Department of Ecoscience, Aarhus University, Roskilde, Denmark.
Department of Ecoscience, Aarhus University, Roskilde, Denmark.
Sci Total Environ. 2023 Nov 1;897:165310. doi: 10.1016/j.scitotenv.2023.165310. Epub 2023 Jul 6.
In situ Chl-a data were used to perform empirical calibration and validation of Sentinel-3 level 2 product in Danish marine waters. Comparing in situ data with both same-day and ±5 days moving averaged Sentiel-3 Chl-a values yielded two similar positive correlations (p > 0.05) with r values of 0.56 and 0.53, respectively. However, as the moving averaged values resulted in significantly more available data than daily matchups (N = 392 vs. N = 1292) at a similar quality of correlation with similar model parameters (slope (1.53 and 1.7) and intercept (-0.28 and -0.33) respectively), which were not significantly different (p > 0.05), the further analyses were focused on ±5 days moving averaged values. A thorough comparison of seasonal and growing season averages (GSA) also showed a very good agreement, except for a few stations characterized by very shallow depth. Overestimation by the Sentinel-3 occurred in shallow coastal areas and was attributed to the interferences from benthic vegetation and high levels of Colored Dissolved Organic matter (CDOM) interfering with the Chl-a signals. Underestimation observed in the inner estuaries with shallow Chl-a rich waters, however, seen as a result of self-shading at high Chl-a concentrations, reducing effective absorption by phytoplankton. Besides the observed minor disagreements, there was no significant difference when the GSA values from in situ and Sentinel-3 were compared for all three water types (p > 0.05, N = 110). Analyzing Chl-a estimates along a depth gradient showed significant (p < 0.001) non-linear trends of declining concentrations from shallow to deeper waters for both in situ (explaining 15.2 % of the variance (N = 109)) and Sentinel-3 data (explaining 36.3 % of the variance (N = 110)), with higher variability in shallow waters. Furthermore, Sentinel-3 enabled full spatial coverage of all 102 monitored water bodies providing GSA data at much higher spatial and temporal resolutions for good ecological status (GES) assessment compared to only 61 through in situ sampling. This underlines the potential of Sentinel-3 for substantially extending the geographical coverage of monitoring and assessment. However, the systematic over- and underestimation of Chl-a in shallow nutrient rich inner estuaries through Sentinel-3 requires further attention to enable routine application of the Sentinel-3 level 2 standard product in the operational Chl-a monitoring in Danish coastal waters. We provide methodological recommendations on how to improve the Sentinel-3 products' representation of in situ Chl-a conditions. Continued frequent in situ sampling remains important for monitoring as these measurements provide essential data for empirical calibration and validation of satellite based estimates to reduce possible systematic bias.
利用原位 Chl-a 数据对丹麦海洋水域的 Sentinel-3 二级产品进行了经验校准和验证。将原位数据与同日及±5 天的移动平均 Sentinel-3 Chl-a 值进行比较，得到了两个相似的正相关（p>0.05），相关系数分别为 0.56 和 0.53。然而，由于移动平均值产生的可用数据量明显多于每日匹配（N=392 与 N=1292），并且具有相似的相关性质量和相似的模型参数（斜率（1.53 和 1.7）和截距（-0.28 和-0.33）），这些参数没有显著差异（p>0.05），因此进一步的分析集中在±5 天的移动平均值上。对季节性和生长季节平均值（GSA）的全面比较也显示出非常好的一致性，除了少数几个以极浅水深为特征的站点外。Sentinel-3 在浅海岸地区出现高估，这归因于底栖植被的干扰和高水平的有色溶解有机物（CDOM）对 Chl-a 信号的干扰。然而，在内陆河口地区，由于高 Chl-a 浓度导致自遮蔽，有效吸收减少，观察到低估。除了观察到的微小差异外，当比较所有三种水质类型的原位和 Sentinel-3 的 GSA 值时，没有显著差异（p>0.05，N=110）。沿着深度梯度分析 Chl-a 估计值表明，原位数据（解释了 15.2%的方差（N=109））和 Sentinel-3 数据（解释了 36.3%的方差（N=110））均存在显著（p<0.001）的非线性浓度下降趋势，从浅水区到深水区，浅水的变异性更高。此外，Sentinel-3 能够对所有 102 个监测水体进行全空间覆盖，提供更高时空分辨率的 GSA 数据，从而更有利于生态状况（GES）评估，而仅通过原位采样获得 61 个。这突显了 Sentinel-3 在大幅扩展监测和评估的地理覆盖范围方面的潜力。然而，Sentinel-3 对浅营养丰富的内陆河口的 Chl-a 存在系统的高估和低估，需要进一步关注，以实现 Sentinel-3 二级标准产品在丹麦沿海水域常规 Chl-a 监测中的常规应用。我们提供了关于如何改进 Sentinel-3 产品对原位 Chl-a 条件的代表性的方法建议。持续频繁的原位采样仍然很重要，因为这些测量提供了用于卫星估算的经验校准和验证的基本数据，以减少可能的系统偏差。
