Constraining the carbon tetrachloride (CCl₄) budget using its global trend and inter-hemispheric gradient

Abstract

Carbon tetrachloride (CCl₄) is a major anthropogenic ozone-depleting substance and greenhouse gas and has been regulated under the Montreal Protocol. However, the near-zero 2007–2012 emissions estimate based on the UNEP reported production and feedstock usage cannot be reconciled with the observed slow decline of atmospheric concentrations and the inter-hemispheric gradient (IHG) for CCl₄. Our 3-D model simulations suggest that the observed IHG (1.5 ± 0.2 ppt for 2000–2012) is primarily caused by ongoing current emissions, while ocean and soil losses and stratosphere-troposphere exchange together contribute a small negative gradient (~0 – −0.3 ppt). Using the observed CCl₄ global trend and IHG, we deduce that the mean global emissions for the 2000–2012 period are  Gg/yr (~30% of the peak 1980s emissions) and a corresponding total lifetime of  years.

1 Introduction

Carbon tetrachloride (CCl₄) is primarily used as a feedstock or processing agent but has also been used as a cleaning agent and solvent [CTOC Report, UNEP, 2011]. CCl₄ is recognized as both an ozone-depleting substance (ODS) and greenhouse gas. As of 2008, CCl₄ accounted for ~11% of total tropospheric chlorine [WMO, 2011]. The ozone depletion potential (with respect to CFC-11) is 0.82 [WMO, 2011], and the 100 year global warming potential is 1400 [WMO, 2011]. In 1987, the Montreal Protocol (MP) included CCl₄, and production and consumption were eliminated for developed countries in 1996. Developing countries (i.e., Article 5 countries) were allowed a delayed reduction, but CCl₄ was fully banned in 2010. CCl₄ continues to be legally used as a contained feedstock, e.g., for hydrofluorocarbon production, since feedstock uses are not regulated.

CCl₄ primary sinks include stratospheric photolysis and ocean and soil degradation. The current best estimate of total lifetime (τ_CCl4) is  years [SPARC, 2013]. Partial lifetimes due to atmospheric photolysis (τ_atmos), the ocean sink (τ_ocean), and the soil sink (τ_soil) are  years [SPARC, 2013],  years [SPARC, 2013], and  years [WMO, 2011], respectively (subscripts and superscripts denote 1-σ range).

MP controls have led atmospheric CCl₄ to decline at ~ 1% per year [WMO, 2011]. Current bottom-up emissions, estimated based on reported production and feedstock usage, were zero after 2007 [WMO, 2011]. There are no known substantial stocks in existing equipment or storage containers. A τ_CCl4 ~ 25 years implies a 4% per year decrease rather than the observed 1%. The 2007–2012 top-down emissions estimate derived using atmospheric observations and the current τ_CCl4 estimate was upward of 50 Gg per yr (Gg/yr) [WMO, 2011]. This very large emissions estimate difference is equivalent to ~1600 railroad tank cars of liquid CCl₄. Fraser et al. [2014] suggested that contaminated soils, toxic waste treatment facilities, and possibly chloro-alkali plants emissions could contribute 10–30 Gg/yr. De Blas et al. [2011] also observed excess CCl₄ above background in Bilbao, Spain, and attributed this to an unidentified source near the measurement site. Odabasi [2008] found that mixing bleach with surfactants or soap could form CCl₄. Unintentional CCl₄ feedstock emissions are highly uncertain [TEAP, 2011] but have been estimated to be approximately 0.5% of the total feedstock used (equivalent to 5 Gg/yr for 2011 production) [Miller and Batchelor, 2012]. None of these potential sources alone can fully explain the 50 Gg/yr discrepancy between top-down and bottom-up emissions estimates.

In this paper, we use available source and sink data in a global 3-Dimensional (3-D) Chemistry Climate Model (CCM) to test existing emissions and lifetime estimates against CCl₄ mixing ratio observations. The model simulations used herein are described in section 2. Section 3 describes the use of the observed inter-hemispheric gradient (IHG) to estimate likely emissions and then the corresponding lifetime to match the observed trend. A discussion on current gaps between bottom-up and top-down emissions and lifetime estimates is in section 4 with conclusions in section 5.

