Temperatures in Transient Climates: Improved Methods for Simulating with Evolving Variability

The societal impacts of future climate change depend on temperatures not only through changes in their means, but also through changes in their variability. Because variations on different timescales -- day-to-day temperature fluctuations vs season-long heat waves -- affect society differently, it is important to distinguish between timescales of variation. Moreover, most available model runs are of transient climates, wherein temperatures are statistically nonstationary. These facts combined mean that assessing changes in variability in future climate model runs requires care. We present a statistical model for characterizing variability changes in transient climates, and results from applying it to temperatures from an ensemble of climate model runs under varying future CO2 trajectories. We find that temperature variability changes can be broadly characterized according to the corresponding changes in regional mean temperature, where variability typically decreases as mean temperatures increase and variability decreases tend to be stronger for high frequency variations. While variability changes are broadly characterized by their associated mean changes, the transient response also depends on the rate of change of mean warming and is typically larger than that in modeled equilibria climates with comparable changes in mean temperature. Understanding variability changes in model output is important in itself, but impacts assessments researchers may further require full simulations of future temperatures for use as inputs in their models. Raw climate model output is inappropriate for this purpose, because climate models do not reproduce the full distribution of temperatures under historical forcing scenarios. We therefore address an ensuing need for simulations of future temperatures that combine both the observational record and model projections of changes in means and variability. Our simulation method is based on transforming observations to account for climate model projected changes, and overcomes some shortcomings of other popular approaches for creating such simulations.