Estimasi Curah Hujan Bulanan Berbasis Machine Learning dalam Mendukung Ketahanan Pangan di Sumatera Utara

Abstract

The monthly pattern of rainfalls in North Sumatra region is affected by various factors of global climate events, including El-Nino Southern Oscillation (ENSO), Indian Ocean Dipole (IOD) dan Madden Julian Oscillation (MJO). The occurrence of these events has caused different monthly rainfall patterns in every area, thus; causing significant impacts to the rainfall frequencies. This study aims to produce a monthly rainfall models in addition to determine the climate types and cropping pattern to support farming preservation in North Sumatra. The methods used were by comparing the observational results of rainfalls towards results yielded by Machine Learning (ML), such as Artificial Nerves Network (ANN) technique, and Mean Absolute Error (MAE), Pearson’s correlation, and Oldeman climate-types classification. The research results indicated a strong correlation value and varied MAE (Mean Absolute Error) values. The global climate phenomena factor that most influenced monthly rainfall in North Sumatra is the model with a combination of IOD (Indian Ocean Dipole) and SOI (Southern Oscillation Index). The climate type in the North Sumatra region falls into category A1 and C1, indicating favorable planting patterns for rice and crops. However, in the western slopes and mountains, the climate type is E1, indicating that the area is too dry for rice cultivation and only suitable for growing other secondary-or-third crops.