RealClimate logo

Yet more aerosols: Comment on Shindell and Faluvegi

Guest post from Drew Shindell, NASA GISS

Our recent paper “Climate response to regional radiative forcing during the twentieth century”, has generated some interesting discussion (some of it very ‘interesting’ indeed). So this post is an attempt to give a better context to the methods and implications of the study.

First, some history. Global model responses to aerosols have been looked at since the early 1990s (Taylor and Penner, 1994; Mitchell et al, 1995, Santer et al, 1995). These studies and subsequent ones have shown that when a forcing is spatially concentrated, the regional climate response does not closely follow the spatial pattern of the forcing. These two figures show an example of that from two recent models (GISS ModelE and GFDL). Despite extremely large localized forcings over the industrialized areas, the climate response is spread out much more broadly in the zonal direction. Similarly, although forcing is extremely large over India and Southeast Asia, those areas show only very weak warming. In particular, the Arctic climate response can be quite different from what the local forcing would imply.

Figure 1. Ensemble mean annual average 1880–2003 radiative forcing (Fs, the top-of-the-atmosphere forcing with fixed SSTs and sea-ice, left column) and the surface air temperature (SAT) response (ºC, local linear trends, right column) from 5-member ensemble simulations driven by tropospheric aerosols including their direct radiative effect only (top row) and both their direct and indirect (via cloud cover) effects (bottom row). [Shindell et al., 2007].

Figure 2. Annual mean-adjusted radiative forcing (W/m2) between years 2100 and 2000 from tropospheric aerosols and ozone changes simulated under an A1B scenario (top) and annual surface air temperature change (°K) from the 2000s (years 2001–2010) to the 2090s (years 2091–2100) due to those same short-lived species in the GFDL model [Levy et al., 2008].

In our paper, we wanted to characterize the geographic forcing/response relationship more clearly. Prior studies had looked at particular scenarios or time periods when forcings were typically changing over much of the world (albeit most strongly in certain regions). So we put idealized forcings from GHGs, aerosols, and ozone in the tropics, mid-latitudes and polar regions to see what would happen. The results showed that the temperature response in the tropics, like the global mean, is only mildly sensitive to the location of forcing. That is, you get an enhanced tropical response to forcing in the Northern Hemisphere extratropics (where you can activate strong positive feedbacks like snow/ice albedo), but the enhancement is only 40-50% over that found with forcings applied elsewhere. In contrast, the extratropical zones are much, much more sensitive to local radiative forcing than to tropical forcing or to forcing in the opposite hemisphere. So to quote from the paper

“global and tropical mean temperature trends during the twentieth century would have been quite similar if short-lived-species radiative forcing had been distributed homogeneously rather than being concentrated in the northern extratropics. Regional concentration of forcing contributed to the departures of Northern Hemisphere mid-latitude and Arctic temperature trends from the global or Southern Hemisphere extratopical means, however.”

We then used the regional forcing/response relationships to derive the aerosol forcing needed to explain the observed global and regional temperature trends. Our results have a substantial uncertainty range which arises primarily from the influence of unforced, internal variability. The global mean preindustrial to present-day aerosol forcing we calculate is -1.31 +- 0.52 W/m2, consistent with the IPCC AR4 range of -0.6 to -2.4 W/m2.

We also estimated aerosol forcing for the tropics and Northern Hemisphere mid-latitudes for several time periods, and compared with historical emissions estimates to tie the forcings to sulfate or black carbon (BC) aerosols when possible. The results show, for example, that nearly all CMIP3 models require strong aerosol cooling at Northern Hemisphere mid-latitudes during the 1931-1975 period to capture both the global mean trends and the NH mid-latitude versus Southern Hemisphere extratropics temperature trends (many CMIP3 models had both sulfate and BC, but not necessarily the correct amounts as modeling their forcing directly is quite uncertain, hence we compared the CMIP3 models’ responses to non-aerosol forcings with observations to see how well they could do without aerosols). During the last 3 decades (1976-2007), the best fit to the temperature responses in the models require negative forcing from tropical aerosols but positive forcing from Northern Hemisphere mid-latitude aerosols. It’s the latter, the positive Northern Hemisphere mid-latitude aerosol forcing that leads to the strong warming impact on the Arctic as well, as the Arctic responds to mid-latitude and local forcing, but the local forcing is primarily driven by mid-latitude emissions that are transported to the Arctic, so the overall climate response ends up being closely tied to Northern Hemisphere mid-latitude emissions. Given the strong sensitivity of the Northern Hemisphere extratropical zones to aerosol forcing, it’s then understandable that those areas could have cooled during the mid 20th century when the aerosol forcing we calculate was substantially larger than greenhouse gas forcing (in absolute magnitude).

