(1) Surface temperature is rising. Because global warming is a long-term process, which can be masked by year-to-year climate variability, it is most clearly revealed in the longest available record of global temperaturei.e., that of surface temperature (Figure 2.1), which is based on stations determined not to have been substantially impacted by urbanization. In the opinion of the panel, the disparity between satellite and surface temperature trends during 1979–98 in no way invalidates the conclusion of the Intergovernmental Panel on Climate Change report (IPCC, 1996) that global surface temperature has warmed substantially since the beginning of the twentieth century. Accelerated warming during the late 1990s has raised the estimated warming to 0.40–0.8 °C in the past 100 years. The warming of surface temperature that has taken place during the past 20 years is undoubtedly real, and it is at a rate substantially larger than the average warming during the twentieth century.
(2) Based on current estimates, the lower to mid-troposphere has warmed less than the earth's surface during the past 20 years. For the time period from 1979 to 1998, it is estimated that on average, over the globe, surface temperature has increased by 0.25 to 0.4 °C and lower to mid-tropospheric temperature has increased by 0.0 to 0.2 °C.9break
9 The range of these trend estimates is determined by applying different trend algorithms to the different versions of the surface and tropospheric data sets. Further discussion of the uncertainties inherent in these estimates is provided in chapters 6–9.
(3) Current estimates of surface and lower to mid-tropospheric temperature trends are subject to a level of uncertainty that is almost as large as the apparent disparity between them. The factors contributing to this uncertainty are:
• the uncertainty inherent in temperature trends determined over relatively short time periods;
• the complexity of the algorithms for processing the satellite data, and the limited validation that has been performed on them thus far;
• the possibility of biases remaining in the data sets that have not yet been recognized and corrected;
• the uneven and, in some places, sparse spatial coverage of radiosonde observations and, to a lesser extent, surface observations; and
• the inherent difficulties in correcting for changes in instrumentation and in the siting of radiosonde and surface stations.
(4) The observed trends have been partially, but not fully, reconciled with climate model simulations of human-induced climate change. The simulated three-dimensional spatial pattern of the temperature changes induced by increasing concentrations of a well-mixed greenhouse gas, such as carbon dioxide, is complicated and varies from model to model, but one common aspect is the tendency for the lower to mid-troposphere to warm more rapidly than the surface, except over high latitudes. More realistic model simulations that take into account radiative forcing with combined changes in human-induced and natural factors, including the three-dimensional structure of the changing distribution of ozone, are in better agreement with the observed changes, but they still predict that the lower to mid-troposphere should warm at least as rapidly as the earth's surface. The models used to perform these simulations are subject to uncertainties and subject to change as more realistic treatments of physical processes are incorporated into them.
(5) The record of satellite observations of lower to mid-tropospheric temperature is still short and subject to large sampling fluctuations. Recent experiments with a number of different climate models indicate that the inclusion of natural climate forcings such as volcanic eruptions, stratospheric ozone depletion, and solar variability can lead to a broad spectrum of simulated 20-year surface and lower to mid-tropospheric temperature trends. In light of this new information, thecontinue
panel cautions that trends in such short periods of record with arbitrary start and end points are not necessarily representative of how the atmosphere is changing in response to long-term human-induced changes in atmospheric composition. Given reliable measurements, as outlined in Recommendation #1, the level of confidence that can be attached to the trends will increase as the period of record of upper air measurements lengthens.
(6) It is not currently possible to determine whether or not there exists a fundamental discrepancy between modeled and observed atmospheric temperature changes since the advent of satellite data in 1979. Measurement uncertainties (Finding #3), modeling uncertainties (Finding #4), and sampling uncertainties (Finding #5) were all considered by the panel as possible causes of the disagreement between models and observations. None of them can be singled out as the dominant factor, nor can any one of them be shown to be unimportant. Surface temperature and lower to mid-tropospheric temperature are different entities, which should not be expected to vary in precisely the same manner in response to human-induced and natural climate forcings during a particular 20-year period of record. Hence, the panel concludes that at least part of the observed disparity between the 20-year changes in surface and mid-tropospheric temperature is probably real, but the measurement, modeling, and sampling uncertainties alluded to above make it difficult to precisely attribute the disparity to any particular sources. A more definitive reconciliation of modeled and observed temperature changes awaits the extension and improvement of the observations and the algorithms used in processing them, better specification of the natural and human-induced climate forcings during this period, and improvement of the models used to simulate the atmospheric response to these forcings.break