Earth system models: an overview

Schematic of traditional approach to climate projection, in which socioeconomic assumptions lead to emissions estimates, to concentration scenarios, which are used to drive physical climate models (upper set of diagrams), and the new strategy in which concentrations are prescribed and an ESM simulates climate change and the corresponding emissions, the latter of which are interpreted to infer corresponding socioeconomic changes. (Reprinted with permission from Ref 54. Copyright 2007 World Climate Research Programme)

Comparison of simulated ocean primary productivity and an estimate made using a model constrained by satellite‐based chlorophyll observations. (Reprinted with permission from Ref 27. Copyright 2009 American Meteorological Society)

Net primary productivity (NPP) for the terrestrial ecosystem as simulated by the Canadian Earth System Model (CanESM1—upper left), by an off‐line version of the CanESM1 terrestrial ecosystem model driven by observationally based meteorology (upper right) and the average from 17 different dynamic vegetation models.51 (Reprinted with permission from Ref 27. Copyright 2009 American Meteorological Society)

Zonal mean CO₂ concentration from 1991 to 2000 from observations (left) and from the Canadian Earth System Model (CanESM1). (Reprinted with permission from Ref 27. Copyright 2009 American Meteorological Society). The annual cycle, larger in the northern hemisphere and smaller in the southern hemisphere is clearly visible, as is the background human‐forced increase in CO₂ concentration.

Model errors in precipitation, sea‐level pressure and surface air temperature for models participating in the CMIP2 intercomparison (ca 2000) and the CMIP3 intercomparison (ca 2005). Models are segregated into two groups: those that employ ‘flux adjustment’ and those that do not (see text for details). From Gleckler et al.38

Root mean square (RMS) errors for precipitation (upper panel) and surface pressure (lower panel) computed for 21 different global climate models participating in the CMIP3 intercomparison project. Also shown is the RMS error for the ensemble mean and median. The different symbols illustrate the sensitivity to different analysis choices, with (*) indicating the standard analysis, (o) representing the result when using an alternate reference data set, (a) indicates an alternate averaging period, (+) indicates different target grid resolution and (–) indicates different ensemble members. From Gleckler et al.38

Near surface air temperature error (model minus observations) in °C for the control integration of the Canadian Earth System Model (CanESM243). Red colors indicate areas where the model is warmer than observed, blue colors where the model is colder than observed.

Difference in vegetation, represented as the mass of stored carbon per unit area (kg C m⁻²) between 2000 and 2090 as simulated by the Canadian Earth System Model (CanESM127) for a particular CO₂ emission scenario. Red colors indicate increases in vegetation whereas blue colors indicate a decrease.

Difference in atmospheric CO₂ concentration time series between ‘coupled’ and ‘uncoupled’ experiments with climate models that have a representation of the carbon cycle. Each curve represents the result of a different model having undertaken both the coupled and uncoupled experiments. In all cases the difference is positive, indicating that the carbon cycle feedback is positive (i.e., results in higher concentrations, and hence greater warming, for a given emission scenario). (Reprinted with permission from Ref 19. Copyright 2006 American Meteorological Society)

Sketch of energy flows in the climate system, based on observations from March 2000 to May 2004. Units are W m⁻². (Reprinted with permission from Ref 6. Copyright 2009 American Meteorological Society)