Investigating WetSpa model application in the Distributed Model Intercomparison Project (DMIP2) using NEXRAD radar rainfall data
Download as PPT, PDF1 like279 views
Ali Safari conducted PhD research from 2005-2009 investigating the WetSpa hydrological model. The research aimed to apply WetSpa to basins in the US using radar rainfall data, calibrate WetSpa parameters using PEST, improve high flow and peak predictions, and modify WetSpa to better predict flows in nested sub-basins. Key findings were that calibration improved WetSpa performance significantly but calibration of parent basins did not guarantee good performance in sub-basins, and WetSpa needed improvement in simulating high flows and peaks which was achieved by modifying surface runoff calculation.
Investigating WetSpa model application in the Distributed Model Intercomparison Project (DMIP2) using NEXRAD radar rainfall data
- 1. Brainstorming Short research plan by Alireza Safari asafarim@vub.ac.be 24th June 2010 1
- 2. Ali Safari • Home country: Iran • Promotor: Prof. Dr. De Smedt • Scholarship: University of Tehran (Iran) • Ph.D. timing: 2005-2009 • Ph.D aim: • Investigating WetSpa model application in the Distributed Model Intercomparison Project (DMIP2) using NEXRAD radar rainfall data, • WetSpa model calibration and predictive analysis using PEST parameter estimation program in conjunction with Box-Cox transformation and ARIMA error model, • WetSpa model improvement in predicting high flows and major peaks, • Improvement of the WetSpa model predictions for nested subbasins using nonlinear Boussinesq equation. • Study area: • DMIP2 experiment basins (Oklahoma, USA) • Keywords: • WetSpa, River flow simulation, Model calibration, ARIMA error model, Box-Cox transformation, PEST program, Time-variant surface runoff, DMIP2, Baseflow recession coefficient, Boussinesq equation Pag.2
- 3. Ali Safari WetSpa model application to the DMIP2 basins and the nested sub-basins done in two ways (explicit calibration at the interior points was not allowed): • a. using default parameter sets (e.g. uncalibrated model run) • b. using PEST optimized parameter sets (e.g. calibrated model run), Results (based on a developed statistical measure to assess model performance): o Calibration of the model improves the model performance significantly. o Calibration of the model for the parent basin is no guarantee for good performance for the nested subbasins. Calibration and predictive analysis of the WetSpa model predictions are done using PEST in conjunction with Box- Cox transformation (to stabilize error variance) and ARIMA error model (to remove autocorrelation in the error series) to lead the modeler to a better understanding of parameter sensitivity issue, and consequently a more reliable inference about the model parameters. WetSpa model predictive analysis reveals that the model is not capable of simulating high flows particularly those that are leading to flooding accurately (WetSpa predictions are not within the margin of uncertainty). Improvement on surface runoff calculation of the model created a new version of the WetSpa model, which is capable of reproducing high flows and major peaks accurately. Using a simple linear groundwater equation in WetSpa to calculate groundwater flow is not enough when the calibrated model parameter set at the watershed outlet is applied to the interior subwatersheds. This is found to be due to applying a sensitive parameter, base flow recession coefficient that is obtained for the calibration outlet for the nested subbasins. This problem of the model is solved by calibrating a much less sensitive parameter in the Boussinesq equation to calculate base flow recession coefficient (Kg). Hence, instead of adjusting a high sensitive parameter (Kg), we need to adjust a much less sensitive parameter, aquifer transmissivity, in the modified version of the WetSpa model. Pag.3