2 Model and Simulations

Our simulations are conducted with the NASA 3-D GEOS Chemistry Climate Model (GEOSCCM) Version 2 (see supporting material A), which couples the GEOS-5 GCM [Reinecker et al., 2008] with a detailed stratospheric chemistry module [Douglass and Kawa, 1999]. A CCM comprehensive evaluation shows that GEOSCCM agrees well with meteorological, transport-related, and chemical diagnostic observations [Eyring et al., 2006]. Of particular importance, GEOSCCM represents the mean atmospheric circulation as demonstrated by its realistic age-of-air, and further, realistic loss and ODS lifetimes [Waugh et al., 2007; Douglass et al., 2008; Chipperfield et al., 2014]. GEOSCCM also reasonably simulates inter-hemispheric transport and the observed ODS IHG [Liang et al., 2008; Chipperfield et al., 2014]. A GEOSCCM simulation with flux-based CCl₃F (CFC-11) and CCl₂F₂ (CFC-12) shows that the model CFC IHGs between 1995 and 2012 compare well with the National Oceanic and Atmospheric Administration—Global Monitoring Division (NOAA GMD) observations [Montzka et al., 1999; Thompson et al., 2004] (see supporting material A).

Five GEOSCCM CCl₄ simulations were performed using geographically resolved emissions from Xiao et al. [2010] (Table 1). Baseline Run A is a 58 year simulation (1960–2017) with the SPARC [2013] photochemistry and soil lifetime, and the WMO [2011] ocean lifetime. Note that our modeled lifetimes differ slightly from the recommendations due to the geographic inhomogeneity in surface concentrations as well as how lifetimes are calculated, i.e., model vs. model and observation combined (supporting material A). Four 18 year (1995–2012) runs, B–E, are initialized with January 1995 Run A initial condition but with varying lifetimes, global emissions, and emission distributions, to examine their impact on the CCl₄ budget. The CCl₄ global annual emissions used in all runs, except Run E, are top-down emissions estimates derived with the global one-box model used in recent WMO Ozone Assessments [Velders and Daniel, 2014]—yielding CCl₄ declines that match the GMD observations (Figure 1a). The GMD data are from in situ measurements and flasks [Hall et al., 2011]. Run E uses reduced global emissions ~ 35 Gg/yr for 1995–2012 and a lifetime of 26 years based on SPARC [2013]. Its CCl₄ decreases at ~ 2.2 ppt/yr—double the observed rate (Figure 1a). This result confirms the existing CCl₄ budget gap: low emissions and the current 25 year lifetime yield a global trend inconsistent with observations. We ran the last five years (2013–2017) of Run A with zero emissions, to quantify the part of the IHG due to processes other than emissions, i.e., inter-hemispheric differences in ocean and soil losses and stratosphere-troposphere exchange (STE).

Table 1. A Description of the Five 3-D GEOS Chemistry Climate Model (GEOSCCM) CCl₄ Simulations Used in This Work

	Description	Simulation Period	Average Emission 1995–2012 (Gg/yr)	Northern Hemisphere (NH) Emission Fraction (EFn)	Lifetime τ (year)	Partial Lifetimes (year)			a (ppt Gg/yr)	bg^g The combined contribution of ocean and soil losses and stratosphere-troposphere exchange (STE) to CCl4 IHG. (ppt)	τns (yr)
						τatmos	τocean	τsoil
Run A	Baseline simulation.	1960–2017	64a^a The global one-box model top-down emissions estimate for τ ~ 25.8 years.	0.937	25.8	47	79	201	0.047 ± 0.003	−0.23 ± 0.19	1.37
Run B	Decreased ocean loss. Decreased atmospheric loss forced by reduced photolysis rate.	1995–2012	35b^b Inter-hemispheric gradient (IHG)-based annual emissions calculated using the average of observed IHG from the NOAA GMD network and the Advanced Global Atmospheric Gases Experiment (AGAGE) network.	0.937	36.5c^c τ, τatmos, and τocean are determined using the global one-box model with the IHG-based emissions and the observed global trend.	62c^c τ, τatmos, and τocean are determined using the global one-box model with the IHG-based emissions and the observed global trend.	160c^c τ, τatmos, and τocean are determined using the global one-box model with the IHG-based emissions and the observed global trend.	201	0.044 ± 0.005	−0.16 ± 0.17	1.27
Run C	Repartitioning of emissions into the Southern Hemisphere (SH) with reduced global emissions.	1995–2012	50e^e The global one-box model top-down emissions estimate for τ ~ 30.7 years.	0.882	30.7d^d For Run C, τocean is determined using the global two-box model by matching the observed IHG.	47	160d^d For Run C, τocean is determined using the global two-box model by matching the observed IHG.	201	0.041 ± 0.003	−0.26 ± 0.17	1.35
Run D	f^f The latitude-dependent ocean loss rates used are 1/288 yr^−1 for 45°N –90°N, 1/222 yr^−1 for 0°N–45°N, 1/122 yr^−1 for 0°S–45°S, and 1/75 yr^−1 for 45°S–90°S. The relative strength of latitude-dependent loss rates are provided by Shari Yvon-Lewis (personal communication) and then scaled to give a 135 year τocean. As Run C, but with latitude-dependent ocean loss rates with faster degradation in the SH.	1995–2012	50	0.882	29.5	47	135	201	0.044 ± 0.003	−0.09 ± 0.16	1.44
Run E	Same lifetimes as in Run A but with reduced emissions as in Run B. This simulation does not match the observed CCl4 decline.	1995–2012	35	0.937	25.8	47	80	201	0.044 ± 0.005	−0.09 ± 0.18	1.28

