相对反射深度的秦皇岛海域叶绿素浓度反演

海水中的叶绿素浓度是描述海洋初级生产力、获知浮游植物丰度及变化规律、评估环境质量、预报生态灾害的主要参数。国内外卫星遥感叶绿素产品的通用反演模型是利用不同波段上遥感反射光谱的强度比值来构建的OCx（x=2~6）算法，应用在一类水体中，全球尺度上的平均相对误差在35%左右。但对于固有光学特性复杂且具有较大区域差异性的二类水体，OCx算法误差较大甚至失效。现有研究成果表明，光谱的相对高度有利于特征波段信息提取及水色信号信噪比的提高。但基于相对高度构建反演模型，目前尚存在波段单一、应用面窄等问题。在我国近岸水体中，相对高度模型的构建方法及应用效果尚需进一步研究和验证。在对秦皇岛近岸海域的叶绿素浓度和表观光学参量进行原位测量的基础上，对高光谱数据进行了规范化处理，选取了特征波段并利用特征波段的相对反射深度构建了反演模型。模型反演值与实测值的相关系数为0.883 58，平均相对误差为28.33%；将模型与OCx等算法进行比较，平均相对误差均降低了27%～50%；模型验证估算值的平均相对误差为31.17%。在此基础上，对我国海洋卫星HY-1C水色水温扫描仪的多光谱数据及实测叶绿素浓度进行了相关分析，并基于443及520 nm处的相对反射深度建立了反演模型，模型估算值的平均相对误差比同期L2B产品降低了53.44%。结果表明，基于相对反射深度构建反演模型，可充分利用叶绿素特征波段信息、降低数据敏感性、提高水色要素的信噪比，进而大幅提高模型的反演精度及稳健性。对于水色要素的高光谱及多光谱反演模型构建、水体光学参量测量、卫星产品普及应用、初级生产力估算、生态环境监测、水动力过程研究等领域具有重要的科学意义及较强的应用价值。

Inversion Method of Chlorophyll Concentration Based on Relative Reflection Depths

Chlorophyll concentration in ocean waters is the main parameter for describing marine primary productivity, estimating phytoplankton abundance and variation, assessing environmental quality and forecasting ecological disasters. The general inversion model of chlorophyll products used by satellite remote sensing at home and abroad is the OCx (x=2~6) algorithms based on the intensity ratio of remote sensing reflection spectra in different bands. When applied to case-1 glasses of waters, the mean relative error on a global scale is about 35%. However, for case-2 waters with complex inherent optical properties and large regional differences, OCx algorithms have large errors or even fail. The previous research results show that the relative spectral height is beneficial to extracting the feature information and improving the signal-to-noise ratio of ocean color. However, the inversion model based on relative height still has problems, such as single band selection and a narrow application range. In China coastal, the construction method and application effect of the relative height model need to be further studied and verified. Based on in-situ measured chlorophyll concentration data and apparent optical parameters in Qinhuangdao coastal waters, after normalizing hyperspectral data and selecting characteristic bands, the inversion model has been constructed based on relative reflection depths of characteristic bands in this paper. The related coefficient between the inversion and the measured values is 0.883 58, and the mean relative error is 28.33%. Compared with the OCx algorithms, the average relative errors are reduced by more than 27%～50%. The model is verified, and the mean relative error is 31.17%. On this basis, correlation analysis was carried out on the multi-spectral data of HY-1C China Ocean Color & Temperature Scanner and the measured chlorophyll concentration, and the inversion model was established based on the relative reflection depths at 443 and 520 nm. The mean relative error of the model was reduced by 53.44% compared with that of the L2B product at the same time. The results show that the inversion model based on relative reflection depths can make full use of the information of chlorophyll characteristic bands, reduce the sensitivity to noise, and improve the signal-to-noise ratio of ocean color constituents, thus greatly improving the inversion accuracy and robustness of the model. This research has important scientific significance and substantial application value for constructing hyperspectral and multi-spectral inversion models of ocean color elements, measurement of water optical parameters, popularization and application of satellite products, estimation of primary productivity, ecological environment monitoring, hydrodynamic process research and other fields.