Wildfire Environmental Monitoring Datasets (2020–2025): Processed Data for Portugal and Feature-Engineered Datasets for Montesinho Natural Park

This collection consists of three processed datasets: allfeat_portugal.csv, Montesinho_OriginalFeatures_complete.csv, and montesinho_processed.csv. allfeat_portugal.csv The dataset allfeat_portugal.csv contains “raw” data. It is the result of merging four harmonized geospatial datasets (NDVI, soil moisture, thermal anomalies, and weather conditions) that cover the area of Portugal (continental territory and islands), distributed over a spatial tessellation. The tessellation consists of 1405 tiles, each with side length of approximately 10 km. The timespan is from 01.01.2020 to 30.03.2025. The dataset dimensions are (2,691,980 rows × 20 columns). There are 19 features. The complete list of features can be found in the file list_features_dataset_Portugal.ipynb in the associated GitHub repository (https://github.com/EchoFire/FireFeatures). This dataset was created using the code in Merging_data_sets.ipynb in the GitHub repository. Montesinho_OriginalFeatures_complete.csv The file Montesinho_OriginalFeatures_complete.csv contains features for the 32 tiles corresponding to the area of Montesinho Natural Park. It consists of 61,312 rows and 13 columns. The index corresponds to the tile_date_id. The features in this dataset are: 0 date1 max_T212 tile_id3 avg_NDVI4 soil_moisture_am5 soil_moisture_pm6 tavg7 tmin8 tmax9 prcp10 wdir11 wspd12 pres Missing values were filled following a rule-based procedure, and the resulting dataset is complete, with no missing values.The notebook used to fill missing values is CompletingMissingValuesMontesinho.ipynb in the GitHub repository. montesinho_processed.csv The dataset montesinho_processed.csv contains data for the area of Montesinho Natural Park aggregated spatially by tiles on the same date. It consists of 1,916 rows and 61 columns. The dataset is complete, contains no missing values, and includes engineered features. The temporal resolution is daily. The notebook used to create this dataset is creation_data_set_fires_Montesinho_park.ipynb in the GitHub repository.The file List_features_montesinho_processed.ipynb in the same repository contains the complete list of features. Filling missing values After exploratory data analysis, selected features were retained. The rules used to complete missing entries are: Compute daily averages over the 32 tiles. Compute an overall average to use as fallback. Rule 1: If a daily average exists, fill the missing value with this value. Rule 2: If not, use the average of the previous and next day. Rule 3: Fallback to the overall average. Missing values of avg_NDVI were filled using forward filling with a limit of 15. Feature Engineering Created 7-, 14-, and 30-day lag features. Added temporal and spatial context features (month, week). Created rolling features with windows of 7 and 14 days. Lag and rolling features were computed per tile before aggregating tiles by date.Missing values introduced by lag and rolling operations were filled using backward fill. The dataset for Montesinho Natural Park was then aggregated by date using the following rules: Maximum values for max_T21, avg_NDVI, and tmax Minimum values for soil moisture and tmin Averages for the remaining variables Citation If you use these datasets, please cite this Zenodo record. Intended Use These datasets are intended for wildfire modeling, environmental monitoring, and spatiotemporal analysis at national and regional scales. Reproducibility All datasets were generated using the code available in the associated GitHub repository.