Meteorological Drought Prediction Based Long ShortTerm Memory Algorithm

Rainfall distribution variation across regions and periods has influenced local ecosystems and human life. Lack of precipitation in a certain time may result in a temporal dry condition known as a meteorological drought while exceeded rainfall may lead to flooding. Meterological drought, in general, is a complex natural phenomenon driven by climatic aspects and extremes (such as El Nino or La Nina). Thus, it is difficult to make an accurate forecasting rainfall pattern through out a period of time for effective water resource management and risk mitigation. The Long Short Term Memory (LSTM) algorithm is a type of Recurrent Neural Network known for its ability to capture long-range dependencies and temporal patterns. It offers promising capabilities for predicting meteorological drought events. This study applies LSTM to historical meteorological data assessment to develop a forecasting model capable of accurately predicting drought occurrences over various time horizons. The methodology involves the preprocessing of weather records with four main indicators: rainfall, temperature, wind speed and humidity, caculating the Standardized Precipitation Index (SPI) as inputs for LSTM model training, and validation using historical drought records and. The preliminary results demonstrate the effectiveness of the proposed approach in accurately forecasting rainfall, thus providing valuable insights for further studying on meteorological drought based the SPI.