The growth of the atmospheric concentration levels of greenhouse gases in the Industrial Age is considered the major cause of the observed warming of the Earth’s surface [1
]. Carbon dioxide (CO2
) is the primary gas that is contributing to the enhanced greenhouse effect, and the largest contribution to total radiative forcing is caused by the increase in the atmospheric concentration of this atmospheric component [1
]. For this reason, the CO2
concentration in the atmosphere is considered a key quantity to study and understand the global greenhouse effect and the ongoing climate change, and it is constantly monitored at several locations around the world, and in particular at remote sites [3
]. These locations are at a great distance from significant anthropogenic sources of air pollutants, so that the CO2
background concentration measurements are poorly influenced by local sources or transport. The long-term observations of the CO2
background atmospheric concentration are of fundamental importance to investigating ongoing climate change and represent a useful tool to evaluate the success of mitigation activities against climate change.
concentration has been measured at the Plateau Rosa site, located in the westerly Italian Alps near Mt. Cervino in Italy, at the altitude of 3480 m a.s.l., since 1989 (and continuously since 1993). The geographical position of the station, that is, at a high altitude and far from urbanised and industrialised zones, allows representative measurements of the atmospheric CO2
(methane) and O3
(ozone) background values to be obtained frequently [7
Several analyses have been carried out since the beginning of the CO2
monitoring activity. In the first studies about the CO2
concentration time series at Plateau Rosa, the daily concentrations were shown for the 1989–1992 period [8
] and the connections between the measured CO2
values, which were available at that time, were analysed, together with the meteorological circulation patterns [8
]. The deterministic backward trajectories were computed every 6 h and a cluster analysis was used for this purpose. Later on, the monthly mean data for the 1989–1999 period were shown by Ferrarese et al., (2002) [10
], who investigated the source and sink areas over Europe and the Boreal Atlantic Ocean by means of a source-receptor model [11
] on the basis of backward trajectories starting from at Plateau Rosa. The monthly data averages for the 1989–2001period were published by Apadula et al. (2003) [12
], who computed a growth rate of 1.52 ppm/year. The column concentration fields for 1996–1997 were computed using the same source-receptor model, the backward trajectories and the CO2
concentrations from three monitoring sites at high altitudes in Europe: Plateau Rosa, Zugspitze and Monte Cimone. The 20-year-long monthly data (1993–2013) were presented by Ferrarese at al. (2015) [13
], who evaluated the growth-rate of that period (1.98 ± 0.04 ppm/year). A detailed analysis of atmospheric circulation was performed using a regional meteorological model to investigate the source areas of the highest concentration event in the complete series available at that time.
Moreover, subsets of CO2
concentration data have been used in studies on a global scale and a regional scale [14
The CO2 concentration series is now thirty years long, which is a significant period of time from a climatic point of view, because the climate is frequently defined as covering a 30-year time period. The background Plateau Rosa CO2 monthly concentration data series is here presented and analysed, the main distinctive parameters are pointed out and the method used to obtain the background data is illustrated.
The Plateau Rosa station (Section 2
) is described in this paper, together with its monitoring instrumentation (Section 3
). The filtering method used to select the background data is detailed (Section 4
) and the main features of the CO2
monthly concentration time series are discussed (Section 5
). The inter-annual variability is analysed in Section 6
, and the conclusions are summarised in Section 7
2. Plateau Rosa Monitoring Station
The Plateau Rosa monitoring station (Global Atmosphere Watch (GAW) Identification Code: PRS) is located in the north-western Alps, in the Italian Aosta Valley region, in the municipality of Valtournenche, near Mt. Cervino. The PRS station (Figure 1
) is located inside the Testa Grigia Laboratory, which belongs to the CNR (Italian National Research Council).
The station (45.93436° N, 7.70778° E) was installed at an altitude of 3480 m a.s.l., upon a large snow-clad mountain plateau far from urban and polluted zones. PRS is one of the highest monitoring stations of the World Meteorological Organization GAW Programme [7
]. Owing to its high altitude and position, it is very often located above the planetary boundary layer, and is thus suitable for the background measurement of greenhouse gases.
