Internal climate memory in observations and models

Attribution of cause of climate change is hindered by our ability to separate internal low-frequency variability from the forced response in the climate system. We characterize the spatiotemporal characteristics of internal variability by comparing ensemble averages of in-band fractional variances in Coupled Model Intercomparison Project Phase 5 (CMIP5) preindustrial control simulations to estimates from observations and reanalyses. For sea surface temperature and sea level height anomalies both models and observations show that variability on time scales less than 5 years is predominantly in the tropics and has the spatial signature of El Niño–Southern Oscillation. On progressively longer time scales the variance moves to the extratropics and from middle to higher latitudes while displaying spatially coherent features. The CMIP5 models show good agreement in the spatial and temporal apportioning of in-band variance when the variances are normalized.