Visible-NIR reflectance spectroscopy as a potential tool for management of soil phosphorus fertility in legume-based pasture in New Zealand

Abstract

Increasing the current precision of nutrient management will need analytical tools that aid in collecting site specific data. A technology with potential is hyperspectral remote sensing. Modern, portable spectroradiometers permit reflectance data in the spectral region between 350 and 2500 nm to be collected quickly. With the limited sampling, handling, and processing required the technology also offers rapid turn around times, if calibrations can be developed. In this paper the findings of a pilot study examining the use of hyperspectral reflectance spectra of pasture to indirectly assess the phosphorus (P) status of the soil are presented and discussed. Spectral data were collected in spring 2004 and again in summer 2006 from a small area of each of 30 legume-based sheep grazed pasture plots that varied in soil P fertility (Olsen P 6-68 μg/ml). Significant (P<0.05) relationships were found with two commonly used vegetation indices and pasture P content (R2 >0.80). In an exploratory analysis using all the spectral waveband data, several paired-bands with high coefficients of determination (R2) were detected for pasture P and K content, but not for pasture N content, pasture growth rate or pasture dry matter content. The differences detected in pasture P content were consistent with the differences in soil P fertility measured by the Olsen P soil test, as indicated by the relationship between pasture P content and soil Olsen P in both 2004 (R2 = 0.90) and 2006 (R2 = 0.86). This pilot study needs to be broadened to examine other methodologies for interpreting the spectral data and extended to other pasture types and soil groups of varying soil fertility. Keywords: Olsen P, soil fertility, remote sensing, hyperspectral imaging, spatial variability, soil phosphate, plant phosphate