Spectroscopic Detection of Nitrogen Concentrations in Sagebrush: Implications for Hyperspectral Remote Sensing
- Examine differences between sagebrush N concentration and spectral response at leaf and shrub scales
- Identify transformed bandwidth intervals most closely related to N concentrations
- Determine if the strength at which narrow absorption features are expressed using a field radiometer warrants extending the investigation to an airborne platform
- Acquire and ground-validate airborne hyperspectral imagery to determine if sagebrush canopy signals are strong enough to support detection of sagebrush N at the canopy scale.
Encouraging final results were obtained from applying spectroscopic methods to estimate sagebrush N concentrations in the laboratory using dry leaf material and in the field using individual live shrub canopies. The project was extended to the canopy scale using hyperspectral imagery we had the opportunity to acquire on August 13, 2010, using an airborne HyMap sensor. Plot level field data was collected from August 9-10, 2009, to support the HyMap collection. A total of 35 plots were sampled for percent cover (sagebrush, shrubs other than sagebrush, grass/herbaceous, bare ground, and dead wood), and average sagebrush height and foliar N content. Foliar N content was analyzed by collecting green-leaf samples from four randomly selected sagebrush individuals in each plot. Leaf area measurements were then used to scale N concentrations from leaf level to plot level. Plot level sagebrush N estimations were then related to corresponding refl ectance data in the HyMap imagery by extracting pixels located within the plot boundaries.
Sagebrush dry leaf spectra produced models that could predict N concentrations within the dataset more accurately than models generated from live shrub spectra because noise (e.g., soil, atmosphere, sampling error, leaf water) was minimized. Including wavelengths associated with leaf water appeared to improve these results. Since leaf water plays an obvious role in estimating N at the shrub scale, August may be an optimal time for acquiring additional sagebrush reflectance spectra because leaf water is lowest in late summer.
HyMap spectra transformed using standard derivative analysis was capable of detecting sagebrush canopy N concentrations using regression with an R2 value of 0.72 and an R2 predicted value of 0.42. Subsetting the HyMap dataset to minimize the influence of bare ground increased R2 to 0.95 (R2 predicted = 0.42). The results of this study represent an important step in addressing the confounding infl uence of bare ground, which was found to be a significant challenge in remote sensing of foliar N in semi-arid landscapes, possibly more so than leaf water.