This website uses cookies, which are small text files that are used to make websites work more effectively. In order to continue using this website, you will need to accept the use of cookies.

In our previous work Melia et al (2016) we showed how trans-Arctic shipping routes would become more available through the 21st century as sea ice declines, using multiple Climate models calibrated to current sea ice observations, eg Figure 1. But sea ice will continue to close shipping routes to open water vessels through the winter months for the foreseeable future and the availability of open sea routes will vary greatly from year to year. In a new paper just published Melia et al (2017) looks at whether the shipping season period (when sea routes are open) can be predicted in seasonal forecasts, again using several climate models, and testing both perfect and imperfect knowledge of the initial sea ice conditions.

 

We find skilful predictions of the upcoming summer shipping season can be made from as early as January, although typically forecasts may show lower skill before a May ‘predictability barrier’. Focussing on the northern sea route (NSR) off Siberia, the date of opening of this sea route is twice as variable as the closing date, and this carries through to reduced predictability at the start of the season. Under climate change the later freeze-up date accounts for 60% of the lengthening season. Both these features are illustrated in Figure 2 from the Canadian CanCM4 model, although its current sea ice is too low, more akin to mid-21st century conditions.

We find that predictive skill is state dependent with predictions for high or low ice years exhibiting greater skill than for average ice years. Forecasting the exact timing of route open periods is harder (more weather dependent) under average ice conditions while in high and low ice years the season is more controlled by the initial ice conditions from spring onwards. This could be very useful information for companies planning vessel routing for the coming season. We tested this dependence of the forecasts on the initial ice conditions by changing the initial ice state towards climatologically average conditions and show directly that early summer sea-ice thickness information is crucial to obtain skilful forecasts of the coming shipping season. This strongly suggests that good sea ice thickness observations should become a key component of the future Arctic observing system.
This study used available climate model ensembles from several research centres. A wider range of climate model simulations will be performed in APPLICATE allowing more and better targeted studies of many other important impacts of a changing Arctic.

  1. Melia, N., K. Haines, and E. Hawkins (2016), Sea ice decline and 21st century trans-Arctic shipping routes, Geophys. Res. Lett., 43, 9720 –9728 doi:10.1002/ 2016GL069315. http://onlinelibrary.wiley.com/doi/10.1002/2016GL069315/pdf
  2. Melia, N., K. Haines, E. Hawkins and J.J. Day (2017), Towards seasonal Arctic shipping route predictions. Env. Res. Lett. 12(8), https://doi.org/10.1088/1748-9326/aa7a60 

sea flow

 Figure 1: Fastest available September trans-Arctic routes in the mid-21st century from multi-model projections. Routes for low emissions, RCP2.6, and high emissions, RCP8.5, each contain 15 consecutive Septembers, from 5 GCMs each with 3 ensemble members, equating to 225 simulations per panel. Cyan lines represent Open Water vessels (OW) and pink line represent Polar Class 6 vessels (PC6); line weights indicate the number of transits using the same route. From Melia et al (2016) where more details can be found.

 

open water vessels

 

Figure 2: Ten member ensemble “idealised predictions” of the NSR season length for open water vessels, using simulations starting each January, showing ranked ensemble member opening/closing dates and ensemble mean trend. The CanCM4 model is used with historical climate forcings over 30 years, although this model has a low sea ice bias more similar to expected mid-21st century conditions displayed in Fig 1. Figure taken from Melia et al (2017)