Are Models Overly Sensitive to Climate Change?

Feedbacks are a natural part of the complexity of our climate system. They represent non-linear processes within a system that are the result of the interaction between two variables or the interaction of a variable with itself (or its changes - "self" interaction). Generally, these processes can only be parameterized, or represented in an empirical way. Because of our inability to represent feedbacks precisely in a model, the model output may not be reasonable.

Temperature change in the climate system can be represented simply as a "forced-dissipative" relationship, for example by non-radiative and radiative processes as well as a net radiative restoring force, which is the feedback. As an example of this feedback process, if the system warms radiatively or non-radiatively, then the net radiative force may become negative (long wave out increases and becomes larger than short wave in). Thus, the net radiative force will act to counter the warming, and a larger restoring force would represent a less sensitive climate. The opposite can be argued for the system cooling. These principles here can be represented by a simple differential or mathematical equation.

A recent paper by Spencer and Braswell (2011) explored the sensitivity of the surface temperature response to a forced radiative imbalance. They used observed shortwave and longwave radiation gathered from satellite measurements and calculated the net atmospheric radiation. They also used observed surface temperatures from the 2000-2010 time frame around the globe and calculated global monthly temperature anomalies relative to the average over the ten year period. Using these two time series, they correlate one versus the other using different time lags.

Then they compared these observed values to the same variables gathered from the 20th century runs of the World Climate Research Programme (WCRP) Coupled Model Intercomparison Project (CMIP) phase 3 multimodel data set. The authors choose the three most and least sensitive model runs, rather than using all of them. They also use the observed and model data to investigate the mathematical relationship for temperature change described above.

When Spencer and Braswell (2011) set a 'feedback' parameter as described above, they demonstrate that only for pure non-radiative forcing and with no time lag can the parameter be accurately diagnosed in the model (Fig. 1). With radiative forcing and a 70/30% mix of radiative versus non-radiative forcing, the response was quite different. "In this case, radiative gain precedes, and radiative loss follows a temperature maximum, as would be expected based upon conservation of energy considerations." They also find, more importantly, that the pure radiative forcing curve in Fig. 1 looks more like one produced using the data from the climate models, while the mixed curve looks more like that produced using the observed data.

The authors then point out, that "we are still faced with a rather large discrepancy in the time-lagged regression coefficients between the radiative signatures displayed in the real climate system in satellite data versus the climate models." The discrepancy means that the climate system possesses less sensitivity that the climate models. This means climate models may overestimate the temperature change forced by a certain process.

There are, however, other processes not accounted for that make determination of the feedback parameter difficult. Once again, we can see that the climate models being relied on to produce future climate scenarios may be inadequate to represent an important process such as radiative feedbacks. Then these scenarios are the ones being used to policy.