Post on 16-Apr-2018
1
Time-varying analysis of parameter sensitivity at
multiple evaluation scales for hydrologic and
sediment modeling
Presented by: Hui Xie
Authors: Hui Xie, Zhenyao Shen*, Lei Chen, Guoyuan Wei
Beijing Normal University
July 2016
Outline
2
3
BackgroundIntroduction
1
Methodology
Results and discussion
Objectives
Watershed descriptionModel setup
Hydrology sectionSediment section
FAST+Moving windows
3 Summary
Background
Supports model calibration
Background
1
Multiple evaluation scales2
Which period are helpful for identification?
Under which condition a specific process matters?
Time-varying analysis
Different processes act over different time scales
Maximize information content
Guide measurement campaigns
No previous research concerned
Background
3
Objective4
Support management plan
Time-varying + multiple evaluation scales for
SEDIMENT modeling
Improve the understanding of hydrologic and sediment
processes
Methodology
Study site
Methodology
1
Zhangjiachong
watershed
Located at the head of
TGRA, 1.6km2
Model setup2 HSPF
Methodology
IMPERLND
PERLND
RCHRES
Watershed
Hydrology Section
Flows & Storages
Model setup
Model setup
Flows & Storages
Hydrology Section
Detachment & Washoff
Sediment Section
Scour & Deposition
Parameter Description Unit Possiblerange
LZSN Low zone nominal soil moisture storage in. 2.0-15.0
INFILT Index to mean soil infiltration rate in./h 0.001-0.50
KVARY Parameter to describe non-linear groundwaterrecession rate in-1 0.0-5.0
AGWRC Groundwater recession rate day-1 0.85-0.999
DEEPFR Fraction of infiltrating water which enters deepaquifers none 0.0-0.50
BASETP Fraction of potential evapotranspiration whichfulfilled only as outflow exists. none 0.0-0.20
AGWETP Fraction of remaining evapotranspiration thatbe met from active groundwater storage. none 0.0-0.20
CEPSC Interception storage capacity. in. 0.01-0.40
UZSN Nominal upper zone soil moisture storage. in. 0.05-2.0
INTFW Interflow inflow parameter. none 1.0-10.0
IRC Interflow recession parameter. day-1 0.3-0.85
LZETP Index to lower zone evapotranspiration. none 0.1-0.9
Hydrologic parameters1
Parameter Description Unit Possiblerange
SMPF Management Practice (P) factor from USLE none 0.0-1.0
KRER Coefficient in the soil detachment equation in./h 0.05-0.75
JRER Exponent in the soil detachment equation in-1 1.0-3.0
AFFIX Daily reduction in detached sediment day-1 0.01-0.50
COVER Fraction land surface protected from rainfall none 0.0-0.98
KSER Coefficient in the sediment washoff equation none 0.1-10.0
JSER Exponent in the sediment washoff equation none 1.0-3.0
TAUCD1 Critical bed shear stress for deposition in. 0.001-1.0
TAUCS1 Critical bed shear stress for scour in. 0.05-2.0
TAUCD2 Critical bed shear stress for deposition of silt none 1.0-10.0
TAUCS2 Critical bed shear stress for scour of day-1 0.3-0.85
Sediment parameters2
Sensitivity analysis
Methodology
3
… • Global method
…
• Variance decomposition
… • Non-linear models
… • Computationally
efficient
Fourier Amplitude Sensitivity Test (FAST)
Saltelli et al. (1999)
Sensitivity analysis
Methodology
3
Fourier Amplitude Sensitivity Test (FAST)
… • Global method
…
• Variance decomposition
… • Non-linear models
… • Computationally
efficient
50,000 samplings for hydrologic
parameters
20,000 samplings for sediment
parameters
Parameters uniformly distributed
Multiple moving windows
Methodology
4
… •Objective function
… •Moving window sizes
2, 4, 8, 30, 60, 180, 360, 540, 720 days
(Massmann et al., 2014)
Results and Discussion
17
Results and discussion
1
NSE=0.82
R2=0.83NSE=0.59
R2=0.61
Model calibration for daily streamflow and sediment
y = 0.6199x + 78.279
R² = 0.60737
0
500
1000
1500
2000
2500
0 500 1000 1500 2000 2500 3000
18
Results and discussion
2 Time-varying sensitivity of HYDROLOGIC parameters
Parameter
groups
1
2
3
4
AGWRC, KVARY
Base flow
INTFW, IRC
Interflow
Moisture
redistribution
LZSN, UZSN, INFILT
ET, water loss
LZETP, BASETP,
DEEPFR, CEPSC
19
1 Base flow
1 Base flow
High sensitivities in the full period
Especially high in dry periods
More than 180 days for identification
Base flow is critical
Should be calibrated
Long length of observation
2 Interflow
2 Interflow
Highest sensitivities in wet period
Monitoring plan arrangedaccording to the sensitive areasStrong correlation with precipitation
2 – 720 days for identification
3 Moisture redistribution
3 Moisture redistribution
LZS more steady
UZS active in the first year
Moisture distribution3
Moisture distribution
Active in the first year when water fill up the lower storages Percolation & Infiltration more important than water
movement into IFS
3
4 ET
4 ET
4 ET
ET from LZS is critical Long length of monitoring required Impact on wet periods
30
Results and discussion
2 Time-varying sensitivity of SEDIMENT parameters
Parameter
groups
1
2
3
4
KRER, JRER
Detachment
KSER, JSER
Washoff
Scour/deposition
TAUCD, TAUCS
Controllable
SMPF, COVER
Detachment1T
SS
(mg
/L)
Detachment1
Erodibility of the Soil Type
Strong correlation with Precipitation
2 WashoffT
SS
(mg
/L)
2 Washoff
Effects of surface condition and flow Sensitive in most of the time 45 days or more required for identification Management practices’ impact on KSER
2 Scour/depositionTA
U
3 Scour/depositionTA
U
3 Scour/deposition
Determine the Sediment dynamics in reach Sensitivity and evaluation scales Varies as Bed
Shear Stress changes Practices in the channels make sense
TA
U
4 ControllableT
SS
(mg
/L)
4 Controllable
Similar to P and C factors in USLE High sensitivity during the wet season
Management practices are effective when evaluatedin scales of more than 60 days
Summary
Summary
At what time period specific parameters are sensitive is defined.
Over what length of measurement to maximize the sensitivity
signals is defined.
The results help support measurement campaigns, calibration,
and management plan.
THANK YOUAny question and suggestion?
Contact: bnuxiehui@hotmail.com