First Institute of Oceanography, Ministry of Natural Resources, Qingdao 266061, China
Laboratory for Regional Oceanography and Numerical Modeling, Pilot National Laboratory for Marine Science and Technology (Qingdao), Qingdao 266237, China
Key Laboratory of Marine Science and Numerical Modeling, Ministry of Natural Resources, Qingdao 266061, China
Department of Atmospheric and Environmental Sciences, University at Albany, State University of New York, Albany, New York 12222, USA
College of Ocean and Meteorology, Guangdong Ocean University, Zhanjiang 524088, China
National Marine Environmental Forecasting Center, Beijing 100081, China
The National Key Research and Development Program of China under contract Nos 2018YFC1407205 and 2018YFA0605901; the Basic Scientiﬁc Fund for National Public Research Institute of China (ShuXingbei Young Talent Program) under contract No. 2019S06; the National Natural Science Foundation of China under contract Nos 41821004, 42022042 and 41941012; the China-Korea Cooperation Project on Northwestern Pacific Climate Change and its Prediction.
To improve the Arctic sea ice forecast skill of the First Institute of Oceanography-Earth System Model (FIO-ESM) climate forecast system, satellite-derived sea ice concentration and sea ice thickness from the PIOMAS (the Pan-Arctic Ice-Ocean Modeling and Assimilation System) are assimilated into this system, using the method of localized error subspace transform ensemble Kalman ﬁlter (LESTKF). Five-year (2014–2018) Arctic sea ice assimilation experiments and a 2-month near-real-time forecast in August 2018 were conducted to study the roles of ice data assimilation. Assimilation experiment results show that ice concentration assimilation can help to get better modeled ice concentration and ice extent. All the biases of ice concentration, ice cover, ice volume, and ice thickness can be reduced dramatically through ice concentration and thickness assimilation. The near-real-time forecast results indicate that ice data assimilation can improve the forecast skill significantly in the FIO-ESM climate forecast system. The forecasted Arctic integrated ice edge error is reduced by around 1/3 by sea ice data assimilation. Compared with the six near-real-time Arctic sea ice forecast results from the subseasonal-to-seasonal (S2S) Prediction Project, FIO-ESM climate forecast system with LESTKF ice data assimilation has relatively high Arctic sea ice forecast skill in 2018 summer sea ice forecast. Since sea ice thickness in the PIOMAS is updated in time, it is a good choice for data assimilation to improve sea ice prediction skills in the near-real-time Arctic sea ice seasonal prediction.
Figure 1. Ensemble spread (one standard deviation) of simulated sea ice concentration (a–d) and sea ice thickness (e–h) on the first day of March (a, e), June (b, f), September (c, g), and December (d, h) 2014 in the FIO-ESM climate forecast system.
Figure 2. Numerical experiment processes.
Figure 3. Arctic sea ice extent (SIE) biases during 2014 to 2018. Black, blue and red lines are from EXP1, EXP2 and EXP3, respectively. The biases are the results between simulations and observations from NSIDC.
Figure 4. Arctic sea ice concentration observations from NSIDC (a, e) and model biases (b–d and f–h) in March (a–d) and September (e–h) during 2014 to 2018.
Figure 5. Root mean square error (RMSE) of Arctic sea ice concentration during 2014 to 2018. (a), (b), and (c) are from EXP1, EXP2, and EXP3, respectively.
Figure 6. Arctic sea ice volume (SIV) during 2014 to 2018 in the PIOMAS, EXP1, EXP2, and EXP3.
Figure 7. Root mean square error (RMSE) of winter season Arctic sea ice thickness during 2014 to 2018. (a)–(c) are calculated based on the CryoSat-2 and SMOS (CS2SMOS) merged sea ice thickness. (d)–(f) are calculated based on PIOMAS sea ice thickness.
Figure 8. Location of IceBridge observations during 2014–2018 (a) and the probability density function (PDF) distribution of modeled sea ice thickness biases compared with IceBridge observations in EXP1, EXP2, and EXP3 (b). The legend in a is the time of the observations.
Figure 9. Time series of spatial correlation coefficients between PIOMAS and modeled Arctic sea ice thickness during 2014 to 2018. Green, blue and red lines are for EXP1, EXP2 and EXP3, respectively.
Figure 10. Arctic integrated ice edge error (IIEE) from different near-real-time forecasts in the early of August 2018. Observation persistence represents observation-based benchmark based on the observed sea ice conditions in the 1st August 2018. Climatology represents the climatology-based benchmark based on the observed sea ice conditions during 1989–2018. The observations are from the same datasets used in data assimilation. CMA represents China Meteorological Administration. ECMWF represents European Centre for Medium-Range Weather Forecasts. KMA represents Korea Meteorological Administration. Meteo France represents Météo France. NCEP represents National Centers for Environmental Prediction. UKMO represents UK Met Oﬃce. MMM represents multimodel ensemble mean. FIO-ESM EXP1, FIO-ESM EXP2, and FIO-ESM EXP3 represent the forecasts based on EXP1, EXP2 and EXP3 in this study, respectively.
Figure 12. Forecasted Arctic sea ice concentration for 25 August 2018 based on different forecasting systems. They are the forecasts from the early of August 2018. The red line is the sea ice edge from satellite observations.
Figure 11. Forecasted Arctic sea ice concentration for 15 August 2018 based on different forecasting systems. They are the forecasts from the early of August 2018. The red line is the sea ice edge from satellite observations.