Global average surface temperature has increased about 3/4 °C (1.35 °F) during the period of extensive instrumental measurements, i.e., since the late 1800s. Most of the warming, about 1/2 °C (0.9 °F), occurred after 1950. The causes of observed warming can be investigated best for the past 50 years, because most climate forcings were observed then, especially since satellite measurements of the sun, stratospheric aerosols and ozone began in the 1970s. Furthermore, 70% of the anthropogenic increase in greenhouse gases occurred after 1950.

Changes in known climate forcings since 1950 are shown in Figure 4. The largest forcings are the positive forcing by greenhouse gases and the negative forcing by aerosols. Stratospheric aerosols, which are sulfates from occasional volcanic eruptions, are well measured. However, human-made aerosols, which have multiple sources and compositions, are poorly measured.

These forcings have been used to drive climate simulations for 1951–1998 with the NASA Goddard Institute for Space Studies SI2000 climate model (Reference 1b). This model has a sensitivity of 3/4 °C per W/m², consistent with paleoclimate data and typical of other climate models. The largest suspected flaws in the simulations are the omission of poorly understood aerosol effects on cloud drops and a probable underestimate of black carbon changes. The first of these is a negative forcing and the second is positive, so these flaws should be partially compensating in their effect on global temperature.

Simulated climate changes are compared with observations in Figure 5. The five model runs differ only because of unforced ("chaotic" or "weather") variability, which is an inherent characteristic of complex coupled dynamical systems. In the model, the stratosphere cools, mainly as a result of ozone depletion, but then warms after volcanic activity as the aerosols absorb thermal radiation. The troposphere and the surface warm as a result of increasing greenhouse gases, with brief cooling intervals caused by large volcanoes. These changes accord with observations, as illustrated. However, it would be a mistake to take this agreement as quantitative confirmation of the principal model parameters and assumptions. A larger (smaller) value for the net climate forcing could yield comparable agreement with the observations, if it were combined with a smaller (larger) value of climate sensitivity. Also, unforced (chaotic) variability in this specific version of the GISS model is probably less than unforced variability of real world climate.

The most important quantity is the planetary energy imbalance (Figure 5d). This imbalance is a consequence of the long time that it takes the ocean to warm. We conclude that the Earth is now out of balance by about 0.5 to 1 W/m², i.e., there is that much more solar radiation being absorbed by Earth than heat being emitted to space. One implication of this imbalance is that, even if atmospheric composition does not change further, the Earth's surface will eventually warm another 0.4–0.7 °C.

The Earth's energy imbalance is a vital statistic, because it is the residual climate forcing that the planet has not yet responded to. It is too small to be measured directly, but we can verify its value because the only place that the energy can be going is into melting ice or heating the air, land and ocean. It is worth examining simple calculations of these energy sinks, because, as we show in the next section, this provides insight about prospects for future global changes.

As summarized in Box 4, most of the energy imbalance has been heat going into the ocean. Sydney Levitus (Reference 4) has analyzed ocean temperature changes of the past 50 years, finding that the world ocean heat content increased about 10 watt-years, consistent with the time integral of the planetary energy imbalance. Levitus also found that the rate of ocean heat storage in recent years is consistent with our estimate that the energy balance of the Earth is now out by 0.5 to 1 W/m². Note that the amount of heat required to melt enough ice to raise sea level 1 m is about 12 watt-years (averaged over the planet), energy that could be accumulated in 12 years if the planet is out of balance by 1 W/m².

The agreement with observations, for both the modeled temperature change and ocean heat storage, leaves no doubt that observed global climate change is being driven by (natural and anthropogenic) forcings. The current rate of ocean heat storage is a critical planetary metric, because it determines the amount of additional global warming that is already "in the pipeline." It is important for a second, related, reason: it equals the reduction in climate forcings that we would need to make if we wished to stabilize the Earth's present climate.

The goal of the United Nations Framework Convention on Climate Change, produced in Rio de Janeiro in 1992, is to stabilize atmospheric composition to "prevent dangerous anthropogenic interference with the climate system" and to achieve this in ways that do not disrupt the global economy. The United States was the first developed country to sign the convention, which has since been ratified by practically all countries. Defining the level of warming that constitutes "dangerous anthropogenic interference" (DAI) is thus a crucial but difficult part of the global warming problem.

The United Nations established an Intergovernmental Panel on Climate Change (IPCC) with responsibility for analysis of global warming. The IPCC has defined climate forcing scenarios, used these for simulations of 21st century climate, and estimated the impact of temperature and precipitation changes on agriculture, natural ecosystems, wildlife and other matters (Reference 11a; http://www.ipcc.ch). Significant effects have been found, but even with warming of several degrees there are winners and losers. The IPCC predicts a change in sea level as large as several tens of centimeters in 100 years, if global warming reaches several degrees Celsius. Their calculated sea level change is mainly due to thermal expansion of ocean water, with little change in ice sheet volume.

