Spatial Predictor Selection Dataset for Next-Day Minimum Temperature Forecasting (ERA5, Western Europe, 2004–2024)

Dataset description This dataset contains the preprocessed database used for the study: "Spatial Predictor Selection for Next-Day Minimum Temperature Forecasting: An Automated Machine Learning Framework Applied Across European Climate Regimes". The database integrates ERA5 reanalysis data and observational temperature records into a unified SQLite structure designed for efficient access and reproducibility. The database is provided in SQLite format (.db file), not MySQL. The database is split into multiple parts and must be reconstructed prior to use (see README or next section for instructions). Reconstruction After downloading all parts, reconstruct the database: Linux / macOS:cat METEO_daily.db.part* > METEO_daily.db Windows (PowerShell):Get-Content METEO_daily.db.part* -Encoding Byte -ReadCount 0 | Set-Content METEO_daily.db -Encoding Byte For a quick start, see the "Code repository" section below. Content The dataset includes: Daily ERA5 variables over Western Europe (0.25° × 0.25° grid resolution) Observed daily minimum temperature (Tmin) from NOAA NCEI stations Time coverage: 2004–2024 Spatial coverage: Western Europe (approx. 36°N–63°N, 15°W–18°E) Each record is indexed by date and spatial location, allowing direct retrieval of predictor time series for machine learning applications. Data sources and licensing This dataset is derived from publicly available data sources: ERA5 reanalysis data provided by the Copernicus Climate Change Service (C3S) Observational data from the NOAA National Centers for Environmental Information (NCEI) The original data remain subject to their respective licenses and terms of use. Users of this dataset must comply with these original licensing conditions, including attribution requirements. This dataset represents a processed and integrated version of these data, and is distributed under the Creative Commons Attribution 4.0 International License (CC BY 4.0) for the added value components (data processing, structuring, and aggregation). File structure Due to its size (~12.5 GB), the database is split into multiple parts. All parts must be downloaded and reconstructed before use. Data sources ERA5 reanalysis: Copernicus Climate Data Store Observational data: NOAA NCEI Global Summary of the Day Users must comply with the licensing terms of these original data providers. Database structure The database is provided in SQLite format. The schema is documented in the accompanying file: database_schema.sql This file contains the full Data Definition Language (DDL) used to create the database, including table structures and indexing strategy. Code repository The code used to generate and exploit this dataset is available at:https://github.com/grezac/edilia The repository provides:- scripts for data processing and model training- a Docker-based execution environment- a one-command reproducible pipeline (run.py) Usage notes The dataset is intended to be used in conjunction with the code repository associated with the study. It allows direct execution of experiments without requiring data reconstruction. Rebuilding the database from raw sources is possible using the provided scripts, but is computationally intensive. Reproducibility This dataset represents the exact version used in the published study. It is designed to ensure full reproducibility of the results when used with the corresponding code and Docker environment. Keywords ERA5, reanalysis, temperature, machine learning, feature selection, geosciences, SQLite, reproducibility