Table 1.

Data flow for raster data analysis with Reconstructability Analysis (OCCAM). Extraction was conducted using ArcPro 3.3 GIS software, with further processing in Python.

Processing Step	Results
Extract rows that include the DV (center cell Z5) and space–time VNN neighborhood at times 1 through 4 at every location in the study area	~11 million rows of 20 variables generated
Eliminate rows that have all uniform values	~6.7 million rows retained
Select rows that have Evergreen Forest (NLCD code 42) anywhere in the row	~4 million rows retained
Stratify data so that ½ are EFO present ½ are EFO absent, shuffle, and split into train/test sets	500K rows in each train/test set, replicated 3 times
Add headers for OCCAM input file, reclassifying 15 to 5 classes with rebinning code in the variable block, also recoding Z5 to 1 for code 42, and 0 for all other values	Classes collapsed to: Water, Developed, and Agriculture, Shrubs, Grasses, Mixed/Deciduous Forest, Evergreen Forest
Upload data to OCCAM and run Search	Report generated
Select best model from Search and run Fit on it	Report generated
Extract model predictions from Fit output and analyze with R-Studio 2024, and Excel 2021	Final results