been ideal for this kind of work; however, it was necessary to strike a balance between number of data points and number of soil parameters. Within the NSIS dataset, not all samples have been completely analysed, and so it was necessary to take only those points that had sufficient parameterisation to make the analysis possible. Selection was achieved by maximising the product of the number of soil parameters and the
number of available sample points with those parameters, which in this case gave a value of 127 × 44 = 5588. If we had wanted to use a dataset with 500 points, for example, we would have been restricted to 9 parameters giving a product of 4500.