147-8 Soil Survey: Estimating Soil Linear Extensibility Percent Using General Linear Model.

Poster Number 1127

See more from this Division: SSSA Division: Pedology
See more from this Session: Pedology: II (includes student competition)

Monday, November 16, 2015
Minneapolis Convention Center, Exhibit Hall BC

Cathy A. Seybold, National Soil Survey Center, USDA-NRCS, Lincoln, NE
Abstract:
The shrink-swell property of soils affects a wide range of rural and urban land uses, and is a major consideration in soil survey interpretations for construction of foundations, roads and other structures. The objective is to develop and validate a model for estimating the soil linear extensibility percent (LEP; a measure of the shrink-swell potential) using general linear models and readily available soil properties within the soil survey database (National Soil Information System; NASIS) of Natural Resources Conservation Services of USDA. Measured data from the Kellogg National Soil Survey characterization database (Lincoln, Nebraska) were used to develop the prediction models. The development dataset was stratified by taxonomic mineralogy class. Organic matter represents a non-reversible shrinkage portion of the soil. For soil layers with organic matter contents between 5 and 35%, LEP is based primarily on the mineral fraction. Twenty-six LEP prediction equations (general linear models) were developed for the major mineralogy classes or groups of mineralogy classes. Only variables that contributed significantly (P = 0.05) to predicting LEP or contributed significantly to the improvement of the root mean square error (RMSE) were included in the models. Non-carbonate clay and cation-exchange capacity (CEC) or effective cation-exchange capacity (ECEC) were found to explain between 31 and 81 percent of the total variation in the LEP data among the models. Model validation was conducted using an independent dataset of 2,125 soil layers from pedons sampled throughout the USA. The collective model (all models) explained 82.2% of the variation in LEP with a RMSE of 1.63%. The sum of residuals indicated a very small bias. The use of general linear models and stratifying by taxonomic mineralogy class proved useful in predicting LEP for use in soil survey when measured data is not available, which will ultimately improve soil survey interpretations.

See more from this Division: SSSA Division: Pedology
See more from this Session: Pedology: II (includes student competition)