Assessment of different rainfall products in flood simulations

Thesis Type: Doctorate

Institution Of The Thesis: Orta Doğu Teknik Üniversitesi, Faculty of Engineering, Department of Civil Engineering, Turkey

Approval Date: 2017




Floods happening due to heavy rainfall are one of the most widespread natural hazards. To predict such events, accurate rainfall products and well calibrated hydrologic models are essential especially in urban settlements for life savings. With the objective of assessing the rain detection potential of rainfall data products, several hourly rainfall datasets are used as forcing inputs in two hydrologic models. Physic based distributed model, WRF-Hydro, and conceptual based lumped model, HEC HMS, are used to simulate the three catastrophic flood events those occurred in Terme Basin in Samsun. For the calibration of both models, gauge data belonging to 22nd November 2014 flood event are used. Furthermore, stream network density effect in rainfall-runoff modeling is investigated in WRF-Hydro model. In model evaluations, two different flood events with different rainfall datasets are used. The datasets contain weather radar data and satellite rainfall estimates from Hydro-Estimator (HE) as nowcasting products; Weather Research and Forecasting Model (WRF) precipitation data as a forecasting product and gauge-based data. Among these datasets bias correction is applied to the weather radar data by using Kalman Filtering and their use in flood modeling is also evaluated in the simulations. Results show that all products have different limitations and potentials depending on the rainfall type. Among them, the HE product generally indicates poor performance in the simulations in this basin. Whereas, gauge data located in close proximity to the study area is good at representing the flood peak occurrence time but has a weakness in the flood magnitude estimation. WRF precipitation data are superior in detecting the rain with some time inaccuracy but as a forecasted product it can be useful as an early warning system to take initial precautions. Bias corrected radar data using the gauging stations in close proximity to the studied one has an affirmative effect on results especially in frontal rainfall type. Results of the models show that both models are generally close to each other in representing hydrograph shape and peak time. The average value of correlation (r) and root mean square error (RMSE) for all events and rainfall products indicate that WRF Hydro (0.61 for r, 62.6 m3/s for RMSE) showed a slightly better success compared to the HEC HMS (0.59 for r, 67.6 m3/s for RMSE). However, one of the flood event that has mainly convective origin makes the difference between the models. In this event, WRF-Hydro model presents the physical-based model’s ability in showing hydrograph peak discharge and time to peak accurately. The overall results indicate that the use of well calibrated hydrologic model with rainfall data that compound of calibrated radar, WRF precipitation forecast and observations in ungauged or poorly gauged areas can help to take necessary precautions against flooding and provide benefit in saving life and property.