Ease of access to satellite sensor imagery and image products has driven the use of remote sensing data in many disciplines, including landscape ecology, forestry, environmental and wildlife management, agriculture, and epidemiology. A common format of these data is as vegetation indices and of these the most widely used is the Normalized Difference Vegetation Index (NDVI). There are many examples where the NDVI has been assessed as a covariate in spatial models of ecosystem health, faunal habitat, faunal distribution and condition, forage quality and disease risk, to name but a few. Occasionally, such applications detach the NDVI from its theoretical underpinnings; there may be no consideration of surface phenomena affecting the radiation environment and causing change in red and near infrared (NIR) reflectance. In this paper, we revisit classic remote sensing research to determine for which environmental modeling applications the NDVI is likely to be meaningful. We collated data from multiple published research studies to characterize the relationships between NDVI and vegetation canopy variables (leaf area index, biomass, percentage cover, and the fraction of absorbed radiation) over different vegetation types. Using these collated data sets, we discuss the physical basis of the relationships between NDVI and vegetation canopy variables, how and why these relationships differ, and what this means for the usefulness of the NDVI as a proxy for Earth surface or climate variables. We conclude with a critique of the variables that NDVI has been used to replace as a proxy and show its success or failure as a proxy variable relates back to the simple relationships known to exist between NDVI and vegetation canopy variables.  Abstract B42C-01.