Concerns about phytoplankton bloom trends in global lakes
Concerns about phytoplankton bloom trends in global lakes
复制标题
DOI：
10.1038/s41586-021-03254-3
复制
发表时间：
2021-02
期刊：
影响因子：
64.8
通讯作者：
Lian Feng;Yanhui Dai;Xuejiao Hou;Yang Xu;Junguo Liu;C. Zheng
中科院分区：
文献类型：
--
作者：
Lian Feng;Yanhui Dai;Xuejiao Hou;Yang Xu;Junguo Liu;C. Zheng
卫星遥感已被广泛用于监测内陆和沿海环境的水质。利用Landsat 5专题制图仪（L5 TM）的卫星数据，Ho等人1显示，从1982年到2012年，全球71个大型湖泊中有68%的夏季水华高峰强度有所增加。然而，我们质疑他们发现的准确性，至少有两个原因：（1）卫星在单个近红外（NIR）波段的反射率不是衡量水华强度的可靠指标，（2）L5 TM的卫星观测不频繁，很难得出有统计意义的结论。Ho等人1认为，L5 TM估计的水华强度（BNIR）（参见Ho等人1中的等式2）基本上是近红外波段的反射率，代表近地表浮游植物生物量。然而，当检查从中国15个湖泊收集的光谱和叶绿素a（Chla;浮游植物生物量的关键指标2）数据集时，这一论点变得可疑3，这些湖泊具有不同的富营养化状态（Chla范围在1.5和222.6 mg m− 3之间;见扩展数据图1和补充信息）。Spyrakos等人4还利用世界各地的现场数据证明了光谱反射率和叶绿素浓度之间的复杂关系。从理论上讲，在近红外波段的信号可以归因于各种水成分，除了藻华，悬浮沉积物和水生植物的存在可能是两个最常见的扰动在内陆湖泊的贡献。Ho等人（1）试图用色调来掩盖富含沉积物的沃茨，但正如后面所详述的，我们的分析表明，参考文献（1）中定义的色调不能准确地代表水体的颜色。在沉积物丰富的沃茨，水华强度往往被大大高估。Ho等人研究的两个湖泊的实例（图1）表明，在相同的图像中，高浊度、低藻类像素的BNIR值高于藻类像素。对历史图像的检查（通过真彩色图像和光谱特征）表明，L5 TM观测在71个研究湖泊中的至少58个（82%）中捕获了沉积物羽流，由于其高BNIR，这些羽流可能被错误地标记为藻华（见扩展数据图2）。正如先前使用Ho等人1研究的湖泊和其他全球沿海/内陆沃茨数据进行的研究所支持的，浑浊沃茨的近红外反射率可以大幅增强（见扩展数据表1）。在内陆湖泊中，间歇性的气象（例如风和降水）和水文（例如河流流量）事件可以强烈影响沉积物浓度5，如以前在美国伊利湖6和奥基乔比湖7以及中国洪泽湖8（参考文献1中检查的三个湖泊）中的研究所示。因此，应仔细评估水体浊度对BNIR的影响。类似于高泥沙负荷，水生植被的生长可能导致高估水华的严重程度。具有高BNIR的像素-特别是植物生长的沃茨而不是水华区域-也在相同的湖泊中发现（见图1a，B），其中大量的沉水植物先前已被报道9。原因是藻华和沉水植被具有相似的高近红外反射率（见扩展数据图3）。此外，先前对全球不同地区和植物物种收集的数据集的研究也表明，由于沉水植被的存在，近红外反射率显着增加（见扩展数据表2）。事实上，一项文献检索显示，在...
Satellite remote sensing has been widely used to monitor the water quality of inland and coastal environments. Using satellite data from the Landsat 5 Thematic Mapper (L5TM), Ho et al. 1 showed an increase in peak summertime bloom intensity in 68% of the 71 large lakes worldwide from 1982 to 2012. However, we question the veracity of their finding for at least two reasons:(1) satellite-derived reflectance in a single near-infrared (NIR) band is not a reliable proxy for bloom strength, and (2) the infrequent satellite observations from L5TM make it difficult to draw statistically meaningful conclusions. Ho et al. 1 argued that the L5TM-estimated bloom intensity (BNIR)(see equation 2 in Ho et al. 1), which is basically the reflectance in the NIR band, represents near-surface phytoplankton biomass. However, this argument became questionable when examining the spectral and chlorophyll a (Chla; a key indicator for phytoplankton biomass 2) datasets collected from 15 lakes in China 3 with varying eutrophic status (Chla ranging between 1.5 and 222.6 mg m− 3; see Extended Data Fig. 1 and Supplementary Information). The complex relationship between spectral reflectance and Chlaconcentrations were also demonstrated by Spyrakos et al. 4 using in situ data from around the world. Theoretically, the signal in the NIR band can be attributed to various water constituents in addition to algal blooms, and the contributions from suspended sediments and the presence of aquatic plants could be two of the most common perturbations in inland lakes. Ho et al. 1 attempted to mask out sediment-rich waters with the use of hue but, as detailed later, our analysis suggests that the hue defined in ref. 1 does not accurately represent the colour of a water body. Bloom strength tends to be substantially overestimated in sediment-rich waters. Examples from two of the lakes studied in Ho et al. 1 (Fig. 1) show that the BNIR value of the high-turbidity, low-algae pixels was higher than that of the algae-present pixels within the same images. The examination of historical images (through both true-colour images and spectral features) shows that L5TM observations have captured sediment plumes in at least 58 (82%) of the 71 studied lakes, and these plumes could be incorrectly labelled as algal blooms owing to their high BNIR (see Extended Data Fig. 2). As supported by previous studies using data from both of the lakes studied in Ho et al. 1 and from other global coastal/inland waters, the NIR reflectance in turbid waters can be substantially enhanced (see Extended Data Table 1). In inland lakes, episodic meteorological (for example, wind and precipitation) and hydrological (for example, riverine discharge) events can strongly influence sediment concentrations 5, as exemplified by previous studies in Lake Erie 6 and Lake Okeechobee 7 in the USA and Hongze Lake 8 in China (three lakes examined in ref. 1). Therefore, the effect of water turbidity on BNIR should be evaluated carefully. Similar to high sediment loads, the growth of aquatic vegetation can lead to overestimation of bloom severity. Pixels with high BNIR—in particular, vegetated waters rather than bloom areas—were also found within the same lakes (see Fig. 1a, b), where massive submerged plants have previously been reported 9. The reason is that algal blooms and submerged vegetation share similarly high NIR reflectance (see Extended Data Fig. 3). Moreover, previous studies with datasets collected across various global regions and plant species also showed markedly increased NIR reflectance due to the presence of submerged vegetation (see Extended Data Table 2). Indeed, a literature search revealed that of the …