A Comparison Of Methods For Scaling Field Data For Use In Mapping Dryland Ecosystem Vegetation With Airborne Imaging Spectroscopy

This research investigated scaling methods of field data to interpret aerial surveys for foliar N estimation using imaging spectroscopy. Foliar nitrogen (N) is an indicator of vegetative growth, which is related dryland ecosystem services. We compared four field-based methods to scale sagebrush foliar N estimates from shrub to the plot level (10 m x 10 m) for 21 plots collected in a dryland ecosystem in 2014 and 2015. Partial least squares regression related estimates to imaging spectroscopy variables. Results showed sensitivity to scaling method; pretreatment of imaging spectroscopy signals; subdividing the dataset into years; reducing predictor variables to reduce noise; and number of model iterations. The best performing methods used biomass allometry with density counts or cover estimates with leaf thickness with a log transformation and Savitzky-Golay smoothing method. Models selected different wavelengths as predictors. Several relied on wavelengths in the visual range associated with chlorophyll and few relied on wavelengths in the "red edge" of 800-850 nm. The best performing model used biomass allometry and a subset of wavelengths that consistently performed well across model iterations. This was used to map predicted foliar N values across the Reynolds Creek Experimental Watershed and can be used to support rangeland management