We leveraged the Land Change Modeler within the Clark Labs’ TerrSet 2020 Geospatial Monitoring and Modeling Software package to produce vulnerability assessments and land use change products for several land use land cover types (TerrSet 2020, https://clarklabs.org/terrset/). TerrSet is an integrated geospatial software system for monitoring and modeling the earth system for sustainable development.
We examined land use change and persistence across the southwestern US during an 18-yr window using 2001 and 2019 NLCD products reclassified into 9 categories. We investigated 15 potential variables to evaluate the top drivers responsible for the observed change. The targeted land use land cover categories included development, forest, shrubland, and grassland. We investigated several sub-models and developed transition potentials based on statistical relationships to produce probability maps representing vulnerability to change. A Multi-Layer Perceptron (MLP) approach was used to aggregate all transition potentials to produce final transition probability maps (i.e., soft predictions), which are used to forecast land cover change (i.e., hard prediction) at pre-defined time steps (i.e., here we selected one future mid-century timestep, 2050).
Products are broken down into two main categories 1) Projected Land Use Land Cover Changes, a static map of forecasted changes based on the time envelope used (i.e., 2001-2019) and the associated drivers of change; and 2) Vulnerability Assessments, probability maps that represent the vulnerability of any given land use or land cover to transition to a different land use land cover in the future. While the hard prediction can be used to examine overall quantity of change, the results of the soft prediction are preferred for evaluating habitat and other environmental applications because it provides a comprehensive assessment of change potential across the greater landscape.
All data released in this product suite are provisional and subject to revision. Accuracy of final products vary considerably depending on the area evaluated for each land use land cover category. In general, validation procedures show lower accuracy when evaluating land use change across large, regional areas, thus error can be minimized and accuracy increased when evaluating land use change at smaller scales (e.g., county to state level). Data users are cautioned to consider carefully the provisional nature of the information. Information concerning the accuracy and appropriate uses of these data may be obtained from the FWS.
