The 127 data points included organic soils, brown earths, podzols and gleys from a wide geographical distribution across Scotland. This dataset was used to produce two training datasets, one with colour as input and the non-colour parameters as output, and another with the parameters as inputs and soil colour as output. The Munsell colour codes for each sample were converted into RGB and Lab values using a translation table developed that included all soil colours found in the NSIS dataset (a total of 627 different Munsell colours were given). Where more than one colour was present, i.e. for gleyed soils with mottling, we used the Munsell colours of the soil matrix, rather than of the peds or mottles (the Munsell colour of the soil matrix was determined as standard, with additional ped/mottle colour only given in cases where multiple colours were observed in a horizon).