Search

Workshop: Multi-annual to Decadal Climate Predictability in the North Atlantic-Arctic Sector

Climate model studies have shown a wide range of responses in the northern hemisphere to reduced sea ice loss in the Arctic. The most robust is the thermodynamic response with local warming in the region of the sea ice loss. Many studies also find a shift in the tropospheric jet towards the equator and a tendency towards the negative phase of the North Atlantic Oscillation. However, these mid-latitude dynamic responses seem somewhat dependent on the extent and location of sea ice loss as well as model differences. We use new atmosphere-only experiments from the Polar-Amplification-MIP of CMIP6 to further explain how the dynamic responses in the atmosphere relate to the pattern of sea ice loss in the Arctic on multi-decadal time-scales. We find exciting new results that suggest that Arctic sea ice loss can also impact the southern hemisphere, while Antarctic sea ice loss can impact the northern hemisphere, and we propose possible atmospheric mechanisms to explain this linkage. We also highlight the importance of large ensembles to differentiate between robust responses and atmospheric noise.

In this study, we analyse the inter-relationships between the Labrador Sea densities, the boundary currents, the AMOC and, more generally, the wider climate of the North Atlantic across an ensemble of climate models. The study mostly relies on the analysis of two 300-year long high-resolution coupled control simulations (with HadGEM3-GC2 and HiGEM, respectively), and is completed withan ensemble of CMIP5 preindustrial control experiments to assess the model dependence of the results. We have characterised the main modes of Labrador Sea density (LSD) variability across these models. The leading EOF of LSD is reasonably consistent across the control experiments (i.e. HadGEM-GC2, HiGEM and CMIP5). All models show a fairly uniform vertical structure, with maximum positive density values near the surface that slowly decrease with depth. The associated Principal Components of Labrador Sea densities (PC1-LSD) is generally associated with multidecadal variability, and have enhanced variance between 12-30 years. There is also a good agreement on the density evolution in the ocean-only forced experiments; all depict an increase in the Labrador Sea densities from the 60s to the mid 90s, followed by a decreasing trend up to the present. These coherent Labrador Sea density changes are encouraging, but do not appear to translate to coherent changes in the AMOC strength in the forced historical runs. In the control experiments, decadal trends in PC1 do have a strong link with equivalent trends in the AMOC at 45°N, and to a lesser extent with the AMOC at 26°N (which tend to exhibit suppressed variability at interdecadal timescales). Interestingly, no link between PC1-LSD and AMOC at 26°N is observed when the Ekman transport signal is removed. The link with the AMOC at subpolar latitudes can be explained through an effect of LSDs on the western boundary currents. Indeed, PC1-LSD is tightly linked to the boundary densities at 45 and 57°N, but can show important discrepancies across models regarding the depths involved. Larger model discrepancies occur when extending the analysis to 35°N. Encouragingly, all models support a link between the multi-decadal trends in PC1-LSD and the delayed trends (by 3 to 10 years) in upper ocean temperature in the Eastern Subpolar Gyre (ESPG T700). This is a consistent result regardless of the particular model representation of the PC1-LSD links with the AMOC indices and the boundary densities, further supporting the role of the LSDs and the AMOC on the recent cold blob and its associated climate impacts.

Prediction of sea-ice is not only important for shipping but also for weather as it can have a signicant climatic impact. Therefore sea-ice predictions are important for accurate inter-annual to decadal prediction of climate in coupled climate models. However, to our knowledge there has been little work on how predictable sea-ice is in a coupled atmosphere-ocean-ice model. We have studied the predictability of sea-ice in HadCM3 using case study ensemble experiments with external forcing from the late 20th century designed to emphasize the predictability in the climate system due to initial conditions. Here, we will concentrate on spring (maximum) ice-volume in a box in the Greenland Sea (30W to 10E, 68N to 80N). Model climatology from a control-run shows that this region has high inter-annual variability in sea-ice volume. We find that although ice may almost completely disappear from this area in late autumn, the same anomalies re-appear in the following spring for at least the first four years in three out of four case studies. The mechanism for this appears to be related to persistence of ocean heat content in the initial conditions and the state of the meridional overturning circulation and its associated heat transport. In this model, the atmosphere appears to less important than the ocean in determining the predictability of sea-ice volume in the Greenland Sea., Previously curated at: http://cedadocs.ceda.ac.uk/807/ Event type: other. The publish date on this item was its original completed date. This item was not refereed before the publication Main files in this record: poster-NCAS09-HermansonSuttonKeeley.ppt Item originally deposited with Centre for Environmental Data Analysis (CEDA) document repository by Dr Leon Hermanson. Transferred to CEDA document repository community on Zenodo on 24/11/20

It is well established that, based on knowledge of the initial conditions, important aspects of climate are predictable up to a year ahead. This predictability is primarily associated with the El Niño Southern Oscillation (ENSO). But is climate predictable further ahead? To what extent does knowledge of initial conditions constrain longer-term climate forecasts? In this study we investigate whether there is evidence that climate variables are potentially predictable beyond the generally accepted limit of ENSO predictability. Recognising that predictability is likely to be dependent on initial conditions we consider model-based case studies rather than measures of average predictability. The atmosphere-ocean coupled Hadley Centre HadCM3 model was used in this work. Starting from a pre-indstrial state, a single run was integrated with observed changes in greenhouse gases, volcanic aerosols and the solar cycle for the period 1860-1950. Still using observed forcings, an ensemble of five members was started from the 1950 state and integrated for another 50 years. Each case study consists of two further ensembles started with contemporaneous intial conditions from two different members of the five member ensemble. Four case studies have been completed using ensemble members with opposing large, persistent, regional differences in sea surface temperatures and ocean heat content. The results show clear evidence that knowledge of initial conditions can constrain predictions of climate variables beyond ENSO predictability. In the particular case studies considered, the aspects of climate that are predictable include North Atlantic sea surface temperatures, western European and southeastern US 1.5m temperatures, tropical and subtropical precipitation inthe Pacific and tropical precipitation in the Atlantic. However, predictability of climate variables is strongly dependent on the initial conditions. What are the mechanisms that give rise to these predictable signals? In most cases it seems to be simple persistence of ocean heat content anomalies, but there is also evidence of other mechanisms including ocean dynamics and ocean-atmosphere interaction. We are carrying out further work to interpret and clarify these mechanisms., Previously curated at: http://cedadocs.ceda.ac.uk/805/ Event type: conference. The publish date on this item was its original completed date. This item was not refereed before the publication Main files in this record: poster-NCAS08-HermansonSutton.ppt Item originally deposited with Centre for Environmental Data Analysis (CEDA) document repository by Dr Leon Hermanson. Transferred to CEDA document repository community on Zenodo on 24/11/20

This poster presents four case studies where predictability has been found in year 2 of a decadal prediction and the mechanisms which lead to this predictability, Previously curated at: http://cedadocs.ceda.ac.uk/806/ Event type: workshop. The publish date on this item was its original completed date. This item was not refereed before the publication Main files in this record: poster-mechdecpred.ppt Item originally deposited with Centre for Environmental Data Analysis (CEDA) document repository by Dr Leon Hermanson. Transferred to CEDA document repository community on Zenodo on 24/11/20