After a two-month hiatus, I've finally found the time to pick up the thread on doubts about global warming and humans' role in it. In previous installments, people wrote in with the reasons why they were skeptical and I tried to synthesize the responses. In a rough-and-ready poll to gauge which line of reasoning resonated the most, the favorite appeared to be #2: The present warming could be a natural uptick.

There's a lot to be said on this topic, so I'll start by biting off a piece of it. What I'd like to try to do is to get beyond the he said/she said debate that it's all too easy to fall into, and see whether a fresh look at the data can break the logjam.

The Average Argument

The standard argument for anthropogenic warming goes something like this:

  1. Surface temperature readings jiggle up and down, but over the course of the past century, their overall trend is up.
  2. The levels of carbon dioxide and certain other gases in the atmosphere have also been rising. The concentration of CO2 in seawater has increased as well. The total observed increase is consistent with the amount of fuel humanity has burned and the amount of land it has cleared since the industrial revolution began.
  3. Basic theory links these two trends. Because of the gases, a smaller fraction of the thermal radiation emitted by the surface escapes into space. The planet becomes a net absorber of energy. The surface heats up and emits more thermal radiation, until the total amount of escaping thermal radiation again balances the energy deposited by sunlight.
  4. Natural factors, such as variations in the sun's output, have been too small to account for the observed temperature increase.

To put this argument into a quantitative framework, climatologists estimate how much extra heating the addition of these gases has caused per unit area of Earth's surface. Other influences on the climate -- aerosol (particulate) emissions, orbital oscillations, solar variability -- can be measured in the same way. All these influences, or "forcings", can then be treated interchangeably and added together. Subjected to this net total forcing, the climate adjusts -- shifting winds, clouds, precipitation, and other weather patterns, which can either amplify or offset the forcing -- and the global average temperature reaches some new value.

Filling in some very rough numbers, the current forcing is 2 watts per square meter, mostly from carbon dioxide, methane, and low-altitude ozone, minus the cooling wrought by particulates. Of this forcing, the oceans appear to have absorbed half, leaving the atmosphere to take up the rest. According to the geologic record, 1 W/sq m should lead to about half a degree Celsius of warming -- which matches the observed increase. This temperature rise is greater than you'd expect from the energy input provided by the forcing, so the climate system appears to be overreacting; feedback loops are amplifying the direct effect.

In this framework, skeptical arguments #2 and #3 have two prongs: first, that the forcings haven't been properly enumerated; and second, that the climate machine does not respond in the predicted way. Respondents to my original posting put forward various natural forcings that they said might outweigh the anthropogenic ones. Others suggested that the response to forcings gets lost in the noise of intrinsic variability.

Elsewhere, I'll examine these hypotheses, but my goal here is to get away from the common-currency approach altogether. If all the forcings are treated interchangeably, it becomes hard to verify whether the 2 watts per square meter is due to this or that influence, whether a crucial forcing has gone overlooked altogether, or whether the half-degree of warming is due to a large forcing on an unresponsive climate, a small forcing on a hair-trigger one, or the coincidental wiggling of a climate wont to do its own thing.

CSI: Climate

The above argument uses, in essence, two data points: the average annual temperatures at the beginning of the century and today. These barely scrape the surface of the available data. Climatologists have maps and time series showing how a boatload of climate variables -- mean temperature, temperature ranges, air pressure, precipitation, and so on -- vary in time and in space, horizontally across the surface and vertically through the atmosphere. These data sets are a gold mine for resolving ambiguity, because the different forcings leave distinct fingerprints.

For instance, carbon dioxide spreads out evenly, whereas aerosols tend to remain concentrated above polluting regions. Carbon dioxide emissions have been steadily climbing over time, whereas sulfur dioxide emissions, the largest contributor to aerosol loading, recently peaked. Such patterns make it possible to tease out their relative contributions. In fact, when scientists attribute xpercent of warming to greenhouse gases, y percent to solar fluctuations, and so on, they derive these numbers by mixing and matching the various fingerprints to reproduce the observed trends.

One of the handiest things about fingerprints is that they help to make sense of climate models. One of the things that troubles many people, skeptics and otherwise, about models is that they are black boxes whose inner workings are obscure and whose output has to be taken on trust. Although quantifying the fingerprints still requires running a climate model, the response can be understood in terms of basic principles. Wall Street Journal science writer Sharon Begley had a nice article in May on this topic, and below I'll review some of the specific fingerprints.

On the downside, different forcings sometimes have similar fingerprints. Even when they don't, the feedbacks, inertia, and other internal dynamics of the climate system can mute the differences. So using fingerprints involves uncertainties, which take some effort to quantify.

