Use of System Identification Techniques to Explore the Hydrological Cycle Response to Perturbations in Climate Models

Abstract:

Identifying the influence of radiative forcing on hydrological cycle changes in climate models can be challenging due to low signal-to-noise ratios, particularly for regional changes. One method of improving the signal-to-noise ratio, even for short simulations, is to use techniques from engineering, broadly known as system identification. Through this method, forcing (or any other chosen field) in multiple regions in a climate model is perturbed simultaneously by using mutually uncorrelated signals with a chosen frequency content, depending upon the climate behavior one wishes to reveal. The result is the sensitivity of a particular climate field (e.g., temperature, precipitation, or cloud cover) to changes in any perturbed region.

We demonstrate this technique in the Community Earth System Model (CESM). We perturbed surface air temperatures in 22 regions by up to 1°C. The amount of temperature perturbation was changed every day corresponding to a predetermined sequence of random numbers between -1 and 1, filtered to contain particular frequency content. The matrix of sequences was then orthogonalized such that all individual sequences were mutually uncorrelated. We performed CESM simulations with both fixed sea surface temperatures and a fully coupled ocean. We discuss the various patterns of climate response in several fields relevant to the hydrological cycle, including precipitation and surface latent heat fluxes. We also discuss the potential limits of this technique in terms of the spatial and temporal scales over which it would be appropriate to use.