Enhancing weather analysis with data interpolation and time series forecasting

Abstract

Weather analysis plays a crucial role in various domains, from agriculture to urban planning. However, accurate and localized predictions can be challenging in urban environments with limited weather station coverage. In this work, we propose a comprehensive workflow that combines dot rain coverage, interpolation, and machine learning models to address this issue. By establishing a network of weather stations strategically distributed across the city and utiliz- ing their weather variables as input for the interpolation techniques, we generate interpolated data for the entire city grid. This approach enables us to fill the gaps in weather station coverage and provide accurate predictions for locations without direct measurements. Subsequently, machine learning models are trained on the interpolated data to forecast var- ious weather variables. We conducted extensive experiments and hyperparameter optimization to achieve accurate predictions with low evaluation loss. Furthermore, our models demonstrate transferability across different weather stations within the city, enabling localized predictions in previously unmonitored areas. The results highlight the effectiveness of our project in im- proving weather analysis capabilities in urban settings. This work opens avenues for further research in applying these techniques to different regions with diverse weather conditions, ultimately enhancing decision-making processes in various sectors reliant on accurate weather predictions.