Light Use Efficiency Setting

The PAL data were produced using the Global Area Coverage (GAC) data of AVHRR (Smith et al., 1997). One GAC pixel is an average of sampled 4 pixels within pixels of five columns and three lines of the original AVHRR data (Townshend and Justice, 1986). This means that one GAC pixel may not represent the most common land cover of the area covered by original 5 by 3 pixels. Some GAC pixels are arranged in a coordinate system having 8 km grid in the PAL data, and the pixel with maximum NDVI during a month is chosen for the monthly compositions. Therefore, the PAL data are temporal and spatial mixtures of different ground objects and vegetation species, and it is difficult to determine the appropriate LUE for each pixel. Even though the IGBP-DIS vegetation map consisted of basic plant types based on plant physiognomy (Running et al., 1994a; Loveland et al., 2000), the compositions of the plant species in the pixels varied. This suggests that using LUEs, which were measured in the field in the literature (e.g. Gower et al., 1999), would not be appropriate without adjustments. Therefore, Awaya et al. (2004) used a constant LUE, which was determined by analysing the relationship between the climatic potential NPP and the sums of products of NDVI multiplied by SR. On the other hand, the plant species distributions show clear geographical gradients. Thus, it is necessary to take into account both the geographical gradient of species and spatiotemporal mixture of various species in the pixels. Therefore, the constant LUE of 0.494gDWMJ—1 was supposed to be an average LUE of global natural vegetation for the PAL data, and was used as a standard of LUE computation in this study (Table 1). Owing to the nature of the GAC data, the appropriateness of the LUE-setting is unclear.

3.3 NPP Estimation 3.3.1 Inter-annual Change

The highest and the lowest annual global NPP were 58.3 and 62.6petagram (Pg) carbon in 1983 and 1998, respectively, for the multiple LUE-setting case. For the case of the constant LUE-setting, the highest and the lowest NPP were 51.0 and 54.8 Pg carbon in 1983 and 1991, respectively (Fig. 4). The constant LUE-setting case was about 12% lower than the multiple

Figure 4: Inter-annual changes in global NPP. The results of multiple and a constant LUE-settings were compared. NPP of a constant LUE case was about 8% smaller than the case of multiple LUE-setting. The slopes of the regression lines show the annual rate of NPP increase.

Figure 4: Inter-annual changes in global NPP. The results of multiple and a constant LUE-settings were compared. NPP of a constant LUE case was about 8% smaller than the case of multiple LUE-setting. The slopes of the regression lines show the annual rate of NPP increase.

1980 1985 1990 1995 2000 Year

1980 1985 1990 1995 2000 Year

LUE-setting case. This shows that the LUE control has a big influence on NPP estimation. The IPCC working group I cites 2 NPP estimations, which are 59.9 and 62.6 Pg carbon per year, in the latest report (Houghton et al.,

2001). These values were similar to the NPPs estimated by the multiple LUE settings, but greater than NPPs estimated by the constant LUE setting. Intensive validation would be necessary to know the actual terrestrial NPP distribution.

Although the atmospheric CO2 concentration is increasing steadily, its rate of increase varies from month to month. The rate of increase appears to be influenced by the CO2 uptakes of terrestrial and ocean ecosystems. Therefore, the rate of changes in CO2 concentration (hereafter, CO2 change rate) may show a similar inter-annual pattern to NPP. According to the atmospheric CO2 measurements, the CO2 change rate was greater in 1983, early 1988, early 1994 and early 1998 than in other periods. On the other hand, the rate was smallest in early 1992, and was smaller in early 1990, 1996 and 1999 than in other periods (Meteorological Agency of Japan, 1999,

2002). A greater CO2 change rate in these periods than in other periods suggests less uptake of CO2 by ecosystems (smaller NPP), and a smaller change rate suggests the opposite. Valleys of NPP appeared in 1982, 1983, 1988, 1993 and 1996, and peaks appeared in 1985, 1986, 1991 and 1998 (Fig. 4). Although the cases of 1996 and 1998 didn't agree with the atmospheric CO2 change rate, our NPP estimation closely resembled to the inter-annual change trend of the atmospheric CO2.

The regression coefficients in Fig. 4 suggest that global NPP was increasing about 0.1 Pg annually (P<0.1) for the case of multiple LUEsettings and about 0.08 Pg carbon annually (not significant) for the case of constant LUE-setting. These results mean that global NPP increased about 3% during 20 years in the 1980s and 1990s. The carbon fixation by the terrestrial ecosystem was estimated at —0.2 + 0.7 per year in the 1980s and —1.4 + 0.7 Pg per year in the 1990s by atmospheric CO2 analysis (Houghton et al., 2001). This suggests that carbon fixation by terrestrial ecosystem was increasing roughly 0.12 Pg annually, and our NPP estimation showed a similar result especially for the case of multiple LUE-settings. Though our estimates of global NPP seem to be reasonable, the accuracy of the corrected PAL and NCEP/NCAR data should be validated carefully to make more accurate NPP estimation possible.

