Dataset for the models trained and tested in the paper "LSTM and Transformer-based framework for bias 2 correction of ERA5 hourly wind speeds"

This repository contains the complete dataset used in the preprint. The current files included in the repository are: A metadata file on the meteorological stations from Environment and Climate Change Canada (ECCC) (ECCC_Stations.csv) The selected stations from the ECCC database (StationsSelected.csv) Canada shapefile for plotting (Canada_shpfile) Wind speed data from ECCC (ECCC_PARQUET) ERA5 meteorological data interpolated at the ECCC stations (ERA5_PARQUET) HDF5 dataset ready for model training and testing (H5dataset) The ID of the training and test stations used in the study (TRAIN_TEST) The tested parameters of the gradient boosting model (GBParams.json) The results of the permutation feature importance using gradient boosting (GBFeaturesImportance.json) The selected static covariates and the best parameters found for the gradient boosting model using random search (GBSelectFeatures.json) The training stations are divided into six folds for training and validation (kfolds.json) The scaling factor between ERA5 mean wind speeds and ECCC mean wind speeds (OBS_ScalingFactor.csv) The value of the static covariates (topography, surface roughness length, etc.) at the station location (staticFeatures.csv) After downloading, unzip the file and place it in the data folder of the corresponding repository.