面向浮游植物类群遥感的HY-1C/D卫星数据应用初探

Remote sensing estimation of phytoplankton groups using Chinese ocean color satellite data

浮游植物是全球初级生产力的重要贡献者,浮游植物群落结构的变化影响着初级生产力,进而影响着海洋的物质循环与能量转换,因此具体量化分析浮游植物各群落结构的生物量对了解浮游植物群落结构变化,进而了解全球初级生产力极其重要。本文基于2016年与2018年4个渤海航次的实测遥感反射率数据和实测HPLC(High Performance Liquid Chromatography)浮游植物色素数据,通过CHEMTAX(CHEMical TAXonomy)方法将HPLC色素数据转化为相应藻种浓度数据,其中硅藻、隐藻、蓝藻与绿藻对总叶绿素a的占比较大。结合奇异值分解和多元线性回归方法,构建适用于中国近海硅藻、隐藻、蓝藻和绿藻浓度的反演模型;利用留一交叉验证法对模型进行验证,结果表明:隐藻、蓝藻和绿藻模型精度较高,决定系数R2均在0.70及以上,硅藻R2为0.44(p均小于0.001),硅藻、隐藻、蓝藻和绿藻浓度反演模型的中值误差ME各为44.81%、45.34%、51.20%和62.80%。随后,将模型应用于国产HY-1C/D卫星海洋水色扫描仪COCTS(China Ocean Color&Temperature Scanner)的遥感反射率日产品数据,获得渤海4个藻种浓度的空间分布特征,发现与实测浓度的空间分布特征分布一致。进一步分析COCTS与MODIS-Aqua、GOCI-Ⅱ的藻种浓度反演模型精度,发现基于COCTS波段的隐藻浓度反演模型精度高于基于MODIS-Aqua、GOCI-Ⅱ波段模型,硅藻、蓝藻和绿藻浓度反演模型精度和MODIS-Aqua相近且均高于GOCI-Ⅱ。在藻种浓度监测的示范性应用上,COCTS效果更好。综上所述,国产卫星HY-1C/D数据在藻种浓度监测方面具有强大的应用潜力。

Phytoplankton is a significant producer of global primary productivity and influences the ocean’s biological cycle and energy conversion. Understanding and detecting the phytoplankton biomass is important to grasp the variations in the marine environment.However, observing the changes of phytoplankton taxa remains a great challenge on spatial and temporal scales. Recent developments in ocean color sensors have enabled large-scale and long time-series remote sensing retrieval of phytoplankton biomass. HaiYang-1C and HaiYang-1D(HY-1C/D) satellites, as the main members of the Chinese ocean color satellite series, can provide ocean color products with a larger observation range, higher accuracy, and resolution, with great application potential.In this study, we collect in situ data, including the pigment concentration with the high-performance liquid chromatography method(HPLC) and remote sensing reflectance(Rrs), from four cruises in the Bohai Sea and the Yellow Sea from 2016 to 2018. Then, we obtain eight typical phytoplankton taxa concentrations through CHEMTAX(CHEMical TAXonomy) software based on these pigment data. We found the sum of the relative contributions of diatoms, cryptophytes, cyanobacteria, and chlorophytes to total chlorophyll a(TChla)accounted for a large proportion(79%). In addition, the spatial distribution of the CHEMTAX-calculated phytoplankton taxa showed a trend of higher nearshore concentration than offshore by spatial interpolation analysis.We used the singular value decomposition(SVD) method to construct a link between Rrs and phytoplankton concentrations. The matrix U obtained from SVD was used to build four models by multiple linear regression methods, to estimate four phytoplankton taxa concentrations. We carried out validation independently based on the measured and estimated concentrations, and the result showed relatively high consistent between diatoms, cryptophytes, cyanobacteria, and chlorophytes and the measured values(determination coefficients(R^(2)):0.44, 0.70, 0.70 and 0.71(p<0.001);median percent error(ME): 44.81%, 45.34%, 51.20% and 62.80%;Root Mean Squared Error(RMSE):0.23 mg/m^(3), 0.24 mg/m^(3), 0.11 mg/m^(3) and 0.06 mg/m^(3), respectively). The established model was further applied to China Ocean Color &Temperature Scanner(COCTS) Rrs data on the HY-1C/D L1A to demonstrate the spatial distribution of four major phytoplankton taxa in the Bohai Sea. The satellite results are consistent with previous studies that showed decreasing concentrations from nearshore to offshore.Finally, this study applies the same modeling approach(SVD) to MODIS and GOCI sensor bands. A comparison of model performance and satellite applications between the three sensors showed that the new model established by COCTS bands outperformed the GOCI-Ⅱmodel and was similar to the MODIS-Aqua model. Also, the satellite application of COCTS is superior to the other two sensors. Generally,this study can provide a methodological foundation for understanding the spatial-temporal evolution of the phytoplankton community in the Bohai Sea. Meanwhile, this study shows the great potential of HY-1C/D in models establishing and phytoplankton community monitoring.