Mercury is a chemical with widespread anthropogenic emissions that is known
to be highly toxic to humans, ecosystems and wildlife. Global anthropogenic
emissions are around 20 % higher than natural emissions and the amount of
mercury released into the atmosphere has increased since the industrial
revolution. In 2005 the European Union and the United States adopted measures to
reduce mercury use, in part to offset the impacts of increasing emissions in
industrialising countries. The changing regional emissions of mercury have
impacts on a range of spatial scales. Here we report 4 years
(December 2011–December 2015) of total gaseous mercury (TGM) measurements at
the Cape Verde Observatory (CVO), a global WMO-GAW station located in the
subtropical remote marine boundary layer. Observed total gaseous mercury
concentrations were between 1.03 and 1.33 ng m
Mercury is present in the atmosphere in three main forms: gaseous elemental
mercury Hg
Anthropogenic sources of mercury account for around 30 % of the total
amount and include emissions from coal burning, mining, cement production,
oil refining, and waste incineration. One-third of the anthropogenic emissions
are thought to come from deliberate biomass burning with Africa as the single
largest continental source; therefore, in this region there could be an
influence from Sahel African biomass burning during the months of November
through to February (Roberts et al., 2009; De Simone et al., 2015). Hg
Reactions of Hg
Strode et al. (2007) estimated that 36 % of all mercury emissions in the
Northern Hemisphere come from the ocean both through primary emission (ocean
upwelling and mercury-containing rocks) and from re-emission of previously
deposited mercury (as Hg
A community strategy developed by the EU was adopted in 2005 and listed 20
actions to reduce mercury emissions, cut mercury supply and demand, and
protect people against exposure. This strategy had a strong focus on the need
to take a global approach and included actions relating to multilateral
negotiations for the conclusion of a legally binding convention on mercury
(
It has been a source of contradiction that in the Northern Hemisphere, while
both measured atmospheric Hg concentrations and wet deposition fluxes have
been decreasing since 1990 (Soerensen et al., 2012) and 1996–2013 (Slemr et
al., 2011; Weigelt et al., 2015); global Hg emissions during this period were
calculated to be increasing (Pacyna et al., 2010; Streets et al., 2011). Very
recently, however, Zhang et al. (2016), using a revised inventory and the
global model GEOS-CHEM, have shown that global Hg emissions may also be
decreasing. They suggest that a large discrepancy in the emissions data was
from locally deposited mercury close to coal-fired utilities. It is thought
that this source has declined more rapidly than was previously predicted due
to shifts in mercury speciation from air pollution control technology
targeted at SO
Using data from ship cruises, Soerensen et al. (2012) observed a significant
decreasing trend of atmospheric mercury concentrations over the North
Atlantic of
Cape Verde site location. Top right, image from Google Earth:
V7.1.5.1557 (6 July 2016). São Vicente, Cape Verde,
16
Here we report 4 years (December 2011–December 2015) of TGM measurements
at the Cape Verde Observatory (CVO), a clean marine background station
located in the subtropical Atlantic. The measurements presented here are part
of the EU Global Mercury Observation System (GMOS) network. The GMOS network
of sites was established in 2011 with the aim of addressing known gaps in the
spatial and temporal measurement of mercury, as well as improving knowledge
of Hg speciation. The data are being used to validate regional and global
scale atmospheric Hg models in order to improve understanding of global Hg
transport, deposition, and re-emission as well as providing a contribution to
future international policy development and implementation
(
The CVO was established in 2006 as a multilateral project between the UK, Germany and Republic of Cape Verde. Long-term atmospheric measurements include reactive trace gases including ozone, carbon monoxide, nitrogen oxides and volatile organic compounds (National Centre for Atmospheric Sciences (NCAS), University of York, UK), long-lived greenhouse gases (Max Planck Institute (MPI), Jena, Germany), and physical and chemical characterisation of aerosol (Leibniz Institute for Tropospheric Research (TROPOS), Leipzig, Germany). Details of the measurements and characteristics of the station can be found in Carpenter et al. (2010).
The CVO is positioned on the northeastern side of São Vicente
(16.85
Time series (December 2011–December 2015) of TGM data measured at the Cape Verde Observatory.
Air is sampled from the main laboratory glass manifold (10 m height of
inlet, 2 in. diameter, residence time 4 s) and then through a 2 m
length of
Seasonal cycle of TGM at CVO. The bars represent the standard deviation of the monthly averages.
Average TGM concentrations and standard deviation statistics from comparable sites in 2013 and 2014. Data from Sprovieri et al. (2016).
