Assessing the forecast skill of Arctic sea ice extent in the GloSea4 seasonal prediction system

An assessment of the ability of the Met Office seasonal prediction system, GloSea4, to accurately forecast Arctic sea ice concentration and extent over seasonal time scales is presented. GloSea4 was upgraded in November 2010 to include the initialization of the observed sea ice concentration from sa...

Full description

Saved in:

Bibliographic Details
Published in:	Climate dynamics 2015-01, Vol.44 (1-2), p.147-162
Main Authors:	Peterson, K. Andrew, Arribas, A., Hewitt, H. T., Keen, A. B., Lea, D. J., McLaren, A. J.
Format:	Article
Language:	eng
Subjects:	Climate change Climatology Earth and Environmental Science Earth Sciences Geophysics/Geodesy Ice Marine Meteorology Methods Observations Ocean circulation Oceanography Sea ice Weather forecasting
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!

Description
Summary:	An assessment of the ability of the Met Office seasonal prediction system, GloSea4, to accurately forecast Arctic sea ice concentration and extent over seasonal time scales is presented. GloSea4 was upgraded in November 2010 to include the initialization of the observed sea ice concentration from satellite measurements. GloSea4 is one of only a few operational seasonal prediction systems to include both the initialization of observed sea ice followed by its prognostic determination in a coupled dynamical model of sea ice. For the forecast of the September monthly mean ice extent the best skill in GloSea4, as judged from the historical forecast period of 1996–2009, is when the system is initialized in late March and early April near to the sea ice maxima with correlation skills in the range of 0.6. In contrast, correlation skills using May initialization dates are much lower due to thinning of the sea ice at the start of the melt season which allows ice to melt too rapidly. This is likely to be due both to a systematic bias in the ice-ocean forced model as well as biases in the ice analysis system. Detailing the forecast correlation skill throughout the whole year shows that for our system, the correlation skill for ice extent at five to six months lead time is highest leading up to the September minimum (from March/April start dates) and leading up to the March maximum (from October/November start dates). Conversely, little skill is found for the shoulder seasons of November and May at any lead time.
ISSN:	0930-7575 1432-0894