Remote sensing and AI applications in monsoon rainfall forecasting

Abstract

Monsoon rainfall forecasting is also important in monsoon dependent areas as agricultural planning, water management, and reducing disaster risks require accurate information. The proposed research explores the use of both types of multi-source remote sensory data and artificial intelligence (AI) methods in promotion of accuracy and reliability of overall, monsoon rainfall prediction. Satellite-retrieved data, such as precipitation, sea surface temperature, soil moisture, and atmospheric markers, were inputted and existed to 4 AI models: Artificial Neural Network (ANN), Support Vector Regression (SVR), Random Forest (RF), and Long Short-Term Memory (LSTM). Modelling The models were tested with normal operating performance measures throughout several monsoon seasons. The obtained experimental results indicate that the LSTM model remains the most successful in the combination of overall performance, Root Mean Square Error (RMSE) = 5.8 mm/day, Mean Absolute Error (MAE) = 4.2 mm/day, and coefficient of determination (R 2 = 0.85) at 1-day lead time forecasting. In comparison with RF (RMSE = 6.6 mm/day) and ANN (RMSE = 7.8 mm/day), LSTM had a higher ability to capture the temporal dependence and intra seasonal variability. The strength of deep learning-based methods was also supported by comparing the performance of the two prediction technology in terms of forecast lead times and monsoon sounds. The results emphasize the validity of remote sensing data along with AI algorithms as a computationally effective and scalable substitute of conventional forecasting systems, especially in the data-scarce areas.