An Objective Method to Create Continuous Digital Soil Organic Carbon Map. Issue 1 (7th January 2014)

• Record Type: Journal Article
• Authors: Frank, Stephanie L.; Owens, Phillip R.
• Is Part Of: Soil horizons. Volume 55:Issue 1(2014)
• Page Start: 1
• Page End: 8
• Publication Date: 2014-01-07
• Subjects: Soil science -- Periodicals; Soil surveys -- Periodicals; Soil science; Soil surveys; Periodicals; Electronic journals; 631.4
• DOI: 10.2136/sh13-10-0024
• Languages: English
• ISSNs: 2163-2812

Abstract

Abstract : Many soil organic carbon (SOC) estimates are represented as a single value for a mapping unit with discrete boundaries rather than a more representative continuous map of the SOC. This study develops a method to more objectively utilize the Soil Land Inference Model (SoLIM) by incorporating statistical analysis of the input variables. The study area is a 14‐digit hydrologic unit code (HUC) watershed in central Indiana that has an area of just over 5260 hectares. It was glaciated until approximately 18, 000 yr ago and is part of a loess mantled, rolling glacial till plain. Soil organic carbon determination was done according to geomorphic landscape position. Values for SOC were obtained by a dry combustion method for a subset of samples with the remaining values being obtained via spectral analysis. Significant terrain attributes derived from a digital elevation model (DEM) were statistically identified using stepwise multiple linear regression. The k ‐means clustering algorithm was used to help partition each terrain attribute into meaningful clusters relative to total SOC, which were then combined to produce meaningful soil classes for each landscape position. The median value of the terrain attribute for each soil class was considered the centroid value, and the corresponding 50% membership values were determined by iterative optimization using different quantiles of the input variables. A stepwise regression model using 10%|90% quantile intervals ultimately produced the best model for the area. This study found that statistical methods can be employed to aid the knowledge based mapping process.