Multimodel AI Prediction Network (MAPNet) applied to temperature forecasts

In this repository are all the data/scripts necessary to reproduce the results of the article 'Multimodel AI Prediction Network (MAPNet) applied to temperature forecasts.' The files in NetCDF format and nomenclature TYPE_FORECAST_202201_202312.nc are the input data for the neural network, where TYPE (TMAX, TMIN) and FORECAST (F24, F48, F72). These files contain all the predictions from the atmospheric models and the 'observations.' This input data was divided into training, validation, and test periods. The files in NetCDF format and nomenclature CNN_TYPE_FORECAST.nc are the output data of the neural network, where TYPE (TMAX, TMIN) and FORECAST (F24, F48, F72). These files contain the TMAX and TMIN forecasts for the test period. The files in NetCDF format and nomenclature MODEL_TYPE_FORECAST.nc are the forecasts from the atmospheric models (BRAMS, ETA, WRF, SMEC) and the observations (SAMET) for the same test period. This data was used to compare the predictions of each model with the predictions of the neural network. The file CCN_Unet.py is a Python program containing the configurations of the convolutional neural network with a U-Net architecture used in the study. The files in ASCII format (.py, .gs) are scripts used for visualizing the results and generating the graphs. The other files are shapefiles and masks used for generating figures.