Machine Learning Applied to Weather Forecasting

Weather forecasting has traditionally been done by physical models of the atmosphere, which are unstable to perturbations, and thus are inaccurate for large periods of time. Since machine learning techniques are more robust to perturbations, in this paper we explore their application to weather forecasting to potentially generate more accurate weather forecasts for large periods of time. The scope of this paper was restricted to forecasting the maximum temperature and the minimum temperature for seven days, given weather data for the past two days. A linear regression model and a variation on a functional regression model were used, with the latter able to capture trends in the weather. Both of our models were outperformed by professional weather forecasting services, although the discrepancy between our models and the professional ones diminished rapidly for forecasts of later days, and perhaps for even longer time scales our models could outperform professional ones. The linear regression model outperformed the functional regression model, suggesting that two days were too short for the latter to capture significant weather trends, and perhaps basing our forecasts on weather data for four or five days would allow the functional regression model to outperform the linear regression model.