|Working Group I: The Scientific Basis|
|Other reports in this collection|
12.3 Qualitative Comparison of Observed and Modelled Climate Change
This section presents a qualitative assessment of consistencies and inconsistencies between the observed climate changes identified in Chapter 2 and model projections of anthropogenic climate change described in Chapter 9.
Most formal detection and attribution studies concentrate on variables with high climate change signal-to-noise ratios, good observational data coverage, and consistent signals from different model simulations, mainly using mean surface air temperatures or zonal mean upper-air temperatures. To enhance the signal-to-noise ratio, they generally consider variations on large spatial scales and time-scales of several decades or longer.
There are many studies that have identified areas of qualitative consistency and inconsistency between observed and modelled climate change. While the evidence for an anthropogenic influence on climate from such studies is less compelling than from formal attribution studies, a broad range of evidence of qualitative consistency between observed and modelled climate change is also required. In addition, areas of qualitative consistency may suggest the possibility for further formal detection and attribution study.12.3.2 Thermal Indicators
When anthropogenic factors are included, models provide a plausible explanation of the changes in global mean temperature over the last hundred years (Figure 12.7). It is conceivable that this agreement between models and observations is spurious. For example, if a model’s response to greenhouse gas increases is too large (small) and the sulphate aerosol forcing too large (small), these errors could compensate. Differences in the spatio-temporal patterns of response to greenhouse gases and sulphate forcing nevertheless allow some discrimination between them, so this compensation is not complete. On the other hand, when forced with known natural forcings, models produce a cooling over the second half of the 20th century (see Figure 12.7) rather than the warming trend shown in the observed record. The discrepancy is too large to be explained through model estimates of internal variability and unlikely to be explained through uncertainty in forcing history (Tett et al., 2000). Schneider and Held (2001) applied a technique to isolate those spatial patterns of decadal climate change in observed surface temperature data over the 20th century which are most distinct from interannual variability. They find a spatial pattern which is similar to model-simulated greenhouse gas and sulphate aerosol fingerprints in both July and December. The time evolution of this pattern shows a strong trend with little influence of interannual variability. (Note that this technique is related to optimal fingerprinting, but does not use prior information on the pattern of expected climate change.)
Other thermal indicators
Changes in sea-ice cover and snow cover in the transition seasons in the Northern Hemisphere are consistent with the observed and simulated high latitude warming. The observed trends in Northern Hemisphere sea-ice cover (Parkinson et al., 1999) are consistent with those found in climate model simulations of the last century including anthropogenic forcing (Vinnikov et al., 1999). Sea-ice extent in the Southern Hemisphere does not show any consistent trends.
Compatibility of surface and free atmosphere temperature trends
Comparison of upper air and surface temperature data in Chapter 2 shows that the lower to mid-troposphere has warmed less than the surface since 1979. The satellite-measured temperature over a broad layer in the lower troposphere around 750 hPa since 1979 shows no significant trend, in contrast to the warming trend measured over the same time period at the surface. This disparity has been assessed recently by a panel of experts (National Academy of Sciences, 2000). They concluded that “the troposphere actually may have warmed much less rapidly than the surface from 1979 to the late 1990s, due both to natural causes (e.g., the sequence of volcanic eruptions that occurred within this particular 20-year period) and human activities (e.g., the cooling in the upper troposphere resulting from ozone depletion in the stratosphere)” (see also Santer et al., 2000). They also concluded that “it is not currently possible to determine whether or not there exists a fundamental discrepancy between modelled and observed atmospheric temperature changes since the advent of satellite data in 1979”. Over the last 40 years, observed warming trends in the lower troposphere and at the surface are similar, indicating that the lower troposphere warmed faster than the surface for about two decades prior to 1979 (Brown et al., 2000a; Gaffen et al., 2000). However, in the extra-tropical Eurasian winter some additional warming of the surface relative to the lower or mid-troposphere might be expected since 1979. This is due to an overall trend towards an enhanced positive phase of the Arctic Oscillation (Thompson et al., 2000) which has this signature.
Model simulations of large-scale changes in tropospheric and surface temperatures are generally statistically consistent with the observed changes (see Section 12.4). However, models generally predict an enhanced rate of warming in the mid- to upper troposphere over that at the surface (i.e., a negative lapse-rate feedback on the surface temperature change) whereas observations show mid-tropospheric temperatures warming no faster than surface temperatures. It is not clear whether this discrepancy arises because the lapse-rate feedback is consistently over-represented in climate models or because of other factors such as observational error or neglected forcings (Santer et al., 2000). Note that if models do simulate too large a negative lapse-rate feedback, they will tend to underestimate the sensitivity of climate to a global radiative forcing perturbation.
Other reports in this collection