Observation and modeling of vertical carbon dioxide distribution in a heavily polluted suburban environment

BAO Zhongxiu,HAN Pngi,ZENG Ning,LIU Di,CAI Qixiang,WANG Yinghong,TANG Guiqian,ZHENG K, an YAO BoCollg o Atmosphri Sins,Chngu Univrsity o Inormation Thnology,Chngu,China; Stat Ky Laoratory o Numrial Moling or Atmosphri Sins an Gophysial Flui Dynamis,Institut o Atmosphri Physis,Chins Aamy o Sins,Bijing,China; Dpartmnt o Atmosphri an Oani Sin,an Earth Systm Sin Intrisiplinary Cntr,Univrsity o Marylan,Collg Park,Marylan,USA; Stat Ky Laoratory o Atmosphri Bounary Layr Physis an Atmosphri Chmistry,Institut o Atmosphri Physis,Chins Aamy o Sins,Bijing,China; Shool o Atmosphri Physis,Nanjing Univrsity o Inormation Sin an Thnology,Nanjing,China; Mtorologial Osrvation Cntr o China Mtorologial Aministration,Bijing,China

ABSTRACT The vertical distribution of carbon dioxide(CO2)is important for the calibration and validation of transport models and remote sensing measurements. Due to the large mass and volume of traditional instruments as well as supporting systems, in-situ measurements of the CO2 vertical profile within the boundary layer are rare. This study used a miniaturized CO2 monitoring instrument based on a low-cost non-dispersive infrared(NDIR)sensor to measure the CO2 vertical profile and meteorological parameters of the lower troposphere (0-1000 m) in southwestern Shijiazhuang, Hebei Province, China. The sensors were onboard a tethered balloon with two processes:the ascending process and the descending process.The results showed that the overall trend of CO2 concentration decreased with height.Weather conditions and CO2 emission sources caused fluctuations in CO2 concentrations. The CO2 concentration varied from morning to afternoon due mainly to the faster spread of air mass during daytime,with strong convections and the accumulation of emissions at night. The low-cost sensor produced results consistent with the traditional gas chromatography method.The Weather Research and Forecasting model could not capture the CO2 profiles well due mainly to the bad performances in boundary layer height and the potential outdated fossil fuel emissions around the experimental site. This experiment is the first successful attempt to observe the CO2 vertical distribution in the lower troposphere by using lowcost NDIR sensors.The results help us to understand the vertical structure of CO2 in the boundary layer,and provide data for calibrating and validating transport models.

KEYWORDS Low cost sensor;co2 vertical profile;tethered balloon;meteorological conditions;non-dispersive infrared(NDIR)

1. Introduction

With the increased concentrations of greenhouse gases in the atmosphere, global warming has become a major scientific and political issue (Le Quéré et al. 2018;Schneider1989).Amongallgreenhouse gases,carbondioxide (CO2) accounts for the largest share (Solomon et al.2007). The concentration of greenhouse gases is rising due largely to anthropogenic activities.Therefore,accurate assessment of the distribution and changes of CO2concentration in the atmosphere is crucial for the formulation of climate policy and prediction of future climate change.

CO2at the ground surface has been observed for a long time. Since the 1970s, the World Meteorological Organization(WMO)has conducted long-term global monitoring of greenhouse gases and reactive gases,accumulating decades of observed data(WMO/GAW 2016).The China Meteorological Administration has started CO2in-situ measurements at Mt. Waliguan in Qinghai Province as one of the WMO/GAW (Global Atmosphere Watch) global background stations and established three WMO/GAW regional stations since 2009 (Fang et al. 2017). However, in-situ measurements of CO2in its vertical distribution are insuffi-cient (Bischof, Fabian, and Borchers 1980; Brenninkmeijer et al.2007;Chan and Kwok 2000;Tolton and Plouffe 2001).VerticalCO2measurementscan be made usinghightowers,tethered balloons,lidar,aircraft,and stratospheric balloons(Li et al.2014).High tower measurements can provide data from the lowest 100-500 m of the atmosphere of the planetary boundary layer, but the range of measured heights is often restricted by the tower height (generally lower than 500 m)(Davis et al.2010;Inoue and Matsueda 2001).A sampling platform mounted on aircraft and stratospheric balloons can measure CO2at higher altitudes(Deutscher et al. 2010; Mays et al. 2009; Nakazawa,Hashida, and Sugawara 2013). However, although these two types of measurements can obtain CO2concentrations inthetroposphere and stratosphere(up to13 km onaircraft and 35 km on stratospheric balloons)(Li et al.2014),they have a higher cost(Machida et al.2008)and a lower vertical resolution (100-200 m) than a tethered balloon (40 m)(Karion et al. 2010). Compared with other vertical CO2observation methods,tethered balloons are not only lowcost and easy to operate (Esteki et al. 2017), but can also continuously observe the vertical distribution of atmospheric CO2in the boundary layer.

