北极海冰密集度数值预报及验证评估研究

本文系统地评估了国家海洋环境预报中心于我国第七次北极科学考察期间开展的北极海冰密集度数值预报结果。该预报系统基于麻省理工大学通用环流模式，并采用牛顿松弛逼近（Nudging）资料同化方法，计算输出未来1~5 d的北极海冰密集度预报产品。本文将数值预报结果同卫星观测的海冰密集度、再分析资料和"雪龙"号第七次北极考察期间观测的海冰密集度数据进行了对比分析。结果表明，预报的北极海冰密集度小于卫星观测值，24 h、72 h和120 h预报结果的偏差分别为-2.7%、-3.1%和-3.2%；数值产品的预报技巧好于气候态结果和惯性预报，但是在海冰出现快速融化或冻结时，基于Nudging同化的数值预报技巧仍有不足。另外，相比船测数据，数值预报结果在海冰边缘区的偏差相对较大，24 h、72 h和120 h预报结果的偏差分别为8.8%、12.0%和14.5%。

Arctic sea ice concentration numerical forecasting and its evaluation

In this study, we evaluated the 24-120 h Arctic sea ice concentration forecasts provided by National Marine Environmental Forecasting Center during the 7th Chinese National Arctic Research Expedition (CHINARE 2016). The Arctic sea ice forecast system was based on the MIT general circulation model (MITgcm) ice-ocean coupled model with Advanced Microwave Scanning Radiometer 2 (AMSR2) sea ice concentration data Nudged. We compared the numerical forecast products with the satellite data, reanalysis data and ship-based in situ sea ice concentration observations during CHINARE 2016. It was shown that the Arctic sea ice concentration forecasts were smaller than the satellite data. The mean biases between 24 h, 72 h, 120 h forecasts and satellite data were -2.7%, -3.1% and -3.2%. The numerical sea ice concentration forecasts were better than the climatological means and the inertial forecasts. But the forecast skill was required to improve when the Arctic sea ice had surged rapid melting or freezing. Moreover, the forecast biases were larger compared with ship in situ observations in the marginal ice zone. The mean biases between 24 h, 72 h, 120 h forecasts and ship in situ data were 8.8%, 12.0% and 14.5%.