Comparing three approaches of spatial disaggregation of legacy soil maps based on 1 DSMART algorithm

Yosra Ellili (1) , Brendan Philip Malone (2) , Didier Michot (1) , Budiman Minasny (3) , Sébastien Vincent (1) , Christian Walter (1) , Blandine Lemercier (1)

1 SAS - Sol Agro et hydrosystème Spatialisation
2 CSIRO - Commonwealth Scientific and Industrial Research Organisation [Canberra]
3 The University of Sydney

Enhancing the spatial resolution of pedological information is a great challenge in the field of Digital Soil 34 Mapping (DSM). Several techniques have emerged to disaggregate conventional soil maps initially 35 available at coarser spatial resolution than required for solving environmental and agricultural issues. At the 36 regional level, polygon maps represent soil cover as a tessellation of polygons defining Soil Map Units 37 (SMU), where each SMU can include one or several Soil Type Units (STU) with given proportions derived 38 from expert knowledge. Such polygon maps can be disaggregated at finer spatial resolution by machine 39 learning algorithms using the Disaggregation and Harmonisation of Soil Map Units Through Resampled 40 Classification Trees (DSMART) algorithm. This study aimed to compare three approaches of spatial 41 disaggregation of legacy soil maps based on DSMART decision trees to test the hypothesis that the 42 disaggregation of soil landscape distribution rules may improve the accuracy of the resulting soil maps. 43 Overall, two modified DSMART algorithm (DSMART with extra soil profiles, DSMART with soil 44 landscape relationships) and the original DSMART algorithm were tested. The quality of disaggregated soil 45 maps at 50 m resolution was assessed over a large study area (6,775 km2) using an external validation based 46 on independent 135 soil profiles selected by probability sampling, 755 legacy soil profiles and existing 47 detailed 1:25,000 soil maps. Pairwise comparisons were also performed, using Shannon entropy measure, 48 to spatially locate differences between disaggregated maps. The main results show that adding soil landscape 49 relationships in the disaggregation process enhances the performance of prediction of soil type distribution. 50 Considering the three most probable STU and using 135 independent soil profiles, the overall accuracy 51 measures are: 19.8 % for DSMART with expert rules against 18.1 % for the original DSMART and 16.9 % 52 for DSMART with extra soil profiles. These measures were almost twofold higher when validated using 53 3x3 windows. They achieved 28.5% for DSMART with soil landscape relationships, 25.3% and 21% for 54 original DSMART and DSMART with extra soil observations, respectively. In general, adding soil 55 landscape relationships as well as extra soil observations constraints the model to predict a specific STU 56 that can occur in specific environmental conditions. Thus, including global soil landscape expert rules in 57 the DSMART algorithm is crucial to obtain consistent soil maps with clear internal disaggregation of SMU 58 across the landscape.

Domaines: Science des sols

Yosra Ellili, Brendan Philip Malone, Didier Michot, Budiman Minasny, Sébastien Vincent, et al.. Comparing three approaches of spatial disaggregation of legacy soil maps based on 1 DSMART algorithm. 2019. ⟨hal-02362155⟩