Estimation of Thin-Ice Thickness and Discrimination of Ice Type From AMSR-E Passive Microwave Data
Estimation of Thin-Ice Thickness and Discrimination of Ice Type From AMSR-E Passive Microwave Data
复制标题
DOI：
10.1109/tgrs.2018.2853590
复制
发表时间：
2019-01
影响因子：
8.2
通讯作者：
K. Nakata;K. Ohshima;S. Nihashi
中科院分区：
文献类型：
--
作者：
K. Nakata;K. Ohshima;S. Nihashi
用诸如高级微波扫描辐射计-地球观测系统（AMSR-E）等微波辐射计探测薄冰厚度，对于估计海冰的生成量非常有效，因为海冰的生成量造成密集的水推动海洋温盐环流。在以往的薄冰厚度算法中，冰厚度的估计是利用冰厚度和AMSR-E的极化比（PR）之间的负相关性。然而，在这些如履薄冰的算法，关系有很大的分散性。我们认为问题是由于没有考虑冰的种类。利用中分辨率成像光谱仪和高级合成孔径雷达数据，将南极洲周围的薄冰区分为两种类型：1）活动冰碛，包括冰碛和开阔水域; 2）薄固体冰，即相对均匀的薄冰区。对于每一种冰类型，我们研究了AMSR-E PR的36 GHz和冰的厚度之间的关系，表明活动frazil类型有一个小得多的厚度比薄固体冰类型相同的PR。这两种冰类型可以区分由一个简单的线性判别方法在平面上的PR和梯度比AMSR-E，与3%的错误分类。根据这些结果，我们提出了一种新的薄冰算法。首先采用线性判别法对两种冰型进行分类，然后利用经验公式求出每种冰型的冰厚。该算法显著提高了薄冰厚度的精度。
Detection of thin-ice thickness with microwave radiometers, such as the Advanced Microwave Scanning Radiometer-Earth Observing System (AMSR-E), is very effective for the estimation of sea-ice production, which causes dense water driving ocean thermohaline circulation. In previous thin-ice thickness algorithms, ice thickness is estimated by utilizing a negative correlation between ice thickness and polarization ratio (PR) of AMSR-E. However, in these thin-ice algorithms, the relationship has large dispersion. We consider that the problem is caused by not taking account of ice type. We classified thin-ice regions around Antarctica into two ice types: 1) active frazil, comprising frazil and open water and 2) thin solid ice, areas of the relatively uniform thin ice, using Moderate Resolution Imaging Spectroradiometer and Advanced Synthetic Aperture Radar data. For each ice type, we examined the relationship between the AMSR-E PR of 36 GHz and ice thickness, showing that the active frazil type has a much smaller thickness than the thin solid ice type for the same PR. The two ice types can be discriminated by a simple linear discriminant method in the plane of the PR and gradient ratio of AMSR-E, with the misclassification of 3%. From these results, we propose a new thin-ice algorithm. The two ice types are classified by the linear discriminant method, and then empirical equations are used to obtain the ice thickness for each ice type. This algorithm significantly improves the accuracy of the thin-ice thickness.