The measurement of the most important greenhouse gases (excluding water vapour), such as CO2, CH4, and O3, is carried out at the PRS station by RSE (Ricerca sul Sistema Energetico—Italy), and a neighbouring meteorological station, also managed by RSE, was installed to support the greenhouse gas measurements, by collecting, in real time, air temperature, relative humidity, pressure and wind (speed and direction) data.
Since these measurements should not be influenced by local sources, the PRS station is equipped with an electrical heating system and does not use any fossil fuel. During the skiing season, diesel-operated snowmobiles are sometimes used to maintain appropriate conditions for the skiing activity, but they operate at lower altitudes, and thus do not influence the measurements. A refuge and a cable car are located in the vicinity of the measuring station; both only operate during daylight hours and are open for about eight months a year. A meteorological station, managed by the Italian Meteorological Service (WMO code: 16052), is located at a horizontal distance of about one hundred meters from the PRS.
The wind speeds are generally greater than 4 m s−1
, and thereby reduce any eventual impacts of local sources. An analysis of the intensity and direction of the winds gathered in the 1971–2008 period at the Italian Meteorological Service station showed that calm conditions (lower wind speeds than 1 m s−1
) occurred during about 17% of the measurements, while the prevailing winds came from North East (about 25% of the events). The wind rose pertaining to one entire year, with a data coverage of about 98%, as measured at the RSE meteorological station, is shown as an example in Figure 2
. A wind speed of less than 1 m s−1
occurred in 5% of the observations, while it was higher than 4 m s−1
in about 60% of the measurements and higher than 6 m s−1
in more than 40% of the measurements.
The climate at the Plateau Rosa station is typical of a continental alpine location, with relatively large diurnal and seasonal temperature variations, frequent atmospheric pressure variations and strong winds. The climatological mean values (for the 1961–1990 period) of the maximum and minimum temperatures measured at the Italian Meteorological Service station (PRS site) are shown in Figure 3
3. CO2 Measurement Instrumentation
The measurements of CO2 at PRS began in April 1989, with the valuable collaboration of the Italian Meteorological Service personnel working at the Monte Cimone monitoring station, who carried out the analyses of the air samples collected at the PRS station by means of a non-dispersive infrared (NDIR) analyser (ULTRAMAT 3E). The air samples were taken every two weeks, using a couple of electropolished stainless steel flasks, at different hours of the day (usually at 9:00, 11:00, 13:00, and 15:00, local time). All the samples were then transferred to the Monte Cimone laboratory where they were then analysed.
In March 1993, PRS set up its own measurement system, which involved the installation of an NDIR ULTRAMAT 5E analyser, manufactured by SIEMENS. The inlet height of the instrument is about 9 m above ground. Both types of measurements (continuous and flask measures) were carried out for about 4 years, until December 1997, and a good agreement between the two types of measurements was observed.
The analyser system is calibrated regularly. The calibration procedure consists of three different steps: the first one regards the measurement of the air by means of two working gas standards (every six hours); the second one involves checking the secondary standards (every 72 h); the third one makes use of five referenced NOAA (National Oceanic and Atmospheric Administration) primary gas standards (referenced scale WMO X2007 [18
]), and this is performed at least once a year. This type of approach is based on the work of Komhyr et al. (1989) [19
] and Cundari et al. (1995) [20
]. The two working gas standards are compared every 72 h with four well-calibrated secondary station standard gases in order to obtain a better determination of their CO2
concentrations and any possible drifts in the concentration with time (measurement scheme shown in Figure 4
). The working and secondary standards were acquired from S.I.A.D. (Società Italiana Acetilene e Derivati) and the primary standards were purchased from NOAA ESRL/GMD (National Oceanic and Atmospheric Administration—Earth System Research Laboratory/Global Monitoring Division). The working and secondary standards, before being utilised, are referenced to the primary standards in order to determine the values with respect to the WMO X2007 scale. Finally, the current primary standards were supplied by the ICOS (Integrated Carbon Observation System) project [6
]. The NOAA primary standard concentrations range from 369.58 to 404.64 ppm and the ICOS concentrations range from 379.194 to 449.890 ppm.
