Manually categorized initial training data for open-water–sea-ice–cloud discrimination

This data set contains labeled training data for supervised classification of open-water/thin-ice, sea-ice, and cloud pixels from MODIS thermal-infrared satellite data and is created from manual categorization of dimensional reduced and unsupervised clustered co-located Sentinel-1 SAR and MODIS MOD021KM/MYD021KM swaths. All data originates from the Brunt Ice Shelf area in the Antarctic Southeastern Weddell Sea [34degW to 18degW; 73degS to 77degS] resampled to an equi-rectangular grid [445 (rows) x 460 (columns)]. The data is organized as tab-delimited tables per Sentinel-1 reference swath with geolocation (lon/lat) and the compiled predictors for different MODIS swath combinations. This data can be used to retrace the classifier training as described in the reference publication [DOI: https://doi.org/10.5194/tc-15-1551-2021] or used as a basis to create your own classification scheme. Additional information can be found in the provided meta data file.本数据集包含用于监督分类的标注训练数据，可针对中分辨率成像光谱仪（Moderate Resolution Imaging Spectroradiometer，MODIS）热红外卫星数据中的开放式水体/薄冰、海冰及云像素开展分类。其构建方式为：对共配准的Sentinel-1合成孔径雷达（Synthetic Aperture Radar，SAR）与MODIS MOD021KM/MYD021KM条带数据进行降维和无监督聚类后，通过人工分类生成标注。所有数据采集自南极东南威德尔海的布伦特冰架（Brunt Ice Shelf）区域[西经34°至西经18°；南纬73°至南纬77°]，并被重采样至等矩形网格[445行 × 460列]。数据以每条Sentinel-1参考条带为单元，组织为制表符分隔的表格，包含地理定位信息（经度/纬度）以及针对不同MODIS条带组合的编译预测变量。本数据集可用于复现参考文献[DOI: "https://doi.org/10.5194/tc-15-1551-2021"]中所述的分类器训练流程，亦可作为构建自定义分类方案的基础。额外信息可参阅附带的元数据文件。