The Yellow Sea Green Tides: Spatiotemporal Dynamics of Long-Distance Transport and Influencing Factors

1. Introduction

Green tides, caused by the extensive growth and accumulation of unattached green macroalgal, have been increasing in both severity and geographical extent globally, primarily driven by the eutrophication of marine environments [1,2]. The world’s largest green tide, driven by Ulva prolifera, have occurred annually in the Yellow Sea since 2007 [3,4,5]. These floating green tides significantly impact marine ecosystems by disturbing zooplankton communities, altering seawater chemistry, reducing phytoplankton biomass, and even modifying the carbonate system [6,7,8,9,10]. Additionally, the accumulation of algae on beaches and in coastal waters disrupts the balance of coastal ecosystems, leads to the death of cultured organisms, and affects tourism [2,10,11]. As a result, these events have caused significant economic losses. Between 2008 and 2015, the combined costs of mitigation efforts and aquaculture losses due to green tide blooms were estimated to be approximately 350 million U.S. dollars [2,12,13].

Unlike other green tides around the world, floating algal mats in the Yellow Sea do not remain in their original blooming region, but drifted to the open area driven by surface winds and associated currents [4,14,15,16,17]. Eventually, most of these algae wash ashore along the coasts of Shandong and Jiangsu provinces in China, and even the west coast of South Korea [4,16,18]. During their long-distance transport, the green macroalgae rapidly grow under favorable conditions of light, temperature and nutrient [4,19,20,21], forming large-scale green tides characterized by both extensive affected areas and high biomass production [4,21,22].

The entire process of long-distance movement and the scale of Ulva prolifera blooms show strong temporal and spatial variability, which is influenced by changing environmental conditions or human activities, such as wind and current patterns, light and nutrient availability, and the management of Porphyra yezoensis seaweed cultivation [15,19,20,23,24]. Understanding these variations and their underlying causes is crucial for assessing the impact of green tides on biogeochemical systems and for developing effective management strategies. Moreover, the spatiotemporal variation of Ulva prolifera has drawn significant public attention since the first sudden outbreak in 2007, due to its close ties with marine aquaculture and tourism industries.

Satellite tracking methods, which can capture and instantaneously record Earth’s surface information, have provided a comprehensive view of spatiotemporal dynamics of green tides [3,5,12,23,25]. In addition, various attempts, including field and laboratory research combined with ecological modeling, has been devoted to explain the observed patterns [19,20,23,26,27,28,29,30]. This review focuses on the current understanding of the unique characteristics of green tides in the Yellow Sea, particularly their long-distance movement and the interannual variability of bloom size. We also briefly discuss the underlying mechanisms driving these fluctuations.

2. Diversified Macroalgae Monitoring Realized by Remote Satellites

Multiple satellite data are used for macroalgae monitoring, with MODIS optical imagery being the most widely used due to its large swath width (2330 km) and frequent revisit capacity (morning and afternoon passes) [31]. However, its relatively coarse resolution (250m) may miss small macroalgal patches and lead to an overestimation of total macroalgal coverage [32,33]. This limitation also applies to green algal detection using COMS GOCI (500m) imagery. To address this issue, satellites with finer resolution, such as the Landsat series, Sentinel 1A/2A, HJ 1A/1B and GF series, are increasingly used for bloom monitoring [7,34]. In addition to optical imagery, microware data, with their all-day and all-weather observation capability, serve as an effective supplement for monitoring Ulva prolifera blooms.

In satellite optical images, macroalgae can be detected using vegetation index threshold method, including the Normalized Difference Vegetation Index (NDVI) [3], the Enhanced Vegetation Index (EVI) [35] and the Modified Normalized Difference Algae Index (NDAI) [36]. These indices are effective due to the spectral similarity between macroalgae and terrestrial vegetation [37]. However, both modified and simple vegetation indices are often less sensitive to variations in environmental and observational conditions, such as aerosols and changes in solar/viewing geometry [38]. To address these limitations, several specialized algae indices have been developed to improve detection accuracy. These include the Floating Algae Index (FAI) [38], the Scaled Algae Index (SAI) [39], the Virtual-Baseline Floating macroAlgae Height (VB-FAH) index [33], the Difference of Vegetation Index (DVI) [33] and the Floating Green Algae for GOCI (IGAG) index [35,40].

In contrast, macroalgal pixels in microwave imagery can also be extracted using threshold methods, leveraging the distinctive backscattering characteristics of the microwave bands [31,32]. It is important to note that macroalgae often appear in the form of stripes or irregularly shaped patches. As a result, the macroalgae pixels identified in satellite images are frequently mixed with surrounding non-algal features [15,31].