• ^a The global one-box model top-down emissions estimate for τ ~ 25.8 years.
• ^b Inter-hemispheric gradient (IHG)-based annual emissions calculated using the average of observed IHG from the NOAA GMD network and the Advanced Global Atmospheric Gases Experiment (AGAGE) network.
• ^c τ, τ_atmos, and τ_ocean are determined using the global one-box model with the IHG-based emissions and the observed global trend.
• ^d For Run C, τ_ocean is determined using the global two-box model by matching the observed IHG.
• ^e The global one-box model top-down emissions estimate for τ ~ 30.7 years.
• ^f The latitude-dependent ocean loss rates used are 1/288 yr⁻¹ for 45°N –90°N, 1/222 yr⁻¹ for 0°N–45°N, 1/122 yr⁻¹ for 0°S–45°S, and 1/75 yr⁻¹ for 45°S–90°S. The relative strength of latitude-dependent loss rates are provided by Shari Yvon-Lewis (personal communication) and then scaled to give a 135 year τ_ocean.
• ^g The combined contribution of ocean and soil losses and stratosphere-troposphere exchange (STE) to CCl₄ IHG.

Figure 1

(a) The global mean CCl₄ mixing ratios from NOAA GMD stations (thick black line) and the 3-D GEOS Chemistry Climate Model (GEOSCCM) model runs A–E. (b) The scatter diagram of the 1995–2012 model mean inter-hemispheric gradient (IHG) vs. global emissions. Symbols represent annual-averaged values. Thick blue crosses on the y axis are annual mean IHGs for 2013–2017 from Run A with zero emissions. The deviation of the 2013 point from the linear regression line is due to the time needed to adjust to the abrupt change in emissions from ~50 Gg/yr to zero. IHG-emission regression lines are also shown (same color as the symbols for each run, regression slopes in parentheses). The horizontal light (dark) gray shading indicates the 2-σ range of GMD observed IHG for 2000–2012 (2007–2012).

3 Constraining the CCl₄ Budget Using Its Global Trend and Inter-Hemispheric Gradient

3.1 The Inter-Hemispheric Gradient of CCl₄

The IHG is recognized as a qualitative emissions indicator for long-lived compounds [Lovelock et al., 1973]. Our simulations show a compact linear relationship between model annual IHG and annual global emissions for all runs (r = 0.92–0.97 for Runs A–E) (Figure 1b), despite their various emissions and lifetimes. All five runs yield very similar IHG-emissions regression slopes (0.041–0.047 ppt/Gg yr⁻¹) and zero-emission intercepts (−0.09 to −0.26 ppt) (Figure 1b and Table 1).

The rate of change for atmospheric mixing ratios of a long-lived compound, e.g., CCl₄, in the two hemispheres is expressed as:

(1)

(2)

where C_n and C_s are annual mean Northern Hemisphere (NH) and Southern Hemisphere (SH) mixing ratios, respectively, inter-hemispheric exchange timescale is τ_ns, E represents global annual emissions, and EF_n is the NH emissions fraction. The coefficient f (=7.9 × 10⁻² ppt/Gg for CCl₄) is a scaling factor for converting emissions to mixing ratios [Velders and Daniel, 2014]. Ocean (α_ocean) and soil (α_soil) losses have subscripts n and s to denote hemispheres. STE also decreases surface concentrations via dilution of tropospheric air with aged stratospheric air that is depleted of long-lived ODSs [Nevison et al., 2007; Liang et al., 2008]. We express the decrease in surface mixing ratios associated with STE as − C_nα_n,STE and − C_sα_s,STE.

