Page Banner

United States Department of Agriculture

Agricultural Research Service

Research Project: SALINITY AND TRACE ELEMENTS ASSOCIATED WITH WATER REUSE IN IRRIGATED SYSTEMS: PROCESSES, SAMPLING PROTOCOLS, AND SITE-SPECIFIC MANAGEMENT Title: Comparison of model- and design-based sampling strategies for characterizing spatial variablity with ECa-directed soil sampling

Authors
item Corwin, Dennis
item Lesch, Scott -
item Segal, Eran -
item Shouse, Peter
item Skaggs, Todd
item Bradford, Scott

Submitted to: Journal of Environmental & Engineering Geophysics
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: April 15, 2010
Publication Date: July 15, 2010
Citation: Corwin, D.L., Lesch, S.M., Segal, E., Shouse, P.J., Skaggs, T.H., Bradford, S.A. 2010. Comparison of model- and design-based sampling strategies for characterizing spatial variablity with ECa-directed soil sampling. Journal of Environmental & Engineering Geophysics. 15(3):147-162.

Interpretive Summary: Soils vary substantially in their properties from one point to the next. This spatial variation has a significant influence on a variety of agricultural concerns including the movement of contaminants through soil, assessing the quality of a soil for its intended use, and precision agriculture, to mention a few. Maps of apparent soil electrical conductivity (ECa) can be used to direct soil sampling that can be used to characterize the spatial variation in those soil properties that influence ECa. Arguably the most significant step in the protocols for characterizing spatial variability with ECa-directed soil sampling is the statistical sampling design that specifies the location of where soil samples are to be taken, which consists of model- and design-based sampling strategies. The primary objective of this study was to compare model- and design-based sampling strategies to evaluate if one sampling strategy outperformed the other or if both strategies were equal in performance. Results show that the model-based sampling approach is superior since it produces a better model with smaller prediction variances. Even though a model-based sampling design has been less prevalent in the scientific literature, it is concluded from the comparison that there is no reason to refrain from its use and in fact warrants greater consideration.

Technical Abstract: Spatial variability has a profound influence on solute transport in the vadose zone, soil quality assessment, and site-specific crop management. Directed soil sampling based on geospatial measurements of apparent soil electrical conductivity (ECa) is a potential means of characterizing the spatial variability of any soil property that influences ECa including sol salinity, water content, texture, bulk density, organic matter, and cation exchange capacity. Arguably the most significant step in the protocols for characterizing spatial variability with ECa-directed soil sampling is the statistical sampling design, which consists of model- and design-based sampling strategies such as response surface sampling (RSS) and stratified random sampling (SRS), respectively. The primary objective of this study was to compare model- and design-based sampling strategies to evaluate if one sampling strategy outperformed the other or if both strategies were equal in performance. Three different model validation tests were used to verify that the regression equation estimated from the RSS data produced accurate and unbiased predictions of the log salinity levels at the independently chosen SRS sites. The model validation tests show that the RSS design can be reliably used to estimate the ln function between ECa and soil salinity and can provide an accurate and unbiased prediction of the validation sample sites chosen by the SRS design. Design optimality scores indicate that the use of the RSS design should facilitate the estimation of a more accurate regression model; i.e., the RSS approach should allow for better model discrimination, more precise parameter estimates, and smaller prediction variances. Even though a model-based sampling design, such as RSS, has been less prevalent in the literature, it is concluded from the comparison that there is no reason to refrain from its use and in fact warrants greater consideration.

Last Modified: 11/22/2014
Footer Content Back to Top of Page