These moderate climate effects, even with rapidly increasing greenhouse gases, leave the impression that we are not close to DAI. The IPCC analysis also abets the emphasis on adaptation to climate change, as opposed to mitigation, in recent international discussions. Adaptation is required, to be sure, because climate change is already underway. However, I will argue that we are much closer to DAI than is generally realized, and thus the emphasis should be on mitigation.

The dominant issue in global warming, in my opinion, is sea level change and the question of how fast ice sheets can disintegrate. A large portion of the world's people live within a few meters of sea level, with trillions of dollars of infrastructure. The need to preserve global coast lines, I suggest, sets a low ceiling on the level of global warming that would constitute DAI.

The history of the Earth, and the present human-made planetary energy imbalance, together paint a disturbing picture about prospects for sea level change. To appreciate this situation, we must consider how today's global temperature compares with peak temperatures in the current and previous interglacial periods, how long-term sea level change relates to global temperature, and the time required for ice sheets to respond to climate change.

Warmth in the Holocene peaked between 6000 and 10,000 years ago, but subsequent cooling was slight. As shown by the Antarctic temperature record (Figure 2), the polar temperature during the Holocene peak was about 1 °C warmer than it was in the mid-20th century. During the previous (Eemian) interglacial period, polar temperatures were perhaps another 2 °C warmer. However, both paleoclimate data and climate models show that polar temperature change is larger than global mean temperature change by about a factor of two. (The ice core temperature anomalies at the pole refer to the inversion level, where the snow is formed; surface air anomalies are slightly larger (Reference 2d).)

This means that, with the 0.5 °C global warming of the past few decades, the Earth's average temperature is just now passing through the peak Holocene temperature level. Furthermore, the current planetary energy imbalance of about 3/4 W/m² implies that global warming already "in the pipeline," about another 0.5 °C, will take us about halfway to the global temperature that existed at the peak of the Eemian period.

Sea level during the Eemian is estimated to have been 5–6 meters (16–20 feet) higher than it is today. Although the geographical distribution of climate change influences the effect of global warming on ice sheets, paleoclimate history suggests that global temperature is a good predictor of eventual sea level change. The main issue is: How fast will ice sheets respond to global warming?

The IPCC predicts only a slight change in the ice sheets in 100 years. However, the IPCC calculations include only the gradual effects of changes in snowfall, sublimation and melting. In the real world, ice sheet disintegration is driven by highly nonlinear processes and feedbacks. The peak rate of deglaciation following the last ice age was a sustained rate of melting of more than 14,000 km³/year, about 1 m of sea level rise every 20 years, which was maintained for several centuries. This period of most rapid melt, meltwater pulse 1A, coincided, as well as can be measured, with the time of most rapid warming (Reference 2d).

Given the present unusual global warming rate on an already warm planet, we can anticipate that areas with summer melt and rain will expand over larger areas of Greenland (Figure 6) and fringes of Antarctica. This will darken the ice surface in the season when the sun is high, promote freeze–thaw ice breakup, and, via ice crevasses, provide lubrication for ice sheet movement. Rising sea level itself tends to lift marine ice shelves that buttress land ice, unhinging them from anchor points. As ice shelves break up, this accelerates movement of land ice to the ocean.

This qualitative picture of nonlinear processes and feedbacks is supported by the asymmetric nature of glacial cycles (Figure 3) and the high rate of sea level rise associated with rapid warming. Although building of glaciers is slow, limited by annual snowfall rates, once an ice sheet begins to collapse its demise can be spectacularly rapid.

This natural melting process will be accelerated by the human-induced planetary energy imbalance. This imbalance provides an ample supply of energy for melting ice (Box 4), which can be delivered to the ice via ocean currents, atmospheric winds, and rainfall, as well as by icebergs drifting to lower latitudes. Furthermore, this energy source is supplemented by increased absorption of sunlight by ice sheets darkened by black carbon aerosols, as discussed below, and the positive feedback process as melt-water darkens the ice surface.

These considerations do not mean that we should expect large sea level change in the next few years. Preconditioning of ice sheets for accelerated breakup may require a long time, perhaps many centuries. However, I suspect that significant sea level rise could begin within decades, if the planetary energy imbalance continues to increase. Whatever that preconditioning period is, it seems clear that global warming beyond some limit will create a legacy of large sea level change for future generations. And once large-scale ice sheet breakup is underway, it will be impractical to stop. The same inertia of the ice sheets, which discourages rapid change, is a threat for the future. It will not be possible to build walls around Greenland and Antarctica. Dykes may protect limited regions, such as Manhattan and the Netherlands, but most of the global coastlines will be inundated.

I argue that the level of DAI is likely to be set by the global temperature and planetary radiation imbalance at which substantial deglaciation becomes practically impossible to avoid. Based on the paleoclimate evidence discussed above, I suggest that the highest prudent level of additional global warming is not more than about 1 °C. In turn, given the existing planetary energy imbalance, this means that additional climate forcing should not exceed about 1 W/m².

Detection of early signs of accelerating ice sheet breakup, and analysis of the processes involved, may be provided by the satellite IceSat recently launched by NASA. IceSat will use lidar and radar to precisely monitor ice sheet topography and dynamics. We may soon be able to investigate whether or not the ice sheet time bomb is approaching detonation.