The study of fingerprints got going in the 1980s and came into its own in the mid-1990s, when Benjamin Santer of Lawrence Livermore and his colleagues made the first rigorous argument for an anthropogenic contribution to surface temperatures. The IPCC devoted a chapter to the subject in its 2001 report and is preparing an even more extensive discussion for its next major report, due out next year. Over the years, researchers have considered ever more variables besides temperature and ever more forcings besides greenhouse gases. They have merged spatial and temporal patterns, looked at regional as well as global scales, and developed more sophisticated mathematical tools.

A major breakthrough occurred last year when researchers figured out why temperature measurements at the surface were not matching those higher up in the troposphere, the lowest layer of the atmosphere -- a discrepancy that was one of the most powerful points made by climate skeptics. One factor turned out to be a slow orbital decay of the satellites taking the atmospheric measurements, which produced a spurious cooling that masked the warming trend. In May, the two scientists who had identified the mismatch -- John Christy and Roy Spencer of the University of Alabama, Huntsville -- were among the co-authors of a report that laid the problem to rest. Some discrepancy lingers, but the report argued that other observational biases are probably responsible.

One by One

Here are some of the fingerprints. For those people who, like me, hunger to grasp the basic physics behind the summary, I elaborate in an appendix below.

  • Greenhouse gases. Carbon dioxide, methane, nitrous oxide, and chlorofluorocarbons warm the surface and troposphere (the lowest atmospheric layer) and cool the stratosphere (the next layer up). These gases get mixed in quickly, so their effects are nearly symmetrical between the Northern and Southern Hemispheres. At the surface, the warming is greater at night than during the day, during the winter than during the summer, and at high latitudes than at low latitudes. Greenhouse gas forcing has grown slowly and steadily over the industrial era.
  • Aerosols. Sulfate aerosols cool the surface, troposphere, and stratosphere alike. At the surface, the temporal pattern is exactly the opposite that of greenhouse gases: cooling is greater at day than during the night, during the summer than during the winter, and at low latitudes than at high latitudes. Black carbon (soot) warms whichever layer of the atmosphere it is found in. Aerosols exert their strongest effect over the regions where they are emitted, which are mostly in the Northern Hemisphere. The sulfate forcing grew more rapidly than the greenhouse one from 1945 to 1980 and has since been declining.
  • Ozone. The depletion of ozone from the ozone layer cools the stratosphere; the cooling is greatest at high latitudes, where the layer has suffered the most. The depletion also causes a small amount of cooling at the surface and in the lower troposphere. As the phase-out of ozone-destroying chemicals allows the layer to rebuild itself, the stratosphere should rewarm. Ozone near the ground (a component of smog) warms the surface and lower troposphere.
  • Solar variability. An increase in the intensity of the sun warms the troposphere and stratosphere nearly equally.
  • Volcanic eruptions. By hurling dust to high altitudes, explosive volcanoes cool the troposphere and warm the stratosphere. The stratospheric warming lasts one to two years, and the surface and troposphere cooling three to five years. Volcanoes also reduce the global average precipitation.
  • Generic effects. Some fingerprints are not specific to one forcing or another; they occur when the temperature changes for whatever reason. These fingerprints are useful for cross-checking the data. For instance, any temperature changes should be greater for land than for sea, and because the Northern Hemisphere has more land than the Southern, any trends should be more pronounced there. In the tropics, temperature changes should be greater in the troposphere than at the surface, because moist air releases heat as it rises and condenses. Also, if the trends are short-term (driven mainly by, say, volcanic eruptions or the 11-year solar cycle), they should have little effect on temperatures in the ocean depths or in boreholes. It takes a while for heat to penetrate down, so a subsurface temperature profile provides a measure of cumulative changes, whereas the surface and atmosphere mark the present moment, with all its vicissitudes.
  • Internal (unforced) climate variability. By definition, a forcing adds energy to the climate system, whereas natural variability merely redistributes it. So if the recent string of hot years is just a wiggle, then the total heat content of the climate system -- all the oceans plus the ice caps plus the atmosphere -- should be constant. Another difference is that a forcing shifts the range of climate events, whereas variability is the range. If historical data indicate that heat waves have become more common and cold snaps less so, the range itself must have shifted, indicating a forcing rather than variability.

View from the Crime Scene

So what do the data say?

