SHORT COMMUNICATION MODIS NDVI Optimization To Fit the AVHRR Data Series—Spectral Considerations

In early 1999, about when this article will hit the press, INTRODUCTION NASA is scheduled to launch its first Earth Observing The physical and physiological parameters of vegetation satellite System (EOS) with array of instruments for pre(leaf area, biomass, physiological functioning, etc.) may cise, scientific, remote sensing of the Earth land, ocean, be obtained from satellite remote sensing. For these reaand atmospheres. One of the main instruments, MODIS, sons, several of the instruments scheduled to fly on the will monitor daily vegetation dynamics with seven speEarth Observing System are designed for land surface cially selected and highly characterized and calibrated studies (Asrar and Dokken, 1993). These instruments spectral channels in the solar spectrum, with spatial resomeasure solar radiation reflected by vegetation and soils lution from 250 m to 500 m. These MODIS data, toat certain wavelength bands. Several indexes were develgether with multiangular observations from the MISR inoped for remote estimates of various vegetation paramestrument (also on EOS), will generate a much better ters. The broadband red (0.60–0.70 lm) and near-infrapicture of vegetation dynamic than previous satellite data red (0.75–1.35 lm) channels have been found to be most records. But studying changes on the Earth surface takes valuable in remote sensing of vegetation. Indexes that years and decades. To do it in our scientific lifetime, use these spectral channels were suggested for remote there is a need to be able to develop a data series that estimates of vegetation dynamics (Tucker, 1979), fraction combines the traditional NDVI-AVHRR data sets that of absorbed photosynthetically active radiation (Asrar et was used to monitor vegetation dynamics with the new al., 1984), unstressed vegetation conductance, and photoMODIS data with its narrower channels. Here we prosynthetic capacity (Sellers et al., 1992). They were used pose that due to spectral difference between the AVHRR to estimate the effect of vegetation on seasonal atmoand MODIS, a combination of the MODIS red and green spheric carbon dioxide variations (Tucker et al., 1986) channels should be used to represent the AVHRR red and the effect of CO2 and temperature change on the channel in such long term data series, to avoid a step in the NDVI of dNDVI of up to 0.05. We use both leaf specvegetation (Myneni et al., 1997a). Myneni et al. (1995) tra and full canopy spectra to develop and test this optishowed that for dense vegetation, the widely used broadmization. Optimization of the MODIS sensitivity for band red/near-infrared vegetation indexes are a measure monitoring the dynamics of vegetation chlorophyll conof chlorophyll abundance and energy absorption. Theretent is also discussed. Elsevier Science Inc., 1998 fore, for proper application of vegetation indexes, it is important to investigate the sensitivity of the vegetation reflectance in different spectral channels to chlorophyll (Chl) content. * Department for Environmental Physics and Energy Research, J. Blaustein Institute for Desert Research, Ben-Gurion University of the Recently, it was found that the sensitivity of NDVI Negev, Sede-Boker Campus 84990, Israel to moderate to high Chl content (white-green to dark† Laboratory for atmospheres, NASA Goddard Space Flight Cengreen vegetation) is small because the reflectance in ter, Greenbelt, Maryland spectral region near 670 nm saturates for low Chl conAddress correspondence to A. A. Gitelson, Dept. for Environmental Physics and Energy Research, J. Blaustein Inst. for Desert Retent (Buschmann and Nagel, 1993; Gitelson and Merzsearch, Ben-Gurion Univ. of the Negev, Sede-Boker Campus 84990, lyak, 1994 a,b; 1996; Yoder and Waring, 1994; Gitelson Israel. E-mail: [email protected] Received 26 February 1998; revised 8 June 1998. et al., 1996; Myneni et al., 1997b). Thus, NDVI is sensi-