Heliyon (Dec 2023)
Developing an efficient climate forecasting model for the spatiotemporal climate dynamics estimation and the prediction that fits the variable topography feature of the upper Blue Nile basin, Ethiopia
Abstract
The spatiotemporal climate estimation and prediction are challenging and demanding for variable topography feature areas. Spatially, the upper Blue Nile basin (UBNB) is unsatisfactory owing to complex topographical features and the lack of accurate climate prediction and estimation models. Yet accurate information and reliable seasonal climate dynamics estimation and forecasting are essential in the region for controlling reservoir operation and flooding prevention. However, there is a lack of reports regarding the accuracy and predictability of the climate in the UBNB region. Therefore, this article aims to improve the Artificial Neural Network (ANN) model by adopting the Impulse Response Function (IRF) and comparing it to the Regional Climate Model (RCM) and European Centre of Medium-range Weather Forecast (ECMWF) models for spatiotemporal climate dynamics estimation and forecasting, which are reliant on the UBNB topography features. Different atmospheric parameter data are investigated from reanalysis models. A fast Fourier transform is applied to remove the redundancy of the data and avoid the computational cost. The RCM and ECMWF models are used to test the performance of ANN model prediction skills. The IRF model was applied to enhance the ANN model's climate prediction performance. A 12-month spatial variation of precipitation is analyzed. The ANN model showed a satisfactory prediction performance, better than the RCM and ECMWF models by 20 %. After increasing the ANN model performance by IRF, the prediction errors are reduced by 10.2 % for precipitation and by 7.9 % for temperature. Based on the model results, the temperature has increased over the past 40 years and is expected to continue for the coming three decades (30 years). In contrast, precipitation over the past 40 years has decreased, and a slight increment will be expected in the next eight years, from 2024 to 2029. Therefore, this model should be practiced across Ethiopia and the Globe for accurate prediction of climate patterns. Hence, clear awareness should be created for the local community by providing a scientific remedy for future climate conditions to reduce production risk.