To address this challenge, alternative approaches, such as the linear pixel unmixing, image composition and machine learning, have been employed to more accurately estimate algae coverage by integrating both moderate- and coarse-resolution data [23,31,41,42]. Furthermore, a multi-sensor, multi-temporal and multi-spatial (3M) remote sensing strategy has been utilized to monitor macroalgal blooms [20]. Overall, advancements in methodologies and the application of comprehensive analytical tools have significantly improved the understanding and quantification of the spatial-temporal variations of green tides, with remote sensing technologies playing a crucial role in this progress.

3. Spatio-Temporal Variation in the Long-Distance Transport of Green Tides

3.1. Long-Distance Drifting

The most distinctive and well-known feature of the green tides in the Yellow Sea is their seasonal long-distance northward transport (Figure 1). Satellites observations have documented that the floating Ulva prolifera initially originates from the Porphyra yezoensis cultivation area in the Jiangsu Shoal in late April and early May [5,43,44]. After being released into the sea during facility recycling, Ulva prolifera drifts out of Subei Shoal and grows rapidly under favorable light, temperature and nutrient conditions [16,19,21]. Driven by the seasonal monsoon and the associated surface currents [14,45], large floating mats of Ulva prolifera aggregate into long, extensive patches, often distributed as irregularly shaped stripes or clusters [10,15,31]. These mats reach their peak size in mid-June, and then land along the southern coast of Shandong province [46]. Afterward, the Ulva prolifera green tides begin to decline, with floating mats only being sporadically observed in the coastal areas in early August [16].

This seasonal northward drift exhibits significant interannual variability in both the spatiotemporal distribution of Ulva prolifera and the duration of the bloom, as indicated by the sequence of Ulva prolifera locations identified from satellites images annually (Figure 1). For example, comparing the extensive Ulva prolifera blooms of 2016-2018 (Figure 1), the green tides in 2016 clearly had the widest extension and longest duration, in contrast to 2017 and 2018. When focusing on the comparison of the spatiotemporal distribution of Ulva prolifera since 2007, the quantified migration trajectory, obtained through calculation of the geometric or barycenter positions, clearly illustrate the interannual variation such as differences in offshore distance or discrepancies in migration along the north-south or east-west direction (Figure 2a) [46,47,48]. Although there are some discrepancies due to the different approaches used to obtain the spatiotemporal information of green macroalgae, such as the migration trajectory in 2016 (Figure 2b), the floating path consistently exhibits a trend of gradual northward drift. Furthermore, as shown in Table 1, the duration of the bloom during this northward movement varies significantly from year to year (Table 1).

In addition to the primary northward path, rafting green macroalgae has been consistently observed in the eastern Yellow Sea and even off the west coast of Korea (Figure 3), indicating a new eastward migration route for the green tides. Research has confirmed that this eastward migration occurred in 2008, 2009, 2011, 2012, 2015 and 2016 [18,33,35,40,50,51]. Less than 4% of the green tide areas in the western Yellow Sea (except for 2011 with 44%) was transported to the eastern Yellow Sea [18]. Along this eastward transport, the area of green tide masses significantly decreased due to limited reproduction and sinking, which may be attributed to the relatively lower nutrient levels in the middle Yellow Sea [51]. Compared to the main northward drift, the eastward drift of the green algae is characterized by a lower frequency of occurrence and smaller biomass. However, it could lead to problems similar to those experienced in China if the green tide eventually reaches the coast of Korean Peninsula. Little is known about the abnormal eastward transport, including bloom formation and spatiotemporal characteristics of this migration.

3.2. Final Landing and Decomposing Process

The end of the long-distance drift is marked by portions of the floating green algae accumulating along the southern coast of Shandong Province, while the remaining algae in the water sink to the seabed and decompose [53,54]. The washing ashore of Ulva prolifera caused destructive damage to coastal aquaculture, incurred substantial costs for clearing macroalgal biomass, and resulted in indirect negative effects on the tourism industry and environmental problems [16,30,55]. Oxygen depletion, along with the release of sulfide and ammonia during the decomposition of Ulva prolifera, is responsible for the mortality of abalone and sea cucumbers [56]. Additionally, the nutrient release during the decay of green tides is suspected to promote blooms of toxic microalgae, commonly known as “red tides” [57,58]. Thus, the final landing and decomposition processes are closely linked to human activities and livelihoods. Unfortunately, knowledge regarding the spatiotemporal variation of these processes remains scarce.

