A flexible and lightweight deep learning weather forecasting model

Abstract Numerical weather prediction is an established weather forecasting technique in which equations describing wind, temperature, pressure and humidity are solved using the current atmospheric state as input. This study examines deep learning to forecast weather given historical data from two London-based locations. Two distinct Bi-LSTM recurrent neural network models were developed in the TensorFlow deep learning framework and trained to make predictions in the next 24 and 72 h, given the past 120 h. The first trained neural network predicted temperature at Kew Gardens with a forecast accuracy of $$\pm$$ ± 2 $${}^{\circ }$$ ∘ C in 73% of instances in a whole unseen year, and a root mean squared errors of 1.45 $${}^{\circ }$$ ∘ C. The second network predicted 72-h air temperature and relative humidity at Heathrow with root mean squared errors 2.26 $${}^{\circ }$$ ∘ C and 14% respectively and 80% of the temperature predictions were within $$\pm$$ ± 3 $${}^{\circ }$$ ∘ C while 80% of relative humidity predictions were within $$\pm$$ ± 20%. Both networks were trained with five years of historical data, with cloud training times of over a minute (24-h network) and three minutes (72-h).