Estimation of Summer Rainfall over an Arid Area using AMSR-E Measurements: A Case Study in Xinjiang, China

Rainfall estimate in arid region using passive microwave remote sensing techniques has been a complex issue for some time. The main reason for this difficulty is that the high and variable emissivity of land surfaces greatly aggravates the complexity of the signatures from the rain cloud. The Xinjiang area, located in the northwest of China, holds all the typical characteristics of arid climate. A rainfall algorithm has been developed for this region by using the Advanced Microwave Scanning Radiometer for Earth Observing System (AMSR-E) measurements. The algorithm attempts to use all 12 channels on the AMSR-E instrument and a two-step method calibrated over 11 days of hourly rain-gauge data. First, Stepwise Discriminant Analysis (SDA) used to optimally estimate rain pixels based on all 12 channels, although only three channels were found to be necessary. Next, a rain predicator scattering index was used to estimate rain rates. A linear relationship between the rain rates and the scattering index above the threshold of 3.0K was constructed with a simple approximately linear function. The estimated rain rates were compared with the rain-gauge data used to calibrate the method, and a good relationship was found with a root-mean-square error of 2.1mm/h. The numerical calculations and comparisons show that the algorithm works well in the Xinjiang area