Meteorological time series forecasts can help decision-making processes carried out by entities in charge of disaster prevention and early warning generation before the possibility of natural events involving situations which are dangerous for communities. There are a considerable number of methods for these forecasts, ranging from simplistic or Naive methods to those which employ more complex techniques such as those using artificial intelligence. This experimental study worked with a meteorological time series from the station agronomy in the city of Manizales, which provides data on the variables: precipitation, average temperature, sunlight and relative humidity. Forecasts were employed with the Naive approach, with artificial neuronal networks and with neuro-fuzzy networks; also comparing these with a multiple linear regression, with the goal of verifying their precision. The results obtained in this study show firstly that it is possible to refine the models generally used in order to achieve more conclusive results and secondly that they can be extended to other monitoring stations in the region, including new variables, both explanatory and predictive.