Comparison of passive microwave ice concentration algorithm retrievals with AVHRR imagery, in Arctic peripheral seas
Comparison of passive microwave ice concentration algorithm retrievals with AVHRR imagery, in Arctic peripheral seas
复制标题
DOI：
10.1109/tgrs.2005.846151
复制
发表时间：
2005-06-01
影响因子：
8.2
通讯作者：
Meier, WN
中科院分区：
文献类型：
--
作者：
Meier, WN
海冰浓度的准确表示对于业务冰分析、过程研究、模型输入和长期气候变化检测很有价值。被动微波图像，例如来自特殊传感器微波/成像仪 (SSM/I) 的图像，对于监测海冰状况特别有价值，因为它们在所有天空条件下均覆盖盆地规模。使用四种常见算法 [Bootstrap (BT)、Cal/Val (CV)、NASA Team (NT) 和 NASA Team 2 (N2)] 得出的 SSM/I 推导的海冰浓度估算值与通过高级甚高分辨率辐射计 (AVHRR) 可见光和红外图像计算的浓度进行比较。比较是在大约八个月的时间内进行的。北极地区，并重点关注靠近冰缘的区域，这些区域的算法之间的差异可能最为明显。结果表明，CV 和 N2 相对于 AVHRR 具有最小的平均误差。 CV 倾向于高估浓度，而其他三种算法则低估浓度。 NT 的低估程度最大，平均接近 10%，在某些情况下甚至更高。在大多数情况下，SSM/I 算法的平均误差在 95% 显着性水平上彼此显着不同。 BT 算法具有最低的误差标准差，但在大多数情况下，没有发现所考虑的算法具有统计上显着不同的误差标准差。这表明空间分辨率可能是冰边缘附近区域 SSM/I 的限制因素，因为没有一种算法能够令人满意地解析混合像素。按季节、地区、冰况和 AVHRR 场景进行的统计细分总体上与总体结果一致。提出了代表性案例研究来说明统计结果。
An accurate representation of sea ice concentration is valuable to operational ice analyses, process studies, model inputs, and detection of long-term climate change. Passive microwave imagery, such as from the Special Sensor Microwave/Imager (SSM/I), are particularly valuable for monitoring of sea ice conditions because of their daily, basin-scale coverage under all sky conditions. SSM/I-derived sea ice concentration estimates using four common algorithms [Bootstrap (BT), Cal/Val (CV), NASA Team (NT), and NASA Team 2 (N2)] are compared with concentrations computed from Advanced Very High Resolution Radiometer (AVHRR) visible and infrared imagery. Comparisons are made over approximately an eight-month period in three. regions of the Arctic and focus on areas near the ice edge where differences between the algorithms are likely to be most apparent. The results indicate that CV and N2 have the smallest mean error relative to AVHRR. CV tends to overestimate concentration, while the other three algorithms underestimate concentration. NT has the largest underestimation of nearly 10% on average and much higher in some instances. In most cases, mean errors of the SSM/I algorithm were significantly different from each other at the 95% significance level. The BT algorithm has the lowest error standard deviation, but none of the considered algorithm's was found to have statistically significantly different error standard deviations in most cases. This indicates that spatial resolution is likely a limiting factor of SSM/I in regions near the ice edge in that none of the algorithms satisfactorily resolve mixed pixels. Statistical breakdowns by season, region, ice conditions, and AVHRR scene generally agree with the overall results. Representative case studies are presented to illustrate the statistical results.