Multivariate digital soil mapping to support soil quality index mapping in southern France - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2021

Multivariate digital soil mapping to support soil quality index mapping in southern France

Résumé

Pedometricians have spent a lot of effort on mapping soil types and basic soil properties. However, end-users typically need a more elaborate soil quality index (SQI) for land management. Soil quality indices are typically derived from multiple individual soil properties, by evaluating whether specific criteria are met. If this is based on individually mapped soil properties then an important problem is that cross-correlations between soil properties are ignored. This makes it impossible to quantify the uncertainties associated with the mapped indices. Our objective was to map a soil quality index over a 12 125 km2 study region located along the French Mediterranean coast to help urban planners preserve soils of highest quality. The index considered the ability of soils to fulfill four functions: 1) production of a physical habitat for plant growth; 2) production of a chemical habitat for plant growth; 3) retention and transfer of water and pollutants, and 4) carbon sequestration, under five land use scenarios: 1) annual crop; 2) perennial crop; 3) pastures; 4) forest; and 5) shrubland. Each soil function fulfillment for a given scenario was represented by a categorical map defined from a set of conditions involving basic soil properties (CEC, organic carbon, clay, silt, sand, pH, soil depth, and coarse fragments), and was represented by a 0/1 value. The final soil quality index was the sum of these values. A regression co-kriging model was developed that, first, mapped separately the basic soil properties from legacy soil data and spatial soil covariates using a Random Forest algorithm and, then, interpolated the residuals using cokriging and the linear model of coregionalization (LMC). Both correlations between different soil properties and between the same soil property for different depth layers were accounted for. The mapping uncertainties of soil properties were propagated by calculating the soil quality index over 300 stochastic simulations of soil properties derived from the LMCs. The final soil quality index and its associated uncertainty were estimated respectively by the mean and standard deviation across the 300 simulations. All resulting maps were at 25 m spatial resolution. For validation we used a cross-validation approach repeated 20 times. Although the final map was pedologically meaningful, its performance in terms of amount of variance explained (AVE) was low. The simulations were able to reproduce the width of the observed distribution, although the shapes of the distributions differed considerably. However, we envisage some ways for improvement, such as changing soil property thresholds that are hard to predict, creating probability maps instead of the mean from simulations, and changing the prediction support from point to area.
Fichier non déposé

Dates et versions

hal-04201803 , version 1 (11-09-2023)

Identifiants

  • HAL Id : hal-04201803 , version 1

Citer

Marcos Angelini, Philippe Lagacherie, Gerard B. M. Heuvelink, Eva Rabot, Maritxu Guiresse. Multivariate digital soil mapping to support soil quality index mapping in southern France. Pedometrics Webinar, International Union of Soil Sciences, Jun 2021, Online, France. ⟨hal-04201803⟩
23 Consultations
0 Téléchargements

Partager

Gmail Facebook X LinkedIn More