Thursday, 13 July 2006

On-the-go Near Infrared Spectroscopic Assessment of Georgia Soils.

Colin D. Christy, Veris Technologies, 601 N. Broadway, Salina, KS 67401 and David E. Kissel, Agricultural and Environmental Services Laboratories - University of Georgia, 2400 College Station Road, Athens, GA 30602-9105.

This presentation describes the usage of a prototype spectrophotometer for performing diffuse reflectance spectroscopy (DRS) in the near infrared region from approximately 1200 to 2200 nm. The spectrophotometer makes in-situ measurements by transferring incident and reflected light through a sapphire window on the bottom of a shank. The prototype was used on a set of five fields in south-central Georgia averaging 20 hectares in size. The fields were mapped by sensing soil at a depth of approximately 7.5 cm on transects approximately 20 meters apart. Samples were acquired from each field for laboratory analysis of pH (1:1 in 0.01 molar calcium chloride) and soil pH buffer capacity, (described here as lime buffer capacity (LBC) with units of mg CaCO3/kg soil/pH) using a calcium hydroxide titration method. In addition, the dried samples (N=116) were scanned in the laboratory using a scanning monochrometer (Foss/NIR Systems model 6500 ) covering wavelengths from 400 to 2500 nanometers.

Partial least squares (PLS) regression was used to create calibrations for pH and LBC using either the field or laboratory spectra. The prediction capability for pH using the field spectra was much poorer than prediction using the laboratory spectra. However, prediction of LBC using the field spectra produced as good or better results as those using the laboratory spectra. Consequently, it was feasible to apply the PLS model to all spectra collected in the field to characterize the spatial variation of LBC. This resulted in a lime buffer capacity map for each of the five fields which characterizes the spatial variation in LBC.

To further investigate the usage of the prototype spectrophotometer for pH prediction, it was used to scan the dried samples in a laboratory setting. A new PLS calibration was made for pH using these spectra, resulting in a 25% reduction in validation error over usage of the field spectra. However, the validation error was still significantly greater than the error obtained using the laboratory grade spectrophotometer. In summary, the field-mobilized spectrophotometer worked well for the prediction of LBC but was not able to deliver spectra of high enough quality for accurate pH predictions. This was true of the field measurements and to a lesser extent true of the lab measurements of dry soil. On the other hand, the potential for success with this measurement has been demonstrated using the laboratory instrument. Field measurement of pH may eventually be feasible by improving the spectral resolution and/or the signal-to-noise ratio of the field system. Such improvements are currently being made.

Back to 1.5A Diffuse Reflectance Spectroscopy, Soil Sensing, Remote Sensing and Image Analysis - Theater
Back to WCSS

Back to The 18th World Congress of Soil Science (July 9-15, 2006)