Back to Search
Start Over
Mapping Soil Textural Fractions at Regional Scale Based on Local Morphometric Variables Using a Hybrid Approach (Case Study: Khuzestan Province, Iran).
- Source :
-
Arabian Journal for Science & Engineering (Springer Science & Business Media B.V. ) . Jul2024, Vol. 49 Issue 7, p9797-9807. 11p. - Publication Year :
- 2024
-
Abstract
- Local morphometric variables (LMVs) are frequently found as weaker predictors than other environmental covariates in digital soil mapping. This study tested and evaluated the performance of a hybrid approach combining gradient boosted regression trees (GBRT) and regularized regression (RR) algorithms in predicting soil textural fractions using a set of LMVs in Khuzestan province, Iran. Here five LMVs (slope gradient, slope aspect, horizontal curvature, vertical curvature, and contour geodesic torsion) were derived from a spheroidal equal-angular DEM as original predictors. The results demonstrated that the hybrid approach improved prediction accuracy for sand, clay, and silt contents by an average of 56% compared to the GBRT models. The importance analysis revealed the significant contribution of tree-based variables obtained from decomposing GBRT models in predicting soil textural fractions. This approach could be recommended for digital soil mapping, particularly in situations of limited environmental covariates or geomorphometric techniques that cannot be easily applied. [ABSTRACT FROM AUTHOR]
- Subjects :
- *DIGITAL soil mapping
*SOIL mapping
*REGRESSION trees
*PROVINCES
Subjects
Details
- Language :
- English
- ISSN :
- 2193567X
- Volume :
- 49
- Issue :
- 7
- Database :
- Academic Search Index
- Journal :
- Arabian Journal for Science & Engineering (Springer Science & Business Media B.V. )
- Publication Type :
- Academic Journal
- Accession number :
- 178148730
- Full Text :
- https://doi.org/10.1007/s13369-024-08961-3