Submitted to: ASABE Annual International Meeting
Publication Type: Proceedings
Publication Acceptance Date: 4/30/2008
Publication Date: 6/30/2008
Citation: Lee, K., Lee, D., Sudduth, K.A., Chung, S., Kitchen, N.R., Drummond, S.T. 2008. Sampling and Calibration Requirements for Soil Property Estimation Using Reflectance Spectroscopy. In: ASABE Annual International Meeting Technical Papers. American Society of Agricultural and Biological Engineers Annual International Meeting, June 29-July 2, 2008, Providence, Rhode Island. Paper No. 084037.
Technical Abstract: Optical diffuse reflectance sensing is a potential approach for rapid and reliable on-site estimation of soil properties. One issue with this sensing approach is whether additional calibration is necessary when the sensor is applied under conditions (e.g., soil types or ambient conditions) different from those used to generate an initial calibration, and if so, how many sample points are required in this additional calibration. In this study, these issues were addressed using data from 10 fields from 5 states in the north-central USA. Partial least squares (PLS) regression was used to develop calibrations between soil properties and reflectance spectra. Model evaluation was based on coefficient of determination (R^2) and RPD, the ratio of standard deviation to standard error of prediction. When sample data from a field in question were included in the calibration stage (full information calibration), RPD values of prediction models were increased by 0.16 to 1.78 for profile data and 0.02 to 1.58 for surface soil data, compared with results from calibration models not including data from the test field (calibration without field-specific information). Including some samples from the test field (hybrid calibration) generally increased RPD by 0.7 to 1.5 with 6 to 15 sample points, with little further improvement given additional points. Using test field points as a bias adjustment (two-stage calibration) increased RPD by 0.1 to 0.2 with 2 to 3 sample points for profile data, and by 0.3 to 0.7 with 5 to 7 sample points for surface data. These results provide guidance on sampling and calibration requirements for NIR soil property estimation.