Sensitivity analysis of atmospheric low‐frequency variability

• Published in: Quarterly Journal of the Royal Meteorological Society
• Main Authors: Corti, S., Palmer, T. N.
• Format: Article in Journal/Newspaper
• Language: English
• Published: Wiley 1997
• Subjects: North Atlantic; North Atlantic oscillation
• Online Access: http://dx.doi.org/10.1002/qj.49712354413; https://api.wiley.com/onlinelibrary/tdm/v1/articles/10.1002%2Fqj.49712354413; https://rmets.onlinelibrary.wiley.com/doi/pdf/10.1002/qj.49712354413

Abstract The sensitivity of the Pacific‐North‐American (PNA) pattern and the North‐Atlantic Oscillation (NAO) pattern to small perturbations is investigated using adjoint techniques. Analysis is perfomed on integrations of a quasi‐geostrophic (QG) model whose internal low‐frequency variability is realistic. Specifically, the sensitivity of the QG‐model PNA and NAO patterns have been made from a large set of five‐day integrations of the adjoint of the QG‐model, linearized about time‐varying basic‐state solutions of the model. the sensitivity perturbations are initially located well upstream of the pattern of low‐frequency variability and propagate downstream by wave dynamics, gaining energy through both barotropic and baroclinic processes. For the PNA pattern, it was found that variations in sensitivity were associated with fluctuations in the strength of the jet across the North Pacific. For the NAO pattern, it was found that fluctuations in sensitivity were correlated with the strength of the PNA pattern, the more sensitive NAO basic states having negative PNA index. This analysis has identified fairly distinct paths in phase space associated with predictable and unpredictable transitions, qualitatively similar to that found in the three‐component Lorenz model. Predictability of climate‐change problems is also investigated. the results from four different climate integrations in which an additional time‐invariant forcing is added to the model equations are presented. QG simulations show that the influence of a forcing proportional to the mean sensitivity pattern induces a stronger change in the probability distribution of the PNA pattern than an equivalent amplitude forcing which directly projects onto the PNA pattern itself.