Tuesday, November 6, 2007
204-18

Assessment of Soil Phosphorous Adsorption Characteristics Using Hyperspectral VNIR Spectroscopy.

Ali Volkan Bilgili1, Harold M. Van Es1, Quirine Ketterings1, Metin Turan2, and Robert Schindelbeck1. (1) Crop and Soil Science, Cornell University, 1015 Bradfield Hall, Ithaca, NY 14850, (2) Department of Soil Science, Ataturk University, Erzurum, Turkey

The environmental pollution risk of agricultural chemicals via leaching, runoff and erosion is of great concern. Knowledge of the soil adsorption and desorption characteristics of P is necessary for environmental risk assessment and the development of mitigation strategies. Conventional absorption isotherm methods are too time consuming and expensive for many applications. P adsorption characteristics of soils and other parameters affecting P transportation can be predicted by visible-near infrared reflectance (VNIR) spectroscopy as factors strongly related to P adsorption, such as iron-aluminum oxides, carbonates and organic matter, are measurable in the VNIR range. Soil samples representing a wide range of New York farms with different soil texture and mineralogical characteristics were selected and adsorption characteristics were measured according to the batch equilibrium method by adding known concentrations of P to samples. Data were fit to linear Langmuir adsorption isotherms and two important indicators of P transport to water bodies were calculated, soil P adsorption maxima (Smax) (range: 442 – 1615 g kg-1) and adsorption energy coefficient (k) (range: 0.029-0.192 ug ml-1). Soils were scanned and soil reflectance spectra within a range of 350 to 2500 nm (2,100 data point per sample) were collected using an Advanced Spectral Design FieldSpec Pro spectroradiometer. Both raw reflectance values and first derivatives and absorbance transformation (log (1/R)) of reflectance spectra were calibrated with adsorption characteristics of soils, using Partial Least Squares (PLS) and Multivariate Adaptive Regression Splines (MARS). Best estimations were obtained for Smax using raw spectra with PLS providing a R2 value of 0.82 and standard error of prediction value of 202 g kg-1. VNIR spectroscopy showed promise in evaluating potential P loss from soil to water sources as an inexpensive technique.