Integration of airborne imaging spectroscopy and spaceborne-multispectral images for regional scale mapping of woody cover in pinyon-juniper ecosystems of the Colorado Plateau

Distributions of woody cover in pinyon-juniper (P-J) woodlands of the Colorado Plateau have not been well studied compared to other major savanna-type systems in the Southwest U.S. such as mesquite or oak woodlands. Our previous effort suggested that photosynthetic vegetation (PV) fraction derived from an airborne hyperspectral imaging spectroscopy (AVIRIS) was a salient variable for estimating woody cover (r = 0.96) over a small regional extent (228 km²).  However, the spatial coverage of AVIRIS is insufficient for a large-scale woody cover mapping in Southwest P-J systems (~ 500,000 km²). Therefore, the objective of this study is to upscale PV-AVIRIS by linking it to PV fraction extracted from spaceborne multi-spectral Landsat ETM+ data (PV-ETM+). Results showed that the correlation between PV-AVIRIS and PV-ETM+ was significant (P < 0.001) but rather weak (curvilinear; r = 0.57) after adjusting for spatial autocorrelation. PV-ETM+ was consistently higher than PV-AVIRIS when vegetation cover was not dense (e.g., PV < 40%). However, the relationship was erroneous in highly vegetated areas. Disagreement between these two sets of data may result from the noises contributed from background layers of non-photosynthetic vegetation (litters), bare soils, and shadows (from tree canopy and terrain). These variations were indistinct in the ETM+ multispectral space. Results suggest that a regression approach may not be able to fully explain the uncertainty across sensors; alternative approaches were required to decipher the complexity of scaling issues.