Date

2016-12-12

Author

Akyürek, Sevda Zuhal

Metadata

Show full item record

Item Usage Stats

178
views

0
downloads

Cite This

Floods caused by convective storms in mountainous regions are sensitive to the temporal and spatial variability of rainfall. Space-time estimates of rainfall from weather radar, satellites and numerical weather prediction models can be a remedy to represent pattern of the rainfall with some inaccuracy. However, there is a strong need for evaluation of the performance and limitations of these estimates in hydrology. This study aims to provide a comparison of gauge, radar, satellite (Hydro-Estimator (HE)) and numerical weather prediciton model (Weather Research and Forecasting (WRF)) precipitation datasets during an extreme flood event (22.11.2014) lasting 40 hours in Samsun-Turkey. For this study, hourly rainfall data from 13 ground observation stations were used in the analyses. This event having a peak discharge of 541 m3/sec created flooding at the downstream of Terme Basin. Comparisons were performed in two parts. First the analysis were performed in areal and point based manner. Secondly, a semi-distributed hydrological model was used to assess the accuracy of the rainfall datasets to simulate river flows for the flood event. Kalman Filtering was used in the bias correction of radar rainfall data compared to gauge measurements. Radar, gauge, corrected radar, HE and WRF rainfall data were used as model inputs. Generally, the HE product underestimates the cumulative rainfall amounts in all stations, radar data underestimates the results in cumulative sense but keeps the consistency in the results. On the other hand, almost all stations in WRF mean statistics computations have better results compared to the HE product but worse than the radar dataset. Results in point comparisons indicated that, trend of the rainfall is captured by the radar rainfall estimation well but radar underestimates the maximum values. According to cumulative gauge value, radar underestimated the cumulative rainfall amount by % 32. Contrary to other datasets, the bias of WRF is positive due to the overestimation of rainfall forecasts. It was seen that radar-based flow predictions demonstrated good potential for successful hydrological modeling. Moreover, flow predictions obtained from bias corrected radar rainfall values produced an increase in the peak flows compared to the ones obtained from radar data itself.

URI

www.agu.com
https://hdl.handle.net/11511/72925

Collections

Department of Civil Engineering, Conference / Seminar