Current studies indicate that the landing sequence of green tides is highly variable. An et al. [59] documented that the order in which macroalgae landed on the coastal cities of Shandong Peninsula can be roughly divided into two patterns. In one pattern, the macroalgae first landed in Rizhao, followed by Qingdao, Rushan, and Haiyang. In the other pattern, the order was reversed. The largest amount of seaweed biomass has typically been washed ashore along the coasts of Qingdao. Since 2015, the coasts of Lianyungang and Rizhao have been severely affected by unprecedent large stranding biomass [17]. Despite this, other landing variables, such as the duration of the landing and landing volumes biomass, can also vary significantly, however, little is known about these aspects so far.

The final decomposition process of Ulva prolifera typically concludes in August (Table 1). According to the distribution pattern of 28-isofucosterol content in August 2015, the potential settlement region was identified in the sea area southeast of the Shandong Peninsula, around 122.66°E, 36.00°N [54]. Based on spatial distribution of Ulva prolifera during dissipation phase, An et al. [59] expanded the potential settlement region to include the coastal waters of the Shandong Peninsula and the central waters of the South Yellow Sea (33°31’N ~ 35°22’ N, 120°14 ‘E ~ 122°8’ E). This study revealed significant yearly variations in the final settlement regions: offshore settlements were located in the central waters of the South Yellow Sea in 2007, 2010, 2014, 2015, and 2019, while inshore settlements were found in the coastal waters of the Shandong Peninsula in other years. Similarly, based on the spatial distribution of green tides over their dissipation periods, Li et al. [60] suggested that the centers of dissipation of green tides over the years have been located along the coast of the Shandong Peninsula, with a radius of about 80 km, centered around Qingdao. Utilizing remote sensing data with higher spatial resolutions combined with C/N values, Zhang et al. [61] identified the primary decomposition regions are located in the coastal zone south of the Shandong Peninsula (35.2−37°N, 120.3−123 °E). However, without laboratory measurements to verifies findings, the reliability of the satellite-estimated results remains uncertain.

4. Interannual Variability in the Size of Green Tide Bloom

The bloom size, typically expressed by the annual maximum daily macroalgal biomass, can be intuitively displayed through satellite observations. As shown in Figure 4, the spatiotemporal characteristic of green tides, including their geographical distribution and annual extension patterns, are clearly depicted. In addition to these intuitive and visual representations, another common approach involves quantifying the macroalgal coverage or biomass.

4.1. Macroalgae Coverage

Two common methods are used to quantify the bloom size of Ulva prolifera: algae-mixed pixel coverage, which represents a collection of pixels corresponding to Ulva prolifera identified directly from remote sensing images under the pure pixel assumption [10,20], and pure algae coverage, which is derived by correcting for the mixed pixel effect [12,13,31,34]. Using proxies such as the maximum daily Ulva prolifera algae-mixed pixel coverage or pure algae coverage, several studies have been conducted to explore the interannual variations in bloom size [10,23,31,34,63].

As summarized in Figure 5, the bloom size of green tides exhibited significant interannual changes, with notable discrepancies observed across years. When comparing the yearly macroalgae coverage expressed in the two areal forms mentioned above, the algae-mixed pixel coverage was found to be 1.5 to 20.0 times higher than the pure algae coverage, likely due to the mixed pixel effect [31]. After processing the pixel un-mixing, the discrepancies were reduced to no more than 2.0 times for the macroalgae coverage obtained by Cui et al., Qi et al. and Hu et al. from 2008 to 2012 [23,31,34]. However, relatively higher differences, ranging from 2.5 to 3.6 times, were observed in the pure algae coverage from 2013 to 2018. Cui et al. attributed these discrepancies to differences in methodology, such as variations in the indices used, how algae-containing pixels were treated, and the thresholds applied for distinguishing algae from surrounding water during pixel un-mixing [31]. The extraction area obtained from relatively high-resolution images closely approximated the true value [64]. Thus, even without pixel unmixing, the extraction areas derived from high-resolution images exhibited a high degree of consistency between MDB [65] and Zhan et al. [66] since 2015.

