Calibration, performance and sensitivity analysis of the HEC-HMS HMS model Juraj M. Cunderlik Slobodan P. Simonovic Assessment of Water Resources Risk and Vulnerability to Changing Climatic Conditions
Outline Introduction Model structure Model calibration Model performance Model sensitivity Discussion 2
Introduction Both event and cont models calibrated Verification and sensitivity analysis Project Report IV Model applications 3
Meteorologic component RAINFALL Rainfall loss component PERVIOUS SURFACE IMPERVIOUS SURFACE Direct runoff component LOSSES DIRECT RUNOFF Baseflow component River routing component AQUIFER BASEFLOW RIVER CHANNEL Reservoir component Event model structure RESERVOIR OPERATION BASIN OUTLET 4
Meteorological component EVAPOTRANSPIRATION PRECIPITATION Snow component SNOW ACCUMULATION AND MELT Precipitation loss component PERVIOUS SURFACE IMPERVIOUS SURFACE Direct runoff component LOSSES DIRECT RUNOFF Baseflow component River routing component AQUIFER BASEFLOW RIVER CHANNEL Reservoir component Continuous model RESERVOIR OPERATION structure BASIN OUTLET 5
Model calibration Stage A Stage B Stage C 43.6 43.5 43.4 43.3 43.2 43.1 43 42.9-81.5-81.4-81.3-81.2-81.1-81 -80.9-80.8-80.7 6
Start UTRCA data Event model calibration Initial parameter values Manual calibration Initial parameter values HEC-HMS Soft constraints Optimization Hard constraints Performance evaluation End 7
Start Cont model calibration UTRCA data Initial parameter values HEC-HMS Soft constraints Manual calibration Hard constraints Performance evaluation End + snow model calibration + semiannual parameterization 8
Model performance Flow comparison graph Scatter graph Residual graph Objective function graph Goodness-of of-fit fit measures 9
Flow comparison graph 10
Scatter graph 11
Residual graph 12
Objective function graph 13
Goodness-of of-fit fit measures Percent error in peak (PEP) Percent error in volume (PEV) Lag-0 0 cross-correlation correlation coefficient (CORR) Relative BIAS (RBIAS) Relative RMSE (RRMSE) Relative peak-weighted RMSE (RPWRMSE) 14
Model performance Thames River @ Byron
1000 900 Event model performance Modeled Observed Calibration 800 700 600 Q [m 3 s -1 ] 500 400 300 200 100 0 3-Jul-00 5-Jul-00 7-Jul-00 9-Jul-00 11-Jul-00 13-Jul-00 15-Jul-00 17-Jul-00 16
300 250 Event model performance Modeled Observed Verification 200 Q [m 3 s -1 ] 150 100 50 0 29-Jul-00 31-Jul-00 2-Aug-00 4-Aug-00 6-Aug-00 8-Aug-00 10-Aug-00 17
Event model performance Byron@Thames PEPF [%] PEV [%] CORR [-] RBIAS [%] RRMSE [%] RPWRMSE [%] Calibration 3.211 1.405 0.992-15.047 32.076 25.314 Verification 1 1.937 8.743 0.977-17.101 26.166 23.994 Verification 2 2.686 26.919 0.967-40.230 45.421 40.740 Well fitted peaks Quicker recessions Underestimated volumes Reservoir operation 18
900 800 Cont model performance Modeled Observed Calibration 700 600 Q [m 3 s -1 ] 500 400 300 200 100 0 11-Oct-83 19-Jan-84 28-Apr-84 6-Aug-84 14-Nov-84 22-Feb-85 2-Jun-85 10-Sep-85 19-Dec-85 19
1200 Cont model performance Modeled 1000 Observed Verification 800 Q [m3s-1] 600 400 200 0 8-Oct-95 16-Jan-96 25-Apr-96 3-Aug-96 11-Nov-96 19-Feb-97 30-May-97 7-Sep-97 16-Dec-97 20
Cont model performance Byron@Thames PEPF [%] PEV [%] CORR [-] RBIAS [%] RRMSE [%] RPWRMSE [%] Calibration 10.667 8.827 0.946-6.443 44.826 40.707 Verification 13.096 7.607 0.939-7.615 46.488 42.245 Snow accumulation and melt adequately reproduced Good simulation of low flow periods No systematic bias in the seasonal model Performance improved with increasing basin area and spatial detail Underestimated streamflow volumes 21
Model sensitivity Local sensitivity analysis Absolute sensitivity coefficient Factor perturbation method Goodness-of of-fit fit measures used as sensitivity functions 22
Model sensitivity Middle Thames River @ Thamesford
70 Event model sensitivity 60 50 40 Qdif [m 3 s -1 ] 30 20 10 Lr 0 Bi Li Td -10-30 % parameter decrease Tc St Rc -20 03-Jul-00 05-Jul-00 07-Jul-00 09-Jul-00 11-Jul-00 13-Jul-00 15-Jul-00 17-Jul-00 24
Event model sensitivity Flood magnitude Clark s s storage coefficient, time of concentration and the loss parameters Peak volume loss parameters and the Clark s s storage coefficient 25
10 9 Tc St Bs Br Cs Ss If 8 Us Ts Sp Gs Gp Gc 7 PEPF [%] 6 5 4 Cont model sensitivity 3 2 1 0-20 -15-10 -5 0 5 10 15 20 PARAMETER CHANGE [%] 26
Cont model sensitivity Flood magnitude Clark s s storage coefficient and the parameters describing soil physical properties (infiltration rate and soil layer storage) Peak volume & low flows SMA groundwater layer parameters Overall goodness-of of-fit fit SMA groundwater layer parameters and baseflow parameters 27
Discussion 28