Assessing local CO2 contamination revealed by two near-by high altitude records at Jungfraujoch, Switzerland

Remote research stations are guarantor of high-quality atmospheric measurements as they are essentially exposed to pristine air masses. However, in a context of increasing touristic pressure for certain sites, attention should be paid to the local anthropogenic emission related to the infrastructure itself. Among emissions, carbon dioxide (CO2) is the most important anthropogenic greenhouse gas and a major contributor to the current global warming. Here, we compared two years of CO2 dry air mole fraction records from Jungfraujoch (Swiss Alps) measured at the Sphinx Laboratory (3580 m a.s.l.; JFJ) and the East Ridge facility (3705 m a.s.l.; JER; horizontal distance of ∼1 km), respectively. Both stations show an overall increase of the annual mean CO2 mole fraction in line with current global trends. On a daily basis, values during the night (00h00–06h00) show robust coherence with variability ranging within the measurement uncertainties matching the WMO compatibility goal of 0.1 ppm, which we considered to be background air CO2 mole fraction for Central and Western Europe. However, JFJ record shows superimposed short-term variability with diurnal CO2 spikes centered around noon. Whereas the variability occurring during time intervals ranging from days to weeks seem to be driven by inputs of air masses from the planetary boundary layer, we suppose that the super-imposed diurnal CO2 spikes occurring essentially in summer are explained by local emission sources related to the infrastructure (visitors, tourism, etc). Nevertheless, we cannot point to a single triggering cause for those spikes as it probably results from a combination of factors. In order to minimize these local emissions, smooth collaboration between all the involved stakeholders is required.


Introduction
Being built in 1937, there is a long history of research at the Sphinx Laboratory of the High Altitude Research Station Jungfraujoch (JFJ hereafter) located in the Swiss Alps (Balsiger and Flückiger 2016). Currently, it is hosting numerous research projects from international institutes (HFSJG 2019) and constitutes a lead location for atmospheric measurements in Europe. It allows to measure the chemical composition of atmospheric air and to document thereby driven climate change (Vollmer et al 2015, Bukowiecki et al 2016, Schibig et al 2016, Mahieu et al 2017 or to understand cloud formation (Kirkby et al 2016, Tröstl et al 2016. A key feature of the site is the remote setting and its exposure to pristine air masses with only sporadic air mass transport originating from the surrounding lowlands set in the atmospheric boundary layer inducing pollution events (Baltensperger et al 1997, Henne et al 2010, Bukowiecki et al 2016. In December 2014, an additional facility was established at the Jungfrau East Ridge (JER) building that is set on the East Ridge of the Jungfrau Mountain 125 m higher and 1 km westward of JFJ (figure 1). This new location is currently not publicly accessible and represent a valuable location to extend the measurement possibilities of the station . Moreover, JER allows for a quality survey of JFJ due to the lack of human pollution and all activities inherent to the touristic exploitation of the Jungfraujoch area. The number of visitors doubled in the last 10 years and reached 1 million for the first time in 2015. Since, 1 million were also achieved in 2017, 2018 and 2019.
CO 2 (given hereafter in ppm, where ppm corresponds to µmol mol −1 in dry air) is a major greenhouse gas that contributes to a large extent to the radiative forcing of the Earth's atmosphere (IPCC 2013). Since the industrial revolution, the atmospheric CO 2 mole fraction has been increasing continuously from a relatively stable pre-industrial CO 2 mole fraction level of ∼280 ppm to reach a global abundance of 407.8 ppm in 2018 (WMO 2019). The Intergovernmental Panel on Climate Change (IPCC 2013) stated that the atmospheric CO 2 increase is clearly due to anthropogenic activities and due to its greenhouse gas properties, the additional CO 2 contributes to the global warming by about ∼1.0 • C of the mean atmospheric temperature increase. Time series of CO 2 measurements are recorded worldwide and most data (including JFJ records) are available online at the WMO's World Data Centre for Greenhouse Gases (https://gaw.kishou.go.jp). Usually, high altitude sites act as reference sites as signals from specific sources (like anthropogenic emissions from industry, traffic and households, etc) are not directly but smoothly recorded by integration over the time due to atmospheric mixing during advection. At JFJ, CO 2 is continuously measured since 2004 with initial values of ∼375 ppm and an annual increase of ∼2 ppm (van der Laan-luijkx et al 2013, Schibig et al 2016. The JFJ record shows also seasonal variations of around 10 ppm with maximal CO 2 values in March-April and minimal values in August-September. Generally long-term trend CO 2 variations are due to anthropogenic emissions whereas seasonal variability is induced by variations in diurnal solar radiation and related CO 2 uptake and release by the biosphere (IPCC 2013).