A big uncertainty is still the influence of unforced internal variability, which we estimated from coupled ocean-atmosphere climate runs. Though that contribution is large, it was still not large enough to account for many of the mid-latitude and Arctic temperature trends without including aerosol forcing. For many cases, the influence of aerosols and internal variability were comparable in size. Though the influence of internal variability leads to a substantial uncertainty range in our results, they are nonetheless useful as other techniques of estimating aerosol forcing of climate have comparably large or larger uncertainties. These include ‘forward’ modeling from emissions to concentration to optical properties (e.g. see [Schulz et al., 2006]), and various estimates based at least in part on satellite observations (see this previous post).

Some of the most interesting conclusions of the study include those relating to the Arctic. For example, we estimate that black carbon contributed 0.9 +/- 0.5ºC to 1890-2007 Arctic warming (which has been 1.9ºC total), making BC potentially a very large fraction of the overall warming there. We also estimated that aerosols in total contributed 1.1 +/- 0.8ºC to the 1976-2007 Arctic warming. This latter aerosol contribution to Arctic warming results from both increasing BC and decreasing sulfate, and as both were happening at once their contributions cannot be easily separated (unlike several earlier time periods we analyzed, when one increased while the other remained fairly constant). Though the uncertainty ranges are quite large, it can be useful to remember that the 95% confidence level conventionally used by scientists is not the only criteria that may be of interest. As the total observed Arctic warming during 1976-2007 was 1.5 +/- 0.3ºC, our results can be portrayed in many ways: there is about a 95% chance that aerosols contributed at least 15% to net Arctic warming over the past 3 decades, there is a 50% chance that they contributed about 70% or more, etc.

It’s also worth considering how to interpret the effects of decreasing sulfate during the past 3 decades. To try to make sure that the complex role of aerosols wouldn’t be misunderstood, when referring to the recent warming due to aerosols at Northern Hemisphere mid-latitudes and in the Arctic, we stated in the conclusions of the paper:

“much of this warming may stem from the unintended consequences of clean-air policies that have greatly decreased sulfate precursor emissions from North America and Europe (reducing the sulfate masking of greenhouse warming) and from large increases in Asian black carbon emissions.”

So it is incorrect, or at least quite incomplete, to say that that controls on air pollution such as those created under the Clean Air Act in the US have caused the recent warming. In the absence of increasing greenhouse gases, our large historical emissions of sulfate precursors would have led to substantial cooling from sulfate, and the subsequent reduction in emissions would have brought temperatures back towards their previous level. So reduced sulfate does not cause warming in an absolute sense, only relative warming compared to a time when emissions were larger. Over the mid-20th century, sulfate precursor emissions appear to have been so large that they more then compensated for greenhouse gases, leading to a slight cooling in the Northern Hemisphere. During the last 3 decades, the reduction in sulfate has reversed that cooling, and allowed the effects of greenhouse gases to clearly show. In addition, black carbon aerosols lead to warming, and these have increased during the last 3 decades.

