Regional high-resolution fast ice maps - Satellite synthetic aperture radar (SAR) data

This dataset comprises high spatial- and temporal-resolution maps of coastal landfast sea ice (fast ice) distribution in the vicinity of the Cape Darnley Polynya in East Antarctica, in the June-November (winter-spring) periods of 2008 and 2009. The maps were derived from cross-correlation of pairs of spatially-overlapping Envisat Advanced Synthetic Aperture Radar (ASAR) images, using a modified version of the IMCORR algorithm to determine vectors of sea-ice motion (as described in Giles et al., 2011). Fast ice is then distinguished from moving pack ice by the fact that it is stationary. The raw ASAR WSM data (swath width 500 km) were processed using ENVI image processing software to produce geo-referenced images with a 75m pixel size. Use of SAR data ensures coverage uninterrupted by cloud cover or polar darkness. Image pairs were chosen with a time separation between 2 and 21 days. IMCORR processing of the image pairs for mapping fast ice follows Giles et al (2011) – using a reference tile size of 32x32 pixels and a search tile size of 64 x 64 pixels. A land mask was applied to avoid contamination from matches on stationary features over the continental ice sheet. The grid spacing was set to 16 x 16 pixels, so the images were over-sampled by a factor of 2 to provide a more dense set of results.Stationary fast ice vectors were chosen from the IMCORR results using a combination of the cluster search technique and a variation of the z-axis threshold technique as detailed in Giles et al (2011). The cluster search technique was applied to the IMCORR results from each image pair to derive the initial set of valid vectors – this set could contain both stationary fast ice vectors and non-stationary pack ice vectors. Due to registration errors in the image pairs, the stationary vectors will not necessarily be centred around zero, so using a simple window around the zero offset mark to differentiate the fast ice vectors was not possible. To select the stationary vectors, a 2D histogram was constructed from the X-Y vector displacements, and a 2D Gaussian was fitted to this histogram. The fast ice vectors will dominate because of the large image pair time separation and small search tile size, so the Gaussian peak should correspond to the centre of the stationary fast ice vectors. All vectors that are within 5 standard deviations of the Gaussian peak are tagged as valid fast ice vectors. This is a minor modification to the method of Giles et al (2011), who used a simple threshold cut on the z-axis of the 2D histogram to define the fast ice vectors.Data format – one fully annotated (self-describing) netCDF file per image pair containing latitude/longitude coordinates of the stationary fast ice vectors.This technique and dataset complement a lower resolution but longer-term dataset (2000-2014) derived from satellite MODIS visible and thermal infrared data. (AAS_4116_Fraser_fastice_mawson_capedarnley).

本数据集涵盖南极东部丹利角冰间湖（Cape Darnley Polynya）周边海域2008年与2009年6—11月（冬春季）的高时空分辨率沿岸固定海冰（fast ice）分布地图。该地图通过对空间重叠的Envisat先进合成孔径雷达（Advanced Synthetic Aperture Radar, ASAR）图像对开展互相关分析生成，采用改进版IMCORR算法计算海冰运动矢量（详见Giles等，2011）。固定海冰与移动浮冰的区分依据为其处于静止状态。原始ASAR WSM数据（幅宽500千米）通过ENVI图像处理软件处理，生成像素尺寸为75米的地理参考图像。合成孔径雷达（SAR）数据的应用确保了观测不受云层覆盖或极夜干扰。图像对的时间间隔设置为2至21天。用于绘制固定海冰分布的IMCORR图像处理流程遵循Giles等（2011）的方法：参考块尺寸为32×32像素，搜索块尺寸为64×64像素。为避免大陆冰盖静止特征造成的匹配污染，研究中应用了陆地掩膜。网格间距设置为16×16像素，因此对图像进行了2倍过采样以获得更密集的结果集。研究采用聚类搜索技术结合z轴阈值技术变体（详见Giles等，2011），从IMCORR结果中筛选静止固定海冰运动矢量。首先对每个图像对的IMCORR结果应用聚类搜索技术，得到初始有效矢量集——该集合同时包含静止固定海冰矢量与非静止浮冰矢量。由于图像对存在配准误差，静止矢量未必以零偏移为中心，因此无法通过简单的零偏移窗口来区分固定海冰矢量。为筛选静止矢量，研究基于X-Y矢量位移构建二维直方图，并对该直方图拟合二维高斯分布。鉴于图像对时间间隔较大且搜索块尺寸较小，固定海冰矢量将占主导地位，因此高斯峰应对应静止固定海冰矢量的中心。所有位于高斯峰5倍标准差范围内的矢量将被标记为有效固定海冰矢量。这是对Giles等（2011）方法的小幅改进，后者仅通过对二维直方图的z轴进行简单阈值切割来定义固定海冰矢量。数据格式为每个图像对对应一个完整标注（自描述）的netCDF文件，其中包含静止固定海冰矢量的经纬度坐标。本技术与数据集可与另一项基于MODIS卫星可见光及热红外数据生成的低分辨率但长期（2000—2014）数据集形成互补。（数据集编号：AAS_4116_Fraser_fastice_mawson_capedarnley）