The Sensitivity of Polar Ozone Depletion to Proposed Geoengineering Schemes

See allHide authors and affiliations

Science  30 May 2008:
Vol. 320, Issue 5880, pp. 1201-1204
DOI: 10.1126/science.1153966


The large burden of sulfate aerosols injected into the stratosphere by the eruption of Mount Pinatubo in 1991 cooled Earth and enhanced the destruction of polar ozone in the subsequent few years. The continuous injection of sulfur into the stratosphere has been suggested as a “geoengineering” scheme to counteract global warming. We use an empirical relationship between ozone depletion and chlorine activation to estimate how this approach might influence polar ozone. An injection of sulfur large enough to compensate for surface warming caused by the doubling of atmospheric CO2 would strongly increase the extent of Arctic ozone depletion during the present century for cold winters and would cause a considerable delay, between 30 and 70 years, in the expected recovery of the Antarctic ozone hole.

Geoengineering schemes have been proposed to alleviate the consequences of global warming (13) by continuous injection of sulfur into the stratosphere. Volcanic eruptions in the past have shown that strongly enhanced sulfate aerosols in the stratosphere result in a higher planetary albedo, leading to surface cooling (4). On the other hand, the potential for exceedingly high Arctic ozone depletion resulting from the simultaneous presence of high surface area density (SAD) of sulfate aerosols and cold conditions in the polar stratosphere is known (5, 6) but was not quantified in the content of geoengineering (13).

In this report, the impact of enhanced sulfate aerosol (due to geoengineering) on future chemical polar ozone depletion is quantified. Our analysis is based on the past dependence of ozone loss on aerosol content derived via the combination of a model and measurement for both the Antarctic and Arctic and by taking into account the expected future stratospheric halogen loading. We describe this dependence by the empirical relation between observed chemical ozone loss and the potential for the activation of chlorine (PACl). PACl accounts for year-to-year variations in temperature, sulfur burden, and the halogen content of the stratosphere. This relation is shown to be valid based on past observations, including 4 years of volcanically enhanced aerosol loading in the stratosphere.

Severe chemical loss of ozone over the Arctic and Antarctica is caused by anthropogenic halogens. The combination of very low temperatures and increasing sunlight after the polar night results in a strong transformation of chlorine from reservoir forms to reactive radicals, leading to the rapid destruction of polar ozone (7). Since the 1990s, most of the available ozone has been destroyed in the Antarctic lower stratosphere (between 12 and 20 km in altitude), which corresponds to a loss in column ozone of about 120 to 150 DU (or Dobson units, one of which equals 2.687 × 1016 molecules per cm2) (8). In the Arctic, the interannual variability of temperatures, and therefore of ozone depletion, is much larger. Chemical depletion of ozone in recent cold Arctic winters exceeded 100 DU [see the supporting online material (SOM)] (811). Over the next half-century, the stratospheric halogen loading—commonly quantified by a measure referred to as Effective Equivalent Stratospheric Chlorine (EESC)—is projected to slowly decline (12). Close to the year 2070, EESC is predicted to reach values last seen in 1980, a benchmark for the recovery of polar ozone (12, 13). Our work is motivated by the concern that elevated SAD attributable to geoengineering will lead to additional ozone depletion that may delay the recovery of polar ozone.

We demonstrate the importance of chlorine activation on cold, liquid sulfate aerosols for polar ozone depletion by comparing two relations between chemical ozone loss and chlorine activation based on past observations (Fig. 1). The first relation is based on a previous description of polar chlorine activation: the polar stratospheric cloud (PSC) formation potential or PFP (8). PFP describes the fraction of the polar region that is below the threshold temperature for existence of PSCs (TPSC). For cold Arctic winters, more than 10% of the vortex region is cold enough to support PSCs, whereas for warm winters, PFP is close to zero. The derived linear relation is compact, except for values derived for the four winters after the eruption of Mount Pinatubo in June 1991.

Fig. 1.