For an analogy, picture a reservoir. Say that around the 1930s, rainfall into the watershed supplying the reservoir began to increase. However, around the same time, a leak developed in the dam. The lake level stayed fairly constant as the rainfall increased at about the same rate the leak grew over the next few decades. Finally, the leak was patched (in the early 70s). Over the next few decades, the lake level increased rapidly. Now, what’s the cause of that increase? Is it fair to say that lake level went up because the leak was fixed? Remember that if the rainfall hadn’t been steadily increasing, then the leak would have led to a drop in lake levels whereas fixing it would have brought the levels back to normal. However, it’s also incomplete to ignore the leak, because then it seems puzzling that the lake levels were flat despite the increased rain during the first few decades and that, were you to compare the increased rain with the lake level rise, you’d find the rise was more rapid during the past three decades than you could explain by the rain changes during that period. You need both factors to understand what happened, as you need both greenhouse gases and aerosols to explain the surface temperature observations (and the situation is more complex than this simple analogy due to the presence of both cooling and warming types of aerosols).

Hence the implication should not be that cleaning up the air causes warming, but that air pollution plays a substantial role in climate, and we can better understand regional climate changes during the past by taking this into account. Economists have argued that inclusion of a broader array of climate forcing agents leads to more cost-effective strategies to mitigate climate change (e.g. [O’Neill, 2003]), so that taking into account the large impact of air pollution and its ancillary effects on human and ecosystem health may also lead to better solutions for climate change.

114 Responses to “Yet more aerosols: Comment on Shindell and Faluvegi”

  1. 101
    Hank Roberts says:

    Bill Hunter’s pointer isn’t foolish, it’s just utterly off-topic.

    Bill, look at the related papers, in particular e.g. this one:

    Impacts of past climate variability on marine ecosystems: Lessons from sediment records

    Yes, nature goes up and down. Add a forcing past the range within which it’s gone up and down, nature takes an excursion to a new state. Both are true.

    But this is about aerosols. Please try to focus.

  2. 102
    Chris S says:

    Hank, Bill, Ike

    I’ve posted a response in the Friday roundup thread to this with regard to species shifts observed in the English Channel over the last ~100 years & also the North Sea.

  3. 103


    Why do you say that the errors in the observed warming, versus the errors in the calculated warming due to aerosol changes are not independent? Theoretically the actual warming due to aerosol changes could be larger than the total actual warming with natural variability accounting for the difference.

    [Response: Yes they could. But the error bars in the Shindell paper were determined all together – it makes no sense to insist that the maximum in one and the minimum in another could have occurred together in the analysis. In fact it is much more likely that the warming attributable to BC+sulphate changes is positively correlated with the total amount of attributable warming. Thus the high end warmings from aerosols occur in situations where there is a large attributable warming. – gavin]

  4. 104


    First since I don’t have access to the Shindell paper I would like to confirm that they are using the standard observed temperatures. For example GISS or HadCrut and not computing their own new value. Thus the errors in the observed temperatures would be from those independent observations. These would have been determined separately, not together with the errors in the Shindell paper on aerosol warming.

    Would you say that if the true value for aerosol warming were at the high end of the range it is more likely that the true value for actual warming is at the top of the range. In other words that the errors are correlated? This is different than asking if the actual temperature change is correlated with the actual aerosol change.

    In this case then this increases the likelihood that 1.1 of the 1.5 observed value is caused by aerosol changes from what one would guess from simply looking at the overlap of the two error ranges. The delta would tend to persist. Thus if actual warming was 1.8 degrees the likely value for aerosol warming is 1.5 degrees, and if the actual warming was 1.2 degrees then the likely value for aerosol warming is .8 degrees.

    Did I get that right?

  5. 105
    Hank Roberts says:

    > since I don’t have access to the Shindell paper …

    You don’t have US$32, or a library nearby? Ask for an offprint.

    Asking Gavin to do research values his time less than a bus ticket.

  6. 106

    Thanks Hank,

    If I was asking him to do research I apologize. You’re right I could run down to the library. On the other hand he said “the error bars in the Shindell paper were all determined together.” So I thought he would have already had the answer to that question.

    If you’re saying that answering a blog post values his time less than a bus ticket, then that is a different point.

  7. 107
    Hank Roberts says:

    Dr. Shindell initiated the post. He might be able to send you an offprint if you ask for one. Just sayin’.

  8. 108

    I would very much appreciate that.

  9. 109
    Ike Solem says:

    Hank, you seem to be very concerned about topic relevance to the blog post, but you must realize that climate science is very interconnected and nothing really exists in isolation, right?