The uncertainty of the measurements was estimated as 0.07 ppm on the basis of various replicated measurements of some working standards with different concentration values. The uncertainty was basically assumed to be equal to the standard deviation assigned to the average value of the differences between the certified value of the standard and that measured by the instrument (the measurements were related to a period of 9 months and the working standards were measured every six hours).
All the gases are dried, prior to being introduced into the analyser, by passing them through a glass water vapour trap, immersed in a cold alcohol bath maintained at −65 °C by means of a freezer. This approach is particularly important to express the CO2 concentration in dry air. The atmospheric CO2 concentration, with reference to the WMO X2007 international mole fraction scale, is expressed in ppm, i.e., in μmoles/mol of dry air. The native data frequency is 0.2 Hz. The measured data was averaged every 30 min until 2007, and then every 60 min in order to be consistent with international measurement standards. The corresponding standard deviations have been computed.
A PICARRO 2301 analyser, with all the necessary accessories for the real-time measurement of CO2
O, was installed in May 2018 and it was made operational, according to ICOS guidelines [6
], in September 2018. Therefore, since that date, the measurement of the CO2
concentration has been carried out with both monitoring systems (PICARRO and ULTRAMAT), and the homogeneity of the measurements is tested by measuring the concentration of the primary reference standard mixtures with both instruments. The comparison of the measurements has shown a high comparability, as the maximum obtained differences are lower than 0.1 ppm, and are thus compatible with the above-mentioned measurement standard uncertainty.
The standard calibration mixtures adopted for the new measurement system (PICARRO) are provided within the ICOS project. The used measurement scheme is quite similar to that required by ICOS, to which the Plateau Rosa monitoring station has to comply [21
]. The timing of the calibration verification has changed significantly, as result of the stability of the new analyser, compared to the measurement system carried out using the ULTRAMAT analyser. In addition to the four reference primary standards, a Short-Term Target Gas (STTG), with one measurement every 24 h, and a Long-Term Target Gas (LTTG), with one measurement every 30 days, were introduced to check the stability of the measurement in the short and long term. The duration of the measurement of each standard is 15 min, and the calculation of the average value is carried out without considering the first three minutes of the measurements. In agreement with ICOS ATC (Atmosphere Thematic Centre), we are planning to modify the plant by increasing the STTG measurement frequency and making the plant more like the one indicated by Laurent [21
Finally, it is important to point out that, since all measurements in the global monitoring network must be comparable with each other, they refer to a single measurement scale. For this reason, various and systematically inter-comparison campaigns have been conducted and managed by NOAA on behalf of WMO/GAW in the context of several European projects. The CO2 inter-comparability target for WMO is 0.1 ppm.
data collected at the Plateau Rosa station are available at WDCGG (World Data Centre for Greenhouse Gases) and in ObsPack (Observation Package) [5
4. Background Data Selection
The identification of atmospheric trace species measurements, which are representative of well-mixed background air masses, thus unaffected by local conditions, is particularly important to monitor atmospheric composition variations at background sites, and to document the long-term CO2 changes in the atmosphere and spatial gradients at a large scale. However, several monitoring sites are subject to the arrival of air masses contaminated by high values of CO2 concentrations from local or regional anthropogenic sources, or low CO2 concentrations, due the presence of sinks located close the monitoring station.
