Relative Spectral Mixture Analysis: a new multitemporal index of total vegetation cover

High temporal resolution remote sensing provides an opportunity to monitor phenological variability and interannual changes in vegetation cover across diverse. A principal tool in multitemporal vegetation monitoring has been the Normalized Difference Vegetation Index. NDVI provides an index of the depth of the red edge and is usually interpreted as a measure of vegetation greenness and/or green vegetation cover. NDVI as a measure of phenology has several failings, particularly when applied over large areas: 1) NDVI is sensitive to the spectra of the soil background, particularly in partially or seasonally vegetated areas, 2) NDVI does not provide information about the non-photosynthetic portion of standing biomass, and 3) NDVI is very sensitive to the presence of small amounts of snow in a pixel. A new method for measuring vegetation phenology has been developed called Relative Spectral Mixture Analysis. RSMA uses a linear spectral mixture analysis to provide an index of the relative cover of four landscape components: green vegetation (GV), nonphotosynthetic vegetation (NPV), soil, and snow. RSMA uses generalized spectral for GV, NPV and snow, but does not require knowledge of the background soil spectra. This allows RSMA to be applied over very large areas (continental-scale) in which the soil background is highly diverse. RSMA has been implemented in the IDL language and used to analyze MODIS nadir-adjusted reflectance products from 2000 to the present. Our results show that RSMA GV index values are highly correlated with NDVI, except in regions with snow where RSMA outperforms NDVI. As a result RSMA GV indices and total vegetation indices (GV+NPV) can be used to extract information from spectral timeseries such as the onset of greenness, the termination of greenness, maximum vegetation cover, integrated vegetation cover (an index of NPP), length of the growing season, and duration of fodder availability. RSMA snow indices correlate well with other satellite-derived estimates of snow cover. Because RSMA provides a simultaneous index of both green and non-photosynthetic vegetation cover, it is particularly useful for the study of arid and semiarid regions where a significant proportion of standing biomass is not green. It is also particularly useful for the study of senescence in seasonally-vegetated regions where NDVI only provides information on changes in greenness. Because RSMA includes snow as an endmember, estimates of GV and NPV cover are impervious to snow events. As a result, RSMA can be used to study evergreen vegetation during the snowy winters as well as phenology in tundra, alpine, or cold desert environments where snow can fall during the growing season.