Spatial planning dataset and QGIS project for Marxan-based risk-informed conservation prioritisation in the Massaciuccoli Lake basin wetland (Tuscany, Italy)

This geographical collection presents the QGIS project and associated geospatial datasets used for Marxan-based spatial conservation prioritisation within the risk-informed spatial-planning domain of the Massaciuccoli Lake basin wetland, Tuscany, Italy. The dataset supports the identification of priority areas for conservation, management, and restoration within areas affected by the risk of forest and grassland habitat degradation and associated biodiversity loss. The spatial planning analysis integrates planning units, anthropogenic cost variables, conservation features, ecosystem-risk hotspots, Marxan input files and outputs, percentile-based planning scenarios, selection frequency maps, best solutions, and final priority classes. The QGIS project is structured in the WGS84 EPSG:4326 coordinate reference system and is organised into the following main groups of layers: Basin limit – including the basin boundary, Ramsar areas, and regional parks and reserves used as cartographic and spatial reference layers. Spatial planning analysis – including the 7,909 planning units within the risk-informed spatial-planning domain, the dataset of planning units with conservation features, anthropogenic cost, and ecosystem-risk information, the five percentile-based planning scenarios, scenario-specific best solutions and selection frequencies, and the final Marxan Ensemble products, including the cumulative best solution, priority classes, and cumulative selection frequency. Anthropogenic cost – including the spatial representation of the five anthropogenic cost components and the resulting total anthropogenic cost used to guide Marxan prioritisation towards solutions associated with lower anthropogenic pressures and territorial constraints. Risk – Predictive analysis – including the medium- and high-risk cells identified using the Multi K-Means and Variational Autoencoder (MKM-VAE) approach and the resulting ecosystem-risk hotspot surface generated through Kernel Density Estimation. These layers represent areas associated with the risk of forest and grassland habitat degradation and related biodiversity loss and were used to delineate the risk-informed spatial-planning domain. Original variables – including the environmental, ecological, and anthropogenic variables used as input conservation features and anthropogenic cost components for the spatial planning workflow. The Marxan model was calibrated using the 75th-percentile planning scenario as the reference conservation-target configuration. Calibration sequentially addressed Feature Penalty Factors (FPFs), the Boundary Length Modifier (BLM), and the number of simulated-annealing iterations. The final configuration used a BLM of 2250 and 10 million iterations. The spatial prioritisation analyses were performed using Marxan v. 2.43, available from the official Marxan website: https://marxansolutions.org/download/marxan-software-version-2-43/. Five percentile-based planning scenarios were evaluated using the 95th, 90th, 85th, 80th, and 75th percentiles. The percentile thresholds were used to derive cumulative conservation targets from the distribution of conservation features within the risk-informed spatial-planning domain. Higher percentile thresholds defined more selective and restrictive target configurations, whereas lower thresholds defined more inclusive and precautionary configurations. Each scenario was independently run 100 times. The main scenario-specific outputs were the best solution, corresponding to the planning-unit set with the minimum Marxan score across runs, and the selection frequency, indicating how often each planning unit was selected across the 100 runs. The five planning scenarios were subsequently combined into a Marxan Ensemble. The cumulative best solution classifies the 7,909 planning units into six relative priority classes, from Class 0 (Lowest), representing planning units never included in a scenario-specific best solution, to Class 5 (Very high), representing planning units included in all five best solutions. The cumulative selection frequency sums planning-unit selection frequencies across all five scenarios and ranges from 0 to 500, providing a complementary measure of selection recurrence and stability. This updated version of the dataset also includes the Marxan input and output files, the results generated during model calibration and execution of the five final planning scenarios, and the ensemble-prioritisation outputs. The source code and R workflow scripts used for data preprocessing, conservation-feature preparation, anthropogenic-cost construction, percentile-based target definition, and Marxan scenario preparation are available on GitHub at: https://github.com/cybprojects65/MassaciuccoliBayesianSpatialPlanning The dataset is intended to support reproducibility, spatial interpretation, and reuse of the Marxan-based prioritisation workflow for wetland conservation planning, restoration targeting, and ecosystem risk-informed environmental management.