Rainfall forecast in northeast of thailand using modified k-nearest neighbor

Abstract

Since damage from natural disasters have increased due to anomalous global climate, scientists and engineers are interested in studying incorporation of the occurence of natural disasters. Thailand faces with flood in the wet season and drought in the dry season every year. The Northeast of Thailand is a region where found damages from disasters especially. This study developed a statistical model for forecasting rainfall in the Chi River Basin using large-scale atmospheric variables (LAV) as the independent variables to the modified k-nearest neighbor model. The significant LAV were identified over both Indian and Pacific Oceans. The model performance was evaluated using box plot of 3-month rainfall to present how well the model can capture the historical data and likelihood skill score (LLH). From both model evaluation, approximately 62% of historical rainfall data was captured forecasting model. LLH of rainfall ensembles in the Chi River Basin are quite good and better LLH can be found post 2000, especially June-August and July-September rainfall.

Keywords

• Large-scale atmospheric variables
• Modified k-nearest neighbor