LAUSR

Abstract Land conversion and fragmentation threaten the resilience and biodiversity of wetland ecosystems which makes the future of their services to humans uncertain. Remote sensing can provide frequent and consistent data to facilitate wetland monitoring from regional to national scales and support their conservation and adaptive management. However, unique characteristics of wetlands, particularly land cover heterogeneity and background reflectance from soil, water and dead biomass, limit the efficacy of remote-sensing based metrics developed for terrestrial ecosystems. To identify the factors impacting satellite-based measurements of wetland greenness, we tested how six spectral vegetation indices responded to the land surface characteristics and regional climatic and edaphic context of 1,138 wetlands sites surveyed by the U.S EPA's National Wetland Condition Assessment. Spectral vegetation indices (SVIs) were estimated using all cloud-free surface reflectance data captured in 2011 by Landsat 5 TM and 7 ETM+. We tested two annually aggregated metrics —maximum and median greenness— for each SVI to facilitate the analysis of such a large dataset of satellite images. Using univariate and multivariate ordinary least square regression, we assessed how the annual maximum and median of each SVI responded to indicators of vegetation structure and composition, presence of dead biomass, open water, bare soil, and climatic/edaphic variables. Results show that, in the full national-scale dataset, the normalized difference vegetation index (NDVI) and the green normalized difference vegetation index (GNDVI) were most responsive to field-based metrics of vegetation structure and composition. However, the responses of SVIs differed significantly among wetland types, suggesting that their use should be tailored to the specific characteristics of the monitored wetlands. Annually aggregated metrics showed different sensitivity in multivariate models, with median greenness being more sensitive to structure and composition, but also to confounding site variables including litter, open water, and bare soil. This study represents a first-time effort to study relationships between the on-site properties of wetlands and their spectral characteristics at a national scale.