A new, non-flux-corrected, global climate model is introduced, the Kiel Climate Model (KCM), which will be used to study internal climate variability from interannual to millennial time scales and climate predictability of the first and second kind. The version described here is a coarse-resolution version that will be employed in extended-range integrations of several millennia. KCM's performance in the tropical Pacific with respect to mean state, annual cycle, and El Nino-Southern Oscillation (ENSO) is described. Additionally, the tropical Pacific response to global warming is studied.Overall, climate drift in a multicentury control integration is small. However, KCM exhibits an equatorial cold bias at the surface of the order 1 degrees C, while strong warm biases of several degrees are simulated in the eastern tropical Pacific on both sides off the equator, with maxima near the coasts. The annual and semiannual cycles are realistically simulated in the eastern and western equatorial Pacific, respectively. ENSO performance compares favorably to observations with respect to both amplitude and period.
An ensemble of eight greenhouse warming simulations was performed, in which the CO2 concentration was increased by 1% yr(-1) until doubling was reached, and stabilized thereafter. Warming of equatorial Pacific sea surface temperature (SST) is, to first order, zonally symmetric and leads to a sharpening of the thermocline. ENSO variability increases because of global warming: during the 30-yr period after CO2 doubling, the ensemble mean standard deviation of Nino-3 SST anomalies is increased by 26% relative to the control, and power in the ENSO band is almost doubled. The increased variability is due to both a strengthened (22%) thermocline feedback and an enhanced (52%) atmospheric sensitivity to SST; both are associated with changes in the basic state. Although variability increases in the mean, there is a large spread among ensemble members and hence a finite probability that in the "model world" no change in ENSO would be observed.