|Lee, Dong Hoon - SUNGKYUNKWAN UNIV KOREA|
|Lee, Kyou Seung - SUNGKYUNKWAN UNIV KOREA|
|Chung, Sun Ok - CHUNGNAM NATL UNIV KOREA|
Submitted to: Asian Conference on Precision Agriculture
Publication Type: Proceedings
Publication Acceptance Date: July 6, 2007
Publication Date: August 2, 2007
Citation: Lee, D., Lee, K., Sudduth, K.A., Chung, S. 2007. Wavelength Selection for Reflectance Estimation of Surface and Subsurface Soil Properties. In: Proceedings 2nd Asian Conference on Precision Agriculture. August 2-4, 2007, Pyeongtaek, Korea. 2007 CDROM. Technical Abstract: Optical diffuse reflectance sensing is a potential approach for rapid and reliable on-site estimation of soil properties. In this study, reflectance sensing in visible (VIS) and near-infrared (NIR) wavelengths was combined with partial least squares (PLS) regression and stepwise multiple linear regression (SMLR) to estimate surface and subsurface soil properties. Wavelength bands important for estimating soil properties were selected. Soil cores (120 cm deep) from ten fields in five states in the US cornbelt were segmented by horizon and analyzed in laboratory for texture (sand, silt, and clay fractions), cation exchange capacity (CEC), Ca, Mg, K, pH, total and organic C, and total N. Using air-dried, sieved soil samples, reflectance data were obtained from 350 to 2500 nm with a laboratory spectrometer. Over all soil horizons, cross-validated predictions of organic C were good (R2=0.87, RPD [the ratio of standard deviation to standard error of prediction] =2.78 using PLS and R2=0.90, RPD=3.16 using SMLR). Prediction of clay fraction, CEC, and pH were also good (0.80< R2<0.87, 2.23< RPD <2.78) using SMLR and acceptable (0.63< R2<0.79, RPD about 2) using PLS. Although SMLR calibrations were better than those by PLS, additional investigation is needed to insure against overfitting with SMLR. Calibrations restricted to the surface horizon were somewhat better for PLS analysis with R2 values from 0.81 to 0.85 and RPD values from 2.08 to 2.73 for clay fraction, Ca, CEC, and organic C. Important wavelengths were 440-600 nm and 1780-2500 nm for clay; 470-700, 965, and 1750-2500 nm for CEC; and 440-500, 965, 1409, and 1775-2500 nm for organic C. These results will be useful in the design and application of in-situ close range sensors for soil physical and chemical properties.