While technical specifications of available instrumentation along with adequate calibration strategies allow to meet the WMO compatibility goal of 0.1 ppm (WMO 2020), the influence of local contamination in such remote areas needs to be quantified (e.g. Loov et al 2008, Ruckstuhl et al 2012. In addition, several studies focused on the advection of lower atmospheric layers to the elevation of the sampling sites (e.g. Griffiths et al 2014, Tsamalis et al 2014, Ferrarese et al 2015, Fu et al 2016. Moreover, conditions and the degree of impact depend on the specific characteristics of the sampling site and the parameter of interest. For example, measurements performed at Mount Zugspitze, Germany, observed that an increased number of visitors on the weekends do not correspond to higher levels of CO 2 suggesting that CO 2 mixing ratios are not significantly influenced by neighboring touristic activities, at least with only 100-200 visitors per day (Yuan et al 2019). There are numerous high alpine stations running high precision measurements (e.g. Monte Rosa, Schneefernerhaus, Sonnblick, etc). Some of which host touristic infrastructures in parallel (e.g. JFJ, Schneefernerhaus), which constitutes a source of contamination as visitors are breathing, smoking and consuming onsite. Moreover, transport to the site requires helicopter flights, trains, cabins as well as infrastructure that all need regular maintenance or construction work. Together with the companion paper about aerosol measurements (Bukowiecki et al 2021), the greenhouse gas observations presented here allow for an investigation of anthropogenic contamination, but also an insight into the tourism increase issue.

Settings and method
The Jungfraujoch (46.5475 • N/7.9852 • E ) is a saddle in the Swiss Alps separating the Jungfrau (4158 m a.s.l.) and the Mönch (4107 m a.s.l.) mountains. JFJ Sphinx laboratory (3580 m a.s.l.) is set at the Jungfraujoch, whereas JER (3705 m a.s.l.) is located SW on the ridge leading to the Jungfrau summit (figure 1). Due to their proximity, the two locations are considered to experience similar regional weather conditions. The Jungfraujoch is often situated in the free troposphere which allows measurements of atmospheric background CO 2 mole fractions. Punctual convective transport of polluted valley air masses occurs mainly during the afternoon and predominantly during the warm season (Sturm et al 2005, Bukowiecki et al 2016, Poltera et al 2017. Local wind direction is driven by the local topography resulting in two prevailing wind directions (NW and SE) and interactions with the mountainous topography may lead to variations between both locations as JFJ is situated on the saddle and the JER building is higher up on the mountain ridge. Therefore, uplifted air masses should generally first cross JFJ before reaching JER.
