Analysis of Machine Learning Performance in Spatial Interpolation of Rainfall Data

The spatialization of precipitation data is crucial for studies on climatology, agriculture, and climate change, as well as for urban and environmental planning. Established spatial interpolation methods such as Inverse Distance Weighting are widely used for this purpose. With technological advancements, machine learning techniques have emerged as more efficient and computationally less costly alternatives for performing spatial interpolation calculations. Therefore, this study aims to analyze the performance of machine learning algorithms, including artificial neural networks, comparing them to the Inverse Distance Weighting method. The accuracy of the models was evaluated using the coefficient of determination, root mean square error, and concordance index, to confirm the effectiveness of these methodologies for maximum daily annual precipitation data. After testing different models for data spatialization and conducting a thorough statistical analysis, we conclude that machine learning models outperformed the Inverse Distance Weighting method, yielding greater variability in accumulated Annual Maximum Daily Precipitation values and improved overall results.