I think you would have to use vegetation 'health' as a proxy for soil fertility (while obviously veg. growth is also correlated with temp. and precip., and soil moisture with topographic position), within a single field that presumably received the same amount of precip and sun, and where the topography is already known, variations in veg. health that are not obviously topographically-caused, must be at least somewhat related to soil fertility). There are several different RGBNearIR vegetation indices (including, but not only, NDVI) you can generate from your multispectral drone images. There would be lots of uncertainty for your client to weigh.........