Deep Learning for Monitoring the Ecuadorian Paramo

In this study, various types of multivariate artificial neural networks have been used to predict and model temperature in the Ecuadorian Paramo, considering climatic variables such as humidity, precipitation, and wind speed. The employed architectures included LSTM, Bidirectional LSTM, Attention-Based LSTM, GRU, CCNN, and LSTM-CNN, which enable capturing complex relationships, long-term dependencies, and focusing on relevant parts of the data. The main objective is to achieve more accurate predictions and understand the interactions between climatic factors and temperature, as well as their influence on the Paramo. The results highlight that LSTM and GRU networks exhibit the best performance, with an RMSE of 0.63 and 0.62, respectively.