Evaluation of TRMM satellite-based rainfall data in southern Haihe RiverBasin and suitability for SWAT model

Abstract：Accurate estimation of regional precipitation plays an important role in hydrologic process evaluation and waterresources management. Southern Haihe River Basin is the region with the highest degree of water resources exploitation andutilization, however, excessive exploitation of surface water and groundwater causes a series of ecological and environmentalproblems, which leads to a serious threat to water security. In this study, the accuracy of 3B42V7 estimation of precipitation inSouthern Haihe River Basin was evaluated on different spatial and temporal scales, and its applicability to hydrological modelSWAT was verified. The daily rainfall data from 28 meteorological stations (2007-2016) and 101 rain gauges (2010-2016)were used to evaluate the accuracy on TRMM 3B42V7. Correlation coefficient, relative bias ratio, mean error and root meansquare error were used to quantitatively evaluate the rainfall accuracy of 3B42V7. Moreover, determinate coefficient andNash-Sutcliffe coefficient of efficiency were used to quantitatively evaluate SWAT simulation results. Two scenarios were setup to drive the SWAT model. In scenario I, the daily rainfall data (2010-2014) from rain gauges and 3B42V7 grid rainfall(2015-2016) were utilized to drive the model. In scenario II, the daily rainfall data (2010-2016) from meteorological stationswere used to drive the model. The results showed 3B42V7 had strong estimation ability in monthly estimation of precipitationwith the root mean square error less than 15 mm and average monthly precipitation error less than 8.5 mm. However, it waspoor in daily precipitation estimation with the correlation coefficient less than 0.6. The number of rainfall stations with therelative bias ratio between - 20% and 20% accounted for 81% and 79% during the summer season and growing period ofmaize, which indicated that 3B42V7 performed better during the wet seasons. In addition, 3B42V7 could well capture therainfall intensity at all levels, however, zero/light rain were underestimated in four zones. In the yearly estimation, the relativebias ratio values in mountainous area of the Daqinghe watershed, plain of the Daqinghe watershed, mountainous area of theZiyahe watershed and plain of the Ziyahe watershed were 2.64%, 9.59%, 7.72%, 20.32%, respectively. It means theoverestimation in plain and mountainous areas, especially in plain areas. Moreover, 3B42V7 well captured the temporal andspatial distribution of extreme precipitation in this study area. The simulated discharges of SWAT driven by data from raingauge and TRMM 3B42V7 were in good agreement with the observed ones. During the validation period, the determinatecoefficient was between 0.56 and 0.96 and the Nash-Sutcliffe coefficient of efficiency was between - 11.09 and 0.94. TheTRMM 3B42V7 provides the possibility to expand the time and space scale of hydrological simulation and can provide datasupport for water resource management and ecological security research.