Rangeland cover component quantification using broad (TM) and narrow-band (1.4 NM) spectrometry

Abstract

Calibrated predictive relationships obtained from simple and multiple regression of thematic mapper or broad-band (BB) and 1.4 nm interval or narrow-band (NB) spectral data were evaluated for quantifying 11 rangeland components (including total vegetation, forb, grass, shrub, litter, and bare soil) and distinguishing among 6 long-term grazing treatments of sagebrush steppe. In general, all 4 data types predicted similar values for each rangeland cover component. Multiple regression models usually had little advantage over simple regression models for predicting cover, particularly for abundant cover components, although this trend was inconsistent among components. Consequently, simple predictive models are recommended for quantifying rangeland indicator components using remotely-sensed data. The use of NB spectral data resulted in lower standard errors of prediction (SEP), although these reductions were inconsistent among rangeland components. Although both data types distinguished among grazing treatments with major plant compositional differences (P < 0.00) using a multivariate analysis of variance (MANOVA), only the NB data distinguished between grazing treatments with minor ecological differences (P < 0.01). These results suggest that in a practical context, NB data are advantageous for quantifying rangeland cover components and distinguishing among grazing treatments under the condition of our study.