[Proposed by Gerard Heuvelink]
We have spent a lot of effort on modelling and mapping soil type and (basic) soil properties. This was time well spent because soil type and soil properties are useful for many purposes, but many end-users require maps of soil functions. As yet we have not paid enough attention to establishing rules and models that derive soil functions from soil properties and other land characteristics. We must work on this with much greater effort and make sure that we also quantify the associated uncertainties. Among others, it requires that pedometricians help define soil functions or measures of soil functions in an unambiguous way.
Mapping soil functions is effectively the way forward. This means moving from monovariate DSM to multivariate DSM, which carries some methodological issues that should be thoroughly addressed in the near future. In the following, I cite three of them, among others: how to find the best inference trajectory between “mapping first the soil properties then combine them” or the reverse (and all the intermediates)? How to account for correlations between soil properties in the DSM models for not creating pedo-chimera that could lead to mistaken estimations of soil functions? How to propagate the uncertainties in mapping soil properties toward the function map? Some of these questions have been already worked out, including recently, but this does not end the story.