Ignitia Dataset : A Multimodal Machine Learning and Simulation Framework for Spatiotemporal Risk Assessment

Wildfires in mountainous regions such as Uttarakhand, India, are driven by complex interactions among topographic, climatic, vegetation, and anthropogenic factors, necessitating high-resolution, region-specific prediction datasets. This work presents the Ignitia Wildfire Susceptibility Prediction Dataset, a curated, analysis-ready geospatial dataset designed to support machine-learning–based wildfire ignition modeling in complex Himalayan terrain. The dataset was constructed using multisource remote-sensing and environmental variables processed within a cloud-based geospatial framework and harmonized at 30 m spatial resolution. It comprises topographic features (elevation, slope, aspect), meteorological variables (humidity, wind), vegetation and land-surface indicators (NDVI, land-cover class, land surface temperature), and spatio-temporal attributes (latitude, longitude, date, and month), along with a binary wildfire occurrence label. All features are provided in cleaned and scaled form to facilitate direct use in supervised learning workflows. This dataset underpins the Ignitia wildfire susceptibility modeling pipeline, where multiple supervised machine-learning models are benchmarked under k-fold cross-validation to quantify feature influence, improve interpretability, and identify key ignition drivers in mountainous ecosystems. Although the broader Ignitia framework integrates probabilistic cellular automata–based fire-spread simulation and interactive web deployment, the present release focuses exclusively on the prediction-ready dataset, enabling reproducible experimentation and comparative evaluation by the research community. The primary contribution of this dataset lies in providing a localized, high-resolution, multimodal feature space for wildfire ignition prediction in mountainous regions, supporting future advances in data-driven fire-risk assessment, explainable modeling, and region-specific wildfire early-warning systems.