Analysis on temporal variability of monsoon rainfall using ARIMA Model and ANN-LSTM

Abstract In this study, machine learning algorithms employed for predicting the weather conditions of the upcoming day with a particular focus on forecasting rain, which serves as a crucial indicator. To identify the most reliable rain-related attributes, most of the researchers utilize line charts, matrix graphs, and scatterplot graphs for visualization and analysis. The investigation reveals several attribute pairs demonstrating a significant degree of similarity and correlation. In the modeling phase, researchers employ elementary models such as decision trees, and regression analysis to assess their fundamental prediction capabilities. The resulting accuracy rate found favourable sometimes on statistics platform. Notably, during the visualization stage, a subtle pattern emerges: samples experiencing rainfall today exhibit a slightly elevated probability of rain the following day. Temporal variability in rainfall refers to the fluctuations and changes in precipitation patterns over time. It encompasses the temporal distribution, intensity, and duration of rainfall events, reflecting the dynamic nature of weather systems. Understanding temporal variability is crucial for effective water resource management, agriculture, and climate impact assessment. Consequently, our attempt to analyze the influence of historical weather on forecasting using ARIMA and ANN-LSTM. The rainfall predictions generated by these models are deemed accurate based on statistical analysis and algorithm simulation.