Characterization of Soil Shrink-Swell Potential Using the Texas VNIR Diffuse Reflectance Spectroscopy Library Katrina M. Hutchison, Cristine L.S. Morgan, and C. Tom Hallmark Texas A&M University, College Station, Texas Conclusion 2454 archived soils from Texas are now part of a spectral VNIR library. Soil COLE values from the Texas soils are highly correlated to total clay content and CEC. Regression using clay content alone and clay content + CEC predict COLE better than using VNIR-DRS with partial least square regression. RPD values show COLE, CEC, and clay content can be predicted effectively using VNIR and that prediction of clay content is the most stable and effective. Methods and Materials Introduction Shrink-swell potential of soil natural fabric is quantified by the coefficient of linear extensibility (COLE). The COLE value is known to be correlated to clay content and clay mineralogy along with other soil properties (McCormack et al. 1975). Visible near infrared diffuse reflectance spectroscopy (VNIR-DRS) has been shown to quantify clay content, clay mineralogy and to be correlated to soil shrink swell potential. (Waiser et al. 2007, Brown 2005, Goetz et al. 2001). The overall goal of this study was to determine if VNIR-DRS is an effective tool for directly quantifying COLE. If so, VNIR- DRS can be used to scan soils in situ to map soil shrink-swell potential in the field. Objectives Create a VNIR-DRS spectral library from archived Texas soils Provide a summary and descriptive statistics of the soils in the spectral library Create predictor models of COLE, clay content, and CEC that might affect COLE using the VNIR-DRS spectrometer References Brown, D.J., K.D. Shepherd, M.G. Walsh, M.D. Mays and T.G. Reinsch Global soil characterization with VNIR diffuse reflectance spectroscopy. Geoderma 132: Goetz, A.F.H., S. Chabrillat and Z. Lu Field reflectance spectrometry for detection of swelling clays at construction sites. Field Analytical Chemistry and Technology 5: McCormack, D.E. and L.P. Wilding Soil properties influencing swelling in Canfield and Geeburg soils. Soil Sci. Soc. Am. J. 39: Waiser, T., C.L.S. Morgan, D.J. Brown and C.T. Hallmark In situ characterization of soil clay content with visible near-infrared diffuse reflectance spectroscopy. Soil Sci. Soc. Am. J. 71: Acknowledgements Thank you to the Texas USDA NRCS Soil Survey for funding this project. Thanks to the Texas Agricultural Experiment Station’s Soil Characterization Lab, including Donna Prochaska and Morgan Arnette for their assistance with data management. Results 2454 soil samples, archived by the Texas Agricultural Experiment Station’s Soil Characterization Laboratory, were transferred to 20 ml vials. Each 20 ml sample was transferred into a borosilicate glass puck and scanned with an AgSpec® Pro (Analytical Spectral Devices, Inc.) that has a spectral range of 350–2500 nm. To calibrate the prediction model, 70% of the soil samples were used and the remaining samples were used for model validation. To compare VNIR to traditional pedotransfer functions, CEC and clay content were used to predict COLE. Models were built using multiple and linear regression, and the same 70/30 calibration/validation data. Table 1. Summary statistics for spectral library UnitsMeanMedianMaxMinN COLEcmcm Clay% Cation Exchange Capacitycmol(+)kg pH Base Saturation% CaCO 3 Equivalent% Organic Carbon% PLS of VNIR-DRS Regression, Pedotransfer Functions COLEClayCEC COLE using clay and CEC COLE using clay RMSD r2r RPD Table 2. Summary of Modeling Results The spectral data were treated by splicing, averaging, and taking the first derivative. Partial Least Squares (PLS) regression was performed with Unscrambler 9.0 to create prediction models to convert spectral reflectance to COLE, clay content, and cation exchange capacity (CEC).