Instruments to make trace gas and meteorological measurements are provided by
the Atmospheric Measurement Facility (AMF), which is part of NCAS. Ozone measurements were made using a
UV photometric analyser (Thermo Electron Corporation). The instrument had a
detection limit of 0.05 ppb and a precision of
Four years of data are presented here obtained between 5 December 2011 and 5 December 2015. In calculating annual statistics, we have used data from 1 December to 30 November. The data were quality-controlled using the central GMOS-Data Quality Management (G-DQM) system (Cinnirella et al., 2014; D'Amore et al., 2015). The G-DQM allows harmonisation of data across the network and is able to acquire and process data in near-real time, allowing immediate diagnosis of issues. It was developed using harmonised standard operating procedures, which had been established over many years by European and Canadian monitoring networks, together with recent literature (Brown et al., 2010; Gay et al., 2013; Steffen et al., 2012). An additional filter has been applied to the data presented here to exclude periods when the relative humidity was higher than 90 %, as the data were prone to increased uncertainties due to water condensing in the instrument. Instrument issues led to some significant data gaps; a lamp failure caused major data gaps between July and August 2012 and May and June 2014, whilst a pump failure caused downtime between October 2012 and January 2013.
The mean TGM concentration over 2011–2015 was
1.191
The data shown in Table 1 illustrate the dominating effect of emissions from
the Northern Hemisphere compared to the Southern Hemisphere, with Mace Head
(53
TGM trends for the full year and then separated by season at the Cape Verde Observatory. The green text shows the slope estimate, with 99 % confidence intervals in brackets.
The CVO TGM monthly mean data show a weak seasonal cycle
(1.289
A smaller seasonal cycle may mean that O
An influence of air masses from the Southern Hemisphere without any
pronounced seasonal variation (Slemr et al., 2015) may be another reason for the smaller amplitude in the
seasonal cycle at the CVO. However, air mass back trajectory analyses show
that the CVO receives very little air representative of the Southern
Hemisphere (
Anthropogenic emissions of mercury affecting the Atlantic region include
emissions from coal combustion, which tend to have maximum impact in
February–March due to a dominance of air from continental regions such as
North America. This is also observed in the seasonal distribution of
anthropogenic combustion tracers such as carbon monoxide (Selin et al.,
2007; Weigelt et al., 2015;
Read et al., 2009). Ocean emissions of Hg
The Theil–Sen function (Theil, 1950; Sen, 1968) was used to evaluate the
4-year dataset inclination based on monthly TGM medians by season. The
results are shown in Fig. 4. In this function the slopes between all
Histograms of observed TGM classified by air mass for the full 2011–2015 dataset.
Over 4 years the data show a weak downward inclination
(0.042
Statistics for the individual air mass classified data.
The seasonal trends calculated here imply that there are differences in the sources of mercury that affect the winter months compared to the summer months, potentially with a smaller decline in emissions over this winter period. This may be because CVO measures Hg coming from the same source region throughout the year but that the emission from that source has not declined as much in winter as it has in summer, e.g. from residential burning. Alternatively the difference could be explained by a difference in air mass between seasons bringing air from different sources that have experienced different trends in emissions over the years. We consider this latter scenario to be the more likely explanation, since air masses originating from continental Africa, which may be influenced by ASGM or biomass burning, frequently reach the CVO in winter (Carpenter et al., 2010), but are more rare in summer. A further alternative explanation for the difference in trends between seasons would be a change in global oxidant concentrations (such as OH) and that this effect had a seasonal dependence, but there is no evidence to support this from studies that estimate OH fields (Hartmann et al., 2013).
A correlation matrix separated by season to show the correlation
between pairs of data, using the corPlot function in Openair (Carslaw and
Ropkins, 2012). The ellipses are visual representations of a scatter plot.
The colour scale highlights the strength of the correlation (red being the
strongest and blue the weakest), and the number is the
In order to understand better the drivers of the TGM behaviour, observations were classified according to the origin and pathways of air masses arriving at the CVO over a 10-day period using the UK Met Office NAME dispersion model in passive tracer mode (Ryall et al., 2001). The air mass classifications have been used previously for evaluating the source regions of reactive trace gases arriving at CVO (Carpenter et al., 2010). For this study eight geographical regions were defined (coastal African, polluted marine, Saharan Africa, Sahel Africa, North America, Atlantic marine South America, and tropical Africa), and from these seven air mass types are classified based on the percentage time spent over each of the eight regions (Fig. 5a and b). These are Atlantic and African coastal (AAC), Atlantic marine (AM), North American and Atlantic (NAA), North American and coastal African (NCA), European (with minimal African influence) (EUR), African (with minimal European influence) (AFR), and European and African (EUR/AFR). The eight regions are shown in Fig. 5a and a trajectory frequency footprint of the trajectories (using all of the data from the measurement period) for each of the seven classifications is shown in Fig. 5b.