  • Vertical patterns. Since 1979, when continuous satellites observations began, the surface and troposphere have warmed and the stratosphere has cooled. Consequently, the boundary between the troposphere and stratosphere has risen by about 170 meters over that period; it has risen more at high latitudes than at low ones. All this is exactly what you'd expect from greenhouse gases, with some limited offsetting by sulfate aerosol cooling. The stratosphere over Antarctica is especially cool, apparently because of the ozone hole. d03;
  • Horizontal patterns. Pretty much the entire surface has gotten warmer, high latitudes more than lower ones. In the early 20th century, land regions and oceans warmed at similar rates, but since the mid-1970s, the land has sped ahead. On the whole, the Northern Hemisphere has warmed more than the Southern. Again, this is just what greenhouse gases would do. d03;
  • Temporal patterns. Nighttime and wintertime temperatures have increased more than daytime and summertime ones, compressing the diurnal and seasonal temperature ranges over land worldwide. The trends are apparent even on a regional level, including North America, albeit with lower statistical significance. There are fewer frost days, fewer frigid extremes, and more warm nights. Greenhouse gases would produce these very effects. d03;
  • Energy balance. An interesting paper last year looked at vertical ocean temperature profiles since 1960 using new data sets and found widespread heating. The upper 700 meters of Atlantic waters and upper 100 meters of the Pacific and Indian Ocean are warmer than the deeper layers. The different depths are exactly what you'd expect, because the Atlantic has stronger convection currents than the other oceans do. All the oceans have warmed; there isn't the zero-sum game of warming and cooling you'd expect from natural variability. All this is consistent with greenhouse gases. Per unit area, the northern seas have warmed less than the southern ones, which makes sense if greenhouse gases have caused an overall warming trend, offset by sulfate aerosols in the northern climes where they are concentrated. d03;

When climatologists run the fingerprinting analysis for different historical epochs, they find that temperature fluctuations prior to the Industrial Revolution were driven primarily by solar and volcanic forcings. In the early 20th century, natural and anthropogenic forcings seem to contribute equally. From midcentury onwards, greenhouse gases rule.

So unless I'm missing something, it seems to me that the case for anthropogenic warming is pretty strong. Is it unequivocal? Well, nothing in life ever is. There might be some low-frequency natural forcing. Plenty of i's in fingerprinting analyses remain to be dotted. Based on the knowledge we have so far, however, I have to call 'em as I see 'em.


Greenhouse gases

These gases absorb infrared radiation and re-emit it in all directions. So they hinder the radiative flow of energy. In the troposphere, the flow is upward: the surface is heated by sunlight, and the temperature decreases with altitude. Adding more of these gases further hinders this upward flow, and the surface and troposphere warm up.

In the stratosphere, the net radiative flow is downward: the ozone layer within the stratosphere is heated by solar ultraviolet radiation, and the temperature increases with altitude. In this sense, the stratosphere is an upside-down troposphere, so the greenhouse effect works in reverse. Adding more of the gases hinders the downward flow, more radiation ends up escaping into space, and the stratosphere cools down.

This combination of troposphere warming and stratospheric cooling is a distinctive fingerprint of greenhouse gases, shared by no other forcing. One way to describe it is in terms of the height of the tropopause, the boundary between the troposphere and stratosphere. This altitude is determined by whether heated air tends to rise in a convection pattern (as in the troposphere) or to stay still and radiate away its energy (as in the stratosphere). The temperature in the troposphere decreases, on average, by 6.5 degrees C per kilometer of altitude. The temperature in the lower stratosphere, determined by ozone absorption higher up, is about -55 degrees C. Therefore, increasing the ground temperature or decreasing the stratosphere temperature by 1 degree leads to a 1/6.5, or 160 meter, increase in the troposphere altitude. Ben Santer et al.'s 2004 paper has a nice chart (figure 6) illustrating the basic principle. These effects are strongest in the tropics, which, being hotter, emit more infrared radiation and are therefore more affected by heat-trapping.

Greenhouse gases also have distinctive effects on the range of temperatures experienced by the surface. At night, the ground cools down by emitting infrared radiation, whereas during the day, the infrared cooling is secondary to solar heating. Because greenhouse gases impede the cooling but not the heating, they exert their greatest influence at night. If they are the cause of global warming, average nighttime temperatures should increase more than daytime ones -- reducing the total daily temperature swing. (To be sure, the trend can be offset by changes in cloud cover and soil moisture.)

For the same reason, greenhouse gases affect wintertime temperatures more than summertime ones, reducing the total annual temperature swing, and high-latitude surface temperatures more than low-latitude ones. These effects are magnified by snow and ice: by reducing snow and ice cover, warming reduces the reflectivity of the ground and allows more solar energy to be absorbed, further increasing the warming; conversely for cooling.

Models with anthropogenic forcings show a variety of other fingerprints that I've omitted from the discussion here because the basic theory is not well understood. For instance, Nathan Gillett of the University of Victoria, British Columbia, and his colleagues have found that during northern winter, the surface pressure increases at mid-latitudes and decreases at high latitudes.


Whereas greenhouse gases affect outgoing thermal radiation, sulfate aerosols affect incoming sunlight: they reflect it back into space, so they tend to cool both the troposphere and stratosphere. The tropospheric cooling tends to lower the tropopause, while the stratospheric cooling tends to raise it; according to more detailed models, the former dominates. Moreover, daytime temperatures cool more than nighttime ones.

