IAP LASG Ranked Top on Arctic Sea Ice Prediction by SPIN 2019 Report
According to 2019 Post-Season Report released by the International Sea Ice Prediction Network (SIPN) in February 2020, among all the two-dimensional spatial (2D) grid prediction data of the Arctic sea ice submitted by 12 global research teams, the IAP LASG total prediction score (the Brier Score, BS) ranked first and the prediction results were close to the actual sea ice coverage area.
The subseasonal and seasonal prediction of Arctic sea ice is of great importance to monitor changes in Arctic sea ice, understand the climate change process, ensure the safety of navigation in Arctic waterways, and realize the exploitation and sustainable development of Arctic natural resources.
The accurate prediction of sea ice has been a great challenge for international science community. Since 2015, SIPN has collected global scientists' predictions of the Arctic sea ice coverage in September from July to August every year and it represents the highest level of subseasonal to seasonal predictions of Arctic sea ice. In 2019, a total of 42 international teams submitted total area prediction results, of which 12 teams further submitted two-dimensional (2D) grid prediction data of Arctic sea ice coverage.
Based on the Arctic sea ice prediction from the FGOALS-f2 S2S dynamic ensemble prediction system, and with the joint support of the Alliance of International Science Organizations (ANSO) and Big Earth Data Science Engineering Project (CASEarth), The IAP LASG team submitted both the total area and 2D grid data prediction results to SPIN in June, July and August, of which the predicted total areas of sea ice in September were respectively 4.12 million, 4.01 and 4.35 million square kilometers while the actual area was 4.32 million square kilometers by observation.
The prediction of the spatial distribution (2D) of Arctic sea ice cover area is more challenging, and its potential value is much higher than that of the total area. The Brier Score (BS) score is used internationally to measure the accuracy of the prediction results. The value range of the BS score is zero to one, the smaller the better. A zero BS score means the best prediction and the prediction is correct. The mean BS score（about 0.04）of IAP LASG FGOALS-f2 team three submissions in 2019 is very close to that of the multi-model ensemble (MME) in terms of scoring skills, superior to the skills of international popular prediction agencies in the United States and Europe.