We have generated a predictive model of soil geochemical variation. The model utilises a training data set of soil geochemical variation and rasters of environmental covariates such as radiometric K, U, Th, DOSE, elevation, mean annual rainfall, mean annual temperature and so on. A genetic programming (GP) approach is used to generate an explicit mathematical equation (model) that explains variation in an observational measure (e.g. soil potassium concentration) as a function of the rasters retained by the GP algorithm. Pre-processing includes normalisation and standardisation of all raster values and all observational data. The main reason for normalising and standardising all inputs to the model is that GP works much better with non-dimensional data. The model output is an equation that is used to generate a single raster of the observational measure (soil K conc). Obviously the output is the product of the transformed rasters and as such the units are not able to be used. My question is whether it is possible to back transform the raster model of soil K into its original units (i.e. ppm)? If it is possible how is this done? Thanks in advance.