The air inlet feeding the CO 2 analyzer at JFJ is placed on the roof of the Sphinx building ∼8 m above a touristic terrace where people can experience an outstanding view and where smoking was not discouraged. Since the end of 2016 panels inform the visitors about sensitive measurements just above their heads and ask them to refrain from smoking. It is relevant when up to 5000 visitors per day reach the site, 5-6 times more in summer than in winter. The JFJ laboratory is set ∼150 m away from the main recreational area that consists of restaurants, shops and a train station. The Sphinx building can be reached with elevators (108 m elevation). The elevator shaft is likely facilitating the transport of air from the underground tunnel complex upwards, which may have an influence on the measurements. It is noteworthy that there are only moderate CO 2 emissions from the infrastructure itself since most units run on electricity, with exceptions only during construction phases. At JER, the CO 2 analyzer was placed in an empty room of the former telecommunication building that is now available for scientific purposes. During the measurement campaign, this facility was visited only occasionally and exclusively by scientists and maintenance staff. JER is reachable by a tunnel starting from the Jungfraujoch main train station, where the air is regularly vented upwards and outside.
We scaled the number of people visiting the Jungfraujoch available as daily totals using the mean annual visitor percentage per hour (arriving onsite) in order to approximate the diurnal cycle of visitors, which do not absolutely mimic reality. The planetary boundary layer (PBL hereafter) data was measured by MeteoSwiss using a ceilometer based at the Kleine Scheidegg (2061 m. a.s.l.) below the Jungfraujoch (Poltera et al 2017). To be noted, PBL influence is likely to occur even if its height is below JFJ elevation due to shallow upslope injections not resolved by the ceilometer. Thus, we use in addition black carbon (BC) as another indicator for advection of boundary layer of air masses, measured as equivalent black carbon (eBC) mass concentrations, see detailed description in the companion paper (Bukowiecki et al 2021).
To perform CO 2 measurements at JER, we use a Picarro G2311-f analyzer based on the Cavity Ring-Down Spectroscopy (CRDS) technology installed in December 2014. At JFJ we use a nondispersive infrared (NDIR) gas analyzer (Sick Maihak model S710; see supplementary information (available online at stacks.iop.org/ERL/16/044037/mmedia)).
Finally, we also used methane (CH 4 ) and carbon monoxide (CO) records measured with Picarro G2401 by the EMPA at both sites between December 2014 and May 2016.

Results and discussion
The JFJ CO 2 in situ record (figures 1 and 2) constitutes the extension of the measurement started in December 2004 (Schibig et al 2016). The record is robust and shows similar results in trend and amplitude compared to JER. Over the 2 years values ranged between 390 and 419 ppm at JFJ and between 390 and 422 ppm at JER. The JFJ CO 2 record (figure 1) shows a mole fraction increase of 2.22 ± 0.04 ppm yr −1 during the interval 2005-2018 as calculated with a Monte Carlo algorithm. This value is significantly higher than the previous published slope of 1.97 ppm yr −1 obtained for the interval 2005-2013 (Schibig et al 2016) and result from strong annual mean CO 2 growth rate over the last decade (Friedlingstein et al 2019, WMO 2019. The seasonality obtained by the difference between the highest (March-April) and the lowest (August) detrended monthly average is 10.72 ± 1.00 ppm, similar as previously determined. The JER record is too short to determine robust trend and annual seasonality.

Comparison JER-JFJ and robust night values
The comparison between NDIR and CRDS records is robust (R 2 = 0.94 in overall and R 2 = 0.96 for night values, slope of ∼1; figure 3(A)) and the agreement is good over the 2 years (figure 2) either in minima or maxima events despite the relative distance (∼1 km) between the two sites and documents a similar annual CO 2 increase. Nevertheless, a mean CO 2 excess (values at JFJ minus values at JER) with normal distribution is observed (figure 3(B)), which is higher than what can be expected due to the different NDIR and CRDS measurement techniques (Schibig et al 2015). When splitting data of the days into two categories named daytime (06h00-18h00) and nighttime (18h00-06h00), the values emphasize the differences between the two sites with +0.49 ppm and only +0.01 ppm, respectively. The mean CO 2 excess (figure 3(B)) is essentially due to the addition of positive CO 2 excursions during daytime observed at JFJ. These excursions occur almost daily during the warm season, essentially May to September between 8h00 and 17h00 when visitors are onsite, whereas from October to April, they are less likely to occur and also less pronounced ( figure 4  and supplementary figure 1). The short-term diurnal maximal offset is generally up to ∼5-10 ppm for a couple of hours around noon. During the night, focusing on the interval 00h00-06h00, values are in excellent agreement independent of the day of the week and within the internal variability of the two instruments (±0.1 ppm) (figures 5 and 6), which supports our expectations that the two sites are experiencing the same meteorological conditions during the night and measure the same air. Since JFJ measurements are considered to be robust CO 2 atmospheric background values during the night, we can assume the same for JER. When visitors are onsite, the daily differences are ranging between 0.3 and 0.7 ppm (figure 5).
