THE ADJUSTMENT OF C-BAND RADAR-RAINFALL ESTIMATES AND ITS IMPACT ON STREAM FLOW SIMULATION UNCERTAINTY: A CASE STUDY FOR JAKARTA URBAN RIVER BASIN, INDONESIA

摘要

The utilization of weather radar has shown the considerable potential for improving the distributed rainfall estimation and prediction of flood. Ciliwung River in Indonesia (476 km~2) has notably contributed to the recurrent flooding in Jakarta. C-band weather radar is available for providing rain information in Greater Jakarta. Yet, as single polarimetric radar, the verification of its observation with point rain gauge measurements is indispensable. In this study, the strategies of radar-rainfall calibration and adjustment for extracting radar estimates that is consistent with the ongoing development of real-time flood prediction in Ciliwung River is designed. Furthermore, the impact of the adjusted radar-rainfall estimates on the rainfall-runoff simulation is investigated. The radar reflectivity is converted to rainfall intensity by Z-R algorithms. Several storm events are selected to evaluate the appropriate Z-R relation by minimizing the mean square error between rain gauges and radar-rainfall in offline scheme. Applying the predefined Z-R relation, the remaining biases in the radar-rainfall estimates are corrected by using bias adjustment factor method calculated as the ratio between the accumulated rain gauge rainfall and accumulated radar-rainfall. The rainfall estimates without adjustment, by Marshall-Palmer Z-R algorithm, calibrated radar-rainfall, and adjusted radar-rainfall are introduced to physically distributed hydrological model in hourly basis. The simulated runoff is evaluated at Katulampa, Depok, and MT. Haryono points representing upstream, middle stream, and downstream outlets. Along with these three inputs, the uncertainties of hydrological model parameter are assessed. Through the calibration process, the algorithm with the lowest error for Ciliwung River basin is obtained. The analysis reveals that the different radar verification methods account for the variation in the performance of simulated flood. The results suggest the importance of correction of rainfall estimated from C-band non-polarimetric radar observation. The proposed framework on radar-rainfall utilization for real-time flood prediction system in Ciliwung River is also discussed.