Spatial prediction of soil particle size distribution in arid agricultural lands in central Iran.

Autor: Zolfaghari, Ali Asghar, Toularoud, Ali Ashraf Soltani, Baghi, Fatemeh, Mirzaee, Salman
Zdroj: Arabian Journal of Geosciences; Oct2022, Vol. 15 Issue 19, p1-13, 13p
Abstrakt: Spatial distribution mapping of particle size distribution (PSD) curve has been rarely investigated. Therefore, the objective of current study was to evaluate Jaky's one parameter model for spatial prediction of complete PSD curve on agricultural lands in arid areas of the Semnan province, Iran. Several 100 soil samples were collected from a depth of 0-30 cm. Remotely sensed (RS) data were considered as environmental predictor covariates. For spatial prediction of the PSD, initially, Jaky one-parameter model was fitted to the measured PSD data and the Jaky P parameter was determined in each soil sample. Two supervised learning approaches, namely random forest (RF) and support vector regression (SVR), were used for digital mapping of Jaky model's parameter (e.g. P), and soil fractions including sand, silt, and clay content. Results indicated that the accuracy of Jaky model for describing PSD according to R2 values ranged from 0.72 to 0.99 with a median of 0.954. Besides, there were a strong positive correlation between Jaky's P parameter and clay (R2=0.72) and silt (R2=0.56), while the relation between sand and P parameter was negative (R2=0.91). Comparisons of SVR and RF showed that the SVR had a better performance than RF for spatial prediction of Jaky's P parameter and sand, silt and clay content. The analysis of correlation coefficient between P, clay, silt, and sand maps indicated that there was a strong positive correlation between P parameter and clay (r=0.83) as well as P and silt content (r=0.65) maps, while the correlation between P and sand maps (r=-0.83) was negative. Consequently, the current research indicated that the Landsat 8 OLI imagery is potentially a valuable environmental covariate for prediction of both complete PSD curve and soil particle-size fractions (e.g., sand, silt, and clay content). Also results showed that the combination of PSD model and digital soil mapping (DSM) techniques can be used to quantify the spatial distribution of the complete PSD curve. [ABSTRACT FROM AUTHOR]
Databáze: Complementary Index