HY-1C/D卫星CZI数据的岛礁浅海水深遥感无控反演能力评估

Retrieval and assessment of island shallow water depth without ground data from the HY-1C/D CZI multispectral imagery

岛礁浅海水深是海洋重要的基础资料。中国南海岛礁远离大陆,水下现场调查效率低、难度大,难以评估水下地形的长期变化,迫切需要大面积、高频次成像的卫星遥感数据。海洋一号C/D(HY-1C/D)卫星双星组网大大提高了覆盖频次,配置的海岸带成像仪CZI(Coastal Zone Imager)可为岛礁水下探测提供快速的业务化遥感数据服务。为充分发掘两颗卫星的岛礁水深反演能力,本文以中国南海永乐环礁为研究区域,以HY-1C/D CZI多光谱遥感影像为数据源,结合半分析模型和对数比值模型开展了不依赖实测数据的水深反演研究,并与GeoEye-1高分辨率卫星多光谱遥感数据反演结果进行对比。结果表明,HY-1C/D CZI多光谱数据在甘泉岛的水深反演结果与实测数据相比,在0—20 m水深范围内的平均绝对误差分别为1.60 m和1.85 m,平均相对误差分别为22.48%和26.23%;HY-1C/D CZI多光谱数据与基于GeoEye-1数据在甘泉岛的水深反演结果相比,平均绝对偏差分别为1.65 m和1.81 m,平均相对偏差分别为22.33%和23.83%。HY-1C/D CZI与GeoEye-1多光谱数据在永乐环礁各剖面变化趋势的交叉对比结果基本一致。具有高频次、宽覆盖特征的HY-1C/D CZI数据在全球岛礁浅海水深反演方面具有广泛的应用潜力。

Shallow water depth of islands and reefs is an important marine element. The islands and reefs in the South China Sea are located far from the mainland, which makes it difficult to assess long-term changes in underwater topography owing to the low efficiency and difficulty in field investigations. Satellite remote sensing imagery capable of large coverage and high frequency is urgently needed. The double satellite network of HY-1C/D greatly improves the coverage frequency. The Coastal Zone Imager(CZI) can provide fast operational remote sensing services for underwater detection of islands and reefs. To fully explore the depth detection capabilities of the two satellites,this study used Yongle Atoll as the research area, with HY-1C/D CZI multispectral remote sensing imagery as the data source. Combined semi-analytical and log-ratio models were used to perform water depth inversion independent of in situ data. The objective is to access the application potential of HY-1C/D CZI imagery for shallow water depth inversion of islands and reefs. This study combined a semi-analytical and logarithmic ratio model( called L-S model) based on satellite remote sensing imagery of HY-1C/D CZI, which includes four bands from visible to near-infrared. The strong linear relationship between water depth and the relevant spectral parameters of the logarithmic ratio model were used to globally restrict the semi-analytical model. After preprocessing the HY-1C/D CZI domestic multispectral imagery, which included geographic projection, geometric precision correction, calculation of the top of atmosphere reflectance, radiometric correction, sun glint correction, and atmospheric correction, a shallow water depth inversion experiment was carried out in Yongle Atoll, independent of in situ water depth or any other priori knowledge based on the L-S model. The water depth inversion results after tidal height correction using the OSU tidal prediction software in Yongle Atoll were compared with the in situ data and cross-compared with the inversion results based on GeoEye-1 remote sensing imagery. Compared with the in situ water depth, the mean absolute errors of HY-1C/D CZI were 1.60 m and 1.85 m, and the relative errors were 22.48% and 26.23%, respectively. The mean absolute error of the water depth inversion result of GeoEye-1 was 0.78 m and the relative error was 10.86%. Compared with the results of GeoEye-1, the mean average absolute deviation of HY-1C/D CZI were 1.65 m and 1.81 m, and the relative deviations were 22.33% and 23.83%, respectively, which were basically consistent in different sate llite sensors. Although the overall accuracy is lower than that of high-spatial-resolution satellite images, the mean absolute error of the inversion results of HY-1C/D CZI can be controlled within 2.0 m, with a high reference value. This solves the problem of the lack of in situ water depth data during shallow water depth inversion at a large scale. In addition, a cross-comparison of the inversion results between HY-1C and HY-1D showed that this method is robust when applied to different satellite sensors. This indicates that HY-1C/D CZI have the advantages of a short revisiting period and large imaging coverage, which can quickly and repeatedly obtain large-scale optical image data of the ocean and perform shallow water depth mapping of islands and reefs, thereby realizing high-frequency monitoring of underwater terrain changes. HY-1C/D CZI imagery and high-spatial-resolution satellite imagery complement each other and compensate for the shortcomings of field measurements. Therefore, based on the HY-1C/D CZI imagery, the water depth information can be retrieved in the range of 0—20 m stably and accurately, which has a wide range of application potential in the shallow water depth inversion of global islands and reefs.