Skip to main content
ARS Home » Research » Publications at this Location » Publication #181186
Title: NEAR-INFRARED REFLECTANCE SPECTROSCOPY FOR ASSESSMENT OF SOIL PROPERTIES IN SEMI-ARID TURKEY

Author
item BILGILI, ALI - CORNELL UNIVERSITY
item VANES, HAROLD - CORNELL UNIVERSITY
item AKBAS, FEVZI - U. GAZIOSMANPASA, TURKEY
item DURAK, ALPER - U. GAZIOSMANPASA, TURKEY
item Hively, Wells - Dean
item DEGLORIA, STEPHEN - CORNELL UNIVERSITY
item OWIYO, THOMAS - CORNELL UNIVERSITY

Submitted to: Soil Science Society of America Annual Meeting
Publication Type: Abstract Only
Publication Acceptance Date: 10/1/2005
Publication Date: 11/7/2005
Citation: Bilgili, A.V., Vanes, H.M., Akbas, F., Durak, A., Hively, W.D., DeGloria, S.D., Owiyo, T. 2005. Near-infrared reflectance spectroscopy for assessment of soil properties in semi-arid Turkey [abstract]. Soil Science Society of America Annual Meeting. Available: http://crops.confex.com/crops/2005am/techprogram/P6996.HTM

Interpretive Summary:

Technical Abstract: Reflectance spectroscopy has been adopted as a nondestructive method for material characterization in a wide range of applications. In this study, near infrared reflectance (NIR) spectroscopy was used to evaluate diverse soil properties for five different soil series of Entisol, Mollisol and Aridisol soil groups in northern Turkey, using 512 soil samples collected in a 25x25 m sampling grid over a 32 ha (800x400m) area . Air-dried soil samples were scanned at 1-nm resolution from 350 to 2500 nm (near-infrared wavelengths) and calibrations between soil properties and soil reflectance were developed by using a cross validation procedure under partial least squares (PLS) regression and multivariate adaptive regression splines (MARS). The two data analysis methods were evaluated and compared by using raw reflectance data and first derivative reflectance data, separately and combined. The strongest correlations (r2 values) between reflectance and soil chemical properties were observed using PLS regression and first derivative data, for exchangeable Ca (0.85), cation exchange capacity (0.88), exchangeable Mg (0.78), organic C concentration (0.84), clay content (0.93), sand content (0.90), and CaCO3 (0.82). Weaker correlations were observed for pH, EC, exchangeable K and Na (0.55 - 0.65). Overall, PLS regression provided larger r2 values than MARS. Also, no improvement was obtained by using both first derivative and raw data combined, nor by increasing the number of soil samples processed from 250 to 512. In conclusion, cross-validated regression models were able to predict soil properties with variable predictive ability, with the best results provided by PLS regression on first derivative reflectance data.