Black carbon (soot) absorbs sunlight. If it is close to the ground, it warms the surface; if it is higher up, it warms higher layers of the atmosphere and reduces the amount of sunlight reaching the surface, cooling it. To be sure, black carbon has been less well studied than other forcings, and it remains something of a wild card.


Ozone absorbs incoming ultraviolet radiation, so its depletion from the ozone layer has cooled the stratosphere, especially at high latitudes, raising the tropopause. The depletion also causes a small amount of cooling at the surface and in the lower troposphere. If the ozone is located near the ground, it warms the surface and adjacent troposphere.

Solar variability

An increase in solar output warms the entire atmosphere nearly uniformly with altitude. Because both tropopause and stratosphere get hotter, the tropopause barely budges. Daytime temperatures increase more than nighttime ones do.


Explosive eruptions inject dust directly into the stratosphere, where it absorbs sunlight and warms the air at the expense of the troposphere. This forcing peaks several months after the eruption, but it takes temperatures several years to return to their pre-eruption values. The surface and troposphere respond more slowly because of the thermal inertia of the oceans. The stratospheric temperature overshoots as it cools, so the sequence is actually warming, cooling, then rewarming, before stabilizing again.

This characteristic S-shaped curve was used by Venkatachalam Ramaswamy and colleagues in a Science paper earlier this year to explain the trend in stratospheric temperatures since satellite measurements began in 1979. Specifically, the overall downward trend (attributed to greenhouse gases) occurs in a series of steps following volcanic eruptions, which makes sense if the downward trend is temporarily offset by the post-eruption rewarming.

Injecting dust into the stratosphere cools the surface, reducing evaporation and therefore precipitation, as discussed by Hugo Lambert and colleagues. Greenhouse gases have the opposite effect, but it is muted because the greenhouse warming is offset by a reduction in the radiative cooling that allows raindrops to form -- another example of how fingerprints depend on whether it is incoming sunlight or outgoing infrared that is affected.

Longer-term effects can arise if the frequency of eruptions changes.

Internal variability

Complicating the picture I gave above, oscillations such as El Niño can have widespread effects, perhaps leading to second-order effects (such as increased cloud cover) that change the energy content of the climate system. But the period of such oscillations is a decade or less, so they should not affect the energy content on longer timescales. 

In practice, climatologists estimate the amount of intrinsic natural variability by running a model in which they have kept the forcings constant. They verify the inferred variability by looking at variations over different timescales. If the models reproduce the wiggles in climate over a decade, they are presumed to handle longer-term ones too.

Generic effects of warming

The warming exerts a stronger effect on land than on sea. When land is heated, it takes a while for the heat to diffuse downward, so it remains concentrated in a thin near-surface layer. When seawater is heated, some of the heat gets carried downward by vertical currents, so the heat is spread out over a larger mass, producing a smaller temperature increase. Moreover, some of the seawater heating goes into increasing evaporation rates rather than increasing the temperature. This land-sea difference is not a fingerprint per se, as it applies to all forcings.

Basic theory also predicts that warming increases both the evaporation rate and the precipitation rate. The former increases more than the latter, because precipitation is limited by the radiative cooling of the troposphere that allows raindrops to form. Therefore, less air needs to rise to keep the hydrologic cycle going, which can weaken certain convection currents. This effect has gotten some recent attention in the context of the Walker circulation. In the equatorial Pacific, winds blow from west to east at high altitudes and east to west along the surface (either weakening or strengthening the east-to-west trade winds, depending on longitude). The circulation is driven by the rise of warm, moist air in the western Pacific. Because winds tend to flow from high to low air pressure, the Walker circulation produces a higher sea-level air pressure in the east than in the west.

Quantitatively, a degree Celsius of warming increases water-vapor concentrations by 7 percent but rainfall by only 2 percent, so there is a 5 percent decrease in the convection flow rate. For the observed warming of 0.6 degrees, you should get a ~3 percent weakening of the Walker circulation. That, in turn, should reduce the difference in sea-level pressure across the Pacific. And in fact, a compilation of barometer readings since the mid-1800s shows just such a decrease.

For more information

Table 1 in the executive summary of the recent U.S. Climate Change Science Program report and Figure 1.3 in chapter 1 give concise summaries of spatial fingerprints. One of the best overviews of recent work on fingerprints is this paper last year by a veritable who's who of experts in this area. This paper by Jim Hansen of the Goddard Institute for Space Studies and colleagues is another useful resources: figure 24 shows the effects of different forcings on surface temperature; figure 20 shows the effects on atmospheric structure; and other diagrams show precipitation, water vapor, winds, and sea-level pressure.