Carbon Dioxide Concentration and Flux in an Urban Residential Area in Seoul, Korea

Carbon dioxide (CO₂) emissions from urban areas play an important role as an atmospheric CO₂ source (Idso et al., 1998; IEA, 2010). The exchange of CO₂ between the atmosphere and underlying surface over urban areas is mostly governed by anthropogenic emissions originated from road traffic, local heating fueled by natural gas, oil, or coal, biogenic processes such as photosynthesis, soil and plant respiration, and meteorological variables such as wind speed, temperature, solar radiation, and atmospheric stability (Florkowski et al., 1998; Day et al., 2002; Koerner and Klopatek, 2002; Clark-Thorne and Yapp, 2003; Pataki et al., 2003; Grimmond et al., 2004; Vogt et al., 2006; Miyaoka et al., 2007; Schimidt et al., 2008; Matese et al., 2009; Kordowski and Kuttler, 2010; Park et al., 2013).

Recent studies on CO₂ flux in urban areas have focused on cities of developed and developing countries in the midlatitudes, such as Chicago (Grimmond et al., 2002), Tokyo (Moriwaki and Kanda, 2004), Vancouver (Walsh et al., 2004), Basel (Rotach et al., 2005), Melbourne (Coutts et al., 2007), Mexico City (Velasco et al., 2005; 2009), Essen (Henninger and Kuttler, 2010), and Beijing (Liu et al., 2012). These studies have found that urban CO₂ fluxes possess diurnal variation, with a maximum during the morning and a minimum at night (Velasco et al., 2005; Coutts et al., 2007) often correlated with traffic patterns near the measurement site (Soegaard and Møller-Jensen, 2003; Miyaoka et al., 2007; Büns and Kuttler, 2012).

South Korea was the 10th largest anthropogenic CO₂ emitting country in 2008 (IEA, 2010). The annual mean CO₂ concentration in the year 2010 at the Anmyeon-do site, one of the World Meteorological Organization's (WMO) Global Air Watch (GAW) sites, was 394.5 ppm (parts per million), with an annual rate of increase of 2.12 ppm yr_-1 since 1999 (KMA, 2011), which is higher than the global mean CO₂ concentration of 388.6 ppm. (Chung et al., 2007) found that the CO₂ concentration in Seoul, the capital city of South Korea, whose population exceeds 10 million, was higher than those in surrounding large cities such as Suwon and Incheon. There have been some estimates of anthropogenic CO₂ emissions using fuel consumption data in Seoul (Lee et al., 2002). However, direct measurements of CO₂ concentrations and fluxes in this megacity are not yet available.

The purpose of the present reported study was to investigate the seasonal and diurnal variations of CO₂ concentrations and fluxes at Boramae, a typical urban residential area of Seoul, Korea using measured CO₂ concentrations and meteorological variables, and to examine the results in the context of fuel consumption, traffic volume, and differential advection of CO₂ due to inhomogeneous emission rates.

The Boramae site [(37^°29.3′N, 126^°56.0′E), 47 m MSL] is located in a residential area of Gwanak-gu District in the southwestern part of Seoul (Fig. 1). It is surrounded by two- to three-story residential buildings, whose heights are 6-10 m. A small area of hilly terrain [peak: Guksabong (179 m) in Sangdo Park] is located 250 m to the northeast, and a 25-m-wide paved road is 50 m to the southwest of the site (Fig. 2). The vegetation cover near the site, except in Sangdo Park, is less than 10% during the summer season. The site is part of Boramae-dong, whose population is 30 321, which had an area of 0.78 km² in 2011 (http://stat.seoul.go.kr).

Figure 1.  The geographical locations of (a) Seoul and the Anmyeondo site in Korea and (b) the Boramae site in Seoul, including the topography. The border of Seoul is indicated by the thick red line.

Figure 2.  (a) Satellite image and (b) land-use types near the Boramae site.

The hourly traffic volume near the site shows a minimum of 890 cars per hour at 0400-0500 LST and a maximum of 2860 cars per an hour during rush hour (0800-0900 LST)(Fig. 3a, http://stat.seoul.go.kr), with an almost constant traffic volume of around 2500 cars per an hour from 0900-2000 LST. The monthly mean daily total traffic volume ranged from 47 000 cars per a month in April to 61 000 cars per a month in November near the site in 2010 (Fig. 3b).

