Skip to main content
ARS Home » Research » Publications at this Location » Publication #285687

Title: A simplified regional-scale electromagnetic induction - Salinity calibration model using ANOCOVA modeling techniques

Author
item Corwin, Dennis
item LESCH, SCOTT - City Of Riverside

Submitted to: Geoderma
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 3/23/2014
Publication Date: 4/13/2014
Citation: Corwin, D.L., Lesch, S.M. 2014. A simplified regional-scale electromagnetic induction - Salinity calibration model using ANOCOVA modeling techniques. Geoderma. 230-231:288-295. doi: 10.1016/j.geoderma.2014.03.019.

Interpretive Summary: Soil salinity is a major agricultural concern throughout the world, particularly on irrigated lands of arid zone soils. The complex spatial variability of soil salinity has stood as a barrier to its measurement and mapping at field scales and larger spatial extents. Even though progress has been made in the area of mapping salinity at regional scales as a result of the combined use of satellite imagery and geophysical techniques, it remains as a challenge since current approaches are complex and technology intensive. Scientists at the USDA-ARS U.S. Salinity Laboratory have developed a simplified methodology for mapping soil salinity at regional scale using a statistical regression technique referred to as an analysis of variance (ANOCOVA) model to calibrate easily obtained apparent soil electrical conductivity measurements to soil salinity over areas ranging from thousands to hundreds of thousands of acres. An evaluation of the approach was made for two case studies: 6000 ha of the Broadview Water District in California's San Joaquin Valley and roughly 250,000 ac of the west side of Kittson County in the Red River Valley of Minnesota. The ANOCOVA model outperformed other statistical approaches and provided estimates for depth predictions of soil salinity that were well within acceptable limits for mapping. The implication of this evaluation is that once ANOCOVA models for each depth are established for a representative set of fields within a regional, then the slope coefficients can be used at all future fields, thereby significantly reducing the need for ground-truth soil samples at future fields, which substantially reduces labor and cost by up to 66% over previous approaches. The methodology has the practical simplicity to allow broad application by government agencies, such as NRCS, with minimal cost and has international implications, particularly for use in countries with limited resources. Land resource managers, producers, agriculture consultants, extension specialists, and Natural Resource Conservation Service field staff are the beneficiaries of regional-scale maps of soil salinity.

Technical Abstract: 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 soil salinity, water content, texture, bulk density, organic matter, and cation exchange capacity. Multi-field ECa survey data often exhibit abrupt changes in magnitude across field boundaries that complicate the calibration of ECa to soil salinity (i.e., ECe, electrical conductivity of the saturation extract) over large spatial extents. The primary objective of this study is to evaluate three regression techniques for calibrating ECa to ECe over spatial scales ranging from a few thousand to a hundred thousand hectares, where ECa was measured using electromagnetic induction equipment. The regression techniques include analysis of covariance (ANOCOVA), field specific regression (FSR), and common coefficient regression (CCR). An evaluation was made by comparing jack-knifed mean square prediction errors (MSPE) of ECe for two case studies: 2400 ha of the Broadview Water District in California's San Joaquin Valley and roughly 100,000 ha of the west side of Kittson County in the Red River Valley of Minnesota. The ANOCOVA model outperformed the FSR and CCR regression models on a prediction accuracy basis with the smallest MSPE estimates for depth predictions of soil salinity. The implication of this evaluation is that once ANOCOVA models for each depth are established for a representative set of fields within a regional-scale study area, then the slope coefficients can be used at all future fields, thereby significantly reducing the need for ground-truth soil samples at future fields, which substantially reduces labor and cost. Land resource managers, producers, agriculture consultants, extension specialists, and Natural Resource Conservation Service field staff are the beneficiaries of regional-scale maps of soil salinity.