At JER, the CO 2 record shows less variability during the raising concentration phase in autumn and winter 2015 and 2016 ( figure 2(A)). From roughly 15th September to 16th February, the reduced variability may be a result of the fact that the Jungfraujoch and East Ridge area is predominantly in the free troposphere, which is supported by PBL data (Poltera et al 2017). JFJ shows a slightly more variable pattern indicated by the spikes in the difference between the two records ( figure 2(A)).
In addition, low values seem to occur mainly in spring and summer at around noon until early afternoon (∼12h00-16h00; figure 2). Thus, during these events JFJ (and to a smaller extent JER as it is higher) is influenced by air masses coming from the boundary layer. Boundary layer air masses in spring and summer are expected to be depleted in CO 2 due to photosynthetic uptake and reduced anthropogenic emissions (e.g. due to no domestic heating). The (negative) differences are likely a consequence of a sampling of different air masses at JFJ and JER. Such events are not unlikely due to the local wind patterns caused by the complex topography.
A few other observations seem to be of recurring nature, yet sometimes equivocal. Synoptic events affect both locations at the same time as seen for instance in supplementary figure 1 (16th and 17th January 2015) and correspond to regional events such as input of polluted air masses from the lower Grindelwald valley or large-scale European patterns such as events driven by wind coming from the south, i.e. South Föhn. These air masses show another CO 2 signature as the well mixed Atlantic CO 2 background values. Generally, Föhn events result in variable CO 2 mole fraction compared to more stable values when air masses are coming from the northern sector. Another observation is that when the PBL is reaching Jungfraujoch (mostly in summer), we may observe a drop in the CO 2 mole fraction in summer and vice versa (see section 4.

Comparison with PBL, eBC, CO and CH 4
We observed no similarities between convective strength from the valley (PBL inputs) as resolved by the ceilometer and CO 2 spikes (figure 4 and supplementary figure 1). As shallow upslope injections cannot be detected by the ceilometer, we further characterize the influences of air masses uplifts based on eBC inputs at both sites (Bukowiecki et al 2021). eBC is mostly produced in the PBL through incomplete combustion of fossil fuel or biomass burning (Petzold et al 2013). Locally, construction works or preparation of the touristic infrastructure on the glacier with snowcat operations also contribute to local eBC emission (Bukowiecki et al 2016). Minor contamination from tobacco smoke was also shown through the correlation between corresponding environmental markers and the number of visitors in August (Frohlich et al 2015). Theoretically, only low eBC concentration should be recorded at JFJ due to the remote location and the fact that JFJ is essentially in the free troposphere.  eBC increase indicates events when JFJ is contaminated by air masses coming from the PBL or by local emissions. To distinguish these two sources, we compare the two site records since local emissions are pre-dominantly seen only at the Sphinx observatory. The comparison between eBC and CO 2 shows that both records evolve often similarly and thus explain a significant part of the short-term variability (days to weeks). A striking observation is shown in figure 7 which highlights the different influence of the PBL input in winter and summer. In February 2015 and 2016, positive excursions in the CO 2 mixing ratio   at both sites occur similarly to the increase in eBC content which means that uplifting of polluted air masses from the PBL (due to heating, traffic, etc) enhance the increase of the CO 2 mixing ratio due to vegetation respiration in winter. On the contrary, in spring and summer (for instance in May and June 2015) when large inputs of eBC occurred, we observe a decrease of the CO 2 mixing ratio due to the uplift of air masses from the valley depleted in CO 2 caused by ongoing photosynthesis (Uglietti et al 2011). While this example shows variation on a day or a couple of days, multi-weeks trends as for instance observed in July 2015, show also such a dependence with slight increase of the eBC content occurring at the same time as a continuous decreasing trend in the CO 2 . Nevertheless, superimposed are still daily variations on most of the summer days that are not correlated to eBC variations as for instance shown for July 2015 ( figure 8(B)). Local emission CO 2 and eBC sources do not have to be related, e.g. breath air in contrast to emissions from construction, snowcat or smoking.
