Related Experiment Video
Updated: Nov 25, 2025
Use of Principal Components for Scaling Up Topographic Models to Map Soil Redistribution and Soil Organic Carbon
Published on: October 16, 2018
Source apportionment of soil heavy metals using robust spatial receptor model with categorical land-use types and
Mingkai Qu1, Jian Chen1, Biao Huang1
1Key Laboratory of Soil Environment and Pollution Remediation, Institute of Soil Science, Chinese Academy of Sciences, Nanjing, 210008, China.
Abstract:
High-density samples are usually a prerequisite for obtaining high-precision source apportionment results in large-scale areas. In-situ field portable X-ray fluorescence spectrometry (FPXRF) is a fast and cheap way to increase the sample size of soil heavy metals (HMs). Moreover, categorical land-use types may be closely associated with source contributions. However, the above information has rarely been incorporated into the source apportionment. In this study, robust geographically weighted regression (RGWR) was first used to correct the spatially varying effect of the related soil factors (e.g., soil water and soil organic matter) on in-situ FPXRF in an urban-rural fringe of Wuhan City, China, and the correction accuracy of RGWR was compared with those of the traditionally non-spatial multiple linear regression (MLR) and basic GWR. Then, the effect of land-use types on HM concentrations was partitioned using analysis of variance (ANOVA). Last, based on the robust spatial receptor model (i.e., robust absolute principal component scores/RGWR [RAPCS/RGWR]), this study proposed RAPCS/RGWR with categorical land-use types and RGWR-corrected in-situ FPXRF data (RAPCS/RGWR_LU&FPXRF), and its performance was compared with those of RAPCS/RGWR with none or one kind of auxiliary data. Results showed that (i) the performances of the correction models for in-situ FPXRF data were in the order of RGWR > GWR > MLR, and the relative improvement of RGWR over MLR ranged from 52.6% to 70.71% for each HM; (ii) categorical land-use types significantly affected the concentrations of soil Zn, Cu, As, and Pb; (iii) the highest estimation accuracy for source contributions was obtained by RAPCS/RGWR_LU&FPXRF, and the lowest estimation accuracy was obtained by basic RAPCS/RGWR. It is concluded that land-use types and RGWR-corrected in-situ FPXRF data are closely associated with the source contribution, and RAPCS/RGWR_LU&FPXRF is a cost-effective source apportionment method for soil HMs in large-scale areas.