In recent years, some low-cost sensors have been proven to be practical,feasible,and suitable for environmental monitoring and have been put into use,such as for pollutant gas (e.g. sulfur dioxide, nitrogen dioxide,ozone) and greenhouse gas (e.g. CO2, methane) measurements (Holstius et al. 2014; Piedrahita et al. 2014;Wang et al. 2015). Among these sensors, the SenseAir®K30 sensor, which is based on non-dispersive infrared(NDIR)technology produced by a Swedish manufacturer(SenseAir),has been verified to be useful for high spatial density CO2monitoring in urban areas (Martin et al.2017;Yasuda,Yonemura,and Tani 2012).

In this study,by using the low-cost K30 sensor carried by a tethered balloon to monitor the vertical CO2distribution,experiments were conducted to understand the vertical distribution of CO2in the boundary layer and provide calibration and validation data for transport models (e.g.the Weather Research and Forecasting (WRF) model) and remote sensing.

2. Materials and methods

2.1. Sampling site

The experiment was conducted at Yuanshi National Meteorological Observing Station, Shijiazhuang (114°30ʹE,37°48ʹN). Shijiazhuang is the capital of Hebei Province,located in the southern part of the North China Plain and at the eastern foothills of the Taihang Mountains,which are low in the southeast and high in the northwest,and it is the economic, cultural, and transportation center of Hebei Province. The experimental site was located 28 km south of the city center with an altitude of 68.4 m, as shown in Figure 1.The mean annual temperature and mean annual precipitation are 13.5°C and 576 mm,respectively.

2.2. Balloon-based CO2 soundings

The CO2vertical profile and meteorological parameters(pressure, temperature, relative humidity) of the lower troposphere (0-1000 m) were measured in the winter of 2019(8-16 January)by a miniaturized CO2monitoring instrument based on NDIR technology. The tethered balloon experiment consisted of two processes(ascending and descending), and the duration of each flight lasted approximately 1-2 h.The average ascending or descending speed was ～0.6 m s−1.The maximum height of the flights was 1000 m(Zhao et al.2019).The experiment procedure is shown in Figure 2.There were two types of tethered balloon experiments: Type 1 involved the balloon rising to a height of 500 m (type 1.1), remaining for about 1.5 h, and then being pulled back to the ground (type 1.2); Type 2 involved the balloon rising to 1000 m (type 2.1) and then being directly pulled back to the ground (type 2.2). During the study period, we carried out a total of 11 experiments, and as the experiments were divided into two processes, i.e. ascending and descending, a total of 22 experiment profiles were obtained. For this paper, we selected 10 typical experiment profiles to explain the vertical distributions and factors controlling such distributions. Table 1 shows the time record of typical experiments.

2.3. Data processing

The CO2concentration was measured by a miniaturized CO2monitoring instrument equipped with a SenseAir K30 sensor with an initial accuracy of ±30 ppm of reading(https://senseair.com/products/flexibility-counts/sen seair-k30; accessed October 2019). After calibration with standard gas and environmental correction,the accuracy of K30 sensor was improved to within ±5 ppm, as compared with the simultaneous and precise greenhouse gas analyzer Picarro G2401 in the laboratory. The potential temperature(θ)was calculated by(Bolton 1980)whereP0equals 1000 hPa,Pis ambient atmosphere pressure,andTkis the Kelvin temperature,andk=0.286.

In experiment 1, the data quality of the miniaturized instrument was verified by comparing with the gas chromatography technology (GC-FID, Agilent 7890A, Santa Clara, California, USA). The GC-measured CO2data were obtained through lab analysis after taking back air samples from air bags onboard the tethered balloon. The planetary boundary layer height (PBLH) was measured by lidar during the experiment period, which was used to verify the potential temperature gradient method.

Figure 1.Location of the experiment site.

