[From Philippe Lagacherie]
I am not sure it is a first-order challenge. If we remember the soil surveyor’s practices, the values of soil properties per horizons are already averaged since they are measured on composite soil samples constituted by soil samples taken at different depths within the horizon. Most often, the legacy measured profiles we have provide no more than three of these averaged soil properties. It is a much too poor dataset for expecting any improvement in the representation of the soil vertical variations brought by more sophisticated models than our good old spline function. I will advise investing time on this subject when I will see substantial progresses in inexpensive methods for producing better data on soil vertical variations. However, another challenge related to soil vertical variation is to be cited: we are not good in predicting the soil properties of the deepest soil layers, especially the saprolite (between the bottom of the pedogenic horizons and the bedrock). Yet, these soil properties are required by many users dealing with forest or vineyard or with superficial aquifers. We should improve this in the future.