The melting of Arctic sea ice and the future of trans-Arctic shipping

By Keith Haines

Most people will be familiar with news of the changing conditions in the Arctic where climate change seems to be at its fastest. The loss of sea ice each summer seems to show a rapidly declining trend, albeit with significant internannual variations. While the decline of summer sea ice in the Arctic may turn out disastrous for the current Arctic ecosystem, there could be positive effects for shipping as an ice-free arctic would provide much shorter trade routes between all north Atlantic ports and the Pacific ports in East Asia. But how can we assess the likely future scenarios for Arctic ice conditions?

2015 12 18 Keith Haines - Figure 1

Figure 1. The route of the Liquefied Natural Gas (LNG) tanker Ob River through the Arctic in Nov 2012, and the ice distribution and concentration it met along the way, accompanied at this time by a Russian icebreaker.

After the September 2007 Arctic sea ice minimum it was quickly recognised that the current generation of climate models were under-representing Arctic variability. Sea ice modelling is now receiving more attention and models are improving, but this is a slow process. Can we use the most recent generation of climate models (IPCC CMIP5, Climate Model Intercomparison Project5) to say anything about the future of Arctic sea ice conditions?  One approach is to calibrate or “bias correct” the current models based on recent observations, and then project this correction forward to obtain future scenarios (Melia et al 2015). Nat Melia at Reading has calibrated both the mean sea ice conditions and the interannual variability by comparing climate model ensemble spread to the detrended interannual variations in the observational record. The relationship (Mean And VaRIance Correction, MAVRIC) is shown below, where the model mean ice distribution  is weighted differently to the interannual anomalies. The observed interannual variability should ensure that recent summer sea ice minima, e.g. 2007, 2011, are potentially represented in each model’s interannual variability:

2015 12 18 Keith Haines - Equation 1

This has allowed a set of calibrated CMIP5 climate models to predict the future mean (shown in Figure 2) and interannual variability of Arctic sea ice conditions.

2015 12 18 Keith Haines - Figure 2Figure 2. The calibrated mean September CMIP5 sea ice thickness predictions, for each decade out to 2045. The current distribution on the left represents the 1979-2014 mean calibrated against the PIOMAS reanalysis record and therefore agrees well with current observations, followed by 2015-24, 2025-34, 2035-44 periods when the declining mean sea ice conditions can be seen. This does not represent what will actually be seen in any given year as there are plenty of interannual variations, which MAVRIC has also calibrated.

In this calibrated climate model framework the uncertainties in the predicted sea ice in any given year can be attributed to (1) interannual variability (orange in Figure 3), (2) variations between the climate models (blue), or (3) emission scenario uncertainty (green), following the approach of Hawkins and Sutton (2009).

2015 12 18 Keith Haines - Figure 3

2015 12 18 Keith Haines - Figure 4a

Figure 3. The percentage uncertainty in predicted sea ice volume out to 100 years based on calibrated CMIP5 data. Initially all the uncertainty is due to interannual variability (orange), as it should be with climate models well calibrated for present conditions. The uncertainty between models (blue) then grows into the future and is dominant again between 30-70 years hence. Emission Scenario uncertainty (green) starts to matter after about 40 years and grows thereafter.

What does this new prediction mean for the future of Arctic shipping? Figure 4 shows the potential Arctic summer shipping routes between the north Atlantic and north Pacific. Almost all current traffic uses the northern sea route through Russian waters (light blue). However, the new ice predictions in Figures 2 and 3 suggest that the more direct route across the central Arctic (green) will likely become navigable by ~2045 while the North West Passage (red) will also be increasingly navigable and preferred for shipping from north American ports. New work is nearing completion analysing the probabilities of each of these sea routes being open to different categories of shipping into the future. The variability and range of possible routes can be displayed using the ensembles of calibrated CMIP products discussed above giving much more information about the risks and rewards of planning Arctic trade routes into the future. Further work will then follow to determine the likely potential to forecast summer ice conditions in a given year along these sea routes on a seasonal timescale (6 months in advance), to allow shipping decisions for the coming Arctic summer to be made more routinely.

2015 12 18 Keith Haines - Figure 4

Figure 4. Potential trans-Arctic summer shipping routes connecting the north Atlantic and north Pacific. Each route has a different probability of meeting sea ice of various thicknesses and this information is critical for planning of voyages that are likely to become much more routine in the coming decades.


Melia, N., K. Haines and E. Hawkins, 2015. Improved Arctic sea ice thickness projections using bias corrected CMIP5 simulations.  The Cryosphere, 9, 2237-2251, 2015

Hawkins, E. and R. Sutton, 2009. The Potential to Narrow Uncertainty in Regional Climate Predictions. Bull. Amer. Meteor. Soc., 90, 1095–1107. doi:

This entry was posted in Climate, Climate change, Cryosphere, Environmental hazards, Numerical modelling, Polar and tagged . Bookmark the permalink.