2.4. Spatial and temporal resolution simulation of CO2 concentration

High spatial and temporal resolution simulations of CO2concentration were conducted by using WRF-CO2for the period of 0600 UTC 7 January to 0600 UTC 9 January 2019 and from 0700 UTC 13 January to 0700 UTC 15 January 2019. Based on the WRF model with chemistry (WRF-Chem), WRF-CO2is a mesoscale, compressible model that provides passive tracer transport networks and mesoscale weather prediction capabilities(Martin et al.2017).It uses the VEGAS(Vegetation-Global-Atmosphere-Soil) model to simulate urban-scale ecosystem carbon emissions/absorption as biosphere carbon flux and GFS (Global Forecast System) grid data as the meteorological boundary field of the CO2nested simulation.It also uses high-resolution(0.1°×0.1°)fossil fuel emissions hourly grid data (Fossil Fuel Data Assimilation System (FFDAS) data (http://ffdas.rc.nau.edu/Data.html)) in 2015 as the prior, which utilizes the Kaya Identity,a method to estimate emissions based on economic factors,and information on national fossil fuel emissions, satellite-derived nightlights, population density, and power plant information to extract each grid point, to simulate the temporal and spatial distribution of CO2at the observation point(Martin et al.2019). We also used GEOS-Chem output as the initial field and boundary field. GEOS-Chem is a global 3D model of atmospheric chemistry driven by meteorological input from the NASA Goddard Earth Observing System(http://acmg.seas.harvard.edu/geos/). The physical process parameterization schemes of WRF were configured as follows: The forecast model was configured with full physics options. The simulation was conducted using the WSM5 class for microphysics, the RRTM scheme for longwave radiation, the Dudhia scheme for shortwave radiation, the MM5 scheme for the surface layer, the unified Noah land-surface model for the land surface,and the YSU scheme for PBL parameterizations.

Figure 2.Diagram of the experiment.For type 1,the balloon stayed at the height of 500 m for ～1.5 h,and for type 2 the balloon was directly pulled back to the ground.

Table 1.Typical tethered balloon experiments.

3. Results and discussion

3.1. Vertical distribution of CO2 and comparison with traditional method

Figure 3 shows typical vertical CO2profiles obtained during the experiment period.The range of CO2concentration from 0-1000 m is 400-600 ppm,and the overall CO2concentration displays a decreasing trend with the increase in height,which is consistent with the study of Esteki et al. (2017). We used the near-surface CO2and PM2.5data of these 10 profiles for linear regression analysis,and found a positive correlation between CO2and PM2.5, so as to infer that source emissions would affect the concentration of CO2. Due to the combined influences of emission sources and meteorological conditions, the level of CO2concentration changed across different days.

This is the first time that we have used miniaturized instruments for vertical monitoring in the boundary layer, and so we compared the results with traditional instruments.Specifically,the results of experiments were compared with the CO2concentration obtained through gas chromatography (GC) through air bag sampling, as shown in Figure 4,and it was found that the two methods produced good consistency.The height of the NDIR sensor and air bag taking samples was basically the same.Within the height of 0-800 m,the CO2concentration measured by the two methods was within 425-500 ppm. The sampling of ground air samples for GC measurements might be influenced by human respiration due to directly pumped-up ambient air near the surface and the changes of wind direction. However,the miniaturized instrument was placed in a box, and thus we speculate that the red dots near the surface higher than the K30 measurement were due to human respiration.

Figure 3.Typical vertical profiles of CO2.The profiles marked with triangles are type 1.1 and type 2.1,and the profiles marked with square signs are type 1.2 and type 2.2.Each point represents the mean CO2 concentration per minute and error bars indicate the range of three K30 s values.

Figure 4. Comparison of CO2 vertical profiles between the low-cost sensor and gas chromatography methods on the afternoon of 8 January 2019.

3.2. Effects of meteorological conditions on the vertical distribution of CO2

The two experimental tethered balloons rose to 706 m on 8 January (Figure 5(a)) and 1019 m on 13 January(Figure 5(d)), and the results of these two experiments were analyzed with particular emphasis because the experimental time was relatively long and the height was relatively high. The potential temperature gradient data on 8 January(Figure 5(c))and 13 January(Figure 5(f)) were processed into minute data in order to better distinguish the position of the boundary layer.The profiles both have an interval where the potential temperature gradient is almost zero.Therefore,it can be judged that the top height of the mixing layer at both ends was about 600 m and 200 m, respectively, which has good consistency with the observed PBLH measured by lidar(Figure 5(c,f),dark blue line).Above the mixing layer,the CO2concentration decreased sharply with the height increase and tended to have a background concentration of 400+ppm.

