Pedometrics › Forums › 10 PM challenges › Can we better understand proximal soil sensing signals › Reply To: Can we better understand proximal soil sensing signals
[by Philippe Lagacherie]
Soil sensing (not only proximal sensing but also remote sensing) is probably our best chance to produce accurate soil maps in the future. This is why I fully support this challenge. The dream is to use the soil sensing without any expensive local measurements that serve for re-calibrating the soil prediction functions. Using radiative models developed by our colleagues involved in remote sensing research for representing the underlying mechanisms of the signal and of its perturbations should be attempted. Using advanced machine learning algorithm, using big datasets that associate measurements of soil properties, soil sensing signal and assessments of potential perturbation factor is another way.
About the two last Gerard’s questions: My first opinion was that I did not believe that skipping the step “prediction of traditional soil properties” would emerge as a better solution. These good old soil properties have two advantages over soil function assessments: 1) their measurements are well known with clear and harmonized protocols, 2) they are closer to the physic of the signal (for example Vis-SWIR spectral signature of clay can be related to well-known chromophores). But recent discussions with my colleagues Cécile Gomez and Mogen Greve led me to a more balanced opinion. Successful attempts of linking directly soil sensing signal with functional soil properties exist in the literature. This is probably because soil sensing can capture some hidden soil properties that are more closely linked to the soil function than the traditional soil properties that are involved in the pedotransfer function.