In addition to the most striking numerical difference, the yearly fluctuations also exhibit considerable variation (Figure 5). For instance, the peak year, either before or after 2012 (which had the lowest algae coverage), varied significantly. Before 2012, studies by Cui et al. [31] and Zhang et al. [10] identified a sudden burst in 2009, while other studies reported the peak year as 2008. This inconsistency is also observed in other researches (Figure 5). After 2012, according to MDB [65] and Zhan et al. [66], the maximum daily coverage was in 2021. The distinct methodological differences and inconsistent thresholds, as mentioned above, are certainly responsible for this inconsistency [67]. Besides, the daily maximum obtained from cloud-free day may be several days or weeks away from the real maximum day, the differences in the identified maximal bloom day across studies may be the main factor contributing to this inconsistency (Table 2) [67]. To address this issue, Qi et al. [41] applied image composition to remove clouds and other artifacts to estimate monthly mean biomass and bloom size.

4.2. Macroalgae Biomass

Total biomass, compared to the coverage area, provides a more reliable indicator of the severity of green tide blooms. However, while extensive research has focused on the macroalgae coverage, studies on biomass estimation remain relatively scarce [12,13,34]. In practice, the biomass of green tides is mostly derived by converting pure algae coverage using specific conversion equations. Thus, the interannual patterns (Figure 6) of macro-algal biomass closely resemble those of macro-algal coverage. Similar to coverage, significant discrepancies exist in these biomass estimates. For instance, the maximum biomass for 2016, estimated by Hu et al. [34] at around 2321±154 kilotons, is nearly twice the biomass estimate (approximately 1190 kilotons) reported by Xiao et al. [13]. Meanwhile, there is also clear increasing trend in the quantified macro-algal biomass, with R2 value ranging from 0.28 to 0.58 (Figure 6).

Significantly, all the algal biomass estimates made so far have neglected the influence of floating algae thickness on estimation accuracy. Lu et al. indicated that the thickness effects contribute approximately 36% to the uncertainty in total biomass estimation, while around 43% of the overall uncertainty is attributed to a few pixels with high MODIS-derived FAI values (FAI >0.05) [69].

5. Factors Leading to the Variations

5.1. Abiotic Factors

Nutrient supply is the most powerful driver of fast growth of marine macroalgae [70]. Coastal eutrophication along the Jiangsu shoal, as indicated by the area-weighted nutrient pollution (AWCPI-NP) index, increased by 45% from 2007-2012 compared to that of 2001-2006. This rise was primarily driven by nutrient loading from river runoff, excessive agricultural fertilization, and intensive mariculture, which have been identified as critical factors contributing to green tides blooms [10,24,71,72]. The increased concentrations of nutrients species, particularly dissolved inorganic nitrogen (DIN), have favored the large extent and prolonged duration of green tides, thereby determining their seasonal pattern [22,72,73,74]. As documented by Li et al. [75], the biomass of Ulva prolifera shows a positive correlation with DIN concentrations (Spearman correlations, r = 0.84, P<0.001) from late March to April. Assuming the Sheyang river is a vital nutrients source for the Yellow Sea, Xiao et al. inferred the river might provide a key nutritional foundation supporting the outbreak of Ulva prolifera [13]. This conclusion was drawn from the significant positive correlation (r = 0.52, P < 0.01) between the annual maximum biomass of Ulva prolifera and the total nutrient load from the Sheyang River. However, rather than river discharge factors, nutrient inputs from human activities contribute more significantly to the interannual variation of nutrients along the Jiangsu coast [72,76]. Therefore, the correlation found by Xiao et al. [13] is insufficient to explain the interannual variation. To date, the role of nutrients in these interannual variation remains unclear due to the lack of longtime nutrient data [23].

Temperature has a significant impact on Ulva germination and vegetative growth [77]. Under nutrient-rich conditions, temperature is the key abiotic factor that influences biomass accumulation and species succession during the initial phase [27,29,78]. The initial biomass of Ulva prolifera showed a strong correlation with SST in March from 2011 to 2016, whereas no significant correlation was found between climatological variables in April [19]. From May to July, water temperature ranges from 15℃ to 25℃, creating suitable conditions for the growth of green macroalgal [58], among which the growth rate of Ulva prolifera was fastest when temperatures ranged from 17℃ to 22℃ [66]. But, extreme high temperature, such as the prolonged heatwave (＞30℃) that lasted for more than a month in 2017, greatly inhibited algal growth [18] and slowed the expansion of the algal blooms [63]. However, changes in SST in the Yellow Sea was not significantly correlated with the interannual variability of the floating Ulva prolifera biomass [23,46].

