Introducing a merged precipitation satellite model using satellite precipitation products, land surface temperature, and precipitable water vapor

Satellite precipitation have a lot of uncertainty and the spatial resolution is still quite coarse. Hence, the purpose of this study is to introduce a new approach based on multivariate linear modelling and Artificial Neural Networks (ANN) for the first time using four multisource satellite precipitation products and some other satellite data. Four different satellite precipitation products with different algorithms, along with Precipitable Water Vapor (PWV) and Land Surface Temperature (LST) data on a daily scale were used (to benefit interactions of atmospheric content). Five linear merge models, including five ANN models were introduced and calibrated using harmony search algorithm. The performance of the models was evaluated over Iran using observational rainfall data of 2014 to 2020. The findings of this study highlighted that the model based on ANN, which merged four rainfall products together with LST and PWV has outperformed individual satellite products and other presented models.