Moreover, a standard methodology that could be applied at each and every monitoring station to identify background concentration measurements of greenhouse gases is still not available. Measured data are often selected by means of filtering procedures [23
], which are essential to estimate the growth rates of greenhouse gas concentrations [24
], to evaluate the regions characterised by CO2
sinks and sources [29
] and also to model the long-distance transport of trace gases [36
Several methodological approaches have been implemented to identify background measurements [23
]. These can be classified as: criteria based on chemical parameters [38
], such as trace gas concentrations or the ratio of trace gases; criteria based on meteorological analysis, obtained by evaluating the transport processes of polluted air masses to the background site [40
] or by evaluating the origin of the air mass by means of the analysis of backward trajectories [41
], or by utilizing Lagrangian particle dispersion models [37
] and, finally, criteria based on statistical methods [20
Among the statistical methods used to derive background values of the CO2
concentrations, the work of Thoning et al. (1989) [24
] is of upmost importance; the developed procedure was initially only applied to Mauna Loa station data, but then later on it was also applied at other remote stations (among others: [51
]). It consists of two steps to control the variability of the hourly data and the difference between consecutive hourly means, and an iterative algorithm that removes any values which differ from the weighted spline curve by more than a given threshold. Yuan et al. 2018 [48
] presented a novel statistical data selection method, named Adaptive Diurnal minimum Variation Selection (ADVS), which is based on the diurnal CO2
patterns that typically occur at elevated mountain stations. Other authors have individuated background values by selecting data collected at remote stations using their own techniques, which usually involve a comparison of the standard deviations of the hourly data with fixed thresholds and an algorithm selected according to the characteristics of the station [20
]. Although several statistical methods could be applied at various measurement stations, in order to ensure a good comparability between the results of the various monitoring stations, the adopted threshold values would need to be specific for each measurement site. In fact, each measurement site has such particular features that makes each one different from other sites.
The methodology applied to Plateau Rosa CO2
data, called BaDS (Background Data Selection), is a statistical method that is based on the consideration that a representative background condition is necessarily characterised by a very little variability within the hourly averages and between the couples of two consecutive mean values. This first phase of the methodology is very similar to the one adopted for the Mauna Loa [24
] and Mt. Cimone [20
] monitoring stations, except for the cut-off threshold values, which are specific of the measurement site.
BaDS, applied to the hourly collected data, basically works in the following way:
The procedure first examines the standard deviation assigned to each hourly average; the datum is flagged if the value exceeds a given threshold (PRS cut-off value σ = 0.7 ppm);
subsequently, each datum (hourly mean value) is compared with the previous one, and the datum is flagged if the difference exceeds a given threshold (PRS cut-off value δ = 0.3 ppm);
a moving median (computed only if there is at least 25% of valid data in the list of the 504 theoretically available hourly data, corresponding to 21 days of hourly measurements) is applied to the data that have passed the previous steps, and each hourly measurement is compared with the corresponding moving median value: if the difference exceeds a given threshold (PRS cut-off value ), the datum is flagged and considered as a no-background datum;
a moving average (computed only if there is at least 10% of valid data in the list of the 504 above mentioned hourly data) is applied to the data that have passed the previous steps, and each hourly measurement is compared with its corresponding moving mean value; the same procedure described in the previous point is applied to identify the background data using the same threshold ;
finally, all the hourly averages flagged in the above descripted steps are readmitted and are considered background data if their residuals from the moving average are less than or equal to .
The results of the BaDS methodology, applied to the Plateau Rosa data for the year 2016, are shown in Figure 5
a as an example of this procedure. In this case, about 21% of the available gathered data was not considered as background data, and was thus rejected. The enlarged plots of a winter (January 2016 in Figure 5
b) and summer month (July 2016 in Figure 5
c) show two detailed examples of the data selection procedure.
The choice of the σ, and parameters is based on experience gained over the years and represents a compromise so that only background measurements are retained and, at the same time, only a small number of hourly values are removed. Basically, the method tends to remove all the data that have excessively negative and positive peak values, compared to the trend curve (moving average).
One possible limitation of this procedure is represented by the number of acquired data. In other words, if there are long periods with missing data during the measurements, the method may not be effective in selecting the background data. For this reason, the curves related to the mobile median and moving average are calculated over a long period (21 days). However, other statistical methodologies also suffer from data deficiency problems.
The diurnal cycles were derived from the complete hourly dataset and from the selected background hourly data. A summer month (July) and a winter month (January), both related to the year 2016, are shown in Figure 6
as examples. The diurnal cycle does not appear in the January 2016 raw data (Figure 6
a), while the effect of the BaDS filter is that of reducing the mean value by 0.42 ppm and the standard deviation by 0.52 ppm (Figure 6
b). The daily cycle in the raw data shows a peak in the morning and a minimum in the afternoon in summer time (Figure 6
c) and application of the BaDS filter reduces the daily cycle and the standard deviation by 0.90 ppm (Figure 6
d). Similar results were also obtained for the other years.
