In this study, we explore the integration of machine learning algorithms into a decision support system for climate finance, focusing on the impact of rainfall on wineries in Italy. Wineries are particularly vulnerable to climate change, and accurate rainfall forecasting is critical to their success; lack of rain can reduce the quantity and quality of grapes, while flooding can damage vineyards. We identify relevant weather characteristics that cause rainfall and predict quarterly rainfall intensity using machine learning techniques. The dataset was collected from the agrometeorological office of the Piedmont region in Italy to measure the performance of three machine learning techniques (Multivariate Linear Regression, Random Forest, and Neural Network). Mean square error and mean absolute error methods were used to measure the performance of the machine learning models. A comparative analysis between precipitation estimation models based on conventional machine learning algorithms and deep learning architectures with models based on Long Short-Term Memory (LSTM) networks is performed. It shows how the Random Forest algorithm presents the best performances, both in the accuracy and explainability of the predictions. Our study contributes to the climate finance literature by showing how machine learning can support decision-makers in managing climate risks in the food chain, specifically in the wine industry in Italy.
Machine Learning to Forecast Rainfall Intensity
Bruni M. E.;Lazzaroli V.;
2023-01-01
Abstract
In this study, we explore the integration of machine learning algorithms into a decision support system for climate finance, focusing on the impact of rainfall on wineries in Italy. Wineries are particularly vulnerable to climate change, and accurate rainfall forecasting is critical to their success; lack of rain can reduce the quantity and quality of grapes, while flooding can damage vineyards. We identify relevant weather characteristics that cause rainfall and predict quarterly rainfall intensity using machine learning techniques. The dataset was collected from the agrometeorological office of the Piedmont region in Italy to measure the performance of three machine learning techniques (Multivariate Linear Regression, Random Forest, and Neural Network). Mean square error and mean absolute error methods were used to measure the performance of the machine learning models. A comparative analysis between precipitation estimation models based on conventional machine learning algorithms and deep learning architectures with models based on Long Short-Term Memory (LSTM) networks is performed. It shows how the Random Forest algorithm presents the best performances, both in the accuracy and explainability of the predictions. Our study contributes to the climate finance literature by showing how machine learning can support decision-makers in managing climate risks in the food chain, specifically in the wine industry in Italy.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.