In 2011, liquefied natural gas (LNG), petroleum, and coal constituted around 90%, 8%, and 2% respectively of energy consumption sources for heating and cooking in residential and commercial regions of Seoul (http://www.kemco.or.kr). For LNG, a total volume of 5.4× 10⁶ m³ was consumed at Boramae-dong from February 2011 to January 2012 (http://www.kogas.co.kr). Among LNG uses, the contribution of domestic heating showed a large seasonal range, with a maximum value of 1.1 × 10⁶ m³ in February and a minimum of 0.03× 10⁶ m³ in August, and whose seasonal variation pattern was negatively correlated to that of air temperature (not shown). That is, LNG use in August was only around 3% of that in February (Fig. 3c). On the other hand, LNG use for cooking showed a smaller seasonal range than for heating, and thus the ratio of the contribution of cooking to total LNG use also showed a seasonal variation, with less than 10% in winter and more than 50% in summer (Fig. 3c).

Figure 3.  (a) Diurnal variation of annual mean traffic volume (car h_-1); (b) the monthly mean daily total traffic volume (car d_-1); and (c) monthly total LNG use for heating and cooking at Boramae-dong from February 2011 to January 2012.

To understand the source area affecting the CO₂ flux at the site, the flux footprint function f(x) at upstream distance x was estimated using an approximate analytical model suggested by (Hsieh et al., 2000):

where k is the von Karman constant (k=0.4). The constants D and p were adopted following the values suggested by (Hsieh et al., 2000); namely:

D=0.97; p=1 for near neutral and neutral conditions;

D=2.44; p=1.33 for stable condition.

The length scale z_u is a function of the measurement height z and the surface roughness length z₀ defined as

The Monin-Obukhov length scale L is defined as

where g is the gravitational acceleration, u_* is the friction velocity, andis the kinematic heat flux.

The surface roughness length was determined as 0.5 m by considering the surrounding building structures and heights (Davenport et al., 2000). The footprint of 80% at sensor height was less than 500 m under neutral and unstable stratification, with a small value in the morning and a large value in the afternoon, which might have been mainly due to the variation in friction velocity.

A 3D sonic anemometer (CSAT, produced by Campbell Scientific Inc.) and an open-path gas analyzer (LI7500, produced by Li-Cor) were installed at a height of 2.5 m above the rooftop of a 15-m-high building at the Boramae site. Because there are no buildings taller than 10 m within a radius of 100 m from this site, and the size of the rooftop is small (3× 7 m²), the wind flow distortion by surrounding environments was negligible.

Three-component wind speed (u: eastward wind speed; v: northward wind speed; w: upward wind speed), sonic temperature T, and CO₂ (C) and H₂O concentrations were sampled at 10 Hz from a sonic anemometer and an open-path gas analyzer for the period from February 2011 to January 2012. Kinematic sensible heat flux (SHF) and CO₂ flux (F _CO2) were then computed using the following procedures:

• Remove soft spikes that are large and short-lived departures from the period means (Schmid et al., 2000);

• Calculate the 30-min mean and standard deviation for each variable;

• Extract the linear-detrended fluctuation value by subtracting a linear mean trend function from the raw values. That is to say, the linear mean trend function for any variable X(t) is determined as a form where the coefficient a is the slope and the coefficient b the value at t=0; then the linear-detrended fluctuation values are calculated as (Lee et al., 2004; Park and Park, 2006).

• Compute the SHF and CO₂ flux (F _CO2) every 30 min using the eddy covariance method (; ). When the flux flows upward, it is defined as positive. The "Webb correction" is applied in order to take into account the effect of density fluctuation during the process of computing a CO₂ flux (Webb et al., 1980; Fuehrer and Friehe, 2002).

• Apply a second phase of quality control. All measured or derived variables are submitted to a plausibility test and are rejected if they fall outside statistically defined constraints for each variable (Schmid et al., 2000).

Table 1 shows the ratio of selected data after the above quality assurance test on all available data for each month. The average ratio was 76% for whole analyzed periods with high values in winter and low values in summer. The low selection ratios in summer were mainly due to the frequent occurrence of rain (KMA, 2012).

