Linear Models: Useful Tools to Analyze GCM Results

Abstract

Using a two-level linear, steady state model, we diagnose the 40-day mean response of a GCM to a tropical sea surface temperature (SST) anomaly. The time-mean anomalies produced by the GCM are simulated as linear response to the anomalous hemispheric distributions of latent heating, sensible heating and transient eddy forcing. Also, the anomalous effect of mountains, caused by anomalies in the zonal mean surface wind is taken into account. All anomalies are defined as the difference between perturbation and control runs. For our analysis, we have taken the tropical Atlantic SST anomaly experiment performed by Rowntree. We have compared the linear model's response in temperature at 600 mb and winds at 400 mb with the same anomalous quantities produced by the GCM. The similarity between the time-mean anomalies of the GCM experiment and the linear model's response is very high. The pattern correlation coefficients are between 0.6 and 0.7 in the region between 30°N and 60°N. The response to each of the anomalous forcings separately is positively correlated with the GCM anomaly pattern. The amplitude of the response to anomalous forcing by transient eddies is a factor of two or three larger than the effects of anomalous sensible and latent heating. The anomalous effect of the orography is negligible. Although intended to be a tropical SST anomaly GCM experiment, the difference between control and perturbation runs does not seem to be directly related to tropical heating near the SST anomaly. Instead, most of the forcing of anomalies in the midlatitudes took place in the midlatitudes itself and, in particular, the remote effects of forcing by tropical latent heat sources were minor.