Water-removed spectra increase the retrieval accuracy when estimating savanna grass nitrogen and phosphorus concentrations

Information about the distribution of grass foliar nitrogen (N) and phosphorus (P) is important for understanding rangeland vitality and for facilitating the effective management of wildlife and livestock. Water absorption effects in the near-infrared (NIR) and shortwave-infrared (SWIR) regions pose a challenge for nutrient estimation using remote sensing. The aim of this study was to test the utility of water-removed (WR) spectra in combination with partial least-squares regression (PLSR) and stepwise multiple linear regression (SMLR) to estimate foliar N and P, compared to spectral transformation techniques such as first derivative, continuum removal and log-transformed (Log(1/R)) spectra. The study was based on a greenhouse experiment with a savanna grass species (Digitaria eriantha). Spectral measurements were made using a spectrometer. The D. eriantha was cut, dried and chemically analyzed for foliar N and P concentrations. WR spectra were determined by calculating the residual from the modelled leaf water spectra using a nonlinear spectral matching technique and observed leaf spectra. Results indicated that the WR spectra yielded a higher N retrieval accuracy than a traditional first derivative transformation (R-2 = 0.84, RMSE = 0.28) compared to R-2 = 0.59, RMSE = 0.45 for PLSR. Similar trends were observed for SMLR The highest P retrieval accuracy was derived from WR spectra using SMLR (R-2 = 0.64, RMSE = 0.067), while the traditional first derivative and continuum removal resulted in lower accuracy. Only when using PLSR did the first derivative result in a higher P retrieval accuracy (R-2 = 0.47, RMSE = 0.07) than the WR spectra (R-2 = 0.43, RMSE = 0.070). It was concluded that the water removal technique is a promising technique to minimize the perturbing effect of foliar water content when estimating grass nutrient concentrations. 2011 International Society for Photogrammetry and Remote Sensing, Inc. (ISPRS). Published by Elsevier B.V. All rights reserved.