Multi-model seasonal climate hindcast skill and predictability for the Southeast United States

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Tuesday, 19 January 2010
Lydia Stefanova, Center for Ocean-Atmospheric Prediction Studies, Tallahassee, FL; and V. Misra, J. J. O'Brien, and E. Chassignet

We assess the current global dynamical models' prediction skill for the Southeast United States region using a collection of seasonal multi-model ensemble hindcasts of near surface variables. The seasonal hindcasts used here come from several global models' ensemble hindcasts, hosted by the Asia Pacific Economic Cooperation Climate Center (APCC), for the period 1982-2002 (the longest period common to the largest number of APCC models). The model hindcasts are analyzed using measures of potential predictability, anomaly correlation, multi-category equitable threat score, and Brier skill score.

We find that the models' predictability and hindcast skill have considerable seasonality and variation within the Southeastern US domain. The largest potential predictability (signal to noise ratio) of precipitation anywhere in the United States is found in the Southeast, in spring and winter seasons. For two-meter temperature, the regions of potential predictability maxima lie outside of the Southeast domain in all seasons.

In terms of deterministic hindcast skill, the winter hindcasts of precipitation are most skillful. In fact, the wintertime precipitation hindcast skill in the Southeast is larger than the precipitation skill for any other season and region of the US. Most of this skill is clearly ENSO-driven. Atmospheric models forced with observed SSTs, as well as coupled ocean-atmosphere models, are found to have very high skill in winter and no skill in summer. The models using observed SSTs, unlike the coupled models, also show some skill in simulating spring precipitation anomalies. The lack of summer and fall precipitation hindcast skill can be largely attributed to the inability of coarse resolution global models to adequately resolve convection or tropical storm activity.

For the two-meter air temperatures, winter hindcasts are least skillful. Models with observed SST forcing have the largest skill in spring and summer, and less skill in fall and winter. Models with coupled SSTs and those forced with imposed hindcast SST show almost no skill in any season. This suggests that local forcings, such as the SSTs of the Gulf of Mexico and the western Atlantic, whose interannual variability is not well captured in coupled models, are more important than ENSO in regulating the seasonal mean near surface temperatures in the Southeastern US.

Probabilistic and categorical hindcasts conform to the deterministic findings, in that there is excellent skill for winter precipitation, and virtually none for summer. We find that, when skillful, the models are conservative, such that low probability hindcasts tend to be overestimates, while high probability hindcasts tend to be underestimates, although both the reliability and resolution are quite high.