Southern Ocean Sea Ice Predictability

Adapted and condensed by OOI from Cerovecki, et al., 2022, doi: /10.1088/1748-9326/ac7d66.

Forecasting Antarctic sea ice conditions, including specifics such as the position of the ice edge in the Southern Ocean, are substantial challenges. As a part of the Polar Prediction Project, there is a focus on improving coupled air-sea-ice prediction models and determining key sources of forecast errors. In a recent study, Cerovecki et al. (2022) show that sea ice forecast skill is linked to the accuracy of the surface forcing, and in particular, the net surface radiation. The goal of the study was to quantify errors that degrade the skill of Southern Ocean sea ice forecasts during the freezing season. They conclude that accurately modeling the surface downward longwave radiation (DLW) component of the net surface radiation is critical to sea ice prediction over the Southern Ocean.

The authors review prior results indicating that climate models have different behaviors in different seasons relative to ground truth. In spring and summer, the models over-estimate the net surface radiation whereas in winter the models under-estimate the net longwave radiation. Recognizing that these issues relate to representations of cloud cover, which can be diagnosed using DLW, the authors also note that some models showed DLW biases of up to 100 W/m2 compared to ground truth. These results were based on comparisons at McMurdo Station, Antarctica, whereas the authors were interested in processes occurring near the ice edge where few direct observations are available.

Figure 1. Monthly mean downward longwave radiation (DLW) for Jan 2016 – Nov 2018 from the OOI Southern Ocean buoy (solid), and three reanalysis products: ERA5 (black dash), ERA-Interim (long grey dash) and NCEP1 (short gray dash). From Ceroveki et al., 2022.

The OOI Southern Ocean surface mooring provided a rare source of in-situ air-sea flux data for comparison. The study used DLW from the METBK instrument package on the OOI Southern Ocean buoy to compare with results from the ECMWF Interim reanalysis (ERAI), the ERA fifth- generation reanalysis (ERA5), and the NOAA National Centers for Environmental Prediction reanalysis (NCEP1). Despite some data gaps, the 1 min OOI METBK observations for Jan 2016 – Jan 2020, were critical to determining model biases. Hourly mean DLW data from the two METBK packages were averaged together to create the observational record.

Comparison of the observed monthly mean DLW with reanalysis output showed systematic underestimates by the models relative to the observations. The nature of the offsets is shown in Figure 1 – the reanalysis models do a relatively good job of capturing month to month variability, but with a consistent low bias. The mean offsets range from -13 W/m2 for ERAI to -28 W/m2 for NCEP1. These biases are comparable to those diagnosed at McMurdo Station, and suggest that the ERA5 DLW radiation underestimate is of the order of 20–50 W/m2. This is consistent with the finding that coupled model forecast systems over-estimate sea ice growth.

The authors conclude that a significant deficit in reanalysis DLW, related to the accuracy of cloud representation in the models, is a common problem over the Southern Ocean and impacts the skill of sea ice cover prediction. In particular, the ERA5 reanalysis may underestimate DLW by up to 50 W/m^2 during the during the freezing season. The OOI Southern Ocean data, from the furthest south sustained air-sea flux mooring, proved uniquely valuable in codifying these results.


Cerovecki, I, R. Sun, D.H. Bromwich, X. Zou, M.R. Mazloff and S -H.Wang (2022). Impact of downward longwave radiative deficits on Antarctic sea-ice extent predictability during the sea ice growth period. Environ. Res. Lett. 17 084008. DOI: /10.1088/1748-9326/ac7d66.