Jump to Main Content
On model differences and skill in predicting sea surface temperature in the Nordic and Barents Seas
- Langehaug, H. R., Matei, D., Eldevik, T., Lohmann, K., Gao, Y.
- Climate dynamics 2017 v.48 no.3-4 pp. 913-933
- Gulf Stream, advection, climate, climate models, heat, ice, prediction, surface water temperature, variance, Arctic region, Atlantic Ocean, Barents Sea, Northern European region
- The Nordic Seas and the Barents Sea is the Atlantic Ocean’s gateway to the Arctic Ocean, and the Gulf Stream’s northern extension brings large amounts of heat into this region and modulates climate in northwestern Europe. We have investigated the predictive skill of initialized hindcast simulations performed with three state-of-the-art climate prediction models within the CMIP5-framework, focusing on sea surface temperature (SST) in the Nordic Seas and Barents Sea, but also on sea ice extent, and the subpolar North Atlantic upstream. The hindcasts are compared with observation-based SST for the period 1961–2010. All models have significant predictive skill in specific regions at certain lead times. However, among the three models there is little consistency concerning which regions that display predictive skill and at what lead times. For instance, in the eastern Nordic Seas, only one model has significant skill in predicting observed SST variability at longer lead times (7–10 years). This region is of particular promise in terms of predictability, as observed thermohaline anomalies progress from the subpolar North Atlantic to the Fram Strait within the time frame of a couple of years. In the same model, predictive skill appears to move northward along a similar route as forecast time progresses. We attribute this to the northward advection of SST anomalies, contributing to skill at longer lead times in the eastern Nordic Seas. The skill at these lead times in particular beats that of persistence forecast, again indicating the potential role of ocean circulation as a source for skill. Furthermore, we discuss possible explanations for the difference in skill among models, such as different model resolutions, initialization techniques, and model climatologies and variance.