Subtracting equation 2 from equation 1 yields the IHG rate of change:

(3)

where ΔC_n-s is the NH-SH mixing ratio difference. We use ΔC_{n − s,ocean/soil} and ΔC_{n − s,STE} to represent the IHG due to the inter-hemispheric asymmetry associated with ocean and soil losses and STE, respectively, where . STE introduces a negative IHG (ΔC_n-s,STE < 0) at the surface since the mean Brewer-Dobson circulation features stronger horizontal mixing and downwelling in the NH, thus larger NH cross-tropopause flux [e.g., Haynes et al., 1991; Holton et al., 1995]. For CCl₄, the ∂(ΔC_n-s)/∂t term is small (1995–2012 mean ~ −0.03 – −0.07 ppt/yr for all model runs). As a first-order approximation, equation 3 simplifies to:

(4)

where a = τ_nsf(EF_n − 0.5) is the slope, and is the zero-emission intercept (Figure 1b). Hence, from equation 4 the IHG should be linearly proportional to emissions.

STE and the surface ocean and soil losses, together, contribute a small negative Northern vs. Southern IHG, −0.09 – −0.26 ppt, as indicated by the zero-emission intercepts from all five runs. The modeled IHGs for 2013–2017 from Run A with zero emissions (heavy blue pluses on Figure 1b) decrease rapidly from 2013 to 2017 and approach the −0.23 ± 0.16 ppt intercept only a couple of years after emissions cease. This confirms that the zero-emission intercept reflects the IHG due to processes other than emissions. The b value from Run D (−0.09 ± 0.16 ppt) is 0.17 ppt higher than that from Run C (−0.26 ± 0.17 ppt), suggesting that faster loss rates in the SH ocean, particularly in the SH extratropics, led to an increased IHG, but this increase is small (< +0.2 ppt).

Global emissions play the predominant role, and the CCl₄ IHG mainly reflects ongoing emissions. The model mean IHG decreased accordingly from 2.6 ppt for Run A, to 1.7 ppt for Run C, and 2.0 ppt for Run D, and to 1.3 ppt for Run B and 1.4 ppt for Run E, consistent with their relative global emissions strength. The minor IHG differences between runs with same global emissions reflect the secondary impacts of differences in EF_n, τ_ns, and sinks.

Using the model-based regression slope (0.047 ppt/Gg yr⁻¹) and EF_n of 0.94 for Run A, we derive a τ_ns of 1.37 years. The other four runs yield similar τ_ns of 1.27–1.44 years (Table 1). GEOSCCM results from a separate flux-based simulation also show a linear relationship between IHG and global emissions for CFC-11 and CFC-12 and similar τ_ns (1.39–1.40 years). The GEOSCCM τ_ns agrees well with multi-model mean τ_ns ~ 1.39 ± 0.18 years estimated using SF₆ [Patra et al., 2011]. Emission shifts between the higher and lower latitudes in the same hemisphere can have a small impact on τ_ns. Patra et al. [2011] estimated that doubling SF₆ sources between 0°N and 30°N from ~ 1 Gg/yr to ~ 2 Gg/yr (~6 Gg/yr global emissions), led to a < 15% decrease in τ_ns for SF₆. Similarly, loss rate changes also impact τ_ns slightly. For example, the Run D slightly longer τ_ns (1.44 years) is due to its faster ocean loss rates in the SH extra-tropics (45°S –90°S) and thus a longer time needed to communicate this loss with the NH. The small variability in τ_ns among our five runs mainly reflects the surface loss changes and STE dilution due to atmospheric photolysis rate changes.

