Using Normalized Difference Vegetation Index to Estimate Carbon Fluxes from Small Rotationally Grazed Pastures

Abstract

Satellite‐based normalized difference vegetation index (NDVI) data have been extensively used for estimating gross primary productivity (GPP) and yield of grazing lands throughout the world. However, the usefulness of satellite‐based images for monitoring rotationally‐grazed pastures in the northeastern United States might be limited because paddock size is often smaller than the resolution limits of the satellite image. This research compared NDVI data from satellites with data obtained using a ground‐based system capable of fine‐scale (submeter) NDVI measurements. Gross primary productivity was measured by eddy covariance on two pastures in central Pennsylvania from 2003 to 2008. Weekly 250‐m resolution satellite NDVI estimates were also obtained for each pasture from the moderate resolution imaging spectroradiometer (MODIS) sensor. Ground‐based NDVI data were periodically collected in 2006, 2007, and 2008 from one of the two pastures. Multiple‐regression and regression‐tree estimates of GPP, based primarily on MODIS 7‐d NDVI and on‐site measurements of photosynthetically active radiation (PAR), were generally able to predict growing‐season GPP to within an average of 3% of measured values. The exception was drought years when estimated and measured GPP differed from each other by 11 to 13%. Ground‐based measurements improved the ability of vegetation indices to capture short‐term grazing management effects on GPP. However, the eMODIS product appeared to be adequate for regional GPP estimates where total growing‐season GPP across a wide area would be of greater interest than short‐term management‐induced changes in GPP at individual sites.