An analog system to enhance seasonal predictions of sea ice

Given the importance of wind and temperatures for the evolution of sea ice anomalies, we propose to develop a monthly-to-seasonal analog forecasting system for sea level pressures and temperatures over the Arctic. This approach departs from the conventional statistical methodologies (e.g., screening regression) and dynamical model forecasts. Our rationale is that neither of the latter two approaches, widely used in seasonal sea ice outlooks, has shown the ability to capture large year-to-year changes of summer ice extent in recent years. The use of an analog approach to forecasts of atmospheric forcing therefore offers a novel and potentially useful approach to improved seasonal sea ice forecasts of pan-Arctic and regional sea ice extent.

The proposal is submitted to the Climate Variability Program (CVP) competition “CVP – Understanding Arctic Sea Ice Mechanisms and Predictability”. It directly addresses the program element “Mechanisms, predictability and prediction of regional sea ice variation and change”. By targeting improvements in seasonal sea ice prediction for pan-Arctic and regional ice extent, the project responds to the NOAA Next Generation Strategic Plan’s objective #2: “Assessments of current and future states of the climate system that identify potential impacts and inform science, service and stewardship decisions”.

Our analog selection will be based on the current fields of the atmospheric circulation (sea level pressure, 500 hPa geopotential), indices of major atmospheric teleconnections (Arctic Oscillation, El Nino/Southern Oscillation, Pacific North American pattern, Pacific Decadal Oscillation) and the evolution of these fields and indices over the immediate preceding period of ~1 week. The atmospheric fields will be from the reanalysis products of the National Centers for Environmental Prediction (NCEP). The weights used for the different variables in the analog selection process will be based on the errors, spatially aggregated over the Arctic, of seasonal hindcasts for the 1948-2014 period of the NCEP reanalysis. We will experiment with different numbers of analogs to be composited into the seasonal forecast fields; the various analogs will be weighted to produce constructed analogs. Analog forecast systems will be developed for pan-Arctic ice extent and ice extent in 14 Arctic subregions.

The skill of the analog forecasts of the seasonal pressure and temperature fields will be evaluated over the Arctic Ocean and subarctic seas, and will be compared with the corresponding metrics from the Coupled Forecast System (CFS) that is run routinely at NCEP, as well as the North American Multi-Model Ensemble (NMME) that is distributed by the Climate Prediction Center. We will evaluate the skill as a function of season and sector of the Arctic/subarctic marine domain where at least seasonal sea ice occurs. We will document the impact of the analog forecast system on seasonal sea ice forecasts derived from a dynamical ocean-ice model.

We will produce an experimental forecast website displaying the forecasts, updated weekly, during the final year of the project after the system is optimized. The analog visualization site will leverage the user interface and display functions of the Arctic Collaboration Environment (ACE), which recently transitioned to a new home at the University of Alaska, Fairbanks. The display will include both the atmospheric forecasts (presented as actual fields and as departures from normal) superimposed on the current sea ice field and its departures from normal. The ACE system provides the capability for overlaying layers of other variables.

Collaborators will include the National Weather Service’s Alaska Region (Richard Thoman), with whom we will explore additional uses of the analog forecast system. At least one journal publication, in addition to the project’s deliverables.