Precipitation estimation over Mexico applying PERSIANN system and gauge data

Abstract

Developing countries like Mexico mostly rely on rain-gauge networks to obtain the much-needed rainfall data to manage their water resources. Recently, satellite-based rainfall estimation can cover remote areas of the world, such as oceans, mountains and desert, where rain gauges are unable to be installed. The purpose of this study is to develop a technique of combining Mexican rain-gauge data with satellite-based rainfall estimation to provide better rainfall information for Mexico. The Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks (PERSIANN) system is an advanced satellite-based rainfall estimation tool recently developed at the University of Arizona. In order to allow the PERSIANN system to use the gauge data, the point, daily gauge rainfall must be rescaled into hourly, grid-area rainfall. A scheme based on cloud infrared images to distribute rain rates is developed to disaggregate daily, area-averaged gauge data, and the produced high-resolution rainfall is used to train the PERSIANN system. The effectiveness of the disaggregation scheme is evaluated in southwest U.S. where the high-resolution hourly rainfall from NCEP is available for validation. Then the same strategy is applied to Mexico using the Mexican gauge data. The results show that the disaggregation scheme provides reliable high-resolution data for training PERSIANN, improving rainfall estimates over places (such as Mexico) with a lack of high-resolution ground based rainfall data.