Figure 5. Effects of stratification stability on the vertical distribution of CO2. Panels (a-c) show the vertical profiles of CO2, the corresponding vertical profiles of temperature and the corresponding vertical profiles of potential temperature, and the potential temperature gradient of 1404-1431 UTC 8 January 2019,respectively.Panels(d-f)show the vertical profiles of CO2,the corresponding vertical profiles of temperature and the corresponding vertical profiles of potential temperature, and the potential temperature gradient of 1459-1523 UTC 13 January 2019, respectively. Stronger vertical mixing happens due to the vigorous convections and dilutes the CO2 concentration(a,d,c,f).

3.3. Comparison with the WRF-CO2 simulated results and impacts of emissions sources

Accurate vertical distributions of CO2concentration are crucial for model calibration and validation, so we compared the observed data in this experiment with the WRFCO2simulated results.In Figure 6(b,d),the green line is the simulation result with GEOS-Chem as the initial field and boundary field of simulation.The blue line denotes the CO2vertical profile with the WRF model simulated data and triangles denote the CO2vertical profile with the observation data; the orange line is the result of the emissions intensity increased to eight times, and the purple line is the result of the emissions intensity increased to 20 times.By comparing the two experiments’ results with the WRF simulated data(Figure 6),the variation in the CO2concentration profile obtained by default FFDAS emissions in both cases was inconsistent with the observed data, while it showed good consistency when the emissions were increased 8- and 20-fold (Figure 6(b,d)). As can be seen from Figure 6(b), the initial and boundary field data of CO2had little influence on the simulation results, while the concentration of near-surface CO2increased with the increase in emissions intensity.From Figure 6(a,c),the WRF simulated potential temperature profiles show a sharp decreasing trend above 980 hPa,which is different to the observed data. This explains that the simulated CO2concentration only changed below 980 hPa because the PBLH simulated by WRF was not consistent with the observed PBLH. We further tested three more schemes,but they all produced relatively similar results, indicating more efforts are needed for PBLH modeling. At the same time,from the simulated surface CO2concentration distribution (Figure 6(e,f)), the CO2concentration from the default FFDAS emissions at the observation site was 420 ppm,which is smaller than the observed data.One possible explanation is that,when conducting the WRF-CO2simulation, real-time FFDAS data were lacking, so a replaceable FFDAS dataset on 8/13 January 2015 was adopted. This substitute might not reflect the true emissions around the experimental site during the experimental period.Moreover,when increased to 8-fold default emissions,the simulated data were more consistent with the observed(Figure 6(g,h)).

Figure 6. (a, c) Observed and WRF simulated potential temperature, and (b, d) observed and simulated CO2 profiles showing the vertical profiles of CO2 on 8 January and 13 January 2019, respectively. Spatial subplots show CO2 at the ground surface layer simulated by the WRF model.Panels(e,f)show the simulated surface CO2 concentrations using the FFDAS default emissions and 10-m wind vectors on 8 January and 13 January 2019,respectively.Panels(g,h)show the same but with 8-fold FFDAS emissions.The blue star indicates the experimental site.

4. Conclusion

In this study,CO2and meteorological elements in the atmosphere within 0-1000 m in suburban Shijiazhuang were observed by a low-cost miniaturized CO2monitoring instrument through simple and easy-to-operate tethered balloon technology.The results showed that the CO2concentration in most profiles decreased with the increase in height.Compared with the traditional GC method,the data from the low-cost sensor produced consistent results. In addition, the vertical distribution of CO2concentration in the boundary layer was mainly affected by the stability of layering and the emission sources. The WRF model could not capture the CO2profiles well due mainly to errors in the PBLH and the prior fossil fuel emissions with coarse resolutions.A further improvement of the modeling performance might be to obtain a better representation of the PBLH and substitute the outdated prior with updated high resolution inventory data.

Disclosure Statement

No potential conflict of interest was reported by the authors.

Funding

This work was supported by the National Key R&D Program of China [grant number 2017YFB0504000] and the National Natural Science Foundation of China [grant numbers 41705113 and 41877312].The authors are grateful to staffs at the Hebei Province Artificial Impact Weather Office and Dr.SUN Wanqi for their kind help with the field experiments, and Mr.CUI Ming for data analysis.