Enhancing seasonal-to-decadal Prediction Of Climate for the North Atlantic Sector and Arctic

Climate predictions by oceanic conditions EPOCASA has shown that climate in our region may be predicted several years in advance from knowledge of oceanic conditions. The northern hemisphere experiences shifts in climate that can persist for several decades. The recent changes, including a series of harsh winters, have highlighted the need to better understand and predict climate variations from a season to a decade in advance. EPOCASA has assessed the predictability of North Atlantic Sector and Arctic climate variability, by analysing observations and by developing and applying the Norwegian Climate Prediction Model (NorCPM). NorCPM combines advanced Ensemble Kalman Filter data assimilation and the Norwegian Earth System Model (NorESM). EPOCASA took the first steps towards operational seasonal-to-decadal climate prediction in Norway that could benefit Norwegian Society and Economy. EPOCASA has shown that climate in our region may be predicted several years in advance from knowledge of oceanic conditions. Specifically, temperature anomalies in the Norwegian and Barents-Kara Seas influence local climate, through heat loss to the atmosphere. Many of these anomalies can be linked to the propagation of temperature signals along the Gulfstream in the North Atlantic and northwards right through the Norwegian Sea. This propagation takes several years and occured several times during the last 60 years. The temperature anomalies in the Norwegian Sea are therefore predictable. Climate models require high ocean resolution to capture this mechanism. We have also improved understanding of the predictable nature of the North Atlantic Ocean. Three typical timescales exist: an interannual, a multi-decadal, and a centennial. The variability on multi-decadal timescales is of great interest because two-way interaction between the ocean and atmosphere can enhance predictability in the ocean and atmosphere. Climate models need to extend to the stratosphere to capture this mechanism. Extensive development Through the EPOCASA research project, the NorCPM was extensively developed. A variety of data were assimilated for different historical periods extending back to 1950. Two ocean reanalyses were produced: one based only on SST, and one with SST and hydrographic profiles. The benefit of using ensemble based data assimilation in isopycnal coordinate was clearly demonstrated (compared to assimilation at specific depths). Assimilation of sea ice concentration is also skilful in constraining sea ice thickness and upper ocean hydrography. NorCPM can predict El Niño Southern Oscillation (ENSO) events 2-3 seasons in advance, with skill matching the best available predictions from other modelling centres. There is also high-skill in predicting Nordic Sea and for Arctic sea ice cover. Further experiments with the ECMWF model highlight the importance of assimilating snow in order to represent the cold European winter 2009/10. NorCPM decadal prediction experiments show improved skill compared to standard climate projections (i.e., without assimilation) in the North Atlantic. Hydrographic data was crucial, with SST only based predictions exhibiting poor skill. We have also investigated the predictive skill in the Nordic Seas and Barents Sea in three other climate models. All models have predictable aspects, but show little consistency possibly because of different initialization techniques and simulated climatology. Ocean model resolution is likely an important factor. A closer inspection of the ocean part in NorESM shows that an ocean with higher resolution better represents how temperature anomalies propagate northwards along the extension of the Gulfstream. NorCPM experiments show that Pacific decadal variability contributes to explaining the early century warming of the Arctic (1920-1940), adding to the impacts of external radiative forcing and Atlantic multi-decadal variability. We also show that model differences dominate uncertainty in long-term projections of the North Atlantic Ocean circulation, and that it is important to account for volcanic forcing in probabilistic prediction of sea level, ocean circulation, and sea-ice variability, because volcanic activity enhances climate variability on annual-to-decadal timescales. EPOCASA has also formed two statistical prediction models based on observations. One statistical model shows that measurements of ocean heat transport can be used to predict sea ice cover in the Barents Sea in wintertime. More than 50% of the observed sea ice variance can be predicted up to two years in advance. Another statistical model shows that the Norwegian surface air temperature and precipitation can be predicted years in advance based on knowledge of oceanic conditions. We predict that Norwegian air temperature will decrease over the coming years.

During the last decade global surface temperatures rose less rapidly than in the preceding decades. The northern hemisphere saw a spate of harsh winters. The 2012/2013 winter, for example, saw extremely cold temperatures and high snowfalls across Europe a nd eastern North America, and one of the driest March in Western Norway since reliable observations started around 1900. At the same time the Arctic experienced extreme warming and accelerated sea ice loss, culminating in the record-low of September 2012. While anthropogenic global warming may explain much of the recent changes, the northern hemisphere climate fluctuates strongly on timescales of several decades. EPOCASA aims to assess the extent to which these changes are predictable, by developing and applying a dynamical climate prediction system that focuses on the North Atlantic Sector and Arctic. EPOCASA is motivated by several recent advances: the demonstration that subpolar North Atlantic sea surface temperature (SST) are predictable on seasonal- to-decadal timescales, with observed links to the Nordic Seas and Arctic; potential role of stratosphere-troposphere interactions in the atmospheric response to tropical and extra-tropical SST, sea ice, and snow cover variations; as well as the need to re duce model systematic error and develop more advanced forecast initialisation techniques. We will analyse observations and perform extensive model simulations to assess the potential of these factors to enhance climate prediction in the region. The EPOCA SA prediction system couples an advanced data assimilation method (the Ensemble Kalman Filter) to the current version of the Norwegian Earth System Model, building upon developments initiated at the Center for Climate Dynamics at the Bjerknes Center. It w ill be first seasonal-to-decadal climate prediction capability in Norway, paving the way for operational climate prediction that will be of direct benefit to norwegian society and economy.