Keywords: NDVI, CO₂ concentration, Correlation analysis, Back trajectory of air mass

Abstract
Variations in amplitude and time of a seasonal cycle of carbon dioxide (CO₂) concentration show a relation with a seasonal change in distribution to photosynthetic activities of vegetation. However, the quantitative relation between them has not been clarified yet. In this study time series CO₂ concentration data observed at Hateruma, monitoring station, Okinawa. Japan is analyzed together with the series NDVI data derived from NOAA/AVHRR around East Asia to investigate the quantitative relation between their seasonal variations. The results show that the CO₂ conectionation at hateruma has correlation with the NDVI values averaged around hateruma and also with the NDVI values averaged along the back trajectories of air masses to hateruma.

1.Introdution
It is well known that the seasonal cycle of carbon dioxide (CO₂) concentration has a seasonal change in vegetation distribution due to photosynthetic activities of vegetation. In northern hemisphere, for example, CO₂ concentration is lower in summer since vegetation activities are high in summer. This relation is confirmed in global scale, however, in regional or local scale the relation between CO₂ concentration and vegetation activities are not quantitatively verified yet.

In this study, time series CO₂ concentration data observed at hateruma monitoring station, Japan (Longitude 123.8, Latitude : 24.0) is analyzed together with the time series NDVI dta derived from NOAA/AVHRR around the station in order to investigate the quantitative relation between them. The NDVI (Normalized Difference Vegetation Index) derived from satellite is well known to have a correlation with the fraction of photosynthetically active radiation absorbed by vegetation, and as a result, with photosynthetic activities of vegetation. Statistical correlation analysis was performed for monthly CO₂ data averaged from daily Data and monthly NDVI data averaged over the selected areas. Two cases are tried for NDVI averaging. First, NDVI values are averaged uniformly around the monitoring station, and next NDVI are selectively averaged along the back trajectories of air mass to the station corresponding to the wind vector (Fig. 1)

Fig.1 Back trajectory analysis
2. Data used in the study

2.1. Green House Gase data
Time series GHG concentrations including CO₂ , CH₄, O₂ etc have been observed at hateruma island, Okinawa and at Ochiishi, Hokkaido in Japan as base line data for GHG by National Institute for Environmental Studies (NIES). In this study CO₂ data at hateruma station was used for the correlation analysis with NDVI distribution. Gas monitoring is carried out hourly basis, however, in this study original data was averaged in each month to get monthly data to compare with monthly composite NDVI data. Figure 2 shows an example of time series CO₂ concentration at hateruma station in 1997 which shows typical characteristics of seasonal change.

Fig 2. CO₂ concentration change in 1997
2.2 NDVI data
Time series NDVI images of 1996 and 1997 was obtained from the NOAA/AVHRR data received at two stations operated by the NIES (Kuroshima, Okinawa and Tsukuba. Ibaraki in Japan). They can cover most of East Asian region. In this study monthly composite NDVI data was used for correlation analysis with CO₂ data. Spatial resolution of NDVI data is around 1.1 km and each pixel has a NDVI value scaled from 0 to 255.