There is a natural tendency to produce models at finer spatial scales that
include both a wider array of processes and more refined descriptions. Higher
resolution can lead to better simulations of atmospheric dynamics and hydrology
(Chapter 8, Section 8.9.1), less diffusive oceanic
simulations, and improved representation of topography. In the atmosphere, fine-scale
topography is particularly important for resolving small-scale precipitation
patterns (see Chapter 8, Section 8.9.1). In the ocean,
bottom topography is very important for the various boundary flows (see Chapter
7, Section 7.3.4). The use of higher oceanic resolution also improves the
simulation of internal variability such as ENSO (see Chapter
8, Section 8.7.1). However, in spite of the use of higher resolution, important
climatic processes are still not resolved by the model’s grid, necessitating
the continued use of sub-grid scale parametrizations.
It is anticipated that the grids used in the ocean sub-components of the coupled
climate models will begin to resolve eddies by the next report. As the oceanic
eddies become resolved by the grid, the need for large diffusion coefficients
and various mixing schemes should be reduced (see Chapter
8, Section 8.9.3; see also, however, the discussion in Section
8.9.2). In addition, the amount of diapycnal mixing, which is used for numerical
stability in this class of ocean models, will also be reduced as the grid spacing
becomes smaller. This reduction in the sub-grid scale oceanic mixing should
reduce the uncertainty associated with the mixing schemes and coefficients currently
being used.
Underlying this issue of scale and detail is an important tension. As the spatial
and process detail in a model is increased, the required computing resources
increase, often significantly; models with less detail may miss important non-linear
dynamics and feedbacks that affect model results significantly, and yet simpler
models may be more appropriate to generating the needed statistics. The issue
of spatial detail is intertwined with the representation of the physical (and
other) processes, and hence the need for a balance between level of process
detail and spatial detail. These tensions must be recognised forthrightly, and
strategies must be devised to use the available computing resources wisely.
Analyses to determine the benefits of finer scale and increased resolution need
to be carefully considered. These considerations must also recognise that the
potential predictive capability will be unavoidably statistical, and hence it
must be produced with statistically relevant information. This implies that
a variety of integrations (and models) must be used to produce an ensemble of
climate states. Climate states are defined in terms of averages and statistical
quantities applying over a period typically of decades (see Chapter
7, Section 7.1.3 and Chapter 9, Section 9.2.2).
Fortunately, many groups have performed ensemble integrations, that is, multiple
integrations with a single model using identical radiative forcing scenarios
but different initial conditions. Ensemble integrations yield estimates of the
variability of the response for a given model. They are also useful in determining
to what extent the initial conditions affect the magnitude and pattern of the
response. Furthermore, many groups have now performed model integrations using
similar radiative forcings. This allows ensembles of model results to be constructed
(see Chapter 9, Section 9.3; see also the end of Chapter
7, Section 7.1.3 for an interesting question about ensemble formation).
In sum, a strategy must recognise what is possible. In climate research and modelling, we should recognise that we are dealing with a coupled non-linear chaotic system, and therefore that the long-term prediction of future climate states is not possible. The most we can expect to achieve is the prediction of the probability distribution of the system’s future possible states by the generation of ensembles of model solutions. This reduces climate change to the discernment of significant differences in the statistics of such ensembles. The generation of such model ensembles will require the dedication of greatly increased computer resources and the application of new methods of model diagnosis. Addressing adequately the statistical nature of climate is computationally intensive, but such statistical information is essential.
Extreme events are, almost by definition, of particular importance to human
society. Consequently, the importance of understanding potential extreme events
is first order. The evidence is mixed, and data continue to be lacking to make
conclusive cases. Chapter 9, Sections 9.3.5 and 9.3.6
consider projections of changes in patterns of variability (discussed in the
next section) and changes in extreme events (see also Chapters
2 and 10). Though the conclusions are mixed in both
of these topical areas, certain results begin to appear robust. There appear
to be some consistent patterns with increased CO2 with respect to changes in
variability: (a) the Pacific climate base state could be a more El Niño-like
state and (b) an enhanced variability in the daily precipitation in the Asian
summer monsoon with increased precipitation intensity (Chapter
9, Section 9.3.5). More generally, the intensification of the hydrological
cycle with increased CO2 is a robust conclusion. For possible changes in extreme
weather and climate events, the most robust conclusions appear to be: (a) an
increased probability of extreme warm days and decreased probability of extreme
cold days and (b) an increased chance of drought for mid-continental areas during
summer with increasing CO2 (see Chapter 9, Section 9.3.6).
The evaluation of many types of extreme events is made difficult because of
issues of scale. Damaging extreme events are often at small temporal and spatial
scales. Intense, short-duration events are not well-represented (or not represented
at all) in model-simulated climates. In addition, there is often a basic mismatch
between the scales resolved in models and those of the validating data. A promising
approach is to use multi-fractal models of rainfall events in that they naturally
generate extreme events. Reanalysis has also helped in this regard, but reanalysis
per se is not the sole answer because the models used for reanalysis rely on
sub-grid scale parametrizations almost as heavily as climate models do.
One area that is possibly ripe for a direct attack on improving the modelling
of extreme events is tropical cyclones (see Section Chapter
2, 2.7.3.1; Chapter 8, Section 8.8.4; Chapter
9, Section 9.3.6.4, and Chapter 10, Box 10.2).
Also, there is the potential for increased understanding of extreme events by
employing regional climate models (RCMs); however, there are also challenges
to realising this potential (see Chapter 10). It must
be established that RCMs produce more realistic extremes than general circulation
models (GCMs). Most RCM simulations to date are not long enough (typically 5
or 10 years for nested climate change simulations) to evaluate extremes well
(see Chapter 10, Section 10.5.2).
Another area in which developments are needed is that of extremes associated
with the land surface (flood and drought). There is still a mismatch between
the scale of climate models and the finer scales appropriate for surface hydrology.
This is particularly problematical for impact studies. For droughts there is
a basic issue of predictability; drought prediction is difficult regardless
of scale.
A particularly important issue is the adequacy of data needed to attack the
question of changes in extreme events. There have been recent advances in our
understanding of extremes in simulated climates (see, for example, Meehl et
al., 2000), but thus far the approach has not been very systematic. Atmospheric
Model Intercomparison Project 2 (AMIP2) provides an opportunity for a more systematic
approach: AMIP2 will be collecting and organising some of the high-frequency
data that are needed to study extremes. However, it must be recognised that
we are still unfortunately short of data for the quantitative assessment of
extremes on the global scale in the observed climate.
Finally, it is often stated that the impacts of climate change will be felt through changes in extremes because they stress our present day adaptations to climate variability. What does this imply for the research agenda for the human dimension side of climate studies?
Other reports in this collection |