Performance evaluation of a multi-site weather generator coupling maximum entropy resampling for estimating the probability distribution of annual maximum daily rainfall in the Loess Plateau.

Weather generators can serve as an important tool for estimating the probability distributions of hydrometeorological variables to determine extreme rainfall quantiles. It is really of great significance for flood disaster management in data-deficient areas. The annual maximum daily rainfall (MDR) dataset was expanded in this study using a multi-site weather generator model coupled with maximum entropy resampling (MWG-MER), resulting in a sizable resampled dataset of extreme variables. The MDR records from 14 weather stations in the Jing River Basin were chosen to confirm the MWG-MER model's applicability on the Chinese Loess Plateau. There, the best-fitting probability distributions were discovered using the L-moments approach, and the resampled data was utilized to determine the spatial pattern of the probability distribution of MDR. The findings demonstrate that the MWG-MER model implemented a limited extrapolation and offered a larger resampling dataset that only used the first principal component to preserve the primary statistical properties of the observed MDR data. The resampling data had good fitting performance compared to the measured data at all sites and revealed different best-fit probability distributions from the measured data at some sites, implying an improvement in the MWG-MER model's distribution fitting accuracy. With the aid of the resampling data, the spatial pattern of the MDR distribution in the Jing River Basin was discovered, and more logical partitions were provided. This information is crucial for scientifically determining the MDR probability distribution in data-deficient areas and for directing the management of flood disasters in the Loess Plateau and other similar regions. [ABSTRACT FROM AUTHOR]