The mixing height h _m(t) at any time step t was estimated using the equations

under neutral stratification ( K m s_-1), and

under unstable stratification ( K m s_-1), where an empirical coefficient c of 0.14 was used (Arya, 1981); f is the Coriolis parameter; and Γ_f is the lapse rate of free atmosphere. Table 1 shows the monthly mean Γ_f obtained from the nearest WMO upper air measurement site of Osan (37.1^°N, 127.1^°E)

Figure 4 represents the wind roses classified by 16 wind-direction and five wind-speed categories at the site using 30-min averaged wind speed and direction data. Winds whose speeds were less than 0.2 m s_-1 were classified as "calm". Southwesterly to northwesterly winds were dominant (67%) in spring (Fig. 4b), southeasterly to southwesterly winds (59%) in summer (Fig. 4c), southeasterly and westerly to northwesterly winds (57%) in autumn (Fig. 4d), and westerly to northwesterly winds (54%) in winter (Fig. 4e). Annually, the southwesterly to northwesterly winds were dominant (55%) at this site (Fig. 4a). From the wind speed point of view, wind-speed category II (1-2 m s_-1) was dominant in spring (40%), summer (44%), and winter (42%), while wind-speed category I (0.2-1 m s_-1) was dominant in autumn (48%). Annually, wind-speed category II prevailed at this site (Fig. 4a).

Figure 4.  Wind roses for (a) the whole year, (b) spring, (c) summer, (d) autumn, and (e) winter at the Boramae site.

Figure 5 shows the diurnal variations of seasonal mean SHF. The seasonal averaged hourly mean SHF had a large positive value during the day, while a small positive value during the night for all the year. The seasonal mean daily maximum SHF had a maximum of 0.12 K m s_-1 at 1300-1400 LST in spring and a minimum of 0.08 K m s_-1 at 1200-1300 LST in winter. The lower daily maximum SHF in summer than that in spring might have been due to the larger cloud amount and more rain in summer (6.6/10, 1005.7 mm) than in spring (4.7/10, 212.6 mm) in spite of the high solar elevation angle in summer (KMA, 2012). And the small positive SHF during the night might have mainly been due to anthropogenic heat emissions, whose annual mean is 55 W m_-2 in Seoul (Lee et al., 2009), as well as heat storage by underlying buildings during the day and re-radiation to the atmosphere during the night, which is different from the SHF over rural or bare surfaces (Arya, 1999; Liu et al., 2012).

Figure 5.  Diurnal variations of seasonal mean kinematic heat flux measured at the Boramae site in spring, summer, autumn, and winter.

Figure 6 shows a diurnal variation of seasonal mean mixing height estimated using Eqs. (4a) and (4b). The mixing height showed a minimum value of about 700 m at 0500-0700 LST and a maximum value in the late afternoon or early evening. As the SHF was positively large during the day, the mixing height also grew to maximum values of 1460 m, 1430 m, 1210 m, and 1140 m in spring, summer, autumn, and winter, respectively. When the SHF became small or negative, the mixing height fell abruptly. The layer above the mixing height became neutral and turned into a residual layer due to the absence of further energy sources from the surface (Stull, 1988).

Figure 6.  Diurnal variations of seasonal mean mixing height at the Boramae site in spring, summer, autumn, and winter.

Figure 7 shows a seasonal variation of observed monthly mean CO₂ concentration and flux at the site. The annual mean CO₂ concentration was 750 mg m^-3, with a maximum value of 827 mg m^-3 in January and a minimum of 679 mg m^-3 in August. The annual mean value at this site was higher than that of 735.5 mg m^-3 measured at the WMO GAW site of Anmyeon-do in Korea, suggesting that the CO₂ concentration in urban sites is higher than that at rural or background sites. Generally, the seasonal variation of CO₂ concentration was very similar to those from other urban and rural sites (Dettinger and Ghil, 1998; Miyaoka et al., 2007; Park et al., 2013).

Figure 7.  Annual variations of the observed monthly mean (a) CO₂ concentration (mg m^-3) and (b) CO₂ flux (mg m_-2 s_-1) for the period February to December 2011 and January 2012. SD stands for standard deviation.

The annual mean CO₂ flux at this site was 0.45 mg m^-2 s^-1, with a maximum value of 0.91 mg m^-2 s^-1 in January and a minimum of 0.19 mg m_-2 s_-1 in June. The seasonal variation of CO₂ flux might be similar to the seasonal variation of LNG use near the site (Fig. 3c). The CO₂ flux in summer was still positive, implying that anthropogenic and natural CO₂ emissions are greater than CO₂ uptake by vegetation in urban residential areas. Moreover, the annual mean flux was about three times larger than the soil respiration rate (0.14 mg m_-2 s_-1) from an urban forest region in Seoul (Joo et al., 2012), suggesting that the anthropogenic emissions of CO₂ in urban areas could be a very important factor in understanding the recent increase of CO₂ concentrations in Korea.

