ENSO Biases in GCMs and Their Relation to Mean State Biases

(w/ Dan VImont, Univ. Wisconsin Madison)


El Ni–o / Southern Oscillation (ENSO) variability represents the leading source of interannual variability in the tropical Pacific and globally.  Our understanding of ENSO developed rapidly in the 1980Õs and 1990Õs with the development of intermediate coupled models in which ENSO variability operates around a prescribed mean state.  This was a useful approach, as it has been found that ENSO characteristics are very sensitive to details of the tropical Pacific mean state and seasonal cycle.  At the same time, global climate models (GCMs) have improved to the point that ENSO variability exists, in some form, in many of the current generation of GCMs.  Unfortunately, large, and even small, biases in GCM simulations of the tropical mean state lead to large biases in simulations of ENSO variability.  While attempts have been made to relate biases in ENSO variability to biases in the mean state of the tropical climate, analysis has been limited to analysis of existing GCM output, qualitative comparisons between GCM output and coupled dynamical theory, and analysis of modal characteristics using very simple models.


The present proposal outlines a research plan aimed at quantitatively estimating the influence of mean state biases on ENSO biases in the present generation of GCMs.  This will be accomplished by development and application of a linearized version of the intermediate coupled models described above.  This linear ocean / atmosphere model (LOAM) can be tuned around observed or modeled mean states, and once having done so, has been shown to reproduce characteristics of the respective observed or modeled ENSO variability (e.g. amplitude, stability, period, seasonal phase-locking, regularity).  An advantage to the linear model is that it can be used to investigate the sensitivity of ENSO characteristics to specific features in the mean state by tuning model parameters to, say, observations, and substituting individual parameters derived from a model.  A research strategy is described that uses this model to:


1.      Characterize (quantitatively) the spatial and temporal structure of modeled ENSO variability, and ENSO characteristics when the LOAM is linearized around each modelÕs mean state 

2.      Using the LOAM, conduct sensitivity studies to quantify how mean state biases affect bias in ENSO simulation.

3.      Using the LOAM, conduct sensitivity studies to understand changes in ENSO behavior under future climate scenarios


This proposal directly addresses CVPÕs focus area in two ways:  (1) it proposes a strategy for understanding the source of bias in simulated interannual ENSO variability, and (2) it identifies specific biases in the mean state that produce those biases.  By identifying ENSO sensitivity to specific mean state biases, the work will provide quantitative guidance for modeling groups trying to improve ENSO simulation.