Goals:
- Evaluate the differences between models of 5 different resolutions (1m, 3m, 1/3, 1, and 3 arc-second) on both held-out and ground verified data [AUC-ROC, AUC-PR, Sens. Spec. TSS].
- Compare Results from 150 ground truth points along the axis of suitability (0-100%), and species occupancy (0-100% percentile of cells). Analyze as two independent logistic regressions, and together as two multiple logistic regressions.
- Model occupancy of suitable habitat patches as a function of distance from an occupied patch (both euclidean and least-cost), patch (e.g. inner area), and class (e.g. isolation) characteristics. [multiple logistic regression].
- Compare the results of a Mature plant presence versus a Juvenile plant presence model, are the important variables similar? Are patch sizes similar?
- Simulate the effects of sample size on model performance using the data set generated after both rounds of field sampling.
- Simulate the effects of occurrence record geolocation accuracy on model performance using the data set generated after both rounds of field sampling.