Fix covariates

[jonrhodes]

Resolve issues with covariates, including: (1) check we have the right covariates against the DAGs, (2) assess if there are alternatives to the covariates chosen (e.g., are there continuous covariates to replace discrete covariates), (3) check for any errors in the covariate data, (4) reclassify discrete covariates where necessary to reduce the number of classes, and (4) check for collinearity among covariates and remove variables where necessary. Write code to address these issues.

[yfchung]

Issue 3: Check for any errors in the covariate data

1. There are quite a number of properties (SUs) that have NA continuous covariate. We could potentially reduce SUs with NAs by reducing NA cells in continuous raster layers. E.g. SoilNit <- terra::focal(SoilNit, w=5, fun="mean", na.policy="only", fillvalue=NA, expand=FALSE, silent=TRUE) # potentially specifying window (w) with a gaussian matrix for better approximation w <- focalMat(rast(ncols=180, nrows=180, xmin=0) , c(1,5), "Gauss")

2. There are some property with very small area (even one with negative area). Suggest to set a lower limit to property size (e.g. > 10m²).