Understanding long-term seasonal or annual or inter-annual rainfall variability and its relationship with large-scale atmospheric variables (LSAVs) is important for water resource planning and management. In this study, rainfall forecasting models using the artificial neural network technique were developed to forecast seasonal rainfall in May–June–July (MJJ), August–September–October (ASO), November–December–January (NDJ), and February–March–April (FMA) and to determine the effects of climate change on seasonal rainfall. LSAVs, temperature, pressure, wind, precipitable water, and relative humidity at different lead times were identified as the significant predictors. To determine the impacts of climate change the predictors obtained from two general circulation models, CSIRO Mk3.6 and MPI-ESM-MR, were used with quantile mapping bias correction. Our results show that the models with the best performance for FMA and MJJ seasons are able to forecast rainfall one month in advance for these seasons and the best models for ASO and NDJ seasons are able do so two months in advance. Under the RCP4.5 scenario, a decreasing trend of MJJ rainfall and an increasing trend of ASO rainfall can be observed from 2011 to 2040. For the dry season, while NDJ rainfall decreases, FMA rainfall increases for the same period of time.
Incorporating large-scale atmospheric variables in long-term seasonal rainfall forecasting using artificial neural networks: an application to the Ping Basin in Thailand
M. S. Babel, T. A. J. G. Sirisena, N. Singhrattna; Incorporating large-scale atmospheric variables in long-term seasonal rainfall forecasting using artificial neural networks: an application to the Ping Basin in Thailand. Hydrology Research 1 June 2017; 48 (3): 867–882. doi: https://doi.org/10.2166/nh.2016.212
Download citation file: