A New Algorithm Based on the Phytoplankton Absorption Coefficient for Red Tide Monitoring in the East China Sea via a Geostationary Ocean Color Imager (GOCI)

Abstract

Rapid and accurate dynamic monitoring and quantitative analysis of red tide disasters are of significant practical importance to national economic development. Remote sensing technology is an effective means for monitoring red tides. This paper utilizes GOCI satellite data and employs a quasi-analytical algorithm (QAA) to retrieve the spectral curves of phytoplankton absorption coefficients. On the basis of a detailed analysis of the differences in the spectral curves of the phytoplankton absorption coefficients between red tide and non-red tide waters, we establish a red tide identification algorithm for the East China Sea on the basis of phytoplankton absorption coefficients. The algorithm is applied to multiple red tide events in the East China Sea. The results indicate that this algorithm can effectively determine the occurrence locations of red tides and extract relevant information about them.

1. Introduction

A red tide is an ecological anomaly characterized by the explosive proliferation or accumulation of certain phytoplankton, protozoa, or bacteria in seawater under specific conditions, leading to discoloration of the water [1]. This phenomenon has severe impacts on coastal ecosystems, fishery resource production, and public health [2,3,4]. Research on red tides has garnered significant attention globally across relevant fields [5,6], becoming one of the important areas of marine science research. The mechanisms causing red tide are very complex, and monitoring red tide disasters and their associated marine environmental conditions forms the foundation for studying red tide events.

In recent years, with the development of China’s social economy and population growth, the degree of eutrophication has become increasingly severe, leading to frequent occurrences of red tides and escalating disaster losses [3,7,8,9]. The ability to quickly and accurately conduct dynamic monitoring and quantitative analysis of red tide disasters holds significant practical importance for national economic development [3,10,11,12].

The outbreaks of red tides are characterized by randomness, suddenness, and short duration, necessitating monitoring methods that can respond rapidly. Remote sensing technology, with its ability for large-area, synchronous, and rapid monitoring, is an effective means for red tide detection [13,14,15,16,17,18]. Current commonly used remote sensing algorithms for determining red tide water color focus on the overall performance of water spectral data. However, the spectral characteristics exhibited by red tide waters result from the combined effects of various factors, including phytoplankton, colored dissolved organic matter, suspended particles, and even bottom substrates [19,20,21], which limits the algorithms to local biochemical, physical, and hydrodynamic conditions at the time of observation. This results in significant regional applicability limitations and weaker transferability [22,23].

The absorption characteristics of phytoplankton are key components of the optical properties of aquatic environments and are closely associated with the composition and relative proportions pigments in phytoplankton cells [24]. The absorption characteristics of phytoplankton for specific light wavelengths (represented by absorption coefficients) form the fundamental basis for estimating phytoplankton biomass, primary productivity, and other water environmental variables via remote sensing technology [25,26,27,28]. The absorption spectrum of phytoplankton reflects information on light absorption resulting from different pigment compositions and can serve as a characteristic marker to distinguish different algal species and estimate phytoplankton community composition [29].

The GOCI is a water color sensor onboard Korea’s first geostationary meteorological satellite. Its band configuration is similar to that of the Sea-viewing Wide Field-of-view Sensor (SeaWiFS) and Moderate-resolution Imaging Spectroradiometer (MODIS), featuring six visible light bands and two near-infrared bands with a spatial resolution of 500 m. It provides eight repeated measurements per day [30]. Compared with polar-orbiting ocean color satellites, GOCI offers lower data error rates, fewer missing data points, and higher temporal resolution [31], indicating that it has significant advantages in red tide monitoring.

Therefore, this study utilized GOCI satellite data to establish a red tide identification algorithm for the East China Sea on the basis of phytoplankton absorption coefficients. This was achieved through a detailed analysis of the differences in the phytoplankton absorption coefficient spectral curves between red tide waters and non-red tide waters. The algorithm was applied to multiple red tide events in the East China Sea. The research results will have practical significance for establishing automated monitoring systems for red tides.

2. Materials and Methods

2.1. Materials

The data used in this study mainly consisted of two parts: measured data and remote sensing data. The red tide events were selected primarily from recent marine disaster bulletins published by the Ministry of Natural Resources of China. The red tide events were compared with remote sensing data, and only those events for which corresponding remote sensing images were available were selected. The remote sensing data used in this study were GOCI Level 2 data. The data from April 2011 to March 2021 were downloaded from the KOSC water color website (https://kosc.kiost.ac.kr/, accessed on 20 July 2022). The data used in this study were remote sensing reflectance data and chlorophyll data, with land and cloud masking applied to obtain daily datasets.

2.2. Methods

2.2.1. Phytoplankton Absorption Coefficient Inversion

This study employed the QAA algorithm to invert inherent optical properties. The QAA algorithm divides the inversion of inherent optical properties into two steps: the first step involves inverting the total absorption coefficient and backscattering coefficient at a reference wavelength from remote sensing reflectance, which is then extended to other wavelengths. This step does not involve spectral models for the absorption coefficients of phytoplankton and yellow substances. The second step decomposes the total absorption coefficient obtained in the first step into phytoplankton absorption coefficients and absorption coefficients for colored dissolved organic matter and particulate organic matter [32].

To validate the accuracy of the inversion results, Figure 1 presents the inverted phytoplankton absorption coefficient spectral curves. As shown in the figure, the inverted spectral curves exhibit good consistency in shape with the measured spectral curves, indicating that the results obtained via the QAA algorithm are reliable.

2.2.2. Red Tide Extraction Algorithm

(1) Analysis of Phytoplankton Absorption Coefficient Spectral Characteristics

Figure 2 shows the differences in the phytoplankton absorption coefficient spectral curves between red tide and non-red tide waters measured near the Yangtze River estuary in the spring of 2011. From the figure, it is evident that there are significant distinctions between the spectral curves: the phytoplankton absorption coefficient spectral curve for red tide waters is notably higher than that for non-red tide waters. The red tide waters exhibit a distinct absorption peak between 440 and 480 nm, whereas the absorption peak in non-red tide waters is much more subdued in this range. Additionally, the absorption peak at approximately 675 nm in red tide waters is significantly higher than that in non-red tide waters. Therefore, it is feasible to extract red tide information on the basis of the peak in the phytoplankton absorption coefficient spectral curves of red tide waters.

(2) Research on Red Tide Remote Sensing Algorithm

To analyze the differences in the phytoplankton absorption coefficient spectral curves between red tide and non-red tide waters, spectral curves were extracted for red tide locations, nearshore non-red tide locations, and offshore non-red tide locations from historical red tide events between 2011 and 2020. A total of 469 spectral curves from red tide waters, 390 spectral curves from nearshore non-red tide waters, and 453 spectral curves from offshore non-red tide waters were extracted. The phytoplankton absorption coefficient spectral curves for each red tide event exhibited similar characteristics. Figure 3 presents the extraction results of the phytoplankton absorption coefficient spectral curves for the red tide event on 18 May 2011.

From the Figure 3, it is clear that there are significant differences between the spectral curves of the phytoplankton absorption coefficients in red tide and non-red tide waters: the spectral curve for red tide waters shows distinct peaks between 440 and 490 nm and another peak near 680 nm. In contrast, the offshore non-red tide waters present lower values between 410 and 550 nm, with a peak at approximately 660 nm but lower values near 680 nm. The spectral curve for nearshore non-red tide waters resembles that of red tide waters in the blue–green light range, both showing high values between 440 and 490 nm; however, the peak in the absorption coefficient spectral curve for non-red tide waters is much flatter than that of red tide waters, and there is no peak near 680 nm.

On the basis of the aforementioned spectral information characteristics, this study attempts to extract red tide information via the operators a_ph660-a_ph555, a_ph660/a_ph555, a_ph680-a_ph660, a_ph680/a_ph660, a_ph490/a_ph660, and a_ph490-a_ph660. The research findings indicate that using a_ph660-a_ph555 can roughly distinguish between red tide waters and offshore non-red tide waters, whereas a_ph680-a_ph660 can approximately differentiate red tide waters from nearshore non-red tide waters.

Figure 4 displays the scatter distribution of a_ph660-a_ph555. The figure clearly shows that the values for offshore non-red tide waters are mostly above 0. In contrast, the values for red tide waters are all less than 0. Therefore, the criterion a_ph660-a_ph555 < 0 effectively distinguishes between red tide waters and offshore non-red tide waters.

Figure 5 shows the scatter distribution of a_ph660/a_ph555. The figure clearly shows that the value of red tide clusters around a relatively narrow range near 0, and the value of the non-red tide onshore is mainly concentrated near and below 0, with a relatively stable distribution. In contrast, the non-red tide offshore data show a wider spread, with values both above and below 0. Figure 6 shows the scatter distribution of a_ph680-a_ph660. The figure shows that all values for red tide waters are above 0, whereas those for nearshore non-red tide waters are all below 0. All offshore non-red tide waters also have values less than 0. Therefore, the criterion a_ph680-a_ph660 effectively distinguishes between red tide waters and non-red tide waters.

Figure 7 shows the scatter distribution of a_ph680/a_ph660. The non-red tide data points are predominantly clustered in the negative range of the a_ph680/a_ph660 axis. The non-red tide onshore data are mainly concentrated near 1, with some values slightly above and below, indicating a relatively narrow spread. In contrast, the red tide offshore data are more dispersed, with values spanning from negative to positive regions.

Figure 8 shows the scatter distribution of a_ph490/a_ph660. The red tide data points are scattered across a wide range of the y axis, with values predominantly in the positive region and some extending up to approximately 100. The non-red tide onshore data show a more concentrated distribution, mostly approximately 0 with some negative values reaching approximately −80. The non-red tide offshore data cluster around a relatively narrow range near 0.

Figure 9 shows the scatter distribution of a_ph490-a_ph660. The red tide data points are clustered around the value of 0.3, with some minor fluctuations above and below this value. The non-red tide onshore data show a more concentrated distribution near 0.3, with a narrow spread. In contrast, the non-red tide offshore data are more dispersed, with values ranging from approximately −3.0 to 0.0, and they are predominantly located below 0.

Consequently, this study uses the criteria a_ph660-a_ph555 < 0 and a_ph680-a_ph660 > 0 to identify red tides. Figure 10 shows the specific process of red tide identification. This flowchart provides a systematic approach for the identification of red tide events on the basis of the spectral characteristics of phytoplankton absorption coefficients derived from GOCI data. It commences with the input of the GOCI remote sensing reflectance of each band(R_rs(λ)). These data are then processed through the QAA to obtain the GOCI phytoplankton absorption coefficient(a_ph(λ)). Two key absorption coefficient differences are subsequently evaluated: a_ph660-a_ph555 and a_ph680-a_ph660. If a_ph660-a_ph555 < 0, and a_ph680-a_ph660 > 0, the pixel is red tide; otherwise, the pixel is non-red tide.

3. Results

The red tide extraction algorithm proposed in this paper is used to extract red tide information from historical red tide events in the East China Sea. Figure 11 shows the geographical location of the study area.

Figure 12a,c,e show the red tide extraction results at 10:00, 11:00, and 12:00, respectively, on 16 August 2016. To analyze the effectiveness of the extraction results, GOCI Level-2 remote sensing chlorophyll products from the same time were obtained. Figure 12b,d,f show the chlorophyll products at 10:00, 11:00, and 12:00, respectively, on 16 August 2016. The figures show that, except for the nearshore areas, the red tide extraction results correspond well with high chlorophyll contents. When a red tide occurs in the East China Sea, the concentration of phytoplankton increases rapidly, and the chlorophyll concentration increases, which can be easily observed through satellite images. However, the chlorophyll concentration retrieved by satellites is likely to be overestimated because of incorrect atmospheric correction, interference from colored dissolved and particulate organic matter in the nearshore area, and the influence of the shallow sea bottom [33]. Therefore, the results of this paper are credible.

Table 1. Statistics of red tide information in the East China Sea from 2011 to 2020.

Date	Location	Dominant Species	Area in the Bulletin (km2)	Identified Area (km2)	Absolute Error (km2)	Composite Relative Error (%)
13 May 2011–4 June 2011	Shiping Cangnan, Wenzhou, Zhejiang Province	Prorocentrum donghaiensis (PD)	200	310.7826	110.7826	35.65
26 July 2011–7 August 2011	South of Crocodile Island, the middle of Jimei Bridge, and the Wuyuan Bay Bridge, Xiamen	Skeletonema costatum (SC)	105	187.1204	82.1204	43.89
24 May 2012–3 June 2012	Dongtou Island, Wenzhou	Karenia mikimotoi (KM), PD	40	33.597	6.403	16.01
3–7 June 2012	Sheng Si, Zhou Shan	KM, PD	240	254.14	14.14	5.56
18 May 2012–7 June 2012	Xiapu and Fuding, Sansha Bay Ningde	KM	130	112.0516	17.9484	13.81
27 May–8 June 2012	Huangqi Lianjiang, Fuzhou	KM	40	60.75852	20.75852	34.17
30 May 2012–8 June 2012	Donghan Fuqing, Fuzhou	KM	6	9.93257	3.93257	39.59
5–8 June 2012	Jibi to Xinwo in Bili Township, Luoyuan, Fuzhou	KM	10	4.090333	5.909667	59.10
26 May 2012–7 June 2012	Pingtan waters	KM	80	33.20033	46.79967	58.50
30 May 2012–3 June 2012	East of Putian, Kengkou, Stone City, Yanyu, Meizhou Island	KM	32	31.17833	0.82167	2.57
30 May 2012–2 June 2012	Xiaoyu, Huian, Quanzhou	KM	7	4.371415	2.628585	37.55
13–29 May 2013	Cangnan, Wenzhou	PD	450	238.7144	211.2856	46.95
18 May–2 June 2013	Kanmen, Yuhuan, Taizhou	PD	120	228.9723	108.9723	47.59
20–24 May 2013	Southeast of Jiushan Island, Ningbo	PD	140	160.9563	20.9563	13.02
3–9 June 2013	Dongfu Island, Zhoushan	PD	100	129.501	29.501	22.78
21 May 2014–5 June 2014	Shengsi, Zhoushan	PD	170	76.852	93.148	54.79
21 May 2014–3 June 2014	Putuo, Zhoushan	PD	300	41.1944	258.8056	86.27
19 May 2014–11 June 2014	Cangnan, Wenzhou	PD	320	230.5109	89.4891	27.97
26 April 2015–3 May 2015	Yushan Islands	PD	200	92.748	107.252	53.63
12–21 June 2015	Nanji, Wenzhou	Gonyaulax polygramma (GP)	390	384.043	5.957	1.53
26 May 2015–2 June 2015	Guzhen, Xiapu	KM, PD	100	5.752	94.248	94.25
17–20 May 2016	East of the Yangtze River Estuary	PD	820	817.4081	2.5919	0.32
16–21 August 2016	Yangtze River Estuary	PD	2000	2171.377	171.377	7.89
9–12 May 2016	East of Flower and Bird Mountain, Shengsi, Zhoushan	PD	470	521.6767	51.6767	9.91
12–16 May 2016	Southeast of Zhujiajian, Zhoushan	PD	200	280.1863	80.1863	28.62
12–22 May 2016	Yushan Islands to Dantou Mountain	PD	480	347.0258	132.9742	27.70
16–21 May 2016	Shengshan sea area, Zhoushan	PD	120	241.2728	121.2728	50.26
22–30 May 2016	Aojiang River Estuary to Xiaguan, Cangnan	PD	100	158.9718	58.9718	37.10
5–14 July 2016	East sea of Zhujiajian, Zhoushan	Chaetoceros compressus (CC)	100	241.8425	141.8425	58.65
18–21 July 2016	Shengshan to Flower and Bird Mountain, Zhoushan	PD	350	487.8534	137.8534	28.26
24–27 July 2016	Southeast of Zhujiajian, Zhoushan	Mesodinium rubrum (MR)	150	164.2896	14.2896	8.70
8–11 August 2016	Southeast of Shengshan	PD	200	617.1895	417.1895	67.60
20–24 May 2017	Yushan waters	PD	220	41.06433	178.9357	81.33
20–24 May 2017	Yushan Islands to the middle of Tantoushan	Karenia brevis (KB), KM, Gonyaulax spinifera (GS), Scrippsiella trochoidea (ST), Ceratium tripus (CT), Pseudo-nitzschia pungens (PP)	420	372.4239	47.5761	11.33
20–30 June 2017	West of Nanjiushan	KB, KM	120	138.3143	18.3143	13.24
27–30 June 2017	Outside of Luoyu Island and Pishan, Yuhuan Island	KM	300	190.9737	109.0263	36.34
7–11 July 2017	East of Zhujiajiao, Zhoushan	Skeletonema costatum (SC)	100	249.9555	149.9555	59.99
7–9 August 2018	Dasong Estuary to Xihu Port, Xiangshan	Chaetoceros curvisetus (CC)	120	94.89789	25.10211	20.92
9–15 August 2018	East of Zhujiajian to Peach Blossom Island	Skeletonema costatum (SC)	150	231.4371	81.4371	35.19
9–15 August 2018	Huangxing Island to Dongfushan, Zhoushan	Leptocylindrus danicus (LD)	150	