(A) Relationship between chemical loss of Arctic ozone in DU and PFP, averaged between 380 to 550 K potential temperature and mid-December and March, for several winters between 1992 to 2005. A linear fit (black line) was derived, excluding the years 1992 and 1993 after the Mount Pinatubo eruption. (A) is adapted from (6), and the ozone loss value for 2005 is taken from (9). (B) As in (A) but with PACl (14) instead of PFP as the abscissa. PACl includes SAD in its formulation. Data for winters 1992 to 1995 (denoted within the panels), which had high SAD as a result of the eruption of Mount Pinatubo, now fall along the compact, near-linear relation once the effect of SAD on chlorine activation is considered.

The second relation (Fig. 1B) accounts additionally for stratospheric aerosol loading. Tilmes et al. (14) defined the potential for the activation of chlorine, which is similar to PFP but uses a threshold temperature for chlorine activation (TACl) as described by Drdla (15). A description of the calculation of this temperature can be found in (14), which is a function of temperature, ambient H2O, and SAD. For background levels of SAD (e.g., an atmosphere not perturbed by large volcanic eruptions) and present-day values of EESC, PACl is comparable to PFP. After a strong volcanic eruption, PACl is larger than PFP because the resulting sulfate aerosols (fig. S1) provide surfaces that lead to efficient chlorine activation (5, 15). The linear relation between chemical loss of Arctic ozone and PACl (Fig. 1B) is compact, including data collected during the four winters that followed the eruption of Mount Pinatubo (1992 to 1995).

For Antarctica, an empirical relation between chemical ozone loss and PACl cannot be inferred, because available observations do not span an appreciable range of ozone loss. Therefore, the Antarctic relation is defined based on results from the NCAR Whole Atmosphere Chemistry Climate Model 3 (WACCM3), for changing halogen content of the stratosphere between 1960 and the present (fig. S2). The model results are in good agreement with recent data, supporting the validity of the approach (14). Model simulations of Antarctic ozone loss (fig. S2) combined with recent data suggest that the relation between chemical ozone loss and PACl is linear until loss saturation occurs, which supports the validity of the approach used to estimate Arctic ozone loss for the geoengineering scenarios.

The empirical relation between chemical ozone loss and PACl (which incorporates the stratospheric sulfur burden) provides a tool to assess the risk of future ozone loss caused by geoengineering. We consider three different hypothetic future SAD scenarios:

  1. The background case assumes no volcanic perturbation and is based on SAD data from the year 2000 (16).

  2. The geoengineering scenario discussed by Crutzen (1) is termed “GeoEng_Large_Aerosol.” Crutzen estimated that roughly 5.3 Tg (1 Tg = 1012 g) of stratospheric sulfur (S) would counteract surface warming due to doubled atmospheric CO2. He considered volcanic-sized particles that require injections of 2 Tg S/year to maintain. For this case, we estimate SAD by multiplying observed SAD (16) in 1992 by 0.53 [5.3 Tg S (Crutzen)/10 Tg S (observed Pinatubo)] for all years.

  3. The geoengineering scenario presented by Rasch et al. (2) is denoted “GeoEng_Small_Aerosol.” They found that an injection of 1.5 Tg S/year, using particles considerably smaller than those assumed by Crutzen (1), would achieve the same radiative effect. Smaller aerosols are expected to cool more efficiently than large aerosols because of the dependence of backscattering on particle size. Furthermore, smaller aerosols have a smaller effect on stratospheric temperature. The GeoEng_ Small_Aerosol case has a perturbation to SAD three times as large as the GeoEng_Large_Aerosol perturbation, owing to the dependence of SAD on the particle radius (17). The resulting SAD used here is based on the mean of the Rasch et al. (2) perturbations resulting from injections of 1 and 2 Tg S/year, because an injection of ∼1.5 Tg S/year was found to effectively counteract global warming for doubled atmospheric CO2 (2).

