GG 312: Global Climate Change and Environmental Impacts (Fall 2002)
Chapter 4: Climate Model Evaluation
4.1. Introduction
4.2. What is Model Evaluation?
4.3. Atmosphere
4.4. Ocean
4.5. Sea Ice
4.6. Land Surface
4.7. Climate of the 20th Century
4.8. Phenomena
4.9. Extreme Events
4.10. Assessment
4.1. Introduction
The coupled climate system consists of the atmosphere, oceans, cryosphere and the land surface (deserts, vegetation, orography, etc). The climate models studied here
include a three dimensional representation and interaction of these components on a global time-dependent basis. These models when run in conjunction with
specified chemical composition of the atmosphere and surface charactertistics, provide the current basis for understanding and simulating the climate system and its
future changes.
Note that the simulation of the coupled climate system requires integrations of over 100 to a 1000 simulated years; therefore, computational costs restrict model
complexity and resolution in space and time.
Another point of interest is the large range of time scales at which processes in the various components of the climate system operate. For instance, the time scales
for the atmosphere are of the order of weeks; for the land surface, a season, and, for the deep ocean, 100 to 1000 of years. This makes climate model simulations of
future change tricky.
Time Scales
Learn more about climate models
Climate Modeling Centers
NCAR Climate Model Output
GISS 2*CO2 Climate Model Output
Visualize Climate
Visit NCAR
4.2. What is Model Evaluation
Model evaluation is assessing the suitability of models (in particular coupled atmosphere-ocean general circulation models) for use in climate change projection
and in detection and attribution studies.
Even if a model is assessed as performing credibly when simulating the present climate, this does not necessarily guarantee that the response to a perturbation
remains credible. Therefore, we also assess the performance of the models in simulating the climate over the 20th century and for selected palaeoclimates.
4.3. Atmosphere
Coupled climate models simulate mean atmospheric fields with reasonable accuracy, with the exception of clouds and some related hydrologic processes. Problems
in the simulation of clouds and upper tropospheric humidity, however, remain worrisome because the associated processes account for most of the uncertainty in climate
model simulations of anthropogenic change.
Figure 4.1: December-January-February climatological surface air temperature in K simulated by the CMIP1 model control runs.
Figure 4.2: December-January-February climatological precipitation in mm per day (top) simulated by CMIP1 model control runs.
Figure 4.3: Zonally averaged December-January-February total cloudiness simulated by 10 AMIP1 models (a) and by revised versions
of the same 10 models (b). The solid black line gives observed data from the International Satellite Cloud Climatology Project (ISCCP).
4.4. Ocean
Considerable progress has been made since IPCC-1995 in the realism of the ocean component of climate models. Models now exist which simultaneously maintain realistic
poleward heat transports, surface temperatures and thermocline structure. However, there are still a number of processes which are poorly resolved or
represented, for example convection.
Ocean Circulation and Climate
Figure 4.4: Zonal mean air and sea temperature 'errors' in °C (defined here as the difference from the initial model state, which
was derived from observations), for three different coupled models. The models differ only in the parameterisation of lateral mixing used in the ocean component. The
different mixing schemes produce different rates of heat transport between middle and high latitudes, especially in the Southern Hemisphere. The atmosphere must adjust
in order to radiate the correct amount of heat to space at high latitudes, and this adjustment results in temperature differences at all levels of the atmosphere.
4.5. Sea Ice
The sea-ice simulations of 15 global coupled models contributed to CMIP-1 are summarised in Table 4.1 which provides
a comparison of ice extent, defined as the area enclosed by the ice edge, for winter and summer seasons in each hemisphere. The last row of the Table provides an
observed estimate based on satellite data covering the period 1978-87.
Figure 4.5provides a visual presentation of the range
in simulated ice extent, and was constructed as follows. For each model listed in Table 4.1, a 1/0 mask is produced to indicate presence or absence of ice. The 15
masks were averaged for each hemisphere and season and the percentage of models that had sea ice at each grid point was calculated. There is a large range in the
ability of models to simulate the position of the ice edge and its seasonal cycle, particularly in the Southern Hemisphere.
Table 4.1: Coupled model simulations for December, January, February (DJF) and June, July, August (JJA) of sea-ice cover
(columns 2-5) and snow cover (106 km2 columns 6 and 7).
Figure 4.5: Illustration of the range of sea-ice extent in CMIP-1 model simulations listed in Table 4.1: Northern Hemisphere,
DJF (left) and Southern Hemisphere, JJA (right). For each model, a 1/0 mask is produced to indicate presence or absence of ice. The masks were averaged for each
hemisphere and season. The 0.5 contour therefore delineates the region for which at least half of the models produced sea-ice. The 0.1 contour indicates the region
outside of which only 10% of models produced ice, while the 0.9 contour indicates that region inside of which only 10% of models did not produce ice. The observed
boundaries are averages over 1961-1990 period.
4.6. Land Surface
Uncertainty in land surface processes, coupled with uncertainty in parameter data combines, at this time, to limit the confidence we have in the simulated regional
impacts of increasing CO2. In general, the evidence suggests that the uncertainty is largely restricted to surface quantities (i.e., the large-scale climate
changes simulated by coupled climate models are probably relatively insensitive to land surface processes). Our uncertainty derives from difficulties in the modelling
of snow, evapotranspiration and below ground processes.
Figure 4.6: An uncertainty ratio for 10-degree latitude bands; (a) control simulations; (b) difference between the control and
doubled greenhouse gas simulations. E is for evaporation, P is precipitation, Tscr is screen temperature, cld is the percent cloud cover and Sn is the net short wave
radiation at the surface. The units on the Y-axis are dimensionless. An asterisk means the value is statistically significant at 95% and a diamond at 90%.
4.7. Climate of the 20th Century
Several coupled models are able to reproduce the major trend in 20th century surface air temperature, when driven by historical radiative forcing scenarios
corresponding to the 20th century (Figure 4.7). However, in these studies idealised scenarios of only sulphate radiative forcing
have been used. One study that includes both the indirect and direct effects of sulphate aerosols, as well as changes in tropospheric ozone, suggests that the observed
surface and tropospheric air temperature discrepancies since 1979 are reduced when stratospheric ozone depletion and stratospheric aerosols associated with the Pinatubo
eruption are included (Figure 4.8). Systematic evaluation of 20th century AOGCM simulations for other trends found in observational
fields, such as the reduction in diurnal temperature range over the 20th century and the associated increase in cloud coverage, have yet to be conducted.
The inclusion of changes in solar irradiance and volcanic aerosols has improved the simulated variability found in several AOGCMs. In addition, some evaluation studies
aimed at the reproduction of 20th century climate have suggested that changes in solar irradiance may be important to include in order to reproduce the warming in the
early part of the century. Another study has suggested that this early warming can be solely explained as a consequence of natural climate variability. Taken together,
we consider that there is an urgent need for a systematic 20th century climate intercomparison project with a standard set of forcings, including volcanic aerosols,
changes in solar irradiance and land use, as well as a more realistic treatment of both the direct and indirect effects of a range of aerosols.
Figure 4.7: Observed (Parker/Jones) and modelled global annual mean temperature anomalies (degrees C) from the 1961- 1990
climatological average. The control and three independent simulations with the same greenhouse gas plus aerosol forcing and slightly different initial conditions are
shown. The three greenhouse gas plus aerosol simulations are labelled GHG+Asol1,GHG+Asol2 and GHGAsol3 respectively.
Figure 4.8: Radiative forcing (Wm-2)since 1979 due to changes in stratospheric aerosols, ozone, greenhouse gases and
solar irradiance. (b-d) Observed global annual mean surface, tropospheric and stratospheric temperature changes and GISS GCM simulations as the successive radiative
forcings are cumulatively added one-by-one. The base period defining the zero mean observed temperature was 1979-1990 for the surface and the troposphere and 1984-1990
for the stratosphere.
4.8. Phenomena
Recent atmospheric models show improved performance in simulating many of the important phenomena by using better physical parametrisations and using higher resolutions
both in the horizontal and in the vertical domain. An intercomparison of El Niņo simulations, one of the most important phenomena, has revealed an ability of coupled
climate models to simulate the El Niņo-like SST variability in the tropical Pacific and associated changes in precipitation in the tropical monsoon regions, although
the region of maximum SST variability is displaced further westward than in the observations (Figure 4.9).
Figure 4.9: Simulated (March-May) and observed (March-April) averaged Nordeste (northeastern Brazil) rainfall indices for (a) the
original AMIP simulations, (b) the revised AMIP simulations.
4.9. Extreme Events
Attention has been paid to the analysis of extreme events in climate model simulations. This increases our confidence that they may be better reproduced by
high-resolution climate models in the future.