3.2 The Global Two-Box Model

We use a global two-box model from equations 3 and 5 to simulate the IHG and long-term global trend of global mean mixing ratio (C) for CCl₄:

(5)

In equation 3, we use a north-south exchange time τ_ns of 1.37 years—the value derived from the 3-D CCM. The IHG associated with surface losses, ΔC_{n-s,Ocean/Soil}, is calculated in the two-box model as C_n/(τ_atmos + τ_ocean,n + τ_soil,n) − C_s/(τ_atmos + τ_ocean,s + τ_soil,s), where ocean and soil partial times are scaled by their respective hemispheric ocean and soil area, as in the 3-D model. The two-box model shows that, assuming uniform surface loss rates everywhere, the IHG (ΔC_{n-s,Ocean/Soil}) generated purely due to the asymmetry in the NH and SH ocean and land areas is ~ 0.3 ppt for τ_ocean = 80 years and τ_soil = 200 years. Using a reasonably short 80 year τ_ocean [SPARC, 2013], a 1000 year upper-limit τ_soil [WMO, 2011] and an extra +0.2 ppt to account for faster ocean loss rates in the SH, in particular the SH extratropical ocean (section 3.1), we derive that the likely upper limit for ΔC_{n-s,Ocean/Soil} is ~ 0.6 ppt. Using ΔC_{n-s,Ocean/Soil} from the two-box model calculations and b values from the 3-D runs, we estimate that ΔC_n-s,STE is ~ −0.5 ppt for τ_atmos ~ 47 years and present-day CCl₄ levels. The STE contribution is smaller (smaller |ΔC_n-s,STE|) for a longer τ_atmos or lower CCl₄ levels (see supporting material B for information on how the two-box model IHG varies with E, EF_n, and partial lifetimes).

Using the same input parameters for Run A and Run C, i.e., partial lifetimes, global emissions, and EF_n, and ΔC_n-s,STE ~ −0.5 ppt, the two-box model yields very similar IHGs, regression slopes, and zero-emission intercepts as the 3-D runs (supporting material B). This suggests that the global two-box model provides quantitatively consistent information on the IHG and its dependence on sources and sinks as the 3-D CCM.

3.3 Estimates of CCl₄ Global Emissions and Lifetime

We use the observed IHG, global mean trend, and likely CCl₄ emissions distribution in the two-box model to estimate likely ranges for emissions and τ_CCl4. Figure 2 shows the two-box model calculated global trend as a function of E and τ_CCl4. Observed global mean CCl₄ trends from the GMD and the Advanced Global Atmospheric Gases Experiment (AGAGE) [Prinn et al., 2000] networks are −1.0 – −1.2 ppt/yr (2000–2012) with annual trends range between −0.9 ppt/yr and −1.5 ppt/yr. Purple contours indicate the range of E and τ_CCl4 that would match the observed mean IHG ~ 1.5 ppt (1.1–2.0 ppt, 2-σ range), using the current best estimate EF_n of 0.94 [Xiao et al., 2010]. The regime where the −0.9 and −1.5 ppt/yr trend contours intersect the purple contours (dark gray shading) outlines the most likely range of E and τ_CCl4 that can reproduce the observed trend and IHG. Assuming that 94% of the global emissions reside in the NH, the mean observed IHG and global trend averaged between the GMD and AGAGE measurements imply mean global emissions of 39 Gg/yr for the 2000–2012 period (1-σ year-to-year variance of ±4 Gg/yr) and a corresponding τ_CCl4 of 35 years. The IHG-based mean global emissions for the 2007–2012 period are ~ 35 Gg/yr.

Figure 2

CCl₄ global mean trend (ppt/yr) as a function of total lifetime and emissions from the two-box model (gray contours). Purple contours indicate the emissions and τ_CCl4 ranges that yield IHGs within the observed 1.1–2.0 ppt range (2-σ) between 2000 and 2012, using the current best estimate EF_n of 0.94. Red (Advanced Global Atmospheric Gases Experiment (AGAGE)-based) and blue (GMD-based) numbers show emissions and lifetimes derived using the observed IHG and trend for individual years (2000–2012). The dark (light) gray shading outlines the range of emissions and τ_CCl4 that can be reconciled with the observations for EF_n of 0.94 (0.88–1.00). The black diamond symbol shows our current best estimate for τ (thick and thin red bars indicate 1-σ and 2-σ uncertainties, respectively) and the upper limit bottom-up potential emissions for 2007–2012 (thick blue bar shows 1-σ variance) with 1-σ uncertainty shown in black-hatched shading.

