TJ1.6 Prediction of Northern Hemisphere Regional Surface Temperatures Using Stratospheric Ozone Information

Monday, 7 January 2019: 3:15 PM
North 128AB (Phoenix Convention Center - West and North Buildings)
Kane A. Stone, MIT, Cambridge, MA; and S. Solomon, D. E. Kinnison, C. F. Baggett, and E. A. Barnes

Correlations between springtime stratospheric ozone extremes and subsequent surface temperatures have been previously reported for both models and observations at particular locations in the Northern Hemisphere (NH). Motivated by the routine and precise measurements of total column ozone (TCO), here we quantify for the first time the potential use of ozone information for NH seasonal forecasts using both observations and a nine-member ensemble of the The Community Earth System Model, version 1 (CESM1) Whole Atmosphere Community Climate Model, version 4 (WACCM4). The ensemble composite correlations between March total column ozone and April surface temperatures display a similar structure to observations, but with slightly lower correlation magnitudes. This is due to internal variability in the model ensemble members, which slightly shift the correlation patterns from member to member. Using a linear regression model with March TCO averaged over 63–90°N as the predictor, we show that with a regression model created over 1980–2010 in observations, March TCO contains predictability of surface temperatures at locations in Eurasia up to 6 years into the future. Similar results are found in the model ensemble members. Additionally, we create an empirical forecast model to predict the sign of the observed as well as the modeled surface temperature anomalies using March Arctic TCO. Through a leave-three-years-out cross validation method, we show that March Arctic TCO can forecast the sign of the April surface temperature anomalies well in parts of Eurasia that show the lowest model internal variability. In some locations in Russia and southern Asia, predictability is seen through May. However, in other locations, such as North America, model internal variability reduces the predictability, even though individual ensemble members show large correlations.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner