This project created maps of the lands potential to transition from Forest, Grassland, and Shrubland to Agriculture or Settlement. We used ESA CCI Land Cover Data, reclassified into 7 categories, and TerrSett’s Land Change Modeler tool. This project used 9 explanatory variables, listed in the workflow below, and change analysis to determine the transition potential of each land cover type using the TerrSett MLP model.
Using change analysis, we can see how categories are losing area and gaining area. Forests had the highest net loss out of all categories, while settlement had the highest net gain. Shrubland had both very high loss and very high gain of area. Grassland had very little transition between 2008 and 2015. The gain loss data is additionally run through a markov chain to determine change allocation.
The Transition Potential Sub-Models, shown above, were built using MLP. We tested each variable for significance in the model, in order to determine which variables to use in MLP modeling for each of the sub-models. The skill and accuracy of each model is shown, as well as the variables that were used to model transitions. Forest to Agriculture was the hardest sub-model to create, mainly due to the fact that it is the largest transition, and therefore the most difficult. While most of the transition from forest to agriculture is in south-east Mexico, it is also widely spread. Grassland to Agriculture was the transition that had the highest skill score, but is the smallest transition. Transition Potentials for each sub-model are shown below.
Overall Transition Potential and Modeled 2030 Land Cover
The main area with high transition potential is the Yucatan Penisula, which is concerning due to the fact that it is an area with very high biodiversity and holds vast biosphere reserves for this reason. Meanwhile Northern Mexico has very low transition potential. This map can be used to target areas that may be under threat of conversion in order to create conservation areas.