Page 376 - 全球气候变化及其影响Global Climate Change and Its Impacts-185×260

Global Climate Change and Its Impacts

scientific basis for climate change impact assessments and the formulation of adaptation/
mitigation strategies.

IV. Analysis and Management of Uncertainties in Multi-Model
Predictions

While multi-model projections provide abundant information for climate change re-
search, they also carry non-negligible uncertainties. Identifying the sources of these uncer-
tainties and addressing them through effective statistical methods, scenario analysis, and
other approaches – while properly communicating such uncertainties – is essential for accu-
rately understanding and applying multi-model projection results.
The sources of uncertainty in multi-model predictions primarily include the following
aspects. Uncertainty in model structure is one of the significant factors. Different climate
models employ varying parameterization schemes for physical processes and model
frameworks, leading to discrepancies in describing the complex physical, chemical, and
biological processes within the climate system. For example, in simulating the formation
and evolution of clouds, different models employ distinct parameterization approaches for
cloud microphysical processes, resulting in significant variations in the simulation of cloud
radiative effects. This, in turn, impacts predictions of global energy balance and climate
change. Uncertainty in initial and boundary conditions cannot be overlooked. Obtaining
initial conditions (e.g., initial states of atmospheric temperature, humidity, wind fields, etc.)
and boundary conditions (e.g., sea surface temperatures, sea ice distribution, etc.) for climate
models involves certain errors. These errors propagate and amplify during model runs,
leading to increased uncertainty in prediction outcomes. Furthermore, uncertainty in future
socioeconomic development and greenhouse gas emission scenarios constitutes another
major source. The future development of human society is influenced by numerous complex
factors, such as the pace of technological innovation, the formulation and implementation
effectiveness of policies, and changes in the global political and economic landscape.
These factors introduce substantial uncertainty into the assumptions regarding future
socioeconomic development pathways and greenhouse gas emission scenarios, thereby
affecting the outcomes of multi-model climate change predictions.
To address and express these uncertainties, statistical methods are widely employed.
For instance, constructing probability distribution functions enables probabilistic analysis of
multi-model prediction results, providing probabilities for different climate change indicators
to fall within specific ranges at future time points or periods. Monte Carlo simulation meth-
ods are commonly used, employing multiple random sampling to simulate model results un-
der varying initial conditions and parameter combinations, thereby assessing the scope and
magnitude of uncertainties. Scenario analysis examines multi-model prediction variations
under different future development scenarios (e.g., socioeconomic pathways and greenhouse
gas emission scenarios), visually demonstrating relationships between uncertainties and dif-

• 368 •