Prediction of Extreme Precipitation in the Upper Indus Basin Using Time Series Models

Abstract

The Indus River, a transboundary and one of the biggest rivers in the world originates in west ern Tibet, flows northwest through Ladakh, Gilgit-Baltistan, and then south and southeast across Pakistan and empties into the Arabian Sea. Tarbela Reservoir is one of the biggest being con structed on the Indus River and plays a pivotal role for agriculture, water storage, hydropower, and ecosystem stabilization. This study attempts to model and forecast the monthly extreme precipitation events in the Upper Indus Basin (UIB) using time series and machine learning approaches. The observed data about precipitation has been collected from the Pakistan Me teorological Department (PMD) for the duration of 1960-2013 and then the monthly extreme precipitation was calculated by using Excel. Two classes of time series models, i.e., Autoregres sive Integrated Moving Average (ARIMA) whereas LSTM form of Recurrent Neural Network (RNN) form machine learning approach have been used. The performance of these models has been assessed by using various error statistics (e.g., RMSE, MBE, AIC, and BIC) as well as graphically. The findings of this study may help policymakers in water availability, hydropower, infrastructure loss, flooding, and agriculture.