Spatial Variables and Land Use Change Models: A Study on Conditioning Patterns of Natural Vegetation Suppression and Persistence

Land-use change models are formulated by identifying patterns of change and persistence. In modeling software, this step is usually performed by characterizing samples based on spatial variables. Despite the importance of this stage, the evaluation of the change and persistence patterns is often neglected by the scientific community. Thus, this study evaluated the conditioning factors of natural vegetation suppression and persistence in three study areas in different Brazilian biomes. The patterns were investigated for five different time periods, 1995 to 2000 (representing training) and 2000 to 2005, 2000 to 2010, 2000 to 2015 and 2000 to 2020 (representing extrapolation). The spatial variables used to identify the patterns were formulated to represent the environmental context of the training period (1995 to 2000). The method used to analyze the data was Violin Plot graphs. Among the modeling challenges investigated, the following stand out: 1) The ability of variables to explain changes; 2) The variation of change patterns across different time periods; and 3) The variation of change patterns across different study areas and within the same study area. Among the main findings, it was shown that: 1) within the set of analyzed variables, some had a greater ability to differentiate between vegetation suppression and persistence.; 2) the farther the extrapolation was from the training period, the lower the ability of the variables to differentiate the patterns; and 3) Vegetation suppression and persistence in different study areas were described by the variables in distinct ways. As possible recommendations, it is highlighted that modelers analyze patterns of change and persistence using statistical techniques.