Occasionally, a higher eBC increase can be observed at JFJ compared to JER. This could be caused by either local contamination at JFJ or by local wind processes, leading to the situation that JFJ is situated within the PBL while JER remains above.
Comparison with CO and CH 4 records (figure 8) leads to interesting observations. CO is only related to anthropogenic emissions essentially produced by incomplete burning of fossil fuels or biomass (Dils et al 2011). In winter the short-term variability in both absolute values and in the difference between JFJ and JER are very similar, which support our suggestion that most CO 2 short-term variability can be explained by variations in wind regimes, essentially by PBL height variability. For example, in January and February 2015 all four atmospheric compounds evolve similarly on a visual basis (figure 8). And thus, single CO 2 spikes are less likely to be observed. In summer, the situation is slightly different as BC, CO and CH 4 are visually correlated and CO 2 shows negative correlation for the aforementioned reasons. However, in summer superimposed to this variability, single CO 2 spikes were observed with maximal values centered on 12h00 at JFJ that are not seen at JER (figure 8), which we suggest result from local contamination related to the number of visitors and/or-less likely due to the missing correlation with CO-to the infrastructure. The detailed paths of these emissions are to be explained.

Investigation of potential triggers of the CO 2 spikes
Diurnal CO 2 spikes are only observed at JFJ, which means that periodic processes act on JFJ during the day as night values are similar to JER. This provides already an indication that the difference in elevation between the two measurement sites is not likely to play a major role. To further investigate the potential triggers for the CO 2 spikes, we used several meteorological parameters measured at the MeteoSwiss station Jungfraujoch (3580 m. a.s.l.). Among which air temperature, relative humidity, wind speed and sunshine duration show no covariations with the CO 2 spikes. Two other factors may suggest potential links: the global radiation and the wind direction.
Global radiation influences thermal uplift of the air and might influence the sites differently (e.g. Henne et al 2005). As seen on figure 4 and supplementary figure 1 for the summer season the CO 2 is not increasing synchronously and the amplitude appears to be independent from the absolute global radiation when global radiation is increasing early in the morning. Hence this parameter seems not to be the primary cause for the CO 2 spikes. Additionally, since the setting of the two stations is almost similar, global radiation should affect both sites to a similar degree, which is not the case. However, upwind regimes need some time to develop; thus, PBL injections will not appear immediately after sunrise.
Regarding the wind direction, a co-variability observed in January 2016 between wind direction and CO 2 excess is interesting as the direction varies only by ±10 • and may be related to pollution originating from a local source. This co-variability is rarely seen in other periods. During Föhn days, we do not see any variations in the CO 2 excess values but we observe abrupt variations occurring simultaneously at the two locations (supplementary figure 1; section 4.2). Nevertheless, we think that due to its location in the saddle, the local wind regime may affect only JFJ through dilution. For instance, during the day and especially during the warm months, thermal air masses continuously pass over the Jungfrau saddle (i.e. JFJ) transporting polluted air from the surrounding valley without reaching JER (see also section 3.2).