In order to compare the present result with others, the annual mean CO₂ fluxes measured at various urban cities around the world are listed in Table 2. The annual mean CO₂ flux at our residential site in Seoul was smaller than those reported for London (Helfter et al., 2011), Montreal (Bergeron and Strachan, 2011), and Beijing (Liu et al., 2012) city centers, but higher than those reported for a suburban site in Melbourne (Coutts et al., 2007), and urban sites in Essen (Kordowski and Kuttler, 2010) and Lodz (Pawlak et al., 2011). Many studies have shown that CO₂ fluxes in urban residential areas tend to be lower than those in urban city centers.

The diurnal variation of seasonal mean CO₂ concentration (Fig. 8a) showed a maximum at 0700-0900 LST and a minimum at 1400-1600 LST, which might be attributable to the CO₂ flux (Fig. 8b), mixing height (Fig. 6), and traffic volume (Fig. 3a). The seasonal mean CO₂ concentration in winter (807 mg m^-3) was higher than that in summer(685 mg m^-3), and its diurnal range in winter (115 mg m^-3) was also larger than that in summer (57 mg m^-3). And the diurnal variations of CO₂ concentration in spring and autumn showed almost the same patterns. The seasonal mean daily maximum CO₂ concentration was recorded as 868 mg m^-3, 794 mg m^-3, 786 mg m^-3, and 713 mg m^-3 in winter, autumn, spring, and summer, respectively.

Figure 8.  Diurnal variations of seasonal mean (a) CO₂ concentration (mg m^-3) and (b) CO₂ flux (mg m_-2 s_-1) in spring, summer, autumn, and winter at the Boramae site.

The diurnal variation of seasonal mean CO₂ flux (Fig. 8b) showed two maxima at around 0700-0900 LST and 2100-2400 LST, and two minima at 1100-1500 LST and 0300-0500 LST. The seasonal mean CO₂ flux in winter (0.89 mg m^-2 s^-1) was 4.7 times larger than that in summer (0.19 mg m^-2 s^-1), and the diurnal range of CO₂ flux in spring (0.45 mg m^-2 s^-1) was as much as twice that in summer (0.24 mg m^-2 s^-1). The seasonal mean daily maximum(minimum) CO₂ flux was 1.09 (0.72) mg m^-2 s^-1, 0.66 (0.21) mg m^-2 s^-1, 0.51 (0.18) mg m^-2 s^-1, and 0.33 (0.08) mg m^-2 s^-1 in winter, spring, autumn, and summer, respectively.

In order to uncover the contributions of CO₂ flux and mixing height to the diurnal variation of CO₂ concentration, the seasonal mean hourly CO₂ concentration deviation from the seasonal mean CO₂ concentration , was fitted to a regression equation of dC _CO2=α F _CO2+β h _m+χ, where F _CO2 is the seasonal mean hourly CO₂ flux and h _m is the mixing height The coefficients α,β, and χ will be determined empirically.

Coefficients of best-fit regression equations, correlation coefficient (R²), and the CO₂ concentrations contributed by each factor for each season are listed in Table 3. The contributed CO₂ concentration was calculated by multiplying the mean value by its coefficient. The coefficients α and β implied that the CO₂ concentrations were positively correlated with CO₂ flux, and negatively with mixing height for all seasons. Figure 9 shows a scatter diagram between observed monthly mean hourly CO₂ concentration and the best-fit regressed CO₂ concentration with CO₂ flux and mixing height. The low correlation coefficient, especially in winter, might imply that other factors such as horizontal advection or storage are needed in order to parameterize the CO₂ concentration more exactly.

Figure 9.  Scatter plots between observed CO₂ concentration deviation (dCon_obs, mg m^-3) and parameterized CO₂ concentration deviation (dCon_par, mg m^-3) in terms of CO₂ flux (F _CO2) and mixing height (h _m) in (a) spring, (b) summer, (c) autumn, and (d) winter. The parameterized equation and the correlation coefficient are indicated.

To investigate the relationship between CO₂ flux and LNG use, carbon fluxes were computed from the observed CO₂ flux (F _C,OBS) and from LNG use (F _C,LNG) using the equations

and

respectively, where m _C is the molecular weight of carbon (12), m _CO2 is the molecular weight of CO₂ (44), L is the volumetric LNG consumption, ρ is the density of LNG (0.75 kg m^-3), r _C is the average carbon content in LNG (0.757), A is the LNG consumed area, and ∆ t is the data period.

