Daily Rainfall Prediction using Generalized Linear Bivariate Model - A Case Study

The present study focuses on the simulation of daily rainfall series based on atmospheric predictors and historical data using a bivariate Generalized Linear Model. Temperature and precipitation data along with a set of covariates were made use of in generating the simulations. Probability of occurrence of rainfall was predicted using logistic regression models. The amount of rainfall on a rainy day was modelled using a gamma distribution. The covariates in the model comprise of different categories such as site effects for spatial variation, year effects allowing long term trends, month effects for seasonality, day effects with temporal auto correlation and atmospheric predictors. Rainfall series were generated for both future and past periods at multi sites simultaneously using atmospheric predictors. The model developed was applied in a typical catchment in the state of Kerala in India. The model simulations were acceptable on the basis of the performance evaluated using statistical analysis. The model can be used as a weather generator to simulate the daily rainfall series for both past and future periods. (C) 2016 The Authors. Published by Elsevier Ltd.