Correlation Analysis between Carbon Dioxide Concentration and Vegetation Distribution

Mitsugu Sonu¹, Yasumi Fujinuma², Masayuki Tamura², Yosifumi Yasuoka¹
1:Institute of Industrial Science, University of Tokyo
4-6-1 Komaba, Meguro-ku, Tokyo 153-0085 Japna
Tel: 81-3-5452-6417 Fax 81-3-5452-6417
Email: sone@skl.iis.u-tokyo.ac.jp
2: National Institute for Envionmental Studies
16-2 Onogawa, Tsukuba, Ibaraki, Japan

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.

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

3.2 Average NDVI along back trajectory
The NDVI values along the back trajectory of air mass was averaged for the area of the 100km width from the trajectory. Figure 6 shows an example of the averaging area (the area painted black is the land area within the area along the back trajectory) And Figure 7 shows the monthly variation of NDVI averaged along the back trajectory in 1996 and 1997 which shows the typical variation pattern of vegetation activities. The NDVI variations in Fig. 7 show some negative correlation with CO₂ variations compared with Fig.2.

Fig.6 An Averaging area of NDVI values along the back trajectory



Fig. 7. NDVI variation along back trajectory

4.Conclusions
The relation between the CO₂ concentration and the vegetation cover conditions (NDVI) was investigated. As expected, the results imply that the air mass coming from the ocean has the back ground CO₂ concentration representing global concentration, whereas the air mass coming from land areas has correlation with the local NDIV values. For more quantitative analysis it is required to construct the model describing the movement of air mass and CO₂ absorption by vegetation.

Authors would like to express our thanks for the NIES for providing us the GHG data and NDVI data. We also thank Dr. Katsumoto and Dr. hashimoto of the NIES for their kind help for analyzing the GHG data and back trajectory data.

References:

• R.B.Myneni, et.al (1997) : Increased plant growth in the northern high latitudes from 1981 to 1991, Nature, vol. 386, pp.698-702