Long-Short Term Memory Technique for Monthly Rainfall Prediction in Thale Sap Songkhla River Basin, Thailand

Rainfall is a primary factor for agricultural production, especially in rainfed region. Its accurate prediction therefore vital planning and managing farmers’ plantations. plays an important role the symmetry of water cycle, many hydrological models use rainfall as one their components. This paper aimed to investigate applicability six machine learning (ML) techniques (i.e., M5 model tree: (M5), random forest: (RF), support vector regression with polynomial (SVR-poly) RBF kernels (SVR- RBF), multilayer perceptron (MLP), long-short-term memory (LSTM) predicting multiple-month ahead monthly rainfall. The experiment was set up two weather gauged stations located Thale Sap Songkhla basin. development carried out by (1) selecting input variables, (2) tuning hyperparameters, (3) investigating influence climate variables on prediction, (4) multi-step-ahead prediction. Four statistical indicators including correlation coefficient (r), mean absolute error (MAE), root square (RMSE), overall index (OI) were used assess model’s effectiveness. results revealed that large-scale particularly sea surface temperature, significant tropical For projections basin whole, LSTM provided highest performance both stations. developed predictive rain acceptable performance: r (0.74), MAE (86.31 mm), RMSE (129.11 OI (0.70) 1 month ahead, (0.72), (91.39 (133.66 (0.68) 2 months (0.70), (94.17 (137.22 (0.66) 3 ahead.