Deep Learning to implement a Statistical Weather Forecast for the Andean City of Quito

A novel approach leveraging deep learning and low-cost single-board computers achieves accurate temperature forecasts for Quito, Ecuador, with errors as low as 0.74°C. Traditional weather forecasting methods struggle in Quito due to the city's unique microclimates and sensitivity to the Intertropical Convergence Zone. The study validated eight neural networks using the Walk Forward Technique, finding that incorporating temperature data from all available stations yields the most precise predictions. This cost-effective solution offers a promising alternative to expensive conventional weather station expansion, addressing critical forecasting challenges in Andean urban environments.

Canar, R. L., Fontaine, A., Morillo, P. L., & El Yacoubi, S. (2020). Deep Learning to implement a Statistical Weather Forecast for the Andean City of Quito. 2020 Ieee Andescon, 1–6. https://doi.org/10.1109/andescon50619.2020.9272106