Irreducible uncertainty in regional near-term climate projections

Abstract:

Effective adaptation to climate variability and change in the next few decades may require a probabilistic quantification of various possible outcomes. As such, quantifying the irreducible uncertainty due to chaotic climate variability in projections of near-term regional climate is essential. Here we use large initial condition ensembles of the FAMOUS global climate model with a 1%/year compound increase in CO2 levels to explore the role of both atmospheric and oceanic initial conditions in quantifying the range of future temperature projections. These simulations are the largest such ensembles to date - a total of over 17,000 simulated years. It is shown that short-term trends in global temperature can be very diverse, and cooling periods are more likely followed by more rapid rates of change. However, inter-annual temperature variability declines as the climate warms, especially over Europe, likely due to Arctic sea-ice retreat. In addition, a case study of the regional consequences of atmospheric initial condition uncertainty for Europe shows that at every land location, trends in temperature over the first 20 years of the transient simulation range from a cooling of more than 1K to a warming of more than 1K. However, on longer timescales, the signal of a warming climate becomes more clear. An ensemble using a range of oceanic initial conditions produces a larger spread in temperature trends than an ensemble using a single ocean initial condition for all lead times and seasons. Although FAMOUS likely exhibits too strong temperature variability, these results highlight the need for climate simulations to be performed with many more ensemble members than at present if the uncertainty in near-term climate is to be adequately quantified.