Climate change impact studies have been significantly applied in the past few decades, where understanding and implementing the available global datasets at regional and even station levels is inevitable. In the highly complex hydrological phenomena happening globally, the various predictors and their complex relationships with local predictand (precipitation) play a crucial role. Towards this step, numerous predictor selection techniques such as maximum relevance minimum redundancy (MRMR), least absolute shrinkage and selection operator (LASSO), step-wise regression (SWR), and whale optimization technique (WOA) are adapted. Twenty-two predictors measured during a period of 30 years (from 1985 to 2014) from 14 stations distributed throughout the lower Krishna River basin are used as the data set in this study. The selected predictors from each method are tested on simulated historical data from 1985 to 2014 by CMIP6 MPI-ESM1-2-HR global climate model at every station using artificial neural networks (ANN). Applying the best set of predictors and choosing them according to the site, can boost downscaling of precipitation by the ANN model, based on the evaluation results of screening methods. WOA outperformed in finding the optimal predictors over other methods, SWR, MRMR, and LASSO. The most effective predictors found for WOA for all sites are relative humidity at 850 and 1000 hPa, geo-potential height at 500 hPa, u-wind at 850 hPa, v-wind at 10 m and 500 hPa, minimum temperature, mean sea level pressure and mean zonal gravity wave stress.