Until recently, the Advanced Very High Resolution Radiometer (AVHRR) was the only broad scale, globally applicable satellite that provided direct spectral data suited for continual monitoring of vegetation. As such, many studies have successfully used AVHRR normalized difference vegetation index (NDVI) to infer photosynthetic monitor growing season phenology and estimate vegetation. On 18 December, 1999 the first Moderate Resolution Imaging Spectroradiometer (MODIS) was launched on the Terra platform of the Earth Observing System (EOS). The MODIS instrument provides new and improved capability for terrestrial remote sensing intended for global change research including a suite of standard products designed to remove the burden of most data processing requirements. To determine the practical usefulness of MODIS products, we sought to establish a relationship between MODIS leaf area index (LAI), enhanced vegetation index (EVI), and NDVI and the above-ground herbaceous green biomass in a semi-arid grassland ecosystem. This study focuses on relating the MODIS eight - day composite LAI, and 16 - day composite EVI and NDVI, to direct measures of above - ground biomass generated within the growing season during four time periods in two steps: 1) establish a methodology for converting plot level biomass measurements to a regional scale; and 2) characterize the relationship between selected MODIS land products and spatially scaled - field observations of grassland vegetation productivity.
Field data were collected in the Little Missouri National Grasslands (LMNG) of western North Dakota (Figure 1). This 809,380 ha area is managed primarily by the USDA Forest Service for cattle grazing, oil and gas leasing, wildlife habitat, and recreational uses (Jensen et al. 2001). Due to their large geographic expanse and the dominance of federal ownership, the LMNG provided an excellent opportunity for collecting field data and therefore relating MODIS - derived land products to grassland vegetation.
Biomass observations were made during the 2001 growing season at 2,200 plots (473 transects) across four time periods, each five days in length, in the LMNG. All herbaceous biomass within a 0.5 m2 quadrat was clipped at ground level every fifty meters along each transect and percentage of living vegetation was recorded. Herbaceous biomass was subsequently dried at 65 C° for at least 48 hours and weighed. Final estimates of above - ground net primary productivity within each clipped plot were determined by multiplying the percentage of living vegetation by the weight of dried biomass. For scaling, biomass measures were subsequently modeled within the spatial extent of each Thiesson polygon using a multiple regression formula combining ETM + NDVI, accumulated growing degree days (thermal time), and summation of precipitation, of the form: biomass = NDVI (65.0112)+ (pptsum(0.9) - gddsum^2(0.0013)) where biomass is the estimated biomass within each Thiesson polygon, NDVI is the average NDVI for a given polygon, pptsum is the summation of precipitation from 1 January 2001 to the date of ground sampling, and gddsum is the summation of thermal time (TAVGdaily - 0) from 1 January, 2001 to the date of ground sampling where TAVGdaily is the daily average temperature. Scaled biomass measurements were compared with MODIS LAI, NDVI and EVI.
MODIS LAI (Figure 2), EVI (Figure 3) and NDVI (Figure 3) were all closely related to observed biomass. The results of this study present a framework for linking small-scale field observations to MODIS LAI, EVI and NDVI while simultaneously providing much needed insight to the relationship between MODIS land products and vegetation productivity. The high correlation between MODIS land products and observed above - ground green biomass proved that MODIS land products are suitable for monitoring grassland vegetation dynamics in northern mixed grass prairie and appear to offer improved capabilities compared with the AVHRR NDVI relationships presented in previous work.