The above emissions and lifetime estimates depend upon the assumed EF_n, which is not accurately known. The uncertainty of our top-down CCl₄ global emissions and τ_CCl4 estimates can be determined based on the likely range of EF_n. Using the observed IHG and the largest EF_n possible—1.0, we deduce that the minimum average global emissions necessary to reproduce the atmospheric CCl₄ observations are ~ 34 Gg/yr (τ_CCl4 ~ 37 years) for the 2000–2012 period. Similarly, we can derive the upper limit global emissions with a lower limit EF_n estimate. ODS usages generally are thought to scale with population, and previously reported EF_n for major long-lived ODSs are all ≥ 0.90 [McCulloch et al., 2001, 2003; AFEAS, 2001]. For example, 0.96 for CFC-11 in the 1990s and 0.94 in the 2000s, 0.95 for CFC-12 in the 1990s and 0.90 in the 2000s, and 0.97 for CH₃CCl₃—calculated using geographically resolved emissions (courtesy of A. McCulloch and the Global Emissions InitiAtive, http://www.geiacenter.org). The EF_n lower limit (0.88) is estimated by doubling current SH emissions. This yields an upper limit global emissions estimate ~ 45 Gg/yr (τ ~ 32 years) for the 2000–2012 period. The corresponding upper and lower limits for the 2007–2012 global emissions are 31 and 40 Gg/yr, respectively.

Even when the 2007–2012 upper limit bottom-up emissions estimate (14 ± 12 Gg/yr) (section 4.1) is considered, it cannot be reconciled with observations as this magnitude of emissions yields substantially smaller IHG and a fast decline rate of ~ −2.6 ppt/yr (Figure 2).

4 Current Gaps

4.1 The Gap Between Emission Estimates

Bottom-up potential CCl₄ emissions were estimated to be near zero after 2007, as derived from the difference between reported production in excess of amounts used as feedstock and amounts destroyed [WMO, 2011]. An upper limit to potential emissions, estimated using a 2% fugitive emissions rate from feedstock and an assumed 75% destruction efficiency [WMO, 2011], was 34 Gg/yr for 2000–2012 (14 Gg/yr for 2007–2012). With a τ_CCl4 ~ 25 years, the 2000–2012 top-down emissions estimate is ~ 63 Gg/yr (46–80 Gg/yr, 1-σ uncertainty) (56 Gg/yr for 2007–2012)—a 30 Gg/yr discrepancy compared to the upper-limit bottom-up emissions. The IHG-based top-down emissions estimate from this work is  Gg/yr for 2000–2012 (subscript and superscript denote the lower and upper limit estimates), which reduces the gap with bottom-up estimates. However, the 2007–2012 discrepancy remains large (Δ = ~ 20–40 Gg/yr).

The trend from bottom-up emissions estimate cannot be reconciled with the mixing ratio observations either. The observed global mixing ratios show a slow steady downward trend for 2000–2012 (Figure 3a), calculated from a least squares fit to be a 77.7 year exponential decay time scale (supporting material C), in contrast to the sharp bottom-up emissions decline from 100 Gg/yr in 1999 to 10 Gg/yr after 2008 (Figure 3c).

Figure 3

(a) Observed CCl₄ (NOAA GMD stations) and the global mean values (black line). (b) As Figure 3a but for mixing ratio anomalies—the right axis shows the equivalent burden anomalies (Gg). c) CCl₄ bottom-up and top-down emissions. Green line shows the upper limit bottom-up potential emissions estimate from reported production, feedstock usage, and the amount destroyed. The one-box model top-down emissions estimate is derived with τ_CCl4 of  years (thin red line, red shading indicates 1-sigma uncertainty due to uncertainties in lifetime and measurements). Thin blue line indicates the top-down emissions estimate (blue shading indicates lower-upper limit uncertainty estimate) derived in this work using the GMD observed IHG.

We de-trend the GMD surface mixing ratio observations and apply a 25 month ½-amplitude Gaussian low-pass filter to estimate year-to-year variations in anomalous abundances (deviations from the long-term trend). These anomalies reflect interannual variations in loss processes (surface and atmospheric) and/or emissions. The smoothing reveals three periods of change: (1) 1995–2005—steady increase in CCl₄ anomalies (mean ~+0.2 ppt/yr), (2) 2007–2011—small decrease (−0.1 ppt/yr), and (3) 2011–2013—an anomalous jump of +0.1 ppt/yr. Using a mixing ratio to global abundance conversion factor of 3.95 × 10⁻² ppt/Gg, these observed anomalies imply (1) ~ +5 Gg/yr increase in period 1 during 2004, (2) a ~ −3 Gg/yr decrease in 2008, and (3) an increase of +3 Gg/yr in 2012 in abundance anomalies. These observed small variations are inconsistent with the large fluctuations in bottom-up emission estimates (year-to-year changes between 10 and 30 Gg/yr for most of the years between 2000 and 2012). While inaccurate bottom-up emissions estimate is a likely cause of this inconsistency, an alternative explanation is that the year-to-year variations of loss rates may dampen large emission fluctuations. For example, a delayed atmospheric loss response (due to the required transport time from surface to the upper atmosphere, and back down to surface) can potentially offset the impact of surface emission decreases due to increased emissions in earlier years. STE interannual variability can also impact surface abundance fluctuations. This points to the need for a better understanding of the coupled impact of sources and sinks on year-to-year variations in abundances.

