Analysis of time series in meteorological stations for the prediction of precipitation in the city of Manizales, Colombia

In this article, different time series are analyzed using the ARIMA, SARIMA, SARIMAX, Prophet, and Neural Prophet models in order to predict precipitation in the city of Manizales, Colombia, using data provided by SIMAC. Additionally, the results obtained by the models for the predictions of the last 7 days show root mean square error (RMSE) and mean absolute error (MAE) values around 19, indicating a good fit of the predicted values against more robust models such as neural networks. On the other hand, the Prophet model achieved an RMSE value of 19.06313 and a MAE of 16.24064, demonstrating lower errors compared to the stochastic models implemented in this work. Furthermore, the predicted values from the Prophet library can be highly useful for the development of best practices in landslide analysis and risk management in the area. Lastly, based on this analysis, an early warning system based on the A25 is developed.