boundary conditions The SICS model inflow boundary conditions depend on conditions outside the model depend on conditions outside the model domain, so modeldomain, so model--produced values may not be produced values may not be realistic.realistic.
Disadvantages in UCODE Disadvantages in UCODE applicationapplication
Advantages of Using UCODEAdvantages of Using UCODE•• This method creates water management This method creates water management
scenarios rather than using the standard scenarios rather than using the standard method of trial and error simulations of method of trial and error simulations of proposed scenarios.proposed scenarios.
•• Optimizing the decision making process Optimizing the decision making process rather than a subjective approach.rather than a subjective approach.
•• UCODES output statistics define UCODES output statistics define uncertainty in the system.uncertainty in the system.
Alternative ApplicationsAlternative Applications
•• Numerical model output can be used to Numerical model output can be used to define a statistic which would be minimized define a statistic which would be minimized or maximized by UCODE; such as total or maximized by UCODE; such as total variance, difference with optimal value, or variance, difference with optimal value, or hydroperiodshydroperiods
••Management model Management model –– Use largerUse larger--scale scale model instead of SICS so hydraulic structure model instead of SICS so hydraulic structure operations can be optimized for a more operations can be optimized for a more realistic and useful solution.realistic and useful solution.
QUESTIONS?