The impact of these three sulfate loadings on PACl and ozone loss for past meteorological conditions is discussed in the SOM (see also fig. S3).

To assess the risk of future polar ozone loss caused by enhanced sulfate attributable to geoengineering, we select meteorological conditions for a recent very cold Arctic winter, a moderately cold Arctic winter, and a typical Antarctic winter, and apply these conditions to a model constrained by an estimate of future EESC (13). PACl and ozone depletion are quantified for the various cases (Figs. 2 and 3). Our results should be viewed as a “generic assessment” of the impact of geoengineering on future polar ozone, assuming a yearly injection of stratospheric sulfate. We assume geoengineering to begin, hypothetically in the year 2010, with a steady rise of SAD over the first 5 years until saturation is achieved. Results for other start dates can be visualized by simply connecting the background PACl case to the two geoengineering scenarios, for any specific start year or duration of SAD rise time.

Fig. 2.

(A and B) The temporal development of PACl between 2010 and 2050, taking into account changing EESC for two geoengineering cases, with observed temperatures for a very cold Arctic winter (A) and a moderately cold Arctic winter (B). The temporal evolution of PACl for background SAD (solid line), taking into account changing EESC, is also shown. Finally, values of PACl based on observed SAD, temperature, and EESC, are shown (dotted lines). (C and D) Chemical ozone loss versus time, derived from PACl (top panels) for the various SAD cases (dark gray, GeoEng_Large_Aerosol case; light gray, GeoEng_Small_Aerosol case), is shown for meteorological conditions corresponding to a very cold Arctic winter (C) and a moderately cold Arctic winter (D). The ozone loss estimates are based on the linear relationship between chemical loss and PACl for the Arctic.

Fig. 3.

(A) The temporal evolution of PACl between 2010 and 2050, taking into account changing EESC for two geoengineering cases and for background aerosol, for the temperature conditions of a typical Antarctic winter. (B and C) Antarctic chemical ozone loss derived from PACl (A) for the three SAD cases (colors as in Fig. 2), via the linear relationship between ozone loss and PACl for Antarctica. The vertical extent of the Antarctic ozone hole is either assumed to remain fixed at present levels, denoted as “Fixed Saturation” (B) or assumed to extend vertically downward in altitude by ∼2 km, denoted as “Variable Saturation” (C).

The scenario that uses meteorological conditions for a recent very cold Arctic winter characterizes the maximum perturbation to PACl and Arctic ozone loss due to geoengineering (Fig. 2, A and C). PACl in the Arctic would exceed the maximum value of PACl for background conditions until about 2055 for GeoEng_Large_ Aerosol and through the end of this century for GeoEng_Small_Aerosol. The estimate for a moderately cold winter (e.g., winter 2003) illustrates the response to geoengineering for meteorological conditions that are representative of about half of the past 15 Arctic winters (Fig. 2, B and D).

Injection of sulfur in the near future (during the next 20 years) can have a strong impact on Arctic ozone depletion. If small-sized aerosols are used, ozone loss between 100 DU (moderately cold winters) and 200 to 230 DU (very cold winters) can be reached. For very cold winters, which occurred 25% of the time in the past 15 years, the estimated ozone depletion is comparable to the total amount of available ozone in the Arctic lower stratosphere (fig. S2). Under these conditions, the SAD perturbation associated with the GeoEng_ Small_Aerosol case could possibly result in a saturation of chemical loss of Arctic ozone, leading to a drastically thinner ozone layer than presently observed. For the SAD perturbation associated with GeoEng_Large_Aerosol, chemical loss of Arctic ozone could exceed 150 DU during very cold winters and 70 DU for moderately cold winters.

