267-5 Assessment of Soil Quality Using Hyperspectral Spectroscopy In a Western Kenya Chronosequence.



Tuesday, October 18, 2011: 2:00 PM
Henry Gonzalez Convention Center, Room 006D, River Level

Rintaro Kinoshita1, Bianca Moebius-Clune1, Harold van Es1, W. Dean Hively2 and Volkan A. Bilgili3, (1)Cornell University-Crop & Soil Sciences, Ithaca, NY
(2)Eastern Geographic Science Center, USGS, Reston, VA
(3)Soil Science, Harran University, Sanliurfa, Turkey
Visible and near-infrared reflectance spectroscopy (VNIRS) is a rapid and non-destructive method to predict various soil properties. However, its application in evaluating multidimensional soil quality (SQ) assessment has not been widely studied in the tropics. VNIRS (350-to-2500 nm; 1 nm resolution) was employed to analyze 227 air-dried soil samples of Ultisols that were collected in a chronosequence study in Western Kenya for the analysis of sixteen Cornell soil quality indicators. Prediction models were constructed using partial least square regression (PLSR) with cross-validation of the leave-ten-out method. The validated models successfully predicted SQ indicators (R2 > 0.80) including soil texture, soil organic matter (OMLOI), active carbon (Cact), cation exchange capacity (CEC), Ca, and Cu. Poorly predicted indicators  (R2 < 0.50) were soil chemical constituents as well as penetration resistance at 0.15 m (PR15). Raw reflectance was generally superior over 1st derivative reflectance in the prediction of the SQ indicators, whereas the combination of raw and 1st derivative reflectance slightly improved the prediction accuracy. Independent sample model validation by separating the sample set into two subsamples consisting of 70 % (n = 159) and 30 % (n = 68) of total samples still maintained good predictability (R2 > 0.80) for clay, OMLOI, Cact, and Cu even when the smaller subset was used for calibration. Furthermore, VNIRS showed moderate to substantial agreement in predicting Cornell SQ scores and the composite soil quality index (CSQI). The VNIRS was also combined with conventional measurements of pH, EC, PR15, and PR45 in the PLSR procedure. This substantially improved the prediction for K, Zn, and S. Moreover, soil chemical indicator values combined with VNIRS improved the predictability of AWC to above the threshold of RPD > 2. Results indicate that low cost VNIRS has potential for substituting conventional analytical methods for most of physical and biological soil indicators especially when combined with direct measurements of pH, EC, PR15, and PR45, but conventional soil chemical tests may need to be retained to provide an integrated soil quality assessment.
See more from this Division: S06 Soil & Water Management & Conservation
See more from this Session: Impact of C3 (Crop Rotation, Cover Crops, and Conservation Tillage) On Soil Quality: I