Assessing Seasonal Predictability Sources and Windows of High Predictability in the Climate Forecast System Version 2

Seasonal prediction is of considerable socioeconomic value as the products may be utilized across many sectors (e.g. energy, agriculture, public health, etc.). In order to improve our current forecast capabilities, models must be continuously assessed in order to discover where improvements can be made. One particular focus is to identify sources of predictability and evaluate their representations in operational models. Low frequency climate modes are a key source of predictability on the seasonal time scale, and such modes may modulate prediction skill of the extratropical atmosphere. Here, the representation of two low-frequency climate modes, the ENSO and the NAO, are examined in the Climate Forecast System Version 2 (CFSv2) using the reforecasts from 1982-2010.

The predictability of these modes is discussed, and deficiencies in the teleconnections are explored. The ENSO is highly predictable, as expected, however there are some errors in the representation. In particular, an eastward shift in the ENSO warm tongue (convection zone) is evident. This eastward shift leads to an eastward shift in the Pacific North American pattern. Although this may look small on a global scale, it has important implications for regional climate anomalies. We find, as previous studies have found, that the model poorly predicts the NAO. Since statistical and dynamic models have predicted NAO with large skill a season to a year in advance, it led us to question what sources of NAO predictability are not well represented and where a disconnection may be evident within the CFSv2. Here, we analyze the representation of northern hemispheric sea surface temperature and the stratosphere during NAO winters, and large differences between the “truth” and prediction were found in both the sea surface temperature and the stratosphere, leading us to believe that these sources of NAO predictability are broken within the model. We also examine how these two low-frequency climate modes modulate the prediction skill of the northern hemispheric geopotential height fields, as well as a domain over the North Atlantic. Results show statistically significant changes in predictive skill during different phases of ENSO and NAO.

This study aims to analyze a model that is part of several collaborative efforts to improve subseasonal to seasonal prediction, as well as assisting NCEP in making improvements to the CFSv2.