1122 
November 1988, which leads to change in the time of observations and therefore to different conditions of 
measurements. In addition, radiometric calibrations may change rapidly during the few months following the 
launch and gain drift corrections are probably less accurate during this period. 
• In 1989, 1990 and 1991, GVI are rather similar in the northern hemisphere, but in the middle of 1991 they 
appear smaller than in the two other years, for all latitude south of 30°N. A possible explanation for this 
discrepancy is the effect of the Pinatubo aerosols on the NDVl TOA signal (Stowe et al., 1991). In June 1991, 
Pinatubo eruption injected a large amount of aerosols in the stratosphere, particularly in the inter tropical zone. 
The most direct effect of this larger optical depth of the atmosphere is a decrease of NDVP® A for the 
corresponding area, while we do not expect rapid effects on the vegetation growth and/or the climate conditions 
in the few months following the eruption. 
3.1.2 Vegetation changes: 
• In the temperate northern latitudes, a peak of GV1 in the summer, with the length of the growing season 
increasing as we go towards lower latitudes. This corresponds to oold-deciduous ecosystems. 
• In the subtropical northern latitudes, a small peak in winter, in opposition of phase (maximum in December) 
corresponding to monsoon-driven ecosystems in eastern orient 
• In the inter-tropical zone, a peak of NPP with a weaker seasonality, roughly in phase with the temperate 
northern hemisphere peak, corresponding to rain forests and savannas. 
• In the subtropical/temperate southern hemisphere, south of 25°S, a peak of GV1, roughly in opposition of 
phase with regard to the northern one. 
3.1.3 Cumulative and compensating effects: 
The global trends of interannual variations are difficult to interpret because of the small number of years 
studied and of the inputs uncertainties. There was an El Niño event in 1986-1987, but in the years 1986 to 
1988, the sensor NOAA-AVHRR was recording data of decreasing quality, due to drifting orbit This orbital 
drift induced increased solar zenith angles, resulting in larger optical path, decreasing signal, and different 
directional effects. Years 1989 and 1990 are 'good years', satellite-wise, and no special climatic event occurred 
during these years. Year 1991 saw two superimposing events, the start of an El Niño episode in early 1991, and 
the eruption of mount Pinatubo in June, which are thought to have opposite effects on the uptake of CO2 by 
terrestrial ecosystems : El Niño may reduce rainfalls on some of the tropical zones, while Pinatubo aerosols 
may decrease heterotrophic respiration by a decrease of temperatures. In addition, Pinatubo aerosols disturb 
satellite measurements. 
3.1.4 Impact of atmospheric correction on GVI: 
Fig.4 illustrates the relative impact of atmospheric correction on the annual averaged GVI in 1989. We can 
observe that atmospheric decontamination has different zonal impact, with stronger effects in most part of the 
southern hemisphere. Indeed, most of the brightest pixel (low difference) are located in the northern 
hemisphere. However atmospheric correction always increase GVI. 
3.2 NPP: 
3.2.1 Latitudinal, seasonal and interannual NPP variations : 
In the same way as for GVI, we present NPP on a "time-latitudinal" diagram, and it appears that the mean 
patterns shown by Fig.2 display strong similarity with Fig.l. Indeed, we also find a high peak in summer for 
the temperate northern latitudes with a lower one in opposition of phase between 10°N and 30°N, and the inter 
tropical zone also shows a weaker seasonality in phase with the temperate northern hemisphere peak. 
Table 1 presents Annual GPP and NPP of the terrestrial biosphere, as computed by our model (unit: 10^ g of 
carbon). 
year 
1986 
1987 
1988 
1989 
1990 
1991 
GPP 
99.1 
93.1 
94.8 
107.9 
108.2 
96.7 
NPP 
69.2 
64.8 
66.5 
75.3 
75.6 
67.7 
satellite 
NOAA-9 
NOAA-11 
3.2.2 Impact of atmospheric correction on NPP: 
After correction, the world annual NPP estimation is raised by 6%, but with a modified spatial distribution. 
Unlike the impart on GVI, the atmospheric decontamination can of course increase locally the estimated NPP, 
but it can also reduce it If we compare Fig.4 with Fig.5 representing (in 1989) the difference (M a / >TOA - 
^/p/jSURF), we notice a correspondence between zones for which A'PP TOA >A7 > P SURf and zones for which the 
relative impart of correction is very low. Indeed, the linear relationship between NDVf rOA/SURF and f(t) (Eq.2)