2.3 Back trajectory
In order to precisely analyze time series CO₂ concentration is required to know the flow of air mass that carried CO₂ gas to the monitoring station. In this study, the back trajectory of air mass at the height of 1500 m for everyday was calculated based on the meteorological data provided by the ECMWF and the model developed by the NIES. The back trajectory data set includes a set of latitudes and longitudes of the air mass at 73 points from hateruma to the source point three days before the monitoring day. Figure 3 shows an example of back trajectory. Back trajectory data was used to calculate the NDVI distribution along the path of the air mass to hateruma.

Fig. 3 An Example of back trajectory
3.Correlation analysis between NDVI and CO₂ concentration
Relation between CO₂ concentration at the hateruma station and vegetation cover conditions around the station was investigated by correlation analysis between them. As for the vegetation cover condition, the distribution of NDVI was used. First, NDVI values are averaged in a circular region around the monitoring station, and next, NDVI values are selectively averaged along the back trajectories of air mass to the station.

3.1Average NDVI in a circular region
The average monthly NDVI value around the hateruma station was calculated for the circular areas with different radii of 100km, Correlation between the average monthly NDVI value and CO₂ concentration of corresponding month was calculated for each circle to evaluate the global relation between them. Figure 4 illustrates an example of the circular area with a radius of 500km, and Fig. 5 shows the correlation for it (R2=0.585). Also Table 1 summarizes the coefficient of correlation between CO₂ concentration and the average NDVI for each circle.

From these results it is shown that the CO₂ has negative correlation between the NDVI values, and that the CO₂ concentration is low for the highly vegetation areas. Also Table 1 shows that the correlation coefficients are quite low for the cases where the percentage of land cover areas in each circle is low. It implies that the CO₂ concentration has relation primarily with vegetation cover conditions over land. In this analysis, however, it shows that the correlation is not stable. It might be party because the flow of CO₂ gas to the station is not considered. Then the correlation between the CO₂ concentration and the average NDVI was analyzed with back trigectory of air mass.

Fig.4 The average area of NDVI value



Fig. 5 the subtractive correlation


Table 1 the coefficient of correlation (R²) between CO₂ and NDVI in 1996 and 1997 for each radius circle

Radius	100	500		1500	2000
Coefficient ofCorrelation (In 1996)	0.163	0.311	0.361	0.263	0.252
Coefficient ofCorrelation (In 1997)	0.441	0.420	0.585	0.518	0.464
The percentage of land area (in 96 and 97)	1.70%	5.50%	19.55%	25.95%	30.70%