Figure 10 shows the monthly mean carbon flux estimated from LNG use (F _C,LNG) and from the observed CO₂ flux (F _C,OBS) at this site. The two carbon fluxes had similar seasonal variations, with a small value in summer and a large value in winter, but the observed carbon flux was larger than that estimated from LNG consumption throughout the year. Specifically, the carbon flux estimated from the LNG consumption was responsible for around 67% of total carbon fluxes with small contributions in summer and large contributions in winter; and the remaining carbon flux might have been contributed by traffic and other energy uses.

Figure 10.  Monthly mean carbon flux (mg m_-2 s_-1) estimated using the monthly total LNG consumption (F _C,LNG) and observed monthly mean CO₂ flux (F _C,OBS) for winds from the residential region at the Boramae site.

In order to analyze the effect of different land-use types on the CO₂ concentration and flux at the site, the wind direction was divided into two categories: one was from the commercial and road region (120^°-300^°; hereafter CR), and the other from the residential region (300^°-360^° and 90^°-120^°; hereafter RS) (Fig. 2b). Figure 11 shows the monthly mean CO₂ concentration and flux for the wind cases blown from the RS and CR regions. The monthly mean CO₂ concentration for winds coming from the two regions showed almost the same seasonal variations for the period of March to August 2011, but that from the CR region was higher than that from the RS region in the rest of the months by 30-50 mg m^-3 (Fig. 11a). Meanwhile, the monthly mean CO₂ flux for winds from the RS region was more than 1.5 times larger than that from the CR region, except for the summer season. In particular, the monthly mean value for winds from the RS region was as much as 2.8 times that from the CR region in February 2011 (Fig. 11b). The high CO₂ flux for winds from the RS region seemed to reduce the CO₂ concentration at the site (Figs. 11a and b)

Figure 11.  Seasonal variations of monthly mean (a) CO₂ concentration and (b) CO₂ flux for winds from the RS region (residential) and the CR region (commercial/road) at the Boramae site.

To understand the temporal variation pattern of CO₂ concentration and flux due to local inhomogeneity of land-use types, two time periods were selected: the summer season (June-August) and winter season (December-February). Figure 12 shows the diurnal variations of CO₂ concentration, flux, and wind speed for winds from the two different regions in summer. The diurnal variation patterns of CO₂ concentration for winds from the CR region and the RS region were nearly the same (Fig. 12a), but the daily mean CO₂ concentration when the wind was coming from the CR direction (693 mg m^-3) was higher than that from the RS direction (681 mg m^-3). Also, the daily maximum CO₂ concentration for winds from the CR region occurred at 0700-0800 LST when the traffic volume was at a maximum (Fig. 3a), which was 1 h later than that (0600-0700 LST) from the RS region. Similarly, the difference in CO₂ concentration between them was large during the night (>20 mg m^-3) and small during the day (<10 mg m^-3).

On the other hand, the CO₂ flux for winds from the CR direction and RS direction showed two maxima at 0700-0800 LST and 2000-2100 LST and two minima at 1200-1500 LST and 0300-05 LST, but a distinctly different diurnal range of 0.00-0.37 mg m_-2 s_-1 and 0.10-0.31 mg m_-2 s_-1 for winds from the RS and CR regions, respectively (Fig. 12b). In particular, during the middle of the day (1200-1500 LST), the CO₂ flux with winds from the RS region was near zero, suggesting that the anthropogenic CO₂ emitted by domestic heating and traffic volume was removed by photosynthesis of the underlying vegetation, whose fractional cover was at a maximum in summer. Also, the weak easterly winds coming from the RS region resulted in a short footprint and low ventilation effect on the atmospheric CO₂ concentration compared to the case of westerly winds from the CR region (Fig. 12c).

Figure 12.  Diurnal variations of (a) CO₂ concentration, (b) flux, and (c) wind speed averaged during the period from June to August 2011 for winds from the RS (residential) and CR (commercial/road) regions at the Boramae site.

Figure 13 shows the diurnal variations of CO₂ concentration, flux, and wind speed for the winds blown from the two different regions during the winter period. The diurnal patterns of CO₂ concentration for winds from the CR and RS regions were nearly the same (Fig. 13a), but the daily mean CO₂ concentration for winds from the CR region (846 mg m^-3) was higher than that from the RS region (796 mg m^-3). Also, the difference in CO₂ concentration between the two regions was large during the night and morning (>60 mg m^-3), and small in the afternoon (<20 mg m^-3). This diurnal pattern was similar to that in summer (Fig. 13a).