3.3.2 Regional Distribution

NPP was very small in the arid areas (the Sahara, the Gobi and Western Australia), high mountains (the Himalayas, the Tibet plateau, the Andes and the Rockies) and latitudes higher than 60° north as reported in the literature (Chapin III et al., 2002; Larcher, 2003). NPP was obviously greater in Europe than other areas in the same latitudes, and NPP was greater in the coastal regions than inland in low latitudes. NPP appeared to be high in agricultural areas in the eastern USA, eastern China and Europe (Fig. 5). The greatest NPP appeared not in the tropical forests but in the tropical deciduous forests or savannas. The tropical rain forests have the greatest NPP and the savannas have about a half the tropical rain forests according to the literature (Chapin III et al., 2002; Larcher, 2003). Thus our results were different from those obtained in previous ecological studies for subtropical and tropical vegetation for the case of multiple LUE setting. However, the constant LUE approach gave monthly NPP maps with more

-150.0 -120.0 -90.0 -60.0 -30.0 0.0 30.0 60.0 90.0 120.0 150.0

-150.0 -120.0 -90.0 -60.0 -30.0 0.0 30.0 60.0 90.0 120.0 150.0

-150.0 -120.0 -90.0 -60.0 -30.0 0.0 30.0 60.0 90.0 120.0 150.0

-150.0 -120.0 -90.0 -60.0 -30.0 0.0 30.0 60.0 90.0 120.0 150.0

-150.0 -120.0 -90.0 -60.0 -30.0 0.0 30.0 60.0 90.0 120.0 150.0

0 300 600 900 1200 1600 1800 NPP (g Carbon / m2 / year)

Figure 5: The results of NPP estimation by the multiple LUE settings. The minimum appeared in 1983 (above), and the maximum appeared in 1998 (below). The greatest NPP appeared in tropical deciduous forest or savannas. (for colour version; see Colour Plate Section)

0 300 600 900 1200 1600 1800 NPP (g Carbon / m2 / year)

Figure 5: The results of NPP estimation by the multiple LUE settings. The minimum appeared in 1983 (above), and the maximum appeared in 1998 (below). The greatest NPP appeared in tropical deciduous forest or savannas. (for colour version; see Colour Plate Section)

similar seasonal changes to NPP maps by ecological models in Cramer et al. (1999). The geographical distribution of NPP obtained by the constant LUE approach was more similar to that obtained by an ecological model (Ito and Oikawa, 2002) than that obtained by the multiple LUE settings (Fig. 5). Although the LUEs were based on the eco-physiological knowledge, the results seemed less reasonable than the constant LUE. An appropriate LUEsetting method for the PAL data should be developed to make more precise estimation possible as described previously.

3.3.3 Regional Variation of Inter-annual Changes

The inter-annual change rates of NPP varied region by region. However, NPP appeared to be increasing in many parts of the world (Fig. 6). Myneni et al. (1997) reported that plant growth was increasing between 1981 and 1990 due to lengthening of the active growing season, and the greatest increase existed between 45°N and 70°N based on an analysis of NDVI images. Although our analysis showed a great NPP increase in East Europe, the rate was not the highest. The greatest rate of increase appeared in southern India (12°N, 79°E) and in south-western Australia (28.5°S, 125.5°E). On the other hand, the greatest rate of decrease appeared in Bolivia (19.5°S, 60°W) and in the Southern Angola (14.5°S, 20.5°E). Inter-annual changes in T, SW, SR and the corrected NDVI were checked in these four areas. No clear trend was

Figure 6: Regional changes of NPP. The rate of NPP change in each region was estimated by a linear regression analysis for the results of the multiple LUE-setting for 17 vegetation classes. Clear increases and decreases of NPP were detected in some areas around semiarid zones. Grey areas show extremely low NPP and were omitted in the regression analysis. (for colour version; see Colour Plate Section)

Figure 6: Regional changes of NPP. The rate of NPP change in each region was estimated by a linear regression analysis for the results of the multiple LUE-setting for 17 vegetation classes. Clear increases and decreases of NPP were detected in some areas around semiarid zones. Grey areas show extremely low NPP and were omitted in the regression analysis. (for colour version; see Colour Plate Section)

—Bolivia --□--Angola --♦- India ----A---Australia

—Bolivia --□--Angola --♦- India ----A---Australia

' 0 24 48 72 96 120 144 168 192 216 1982 84 86 88 90 92 949 69 8 00 Months since January of 1982 / Year

Figure 7: Inter-annual changes of NDVI in areas with the greatest rate of change of NPP. NDVI decreased consistently in Bolivia and Angola, but increased slightly in India.

' 0 24 48 72 96 120 144 168 192 216 1982 84 86 88 90 92 949 69 8 00 Months since January of 1982 / Year

Figure 7: Inter-annual changes of NDVI in areas with the greatest rate of change of NPP. NDVI decreased consistently in Bolivia and Angola, but increased slightly in India.

observed in the NCEP/NCAR parameters except for SW and T in Angola. However, NDVI has tended to decrease in Bolivia and Angola, and has tended to increase in India (Fig. 7). Long-term trends in NDVI seemed to have a strong influence in the regional variations of inter-annual NPP changes, although no clear relationship between NPP and the parameters was found in south-western Australia. The greatest NPP changes appeared in semiarid areas where precipitation changes have significant impacts on vegetation growth. The data in Fig. 6 suggest that changing precipitation patterns have had a greater impact in semiarid areas than an impact in boreal areas by increasing air temperature on vegetation growth. Although NDVI values are also affected by factors such as land cover change, biomass burning and so on, the NDVI images appear to show the effect of changing precipitation patterns on vegetation growth in semiarid areas, which was unclear in the reanalysis data. Therefore, NDVI appears to be a good indicator of global change in vegetation.

Was this article helpful?

0 0

Post a comment