STATISTICAL APPROACH TO DISCHARGE PREDICTION FOR FLOOD FORECASTS USING TRMM DATA

摘要

Real-time rainfall data are being used for flood forecasting, but they are not often available in many developing countries. Further, rainfall data from sparse gauging stations are usually inadequate to develop representative aeria 1 samples. In this context, developm ent of a suitable technique for application of satellite-d erived rainfall products is useful for flood forecasting.This paper presents the outcome of a study conducted to com pare the Tropical Rainfall Measuring Mission (TRMM) rainfall product (3B 42RT) over Nepal with observed point gauge data and explore the possibility of using it for flood forecasting in Bagm ati river. In this study rainfall data were compared on grid-by-grid basis. Statistical analysis was carried out to find the correlation between the observed data and TRMM gr id data. It was found that in areas with higher number of rain gauges within the grid, TRMM rainfall closely approxim ated the annual average precipitation under the grid dom ain. It was also found that TRMM slightly overestimated the rainfall in relatively less rainfall areas. The TRMM estimated rainfall was also compared with the Automatic Weather Station (AWS) rainfall data and found that TRMM well captured the diurnal variability. W eighted 3-hourly rainfall fields of TRMM of year 2004 were used as independent variables to predict the daily discharge of Bagm ati river at a gauging site using regression techniques. The model predicted the daily peak discharge accurately but erratic fluctuations were obtained during periods of low flows. The output of the study, and the approach in general, could have potential application in operational flood forecasting.