[A-AS08] Multiple scale structure and their interactions in Asian monsoon system

[AAS08-P05] Gridded Precipitation Datasets Comparison against Observation Dataset on Flood Inundation

*Guek Leang Hak¹、Shigenobu Tanaka¹、Kenji Tanaka¹ (1.Disaster Prevention Research Institute, Kyoto University)

キーワード：Flood, Precipitation, RRI model, Stung Sen river basin, Tonle Sap lake

Flooding is the most common and frequent disaster in Cambodia, especially in Stung Sen river basin, which has caused a great loss to human lives and massive destruction to social, infrastructural, environmental, and economic systems. Changing rainfall pattern has been induced by climate change leading to an increase the flood events. Over the years, water resources in the Tonle Sap Lake of Cambodia have been altered by human behaviors such as rapid population growth, urbanization, deforestation, agricultural expansion, and hydropower demand contributing to climate variations. To observe the flood characteristics in Stung Sen river basin, the biggest tributary in Tonle Sap Lake, various precipitation datasets are selected to conduct besides the observation datasets. Precipitation is one of the predominant components in the hydrological cycle and in the Rainfall-Runoff (RRI) Model which allows to access flood inundation information. RRI model is a hydrologic distributed model that is developed to simulate rainfall runoff and flood inundation concurrently. It is best to represent the flow change and flood evaluation considering subsurface and vertical infiltration flow. The observed rainfall data in the target area, which are dominantly acquired in downstream of the basin and no rain gauges available in upstream, are limited to interpret the flood characteristics. Therefore, multiple gridded precipitation datasets, namely APHRODITE, GSMaP, GPCC, and TRMM from different resolutions are used to examine the RRI model performance for river flow and flood inundation (depth and extension) and to compare which precipitation is best for this study. Among the gridded precipitation datasets, TRMM shows a better performance for predicting discharge and flood inundation with statistical indicators of NSE = 0.74 and R² = 0.76 while observation obtains NSE = 0.48 and R² = 0.55. The spatial inundation maps are indicated the flooded depth and extension of each precipitation product. The peak inundated depths vary from 3 meters to more than 5 meters considering the rainfall distribution. The results revealed that the inundation depth from rain gauges and TRMM displayed an extreme flood at the city center more than 5 meters. TRMM is the most compatible for this study out of the other four rainfall datasets. Overall, it is pointed out how the availability of precipitation information is crucially important to interpret the result of flood inundation for the case of Stung Sen river basin. More rain gauges are encouraged to install, particularly in upstream of the basin.