Figure 13.  The same as Fig. 12, except for the period of February, December 2011, and January 2012.

On the other hand, the CO₂ flux for winds from the two different regions did not show a distinct diurnal variation, but that from the RS region (1.23 mg m_-2 s_-1) was much higher than that from the CR region (0.68 mg m_-2 s_-1) during the winter season (Fig. 13b), which was opposite to those during the summer season (Fig. 12b). This large CO₂ flux difference between the summer (Fig. 12b) and winter (Fig. 13b) seasons for winds from the RS region might have been related to the large seasonal difference in LNG consumption (Fig. 3c), while the small difference between the summer (Fig. 12b) and winter (Fig. 13b) for winds from the CR region might have been related to the small seasonal difference of traffic volume on the road near the site (Fig. 3b). Moreover, the relatively strong westerly winds in the afternoon (1300-1500 LST) from the CR region (Fig. 13c) might have reduced the CO₂ concentration in Fig. 13a by increasing the ventilation effect at the site.

In short, the CO₂ concentration for winds from the CR region was higher than that from the RS region, although the CO₂ flux for winds from the CR region was smaller than that from the RS region. This may have resulted from the local horizontal advection from commercial regions, with tall building complexes and heavy traffic, located in the southern part of the site.

The diurnal and seasonal variation of the CO₂ concentration and flux using a 3D sonic anemometer and an open-path CO₂ and H₂O analyzer at Boramae, located in a residential area in Korea for the period from February 2011 to January 2012 were analyzed with respect to traffic volume, LNG use, and wind direction.

The monthly mean CO₂ concentration showed a minimum value of 679 mg m^-3 in August and a maximum of 827 mg m^-3 in January. Meanwhile, the monthly mean CO₂ flux also showed a minimum of 0.19 mg m_-2 s_-1 in June and a maximum of 0.91 mg m_-2 s_-1 in January. High concentrations and fluxes in winter were found to be mainly contributed to by heavy LNG use in the residential area.

Accordingly, seasonal mean hourly CO₂ concentrations showed a maximum at 0700-0900 LST and a minimum at 1400-1600 LST, while seasonal mean hourly CO₂ fluxes showed two maxima at 0700-0900 LST and 2100-2400 LST, and two minima at 1100-1500 LST and 0300-0500 LST. Also, the seasonal mean daily CO₂ concentration range was found to be between a minimum of 57 mg m^-3 in summer and a maximum of 115 mg m^-3 in winter, while the CO₂ flux range was between a minimum of 0.24 mg m_-2 s_-1 in summer and a maximum of 0.45 mg m_-2 s_-1 in spring. In addition, the diurnal variation of CO₂ concentration was found to be positively correlated with the CO₂ flux and negatively correlated with the mixing height, with correlation coefficients (R²) higher than 0.79, while the carbon flux computed by LNG use was responsible for 67% of the observed carbon flux at the site.

To understand the effect of different land-use types near the site on the CO₂ concentration and flux, the data were divided into two categories according to wind direction: one from the residential (RS) region, and the other from the commercial and road (CR) region. Both CO₂ concentrations in summer (June-August) and in winter (December-February) for winds from the CR region were higher than those for winds from the RS region, while the CO₂ flux for winds from the RS region showed a higher daily range than those from the CR region. Also, in winter, the CO₂ concentration for winds from the CR region was higher than that from the RS region, while the CO₂ flux from the CR region was smaller than that from the RS region.

It was also found that the CO₂ concentration was mainly affected by diurnal and seasonal variations of mixing height, CO₂ flux, and surface complexity, while the CO₂ flux was controlled by diurnal and seasonal variations of road traffic, energy use, and differential advection according to the wind direction due to the local inhomogeneous land cover.

Although the eddy covariance method over complex surfaces has many random and/or systematic errors (Aubinet et al., 2012), this study will be a useful reference for determining the various factors affecting temporal variation patterns of CO₂ concentrations and fluxes in residential areas of Seoul. Moreover, the methodologies adopted in this study could be applied to analyze and parameterize the CO₂ or other pollutant fluxes, anthropogenic heat flux, or energy fluxes over complex terrain such as urban or suburban areas. Further studies will be very helpful in order to improve urban-related climate models.