This dataset provides a geospatial classification of large parking lots (greater than 900 m²) in the federal state of Hesse, Germany, evaluating their suitability for photovoltaic (PV) installations.

The parking lot geometries were derived from two authoritative and open sources: the Authoritative Topographic-Cartographic Information System (ATKIS) and OpenStreetMap (OSM). Duplicates were removed, and the geometries were trimmed to match the boundaries of Hesse. Parking lots smaller than 900 m² were excluded to align with current policy guidelines and technical feasibility thresholds for PV canopy installations.

Each parking lot was classified into one of two categories employing machine learning. Class 0 indicates parking lots unsuitable for PV, while Class 1 designates those suitable for PV installations. The classification utilized an XGBoost model trained on more than 1,000 manually labeled parking lots using various input features.

This dataset is formatted as a GeoPackage (.gpkg) and contains a layer named "prediction". It is set within the ETRS89 / UTM Zone32N coordinate reference system (EPSG code specified). The dataset includes two key attributes: (a) "id" as a unique identifier for each feature, and (b) the "prediction_class". The latter attribute indicates the area's suitability, with a classification of 0 for unsuitable and 1 for suitable. This information is helpful for various applications, including environmental studies and land utilization strategies.

Besides, a QGIS style layer is given "prediction_parking_pv_hesse.qml". This file is designed to visually distinguish between parking lots classified as suitable and unsuitable for PV canopy installations in the dataset "prediction_parking_pv_hesse.gpkg". The symbology is defined as follows: class 0 (unsuitable): Red fill, and class 1 (suitable): Green fill.

The style enhances the readability for map viewers and supports the quick visual interpretation of suitability categories. It can be directly applied to the prediction layer in QGIS for consistent thematic mapping.