Premium
Sampling and Data Analysis Optimization for Estimating Soil Organic Carbon Stocks in Agroecosystems
Author(s) -
Sherpa Sonam R.,
Wolfe David W.,
Es Harold M.
Publication year - 2016
Publication title -
soil science society of america journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.836
H-Index - 168
eISSN - 1435-0661
pISSN - 0361-5995
DOI - 10.2136/sssaj2016.04.0113
Subject(s) - kriging , agroecosystem , sampling (signal processing) , environmental science , soil carbon , variogram , simple random sample , topographic wetness index , soil survey , sampling design , statistics , stratified sampling , hydrology (agriculture) , soil science , mathematics , soil water , geography , digital elevation model , computer science , remote sensing , geology , agriculture , population , filter (signal processing) , archaeology , sociology , computer vision , demography , geotechnical engineering
Core Ideas SOC stocks are highly variable at the agroecosystem scale. Optimized sampling and estimation approaches are crucial for low cost SOC assessment. Systematic sampling provided thorough geographic and attribute space coverage. Ordinary kriging outperformed regression kriging, simple mean, and SSURGO approaches. Low‐cost approaches for measuring soil organic C (SOC) stocks are essential for verifying farm management effects on C sequestration in agroecosystems. This study compared sampling and data analysis optimization approaches for estimating SOC stocks of a complex agroecosystem. Soil samples were collected from a 232‐ha area of a working dairy farm with multiple land uses, crop rotations, topographic features, and manure application rates. The SOC stocks were estimated by (i) simple mean, (ii) ordinary kriging, (iii) regression kriging, and (iv) using the USDA Soil Survey Geographic (SSURGO) database. Relationships among sampling schemes, estimation approaches, auxiliary information, and SOC stocks were explored. Slope, elevation, soil type, and land use types displayed a high degree of variation at the study site, yet relationships with SOC stocks were weak or nonsignificant. Systematic sampling provided the best coverage of both geographic and attribute space, allowing reliable variogram estimates and RMSEs that increased little with reduced sample number. Random and stratified random sampling approaches resulted in reduced accuracy. Ordinary kriging had lower RMSEs than either regression kriging or the simple mean, and total SOC stocks estimates fluctuated less when sample sizes were reduced. Using SSURGO overestimated total SOC stocks by up to 5.1% compared with the other approaches, suggesting that this approach may be worthy of evaluation for circumstances with a low budget and low confidence level requirements. Systematic sampling and ordinary kriging can provide an optimal strategy for estimating SOC stocks in agroecosystems with complex topography and land uses.