Abstract:
Global warming entails not only an increase in mean surface temperature, but widespread changes across all components and scales of the climate system. To understand these changes, it is necessary to consider the fluctuations around the mean and the distributions of climate variables, that is, their variability. How this variability depends on the mean climate state and evolves during periods of warming remains, in many respects, unclear. For regional surface temperature variability, models and observations provide conflicting evidence on decadal and
longer timescales that suggests that models likely underestimate variability.
Here, we investigate how the variability of surface climate depends on the timescale and mean climate state. To this end, we examine the moments — standard deviation, skewness and kurtosis — of the distributions of surface temperature and precipitation, as well as their power spectra. We use an ensemble of transient simulations of the last 23, 000 years, from the Last Glacial
Maximum (LGM) to present-day to analyze the influence of forcings and model complexity on simulated variability. This ensemble thus covers Earth’s most recent warming episode, the Last Deglaciation, during which Earth’s global mean surface temperature warmed by about 4–7° C. We compare the variability in the Deglacial ensemble to that found in projections of possible future warming, as well as to reconstructions, reanalyses and direct observations.
Surface climate variability depends on timescale and background state. Annual to millennial variability is enhanced during the transitionary Deglaciation with respect to the quasi-stationary LGM and Holocene. For surface temperature, we find larger variability during the LGM than the Holocene, but the difference is smaller than in observations. For precipitation, the opposite holds, as variability is smaller during the LGM. There are large inter-model differences in skewness and kurtosis for both temperature and precipitation in the Deglacial ensemble and future projections, suggesting considerable uncertainty in the simulation of extreme conditions.
We differentiate the Deglacial simulations based on the complexity of the underlying model, from a two-dimensional Energy Balance Model (EBM) and Earth System Models of Intermediate Complexity (EMICs), to General Circulation (GCMs) and Earth System Models (ESMs). GCMs and ESMs demonstrate complex patterns of surface climate variability in space and time that change with the background state. EMICs, on the other hand, display less variability and more homogeneous patterns, while the EBM simulates at times extreme excursions in variability. Comparison to observational records shows that the complexity of GCMs and ESMs is necessary to adequately simulate variability. Beyond that, boundary conditions and applied forcings become more relevant. Variations in ice sheets, meltwater routing and volcanism affect variability on all examined timescales, from annual to millennial. The inclusion of transient volcanic forcing narrows the gap between simulated and reconstructed variability.
Overall, our results demonstrate that the ability of models to simulate variability of surface climate depends on model complexity, forcing protocol and boundary conditions. We identify requirements for simulating adequate levels of variability, from transient volcanic forcing to a minimal model complexity.