I want to test the effect of elevation on the intensity of burrowing by gopher tortoises in flatwoods [a type of pine savanna characterized by both low (wet) and high (dry) areas]. However, in several of the study areas, there is both a global trend in elevation as well as local topographic relief and both seem to be important in predicting burrowing intensity.
The elevation data were provided as a digital elevation model (DEM) in raster format, and I have figured out how to detrend the raster surface (using a form of trend surface analysis on the x and y coordinates of the cells) and I can use a raster of the residuals from the detrended DEM as a predictor in the point process model to evaluate the effect of local topographic relief on burrowing intensity. My question pertains to the best way to include both the global and local effects of elevation in the point process model. For example, could I potentially use a raster based on the predicted values from the polynomial regression (representing the trend) and a raster of the residuals (representing the local topographic relief) together in the same model to evaluate the global and local effects of elevation?
Is this a valid approach?...or is it problematic because the two variables are not independent (since one is derived from the other)? Is there a better framework to evaluate the global and local effects in one composite model (preferably in the context of point process modeling...here, based on a loglinear model with spatially varying burrowing intensity as the response and the elevation surface(s) as the predictor(s))?