    Aerosols and sea surface temperatures are indeed closely related, indeed that’s something that the Shindell & Faluvegi paper directly relates to, because their results seem to contradict those of another recent Science paper on the aerosol effect on tropical Atlantic SST’s:

    Evan et al. The Role of Aerosols in the Evolution of Tropical North Atlantic Ocean Temperature Anomalies, Science Oct 20 2008″

    Apparently, looking at the Tropical Atlantic in isolation isn’t very useful, at least that’s what I gather. I would like to hear some of the working experts weigh in, though.

    In any case, you quote a paper that refutes your arguments, see this from the abstract:

    In particular, response of higher trophic levels to changes at the base of ecosystems appears to be highly non-linear and shows variations of several orders of magnitude, as exemplified by fish stocks in several areas

    Sea surface temperatures are hardly the only ecological factor controlling the base of ecosystems in the ocean – high sensitivity means that many different factors could be playing roles in any fish-based record – disease, predation, competition, etc.

    And then, you say, “Yes, nature goes up and down. Add a forcing past the range within which it’s gone up and down, nature takes an excursion to a new state. Both are true.”

    That’s a very simplistic statement, isn’t it? Nature is variable, but always up and down? By the same amount? For a specific example, see the Multivariate El Nino / Southern Wobble index

    That’s the best understood ‘natural cycle in the oceans’, and it’s not exactly predictable, is it? Can you predict the decade to come by the decade that passed? Obviously not – sometimes there are several downs before an up, sometimes the opposite, and strength and duration varies all over the place.

    Nevertheless, you have a lot of people running around claiming that the “Pacific Decadal Oscillation is locked into a cool phase” on zero evidence, or the “an enhanced AMO is leading to more hurricanes”, etc.

    Those arguments are not supportable. Yes, there are very periodic natural cycles – the moon’s orbit, say, or the tidal response to the moon’s orbit – which is not quite as predictable as many think, in fact. That doesn’t mean everything in nature is periodic, does it?

    Secondly, any ocean fluctuations or variations are certain to be influenced by global warming energy changes, although we really don’t seem to know how that will play out (with respect to more or fewer El Ninos and La Ninas). In this area, the past is a poor guide to the future.

  10. 110
    Hank Roberts says:

    Nicolas, the contact info for the first author is on the journal page.
    Back in the day, we asked for offprints by sending letters including SASEs or postage stamps. These days, maybe you can do it all with electrons. Try.

  11. 111
    Brian Dodge says:

    Ike Solem Says:28 April 2009 at 9:25 PM
    “That’s the best understood ‘natural cycle in the oceans’, and it’s not exactly predictable, is it?”

    Even simpler systems with fewer degrees of freedom and less complicated feedbacks have long been known to behave chaotically.

    “Common intuition suggests that when a circuit is off synchronization the observed output, although not periodic, will be a sum of periodic (intermodulation) components. In fact, at least for a large class of systems we have studied, the output does not have this relatively simple form but is actually chaotic. This paper studies a simple but realistic model for a large class of triggered oscillators. Theory and experiments both confirm that the output shows the properties of sensitivity to initial conditions, nonperiodicity, broad spectrum, and complicated recurrence, that characterize chaotic motion.”
    Synchronization and chaos
    Tang, Y.; Mees, A.; Chua, L. IEEE Transactions on Circuits and Systems, Volume 30, Issue 9, Sep 1983 Page(s): 620 – 626

    It staggers my mind that a handful of op amps and a few dozen resistors and capacitors can demonstrably exhibit chaotic behavior, but many people believe that far more complicated natural systems have “periodic” or “oscillatory” behavior. “Wobbly” is much better descriptor. It’s also straightforward to convert a sinusoidal waveform (like one would expect from Milankovitch cycles) into a sawtooth waveform (approximately like we observe in the ice age temperature record) with a teeny bit of nonlinear feedback; throw in a little “sensitivity to initial conditions, nonperiodicity, broad spectrum, and complicated recurrence,” and viola – Dansgaard-Oeschger events, LIA, MWP, and so on.

