Regional Analysis of Influential Meteorological Factors in Daily Precipitation Forecasting Using Machine Learning

Severe precipitation conditions create a substantial risk to lives of individuals. For the management of reservoirs, food production and city development, accurate daily precipitation prediction is essential. This study explores the utilization of different Machine Learning (ML) models for forecasting precipitation based on 21 years of meteorological data from Narela, Rohini, Anand Vihar and Okhla of Delhi, India. The dataset contains meteorological features such as WS2M (wind speed at 2 meters), T2M (temperature at 2 meters), T2MDEW (dew point temperature at 2 meters), T2M_MAX, T2M_MIN, RH2M (relative humidity at 2 meters), PS (surface pressure), WS10M (wind speeds at 10m), WS10M_MIN, WS10M_MAX and precipitation, with precipitation as the target variable. The performance of several Machine Learning (ML) models, including Long Short-Term Memory (LSTM), Gated Recurrent Units (GRU), Convolutional Neural Networks (CNN), and hybrid model of Convolutional Neural Network and Long Short Term Memory (CNN-LSTM), is compared. Out of all of these models, the CNN-LSTM model performs the best, showing the lowest RMSE and highest R2 across all 4 locations. The results demonstrate the ability of Machine Learning (ML) models, particularly CNN-LSTM, to significantly enhance the precision of precipitation predictions. These findings emphasize the potential of hybrid Machine Learning (ML) models in providing accurate daily precipitation forecasts, especially in urban areas with complex weather patterns and hence underscore the significance of advanced Machine Learning (ML) methods in improving the accuracy of precipitation forecasts, which are vital for applications such as water resource management and hydrological modelling in urban areas like Delhi. The current research relies on the ability to alter city climatology and planning, enhance making decisions using accurate precipitation predictions, and help to mitigate the effects of disasters.