Robust Forced Signals and Irreducible Uncertainties in Projections of Extremes

Abstract:

Decision makers express a strong need to shift the focus of climate projections from changes at global scale to changes at regional to local scale, from the end of the century to the coming decades, from changes multi-decadal mean temperatures to changes in extreme events. This new focus of climate projections implies an increasing contribution of internal variability to projection uncertainties. Understanding and quantifying the role of internal variability is vital to identify the limits of predictability – the irreducible uncertainty in multi-decadal projections – and the limits of model evaluation and bias correction.

Using different large initial condition ensembles, we demonstrate that models agree remarkably well on the forced signal of temperature and heavy precipitation extremes, the pattern of change in the absence of internal variability. The disagreement between individual model simulations on local to regional changes in extremes primarily arises from internal variability. Thus, in the coming 3-5 decades trends towards more intense hot and less intense cold extremes may be masked or even reversed in some places even if greenhouse-gas emissions rapidly increase. Likewise, despite a trend to more intense precipitation, opposite trends of multiple decades cannot be excluded over most land points. Despite large irreducible uncertainties at local scale, in an aggregated spatial probability perspective projections are again remarkably consistent already for the coming decades.

We argue that it is vital to specify whether model agreement or robustness refers to the forced signal or for individual realization. Likewise, the level of confidence in projections of extremes, often given in assessment reports, depends on whether a statement applies to a single realization of the future, or to the forced signal.

Our findings demonstrate that within all the complexity of non-linear processes controlling temperature and heavy precipitation extremes, a remarkably simple first-order pattern emerges: as global temperatures increase, the forced hot extremes and heavy precipitation intensification is widespread over most of the land region and consistency across models for these changes is high in particular over Eurasia and North America.