The doubling of atmospheric CO2 with respect to preindustrial values is expected to occur between 2050 and 2100 (18). In this time frame, chemical ozone loss attributable to geoengineering reaches 125 to 150 DU for very cold Arctic winters, as compared with 80 DU for the background case. For either geoengineering case, ozone depletion would exceed presently observed values, because the recent spate of very cold Arctic winters occurred for low (i.e., near background) values of SAD. A moderately cold winter would reach 60 to 80 DU of ozone depletion, a value observed in the past only for very cold Arctic winters. Therefore, 75% of all winters would result in ozone loss of at least 60 to 80 DU, and possibly as high as 150 DU, if geoengineering through stratospheric sulfur injections is used to mitigate global warming.

The predicted future evolution of Antarctic ozone and PACl is depicted in Fig. 3. We show results for two different assumptions: either future chemical loss of Antarctic ozone due to geoengineering will continue to saturate at present-day values of 150 DU (8) (Fig. 3B), or else future chemical loss will saturate at higher levels, resulting from the likely strong increase of SAD at altitudes of 10 to 12 km in the Antarctic stratosphere, where ozone loss is presently not saturated (Fig. 3C). Indeed, enhanced Antarctic ozone loss at 10- to 12-km altitudes was observed after the eruption of Mount Pinatubo (19). For Fig. 3C, we assume that geoengineering will lead to a downward extension of ozone loss, adding another 15 DU (the amount between altitudes of 10 and 12 km) to the total column abundance of ozone that could be lost.

The primary effect of geoengineering on Antarctic ozone would be to delay the recovery of the Antarctic ozone hole. The time when Antarctic ozone loss drops below saturation in late winter would be delayed by 15 years for GeoEng_Large_Aerosol and by more than 30 years for GeoEng_Small_Aerosol (Fig. 3B). Assuming a vertical expansion of the ozone hole, the first stage of recovery would be delayed by another 30 years for both cases (Fig. 3C). The recovery of Antarctic ozone to conditions that prevailed in 1980 would be delayed until the last decade of this century by geoengineering, assuming no vertical expansion of the ozone loss region (Fig. 3B). However, if the vertical extent of the ozone hole were to increase, this stage of recovery would not occur within this century (Fig. 3C).

Our estimates of the risk of geoengineering do not consider several important effects. A possible consequence of a stratospheric sulfur injection, which was not considered here, is a strengthening of the polar vortex, because of the stronger temperature gradient between high and low latitudes caused by enhanced stratospheric aerosol (20). This could increase the frequency of cold polar winters, leading to even greater ozone loss than estimated. On the other hand, volcanic aerosols might result in dynamic instabilities, causing earlier major stratospheric warming at the end of Arctic winters, which results in smaller PACl and therefore less ozone loss. Also, enhanced sulfate aerosols could suppress denitrification in the polar vortices in a manner that might affect the linear relation between ozone loss and PACl. Enhanced ozone loss through geoengineering would also likely result in a thinning of the ozone layer at mid-latitudes due to the export of polar processed air, as observed during the period of rising halogen loading (1979 to 1995) (21). Further, the enhanced stratospheric aerosol levels would disturb ozone photochemistry in mid-latitudes, resulting from the suppression of stratospheric NOx, leading to even further ozone depletion (22). Additional uncertainty results from our use of idealized aerosol size and perturbation. Finally, the impact on ozone of any future major volcanic eruption would likely exceed, by large amounts, the ozone loss that was observed after the eruption of Mount Pinatubo and El Chichón, because additional volcanic aerosols will be acting on an already perturbed stratospheric SAD layer. Comprehensive chemistry–climate model simulations, considering these effects and perhaps others, are needed to fully evaluate the impact of geoengineering on atmospheric ozone.

A substantial increase of stratospheric sulfate aerosol densities caused by various geoengineering approaches will likely result in strongly enhanced chemical loss of polar ozone during the next several decades, especially in the Arctic. The expected recovery of the Antarctic ozone hole due to a reduction in stratospheric halogen loading, brought about by the implementation of the Montreal Protocol, could be delayed by between 30 and 70 years by the aerosol perturbation associated with geoengineering.

Supporting Online Material

SOM Text

Figs. S1 to S3


References and Notes

View Abstract

Navigate This Article