  12. 112
    Hank Roberts says:

    Ike, that was a fine attack on those who pretend ENSO/PDO are regular exact repeating pattern to justify argument about climate. Nobody here does. Look for that over at rankexploits.

    Bill Hunter and I just exchanged a few words about sardines. This about sums that up:
    (remove the space after “geo” to make that work — spam filter barfs)

    Yes, it’s all connected; do changes in plankton or sardines affect rainfall in Monterey? Hey, could be.

  13. 113
    Mark says:

    Glad to see people trying to make others more aware of how we can improve environmental damage. I’m trying to bring awareness by spreading the word on the winners of the Tomorrows World video contest:

    They had a competition over videos about water efficiency and flooding. Living off the West Coast, it’s a very real worry of mine. I think the winners did a great job! Check out their work and forward the link if you like it.

  14. 114
    Ike Solem says:

    Hank says: “Ike, that was a fine attack on those who pretend ENSO/PDO are regular exact repeating pattern to justify argument about climate. Nobody here does.”

    What? That argument has been put forth in realclimate threads over and over and over again, on topics from hurricanes (AMO) to drought (PDO/ENSO). It’s also prevalent over at Dot Earth, see the Don Easterbrook comment:

    This is the cool water phase of the PDO and it isn’t going to change for at least 2-3 decades (at least it never has in the past) and it is unaffected by atmospheric CO2 as shown by the three PDO switches this century that occurred before atmospheric CO2 increased significantly.

    The point, Hank, is that claiming that nature acts like a spring that goes “up and down” until forced into a new state is just plain wrong – a gross oversimplification of complex natural processes, which is usually what denialists do – now that has been a steady pattern for some time. Please don’t propagate such nonsense.

    In any case, let’s get back to the topic of aerosols and their effect on sea surface temperatures, and ‘natural oceanic cycles’. Consider two other much-remarked papers on the subjects:

    Shanahan 2009 Atlantic forcing of persistent drought in West Africa, Science

    There, they use a sediment core from Lake Ghana in West Africa to deduce “African monsoon history” over the past three millennia, which they link to sea surface temperatures… how? Their conclusion is that the AMO is very real, and that modern drought is not anomalous. I would wait and see what more recent coral-based SST reconstructions reveal before putting much faith in that, for example:

    “We present the first absolutely dated and annually-resolved multi-centennial record of Atlantic sea surface temperature… Estimates of externally-forced background variability suggest that anthropogenic forcing can account for most of the warming since 1850 A.D. Multidecadal variations superimposed upon this background disappear prior to 1730 A.D. in favor of interdecadal (15-20 year) variability. This suggests observed multidecadal variability is not persistent and may be difficult to predict. – Saenger et al. 2009”

    Yes, that is aerosol related, not off-topic. Take a look at this paper, also recent:

    Evan et al. 2009 The Role of Aerosols in the Evolution of Tropical North Atlantic Ocean Temperature Anomalies

    TO be brief, their results seem to conflict with those in the original post, i.e.

    So we put idealized forcings from GHGs, aerosols, and ozone in the tropics, mid-latitudes and polar regions to see what would happen. The results showed that the temperature response in the tropics, like the global mean, is only mildly sensitive to the location of forcing.

    The model used by Shindell & Faluvegi is probably a bit more realistic than that used by Evan et al. However, the role of aerosol dust is probably still significant – but is there a good record of the aerosol forcing over the Atlantic over the past few decades? Well, there are estimates:

    “Prospero & Lamb 2003 African droughts and dust transport to the Caribbean: Climate change implications”

    Regression estimates based on long-term rainfall data suggest that dust concentrations were sharply lower during much of the 20th century before 1970, when rainfall was more normal.

    So, the increase in dust after 1970 should have led to cooling Atlantic temperatures, not warming Atlantic temperatures, so how does that work out in the Evan et al. model?

    None of this, of course, eliminates the need to reduce particulate black carbon aerosol concentrations – but to do that, we need to focus on shipping and trucking and coal combustion in the industrialized world, and biomass burning and inefficient internal combustion engines in the developing world (as well as coal, the biggest problem).