Landsat-8的舟山近岸海域总悬浮物遥感反演与时空变异研究

总悬浮物(TSM)是海洋水质和水环境评价的主要参数之一。舟山海域位于杭州湾边缘，泥沙含量多，总悬浮物长期处于高浓度状态，其分布与动态变化对近岸水质、水循环以及该区域的渔业、旅游业都有较大影响。Landsat-8卫星影像作为一种应用广泛、空间分辨率高、便于获取的光学信息，可以为舟山海域的TSM观测提供有力的技术手段。利用舟山近岸海域实测TSM吸收系数a_p(440)和水面光谱，开发了基于Landsat-8反演TSM的最优模型，发现基于近红外和蓝色波段比B5/B2的S-函数和线性函数模型反演精度较好，相对于线性模型，S-函数具有更强的鲁棒性，该模型形式为a_p=3.72/(0.009+e-5.249B5/B2)，克服了常用函数模型（如线性、对数、指数函数）应用于实际卫星影像时出现光谱幅度范围受限的困难。海岸带内陆复杂水体水色遥感的另一难点是大气校正，以往的研究往往只采用某种大气校正模型，但该模型不一定适合研究水域，从而使得校正结果并不一定是最优的。在本研究中，验证比较了FLAASH，6S和ACOLITE三种大气校正方法面向Landsat-8水色反演的校正结果，发现ACOLITE方法获取的光谱形状最准确，误差最小，特别在蓝色波段，明显好于FLAASH和6S方法。将最优模型应用于舟山近岸海域2013到2018年的10幅Landsat-8图像。实测数据和反演结果显示：舟山近岸海域的TSM吸收系数a_p(440)的变化范围在1.64～417.04 m-1，均值118.47 m-1，TSM吸收占水体总吸收的90%以上，该海域实测的水面光谱形状呈现典型的河口海岸带复杂浑浊水体的光谱特征，很多光谱曲线具有双峰特征，遥感反射率幅值较大，特别是红色、近红外波段，由于受河口高浓度TSM的影响，遥感反射率高于远海较清洁的海水；衢山、洋山、宁波等近岸区域的TSM浓度明显高于东极、嵊泗等远海区域，随河口羽化区呈现由高到低的梯度变化，在近岸区域分布复杂，外海区域分布较为均匀。在时间分布上具有冬季浓度远高于夏季的特点，其中12月最高，最大值为413.32 m-1，8月最低，最小值为3.69 m-1，5月、10月期间也有TSM的局部峰值。这些TSM时空变异特征表明舟山海域悬浮物的分布和变化一方面受地形、海流、潮汐、季风、台风等自然环境因素的影响，另一方面也受如海运、港口建设、海岛旅游等人类活动的较大影响。

Using Landsat-8 to Remotely Estimate and Observe Spatio-Temporal Variations of Total Suspended Matter in Zhoushan Coastal Regions

Total suspended matter (TSM) is one of the important parameters for ocean water quality and aquatic environment assessment. Zhoushan Islands is located on the edge of Hangzhou Bay, where sediments present very high concentration and TSM shows suspended status for long-term. The distribution and variation of TSM in Zhoushan Islands have great impacts on its coastal water quality, ferry, fishery and tourism. Because of its high spatial resolution and open access, Landsat-8 imagery has been widely used for ocean color studies and hence providing good spectral information for observing variations of TSM in Zhoushan coastal regions. This study used in-situ measured TSM absorption coefficient (a_p(440), m-1) and water surface spectra to develop the TSM remote sensing model based on Landsat-8 imagery, and results show that the S-function using blue and near-infrared bands of Landsat-8 performed better than the other functions. The S-function is with the form a_p=3.72/(0.009+e-5.249B5/B2). This function overcomes the shortcomings of previous inversion functions (e. g., linear, logarithmic, and exponential functions) which performed well for modeling datasets, but often failed when they were applied to real satellite images since spectral values in images are often much larger than those in the modeling dataset. Another problem of coastal and inland water color remote sensing is the atmospheric correction. The previous studies usually used a specific atmospheric correction method, but it might not return the best results. In this study, we tested and compared three types of atmospheric corrections models, FLAASH, 6S, and ACOLITE, and found that ACOLITE was better than the other two methods for ocean color remote sensing using Landsat-8, especially in blue band, where ACOLITE performed better than the FLAASH and 6S methods. The S-model was used to a series of Landsat-8 images covering the Zhoushan Islands from 2013 to 2018. The field measured and image-derived results show that TSM absorption coefficients in Zhoushan Islands were extremely high, ranging from 1.64 to 417.04 m-1, with mean 118.47 m-1, which account for more than 90% of total water absorption coefficients. The measured above-surface spectra demonstrated typical spectral signatures of complex turbid coastal water, with two peaks within green and red bands and relatively high reflectance within red and near-infrared bands. Due to the high concentrations of the riverine discharged TSM, remote sensing reflectance in estuarine and coastal regions were much higher than those in open sea. TSM concentrations illustrated a clear downward gradient from Hangzhou Bay to open sea, and TSM in coastal areas such as Qushan, Yangshan and Ningbo were much higher than those in open sea area such as Dongji and Shengsi. TSM distribution pattern in coastal regions were usually more complex than those in offshore regions. TSM concentrations in winter were usually much higher than those in summer, where the highest was found in December with value of 413.32 m-1, and the lowest was found in August with value of 3.69 m-1. There are also local peaks of TSM during May and October. Our results indicate that the TSM spatio-temporal variations in Zhoushan Islands are not only controlled by many natural environmental factors, such as currents, tides, typhoons, monsoons, but also aresignificantly influenced by many human activities, such as ferry, ocean shipping, harbor projects and island tourism.