Besides nutrients and temperature, several abiotic factors, such as salinity [29,79], light intensity [80,81] and pH [73,75], also influence the productivity of Ulva prolifera. However, none of these factors alone can fully explain the spatial-temporal variability in the growth of Ulva prolifera [29]. Ulva prolifera exhibits a strong capacity to adapt to a wide range of environmental conditions [11]. Therefore, the variation in Ulva prolifera species may be influenced by a range of environmental factors acting in a complex and non-linear manner. Moreover, as reported by Jin et al. [19], different abiotic factors played different roles during different growth phases of Ulva prolifera. Thus, comprehensive and multi-disciplinary field studies are essential for understanding the annual variations.

5.2. The Rapid Expansion and Cultivation of Porphyra Mariculture in Jiangsu Province

Ulva prolifera in the Yellow Sea is believed to have originated from the raft frames used for Porphyra cultivation along the coast of Jiangsu Province [13,33,82]. Annually, approximately 6500 tons of Ulva prolifera detach from the Porphyra rafts, with 62% floating on the surface water, ultimately leading to the formation of massive green tide blooms [16]. The Porphyra aquaculture area showed a sharp increase since 2007, with rapid expansion occurring in the Sansha regions on offshore tidal flats located approximately 30-60 km from the coastline. In these areas, attached Ulva prolifera, once detached, can quickly drift into the Northern Yellow Sea [20,83]. Statistically, the maximum daily Ulva prolifera shows a relatively high coefficient (r = 0.63) with the Porphyra cultivation area in the northern part of the Jiangsu Shoal [20], indicating its significant role in the interannual variation of Ulva prolifera.

In addition to the Porphyra cultivation area, Xing et al. suggested that the timing of Porphyra facility recycling [20], which is related to early macroalgal biomass, may also be a key factor influencing the interannual biomass of Ulva prolifera. Shao et al. [84] documented that geographical features and farming patterns of Porphyra in Subei Shoal triggered the formation of massive floating Ulva prolifera. However, without statistically meaningful evidence, the extent to which it contributes to interannual variation remains unclear.

5.3. Monsoons and Ocean Currents

The southeastward monsoons prevail in the Yellow Sea from May to July, driving the movement of surface ocean particles and thus facilitating the northward migration of green tides [14,85]. Thus, the interannual variation in the local monsoon wind field, which is controlled by or lags behind the climatological variation of atmospheric-oceanic system [85], is one of the key factors influencing the migration direction and distribution patterns of green tides [47]. In addition, the wind-induced Subei Coastal Current, residual tidal cycles, and tidal movements also influence the movement of green tides [86]. Therefore, monsoons and wind-driven current systems help explain the interannual variation in the migration trajectories of macroalgae.

6. Concluding Remarks and Research Perspectives

Green tides have consecutively appeared every summer in the Yellow Sea for the past 16 years. Extensive studies have significantly enhanced our understanding of their spatiotemporal variations, the interannual variability in bloom size, and possible factors driving these fluctuations. However, many questions still remain unanswered.

1) There are still significant gaps in our understanding of the long-distance movement of Ulva prolifera. Satellites monitoring of macroalgal has identified new eastward drifts of green tides in 2008, 2009, 2011, 2015 and 2016. However, more detailed information on these eastward drifts, including their spatial-temporal dynamics, underlying causes, and migration mechanisms, remains unclear. Even for the more common northward movement, there is a lack of a reliable and effective quantitative method to clearly depict the interannual variations in this movement. Additionally, important knowledge gaps persist regarding the final landing and decomposition processes of Ulva prolifera.

2) The quantified Ulva prolifera areal coverage and corresponding biomass estimates vary significantly. The macroalgae are typically distributed in patches of different sizes or in banded stripes with various shapes [15,25,31]. As a result, the pattern of macroalgal distribution, and thus the estimated coverage area derived from images with different resolutions, can differ substantially. This raises the question: which data is reliable? A simultaneous field investigation may be necessary to explore a more reliable and universal method for estimating algae coverage or biomass. Moreover, there is an urgent need for robust automatic macroalgal detection systems that are less influenced by subjective factors. Furthermore, ongoing research into the spatial distribution of the thickness of floating Ulva prolifera will enhance the accuracy of biomass estimates in the Yellow Sea.

3) The dynamic factors outlined above, and their interactions, are responsible for the interannual variability in bloom size, the sudden outbreak in 2007, and the sharp decline in 2017. However, while there is compelling evidence that seaweed aquaculture influences the scale of green tides, other mechanisms driving these variations need to be further explored to develop effective control strategies. Therefore, comprehensive, multidisciplinary field studies, laboratory research, and ecological modeling are essential for a deeper understanding of Ulva biology and physiology, in order to clarify the observed patterns.