4.2 The Gap Between Lifetime Estimates

The current best estimate τ_CCl4 of  years is inconsistent with the observations, as the implied mean global emissions of  Gg/yr for 2000–2012 can produce IHG matching the observations only if EF_n equals —a value much lower than those reported for major ODSs. The observed IHG and trend imply a τ_CCl4 of 35 years (uncertainty range, 32–37 years). This requires much longer partial lifetimes than current best estimates. In Run B, we tested τ_atmos by increasing it from 47 years to 62 years (the upper limit for τ_atmos). Such an increase requires a ~ 60% reduction in CCl₄ photolysis rate—highly unlikely as it greatly exceeds the lab-measured 15–20% cross-section uncertainty [Rontu Carlon et al., 2010; SPARC, 2013]. High-altitude CCl₄ measurements are scarce, but a comparison of Run B with two balloon profiles shows a model high bias in the critical stratospheric photolysis loss region (10–70 hPa) (Figure A3). Keeping τ_atmos unchanged, a τ_CCl4 of 35 years implies a need for significant increases of τ_ocean and/or τ_soil from current best estimates. This suggests a need for new measurements to re-evaluate partial lifetimes, τ_ocean and τ_soil in particular. However, it is important to note that the current 25 year lifetime estimate, even with its 2-σ range, cannot reconcile mixing ratio observations with bottom-up emission estimates.

5 Summary

CCl₄ was increasing in the atmosphere until the early 1990s and is now in decline [WMO, 2011]. This decline results from regulations by the Montreal Protocol. The 1990–2006 decline was caused by an emissions decrease and loss processes [WMO, 2011; SPARC, 2013]. The current CCl₄ decline should be primarily determined by the lifetime, because post-2007 bottom-up emissions are estimated to be near zero. The slow ~ 1%/yr decline cannot be reconciled with our current ~ 25 year lifetime estimate (implied top-down emissions ~ 56 Gg/yr between 2007 and 2012) derived from CCMs and atmospheric, ocean, and soil observations [SPARC, 2013].

We simulate CCl₄ using a fully coupled 3-D chemistry-climate model with a state-of-the-art photochemical loss scheme, along with current estimates of ocean and soil sinks. Our CCM results show that the inter-hemispheric gradient (IHG) and the global trend provide useful information for quantitatively constraining CCl₄ emissions and lifetime estimates:

1. Near-zero emissions from the UNEP reported production and feedstock usage in the recent years cannot be reconciled with the observed CCl₄ decline, year-to-year variability, and the IHG. At a minimum, mean global emissions of 34 Gg/yr after 2000 (31 Gg/yr for the 2007–2012 period) are required to reproduce the observed IHG.
2. Ocean and soil losses contribute at most +0.6 ppt to the IHG, assuming a long soil lifetime of 1000 years (upper limit estimate) and a short ocean lifetime of 80 years (lower limit estimate) and fast loss rates in the SH ocean. Stratosphere-troposphere exchange introduces a negative gradient of about −0.5 ppt, which tends to offset the contribution from surface losses. The observed CCl₄ IHG mainly reflects current emissions and its NH-SH asymmetric partitioning.
3. Using the observed IHG and global trend and our knowledge of emissions distribution, we deduce that the mean global emissions during 2000–2012 are  Gg/yr and a corresponding total lifetime of  years. Subscript and superscript denote the lower and upper (upper and lower) limit estimates for emissions (lifetime). This would necessitate longer partial lifetimes, in particular ocean and/or soil partial lifetimes, than the current best estimates.

Acknowledgments

We thank Mark Schoeberl for helpful discussions and Elliot Atlas for providing balloon measurements. All data presented in this paper are available from the corresponding author upon direct request. The authors thank the referees for their valuable comments.

The Editor thanks two anonymous reviewers for their assistance in evaluating this paper.