PROJECTING FUTURE CLIMATE CHANGE.
Climate scientists also seek to understand how
climate would change in the future if greenhouse
gas emissions and other anthropogenic forcing
factors were to evolve in particular ways. Such
information can be of significant interest to policy
makers, who might choose to implement policies
that push toward some future scenarios and away
from others, and to many other decision makers as
well (e.g., insurance companies deciding which
policies to offer). Climate models have emerged as
an important tool for making projections of future
conditions. For a given scenario, it is common to
make an ensemble of projections using different
climate models or model versions; this is motivated
in part by uncertainty about how to construct a
climate model that can deliver highly accurate
projections (Parker 2006; Betz 2009). This
uncertainty in turn stems both from limited
theoretical understanding of some processes
operating within the climate system as well as from
limited computing power, which constrains how
, existing knowledge can be implemented in models
(see the discussion of parameterization in Section
4.2).
Different types of ensemble study explore the
consequences of different sources of uncertainty in
modeling. A multi-model ensemble (MME) study
probes structural uncertainty, which is uncertainty
about the form that modeling equations should
take as well as how those equations should be
solved. An example of an MME study is the Coupled
Model Intercomparison Project Version 5 (CMIP5),
which produced projections using a few dozen
climate models developed at different modeling
centers around the world (see Flato et al. 2013:
Table 9.1). A perturbed-physics ensemble (PPE)
consists of multiple versions of the same climate
model, where the versions differ in the numerical
values assigned to one or more parameters within
the model. Examples of studies employing PPEs
include the climateprediction.net project (Stainforth
Climate scientists also seek to understand how
climate would change in the future if greenhouse
gas emissions and other anthropogenic forcing
factors were to evolve in particular ways. Such
information can be of significant interest to policy
makers, who might choose to implement policies
that push toward some future scenarios and away
from others, and to many other decision makers as
well (e.g., insurance companies deciding which
policies to offer). Climate models have emerged as
an important tool for making projections of future
conditions. For a given scenario, it is common to
make an ensemble of projections using different
climate models or model versions; this is motivated
in part by uncertainty about how to construct a
climate model that can deliver highly accurate
projections (Parker 2006; Betz 2009). This
uncertainty in turn stems both from limited
theoretical understanding of some processes
operating within the climate system as well as from
limited computing power, which constrains how
, existing knowledge can be implemented in models
(see the discussion of parameterization in Section
4.2).
Different types of ensemble study explore the
consequences of different sources of uncertainty in
modeling. A multi-model ensemble (MME) study
probes structural uncertainty, which is uncertainty
about the form that modeling equations should
take as well as how those equations should be
solved. An example of an MME study is the Coupled
Model Intercomparison Project Version 5 (CMIP5),
which produced projections using a few dozen
climate models developed at different modeling
centers around the world (see Flato et al. 2013:
Table 9.1). A perturbed-physics ensemble (PPE)
consists of multiple versions of the same climate
model, where the versions differ in the numerical
values assigned to one or more parameters within
the model. Examples of studies employing PPEs
include the climateprediction.net project (Stainforth