The PRS data does not seem to be influenced by the vegetation in winter (the diurnal cycle is absent), whereas it may be enriched by regional sources and sinks. The photosynthetic activity in summer is responsible for the diurnal cycle. The application of the BaDS filter reduces the effect of both photosynthetic activity and transport from regional sources and sinks. The absence of a diurnal cycle in the filtered data indicates that the application of the BaDS filter leads to the selection of data that are not affected by local sources and sinks.
In the Zhongshan station in the Antarctic, the monthly averaged deviations of the hourly mean CO2
mole fractions from the daily means are relatively low (about 1 ppm) throughout the entire year, thereby indicating that the observation point is not influenced by regional sources or sinks [50
]. The monthly mean diurnal CO2
cycle at the Lampedusa station in the Mediterranean Sea (Italy) shows no cycle in winter and a very small diurnal cycle in summer, with mean hourly standard deviations of about 1.5 ppm in winter and 2.0 ppm in summer [41
]. The monthly diurnal CO2
cycle at the Kasprowy Wierch mountain station (Poland) is only present in summer, with a peak-to-peak amplitude of about 5 ppm and a lower mean hourly standard deviation than 1 ppm in winter and of about 1 ppm in summer [54
]. Thus, the comparison between the monthly diurnal CO2
cycle at PRS and the ones at other mountainous [54
] or island [41
] sites shows that the PRS observations are characterised by less variability, while the comparison with the Antarctic site shows the influence of the continental source and sink data on the PRS data.
The whole data set of the hourly averages, considered as background data, e.g., those resulting from the application of the BaDS filter, was then used to calculate the daily averages, and from these the monthly averages. As expected, the yearly cycle (year 2016 in Figure 7
) shows a maximum at the beginning of spring (March) and a minimum in late summer (August).
The atmospheric CO2 concentration has been measured at the Plateau Rosa site in the north-western Alps since 1989, and in a continuous mode since 1993. The complete series now covers 30 years, and it is thus suitable for climatological studies.
The measurements have been performed using the most up-to-date time measurement techniques and the data have been compared since the instrumentation was changed. The data have been analysed using a filtering technique, called BaDS, to select background data and to build the monthly PRS background data series.
The main characteristics of the monthly PRS background series are: trend (2.05 ± 0.03 ppm/year), peak-to-peak amplitude of the seasonal oscillation (10.4 ± 0.9 ppm), maximum and minimum times (March-April and August, respectively). The series shows a linear (or exponential) growth rate as a result of anthropogenic emissions, a first harmonic with a period of one year, due to seasonal oscillations, a second harmonic with a 6-month period, and a small variation in amplitude (0.78 ppm in 25 years). The addition of two more harmonics (periods of 4 and 3 months) has not significantly improved the agreement between the modelled and observed values. Furthermore, the complete series of monthly data (1989–2018), including the observations performed with flasks in 1989–1993, is well represented by the curve composed of an exponential term, 4 sinusoidal harmonics and a modulation amplitude term.
Moreover, the historical series also contains traces of climatic variability at a planetary scale: in fact, the annual growth rates of the PRS series are in agreement with those observed at the Mauna Loa station, and are able to identify the ENSO years. The monthly growth rates appear to be correlated with some climatic indexes (SOI and MEI), with a lag of 4 months.
The position of the station, that is, in a mountain environment and at very high altitude, the constancy of the continuous measurements, the appropriate instrumentation, which collects reliable values, and good-practice measurement rules have permitted an important long-term series of background atmospheric CO2 concentrations to be obtained.
The Plateau Rosa measurement station, as well as other measurement stations in international monitoring networks, provides fundamental data that may be used to evaluate how the commitment to containing CO2 emissions, which has been undertaken by policymakers throughout the world, will be effective in containing global warming. At the same time, it provides data of particular importance for the modelling of carbon cycles and for the evaluation of CO2 sources and sinks by means of inverse modelling.