Habitat probability maps for vulnerable marine ecosystems (VMEs) on the Tasmanian Seamounts

This record describes habitat probability maps for vulnerable marine ecosystems (VMEs) on the Tasmanian Seamounts produced by Berry et al. (2025).\n\nHabitat probability maps were produced from two random forest models utilising multibeam sonar derived rasters, acquired as part of RV Investigator voyage IN2020_V10, and annotated video data acquired as part of RV Investigator voyage IN2018_V06. Additionally, fishing impact information was used as a model input, which was sourced from Williams et al. (2020). Input raster data is at a grid resolution of 15m. Continuous predictor variables used in modelling include backscatter, bathymetry, Fine BPI, LS Factor, Plan Curvature, Profile Curvature, Relative Slope Position, Ruggedness. Categorical predictor variables used in modelling include Fishing Impacts and Seamount. The response variable utilised a VME threshold developed by Williams et al. (2020). Model outputs include habitat probability maps for Model 1 and Model 2, and corresponding uncertainty maps (co-efficient of variation).\n\nContained in this record are:\n•\tPredictor variables used in models\n•\tResponse variable used in models\n•\tModel outputs (including habitat probability and uncertainty)\n\nFor more details, please refer to the article:\nBerry, C., Lucieer, V., Williams, A., Althaus, F., Whittaker, J. (2025) Assessing the importance of acoustic backscatter and bottom fishing impact as predictor variables for mapping coral reefs on the Tasmanian seamounts. Frontiers in Remote Sensing (under review).\n