June 2004 / Vol. 54 No. 6 • BioScience 547ArticlesEarth systems science—the integrated analysis ofthe atmosphere, oceans, cryosphere, and biosphere as afunctioning system—had its origins in the early 1980s. Afew years earlier, the introduction of the now famous at-mospheric “Keeling curve” for carbon dioxide (CO2), mea-sured at Mauna Loa, Hawaii, had shown definitively for thefirst time that humans were progressively changing Earth’s total atmosphere (Keeling et al. 1976).Atmospheric scientistshad recently completed the first primitive global models ofatmospheric circulation. Ocean and cryospheric scientistswere taking measurements of ocean temperature and icesheet area, which, although not global, were certainly regional in scope. Unfortunately, however, ecologists (in-cluding the senior author) had no background in addressingecology at global scales. Before the 1980s, biology focused onthe organism level, and ecological studies, at best, embraceda 0.1-hectare (ha) field plot. No global ecological measurementwas even considered, so analyzing seasonal trends in atmo-spheric CO2concentrations was the only option for assess-ing global biospheric activity. In early discussions, ecologistsconsidered measuring vegetation density, canopy height,plant biomass, species classes, and other common ecologicalvariables. Only slowly, by realizing that true global measure-ments could only be made using satellite remote sensing,did ecologists begin to develop a conceptual basis for com-puting a satellite-based estimate of net primary production(NPP).From a theoretical standpoint, NPP marks the first visiblestep of carbon accumulation; it quantifies the conversion ofatmospheric CO2into plant biomass. The earliest attempts toevaluate ecosystem processes such as NPP at a global scale weremade by geographers, and the famous estimates by Liethand Whittaker (1975) of global NPP are still quoted today.These estimates were based on regressions of temperature datafrom about a thousand weather stations, computed to a sim-ple annual measure of actual evapotranspiration (AET) in mil-limeters (mm) per year and then regressed against a handfulof NPP field plots. The resulting equation,NPP = 3000{1 – exp[–0.0009695(AET – 20)]},was used to compute and map the first global estimate of NPP(118 billion metric tons per year of biomass) from essentiallytransformed and coarsely extrapolated meteorology data.Three activities that first evolved in the early 1980s provedto be the foundations of global-scale terrestrial ecology. First,Steven W. Running (e-mail: [email protected]) is a professor and director,Faith Ann Heinsch and Maosheng Zhao are research personnel, Matt Reevesis a graduate research assistant, and Hirofumi Hashimoto is a visiting graduate student in the Numerical Terradynamic Simulation Group,Department of Ecosystem and Conservation Sciences, University of Montana,Missoula, MT 59812. Ramakrishna R. Nemani is a research scientist in theEarth Science Division at the National Aeronautics and Space Administra-tion, Ames Research Center, Moffett Field, CA 94035. © 2004 American Institute of Biological Sciences.A Continuous Satellite-DerivedMeasure of Global TerrestrialPrimary Production STEVEN W. RUNNING, RAMAKRISHNA R. NEMANI, FAITH ANN HEINSCH, MAOSHENG ZHAO,MATT REEVES, AND HIROFUMI HASHIMOTOUntil recently, continuous monitoring of global vegetation productivity has not been possible because of technological limitations. This article in-troduces a new satellite-driven monitor of the global biosphere that regularly computes daily gross primary production (GPP) and annual net pri-mary production (NPP) at 1-kilometer (km) resolution over 109,782,756 km2of vegetated land surface. We summarize the history of global NPPscience, as well as the derivation of this calculation, and current data production activity. The first data on NPP from the EOS (Earth ObservingSystem) MODIS (Moderate Resolution Imaging Spectroradiometer) sensor are presented with different types of validation. We offer examples ofhow this new type of data set can serve ecological science, land management, and environmental policy. To enhance the use of these data by non-specialists, we are now producing monthly anomaly maps for GPP and annual NPP that compare the current value with an 18-year average valuefor each pixel, clearly identifying regions where vegetation growth is higher or lower than normal.Keywords: net primary production, MODIS, carbon cycles, terrestrial remote sensing, biospherethe expansion of atmospheric flask sampling to many addi-tional sites made it possible to draw inferences on a biosphericscale about the photosynthetic uptake of CO2in summer, andthe evolution of CO2in winter, in the Northern Hemisphere(Tans et al. 1990). Second, the global climate models beganincluding simple submodels of land surface processes that in-corporated some basic ecology (e.g., physiological controls bystomata on leaf gas exchange; structural vegetation measuressuch as leaf area index [LAI]). The biosphere-atmospheretransfer scheme, or BATS (Dickinson 1996), laid the foun-dation for all future representations of terrestrial vegetationin global climate models, such as the well-known SiB (sim-ple biosphere) model of Sellers and colleagues (1986). Third,a few ecologists, most notably Compton J. Tucker of the National Aeronautics and Space Administration (NASA),began exploring data from meteorological satellites (i.e.,AVHRR, or Advanced Very High Resolution Radiometer) tosee whether some measure of vegetation was discernible.The cover of Science in August 1985 inaugurated global vegetation analysis, showing the first NDVI (normalized dif-ference vegetation index) image of the African continent(Tucker et al. 1985).In the early 1980s, the Reagan administration discour-aged NASA from using the fledgling Earth remote sensing onlyfor applications like crop yield analysis (the AgriStars and Lacieprograms begun in the 1970s), urging that the technology beused to pursue more “big science”projects.These Reagan ad-ministration policies encouraged NASA to begin a new focuson global ecology. Early workshops were brainstorming sessions on “land-related global habitability science issues”(NASA 1983), exploring what kind of ecology could possiblybe accomplished globally. It was not long before some initialexploration correlated the annual time series of NDVI toannual NPP, the first step in translating the radiometric in-dex of