Figure 6 shows histograms representing the data in each of the seven
classifications and Table 2 details the associated statistics. The lowest
variability in TGM was observed in air that had travelled the longest period
since contact with continental sources even though these would have been
subjected to greatest potential for ocean emissions (AM, NCA, NAA). The
lowest concentrations (1.144
Paired
Biomass burning, of both anthropogenic and biogenic origins, is prevalent in
Africa. In Northern Hemisphere Africa, burning occurs primarily in the
Sahel, moving from the northern to the southern Sahel between November and
February (Roberts et al., 2009). From Fig. 5b it would appear that there are
few trajectories which originate from this region; however, an influence from
biomass burning could be one explanation for the variable, and sometimes
higher, mercury concentrations within AFR air masses. Previous studies have
found a relationship between TGM and carbon monoxide during such episodes
(Slemr et al., 2006; Brunke et al., 2012). Figure 7 shows a correlation
analysis using a matrix method between pairs of data at the CVO for AFR air,
separated by season. O
We next consider episodes when TGM concentrations were enhanced, to
investigate the potential influence of biomass burning on the measurements.
On two occasions TGM exceeded 1.7 ng m
4-year TGM trends using a Theil–Sen function based on seasonal TGM medians separated by air mass. The data for all air masses are shown in the top left panel with the rest of the panels showing the data separated by the seven air mass classifications. The green text shows the slope estimate and the 95 % confidence intervals are in brackets. In each case the solid red line shows the trend estimate and the dashed red lines show the 95 % confidence intervals for the trend based on resampling methods.
A similar analysis was performed for the period 19 September 2015 until the 19 October 2015 and the corresponding plots and trajectories are shown in Fig. 9. In this case the period of elevated concentrations is shorter, with the episode lasting around a week. From the trajectories the air may have been influenced by air from the biomass region in Sahel Africa, which is at its most northern location in the month of October (Roberts et al., 2009). However, CO was not elevated during the period of peak [TGM] suggesting that biomass burning was not the source.
It has been previously established that West Africa is an important source region for ASGM activity (Telmer and Velga, 2009) but it is difficult to determine whether West Africa is a growing source of emissions since data have been limited and are subject to large uncertainty. It is likely, however, that the ASGM emissions are less regulated than anthropogenic emissions from coal combustion, ferrous/non-ferrous metal and cement production from Europe and the US (UNEP, 2013), and so is less likely to be decreasing in source strength.
A further analysis using the Theil–Sen function was performed to evaluate trends in the 4-year dataset within individual air masses, based on seasonal TGM medians (Fig. 10).
The overall data (top left panel) show a decrease of
From an analysis of trajectories it was found that the AFR air was within the
Sahel region outlined earlier for only around 1–2 % of the time and
mostly in October when the burning is at its most northern point (Fig. 5b).
In order to evaluate whether these air masses biased the decrease in AFR air,
a sensitivity analysis was performed on the Theil–Sen analysis with the
October data removed. The data flagged as southerly were also removed. The
updated concentration change was
Weigelt et al. (2015) observed an annual decrease in TGM concentrations at
Mace Head (53
We report a 4-year decreasing trend in total gaseous mercury (TGM)
concentrations over the subtropical North Atlantic of
The data presented here are freely available at the Centre
for Environmental Data Analysis (CEDA) at
Katie A. Read, Lucy J. Carpenter, Alastair C. Lewis, and John Kentisbeer contributed to the preparation of the manuscript. Katie A. Read and Luis M. Neves made the measurements. Zoe L. Fleming ran the NAME trajectories and assigned the air mass classifications.
The authors declare that they have no conflict of interest.
The authors acknowledge the Natural Environmental Research Council (NERC) and the Atmospheric Measurement Facility (AMF), National Centre for Atmospheric Science (NCAS) for their continued funding of the Cape Verde Observatory. The measurements of TGM were initiated due to financial support from the EU FP7-ENV-2010 project “Global Mercury observation System” (GMOS, grant agreement no. 265113). Edited by: R. Ebinghaus Reviewed by: two anonymous referees