Climate model

A climate model is a numerical simulation of the climate. The mathematics describing the complex nonlinear dynamics of the climate system cannot be solved analytically. Simulations run using climate models can provide insight into climate behaviors and modes, and the impact of changes considered external the the climate system, because they cannot yet be predicted or because they are the result of human behavior.

The external changes, called forcings, are inputs to the climate simulations. Climate forcings include variations in solar activity, volcanic emissions of aerosols, particulates and greenhouse gases, human emissions of aerosols, particulates and greenhouse gases, human land use changes, and surface characteristics. Some surface characteristics such as topography are static over the course of a simulation while others may change as a result of simulated internal processes of the climate, such as soil moisture content or snow cover. The latter are examples of climate feedbacks.

Climate feedbacks are internal climate processes that positively reinforce or negatively oppose the direct effect of external climate forcings. For example, the direct effect of increases in solar activity or increases in greenhouse gas levels is a warming effect. The resulting increases in temperature increase evaporation. The increased water vapor, the most abundant greenhouse gas, is expected to further increase the direct warming effect, and represents a positive feedback to either of those forcings.

While humans live in the atmosphere on the land, nearly all the heat capacity of the climate system is in the oceans. Therefore, simulations of the atmosphere, must be coupled to simulations of the oceans for any hope of realistic long term simulations. A warming or cooling of the climate should be reflected in the heat content of the ocean. A global climate model (GCM) that is coupled to the ocean is called an Atmosphere Ocean GCM (AOGCM).

Climate Prediction

The climate is generally not considered predictable, because we cannot yet predict future solar behavior or volcanic activity and not all the internal processes of the climate are well understood. Climate models can be used to attempt to project the climate based on certain assumptions about future forcings such as greenhouse gas emissions scenarios or changes in solar activity. The usefulness of such projections for human decision making is dependent on the skill of the models in representing the internal climate processes involved. Climate scientists attempt to validate or prove the skill of the models by tuning them to match observations of the climate. Currently the most advanced climate models, still disagree by more than a factor of two in their sensitivity to greenhouse gas forcing, and have significant disagreements with observed climate behavior at high latitudes.

Models used by the IPCC AR4

Most of the models used by the Intergovernmental Panel on Climate Change 4th Assessment Report (IPCC AR4) are AOGCMs, and have been extensively studied with results and analyses appearing the peer review literature. Models are a key tool of climate science and are central to attempts to assess the relative contributions of various climate forcings to the recent global warming and to attempts to project the change of climate in response to future greenhouse gas scenarios.

Even though the climate sensitivities (to CO2) differ by over a factor of two, the IPCC has not decided which models are more accurate for purposes of attribution and projection. All the models are claimed to have good agreement with the 20th century temperature trends, yet are acknowledged to have significant errors. The IPCC hopes that by combining the range of models they will bracket the sensitivity of the actual climate, and that the errors in different models will be different and will tend to cancel out when the results are combined. Multiple runs of a model, called an ensemble, are needed to capture a model's climatology, since models are expected to have internal variability just as the climate does. When runs from multiple models are combined together in an attempt reduce or cancel out their errors, this is called a meta-ensemble. Meta-ensembles rely on the assumption that their errors are not correlated, otherwise, they cannot cancel.

Unfortunately the models have been shown to have signifcant correlated error. Roesch (2006) found that all the AR4 models had a positive surface albedo bias, chiefly due to excessive snow cover area and snow water equivilent in the temperate climate zones, a delayed snow melt, and a poor parameterization of the albedo in forests with snow cover. He was able to globally and annually average positive surface albedo bias and compare it to satellite observations. The albedo errors were 0.016 and 0.019 relative to the PINKER and ISCCP-FD satellite observations respectively. A positive albedo bias in the models means that more solar energy is being reflected back into space than in the actual climate. For approximate perspective on the size of this error, the albedo errors can be converted to watts per meter squared by applying a figure for the average downward solar surface flux of 198W/m^2. This gives a globally and annual averaged error of over 3W/m^2. This is over 4 times the decadal mean global warming energy imbalance of 0.75W/m^2 (Hansen 2005)

Stroeve (2007) reported that all the AR4 models were melting the Arctic polar ice but at rates significantly slower than the actual climate. The melting of the ice cap in response to warming replaces the ice with open ocean which has a much lower albedo, thus reflecting less solar energy and absorbing more solar energy, thereby positively reinforcing the warming. This is the ice-albedo feedback. Since, the models all under represent this positive feedback observed in the climate, they are once again biased against solar forcing, reflecting more solar energy into space than the observations. This is more correlated error that cannot be canceled by the IPCC meta-ensembles. It is unclear to what extent this is independent or duplicative of the surface albedo errors reported by Roesch. Roesch did not explicitly report polar ice cap discrepancies with the observations in his study.

References

Hansen, James; Larissa Nazarenko, Reto Ruedy, Makiko Sato, Josh Willis, Anthony Del Genio, Dorothy Koch, Andrew Lacis, Ken Lo, Surabi Menon, Tica Novakov, Judith Perlwitz, Gary Russell, Gavin A. Schmidt, Nicholas Tausnev. "Earth's Energy Imbalance: Confirmation and Implications" (2005). Science 308 (5727): 1431–1435. DOI:10.1126/science.1110252

Roesch, A. (2006). "Evaluation of surface albedo and snow cover in AR4 coupled climate models". J. Geophys. Res.. DOI:10.1029/2005JD006473. “The mean annual surface albedo of the 15 AR4 models amounts to 0.140 with a standard deviation of 0.013. All AR4 models are slightly above the mean of PINKER (0.124) and ISCCP-FD (0.121)." "These substantial deviations are still far too high to meet the required accuracy of surface albedos in GCMs.”

Stroeve, J.C.; M. M. Holland, W. Meier, T. Scambos, and M. Serreze (2007). "Arctic sea ice decline: Faster than forecast". Geophys. Res. Lett 34. DOI:10.1029/2007GL029703. “All models participating in the Intergovernmental Panel on Climate Change Fourth Assessment Report (IPCC AR4) show declining Arctic ice cover over this period. However, depending on the time window for analysis, none or very few individual model simulations show trends comparable to observations." Co-author Scambos from press release: "Because of this disparity, the shrinking of summertime ice is about thirty years ahead of the climate model projections.”