The number of visitors is well in agreement with CO 2 excess spikes in summer. In winter, only disparate CO 2 excess values are observed or no excess at all as for example in January 2016 but also with only 100-200 visitors per hour (figure 4 and supplementary figure 1). In summer with up to 1000 passengers arriving per hour around noon, the anthropogenic influence is likely to be part of the emission source through direct as well as indirect influences. The visitors release CO 2 through breathing and smoke especially on the terrace below the laboratory's air inlet. People are mostly staying outside during sunny days, i.e. when the global radiation is high. Indirectly JFJ might also be influenced through the elevators that can act as chimneys uplifting local polluted air masses from the touristic complex (tunnels) or from maintenance or construction work at the infrastructure. In addition, activity related to the whole infrastructure is also enhanced when more visitors are onsite. It would be tempting to attribute most of the CO 2 excursions to local anthropogenic emission, nevertheless the lack of synchronicity for certain days tends to suggest that other parameters additionally influenced these diurnal excursions. The fact that all parameters (visitors, PBL, global radiation) have a diurnal behavior and also a pronounced seasonality makes it difficult to disentangle which one is acting as main contributor of the CO 2 spikes.

Summary and outlook
The following observations have been made: • The comparison of CO 2 records over a 2 years period shows a good agreement between both sites. Especially in the night (00h00-06h00) values are indistinguishable within the measurement precision and establish that both sites are measuring robust atmospheric values. • Daytime CO 2 spikes are not seen at JER, whereas cycles are often seen at JFJ, especially in summer. • There is a distinct mean seasonal pattern with highest hourly CO 2 excess values of up to 1.3 ppm occurring around noon during the summer months. • PBL injections seem to explain most of the daily to weekly variability. But there is also a possible influence of local anthropogenic contamination on CO 2 measurements at JFJ namely through the high number of visitors especially in summer or activities inherent to the infrastructure of the site. • The low contaminated CO 2 and aerosol records establish JER as an additional site to perform high quality atmospheric measurements.
Yet, information panels about smoking effects on the touristic terrace are having a positive feedback on the observed excessive aerosol emissions documenting that the visitors read our information and react positively to lower emissions and help to improve our sensitive measurements (Bukowiecki et al 2021).
Here, we do not finish the discussion regarding the origin of these CO 2 spikes at JFJ as we cannot point to a single triggering cause. It probably results from a combination of factors, such as PBL variations-or more generally input of polluted air masses from the valley-local wind dynamics as well as the influence of visitors or infrastructure, especially in summer. Other competing factors (not revealed yet) might also likely to be involved. To minimize this influence, a new heated inlet system was placed ∼80 m away from the tourist platform and is presently tested. Further studies may help to disentangle this issue, (a) a thorough investigation of local wind dynamics coupled to CO 2 profiles at both location, (b) 14 C measurements could highlight contamination by fossil fuel combustion emissions (Bozhinova et al 2014), (c) 13 C and O 2 measurements as indicator of biospheric sources may help to identify the origin of the contamination (Guillon et al 2015), (d) the comparison of high-quality water vapor concentration measurements at both location using humidity as a proxy for mountain venting processes (Henne et al 2005) and, thus, as a proxy for boundary layer influence (Cooper et al 2020) and (e) CO 2 monitoring in different locations of the station (touristic and meteo terraces, tunnel system) to see how coherent the variability is.
Finally, the CO 2 and the simultaneous aerosol studies (Bukowiecki et al 2021) highlight punctual anthropogenic pollution events and illustrate the sensitivity of instruments to anthropogenic activities at remote locations, with local contamination likely to be recorded by aerosols (Bukowiecki et al 2016), CO and to a certain extent in CO 2 records. In a rising tourism sector, spotting pollution events is a first step towards the quantification of anthropogenic gas emission related to the infrastructure, which would allow to provide solution for a good collaboration among all the involved stakeholders, i.e. scientists and tourism actors.

Data availability statement
The data that support the findings of this study are openly available at the following URL/DOI: https://gaw.kishou.go.jp. The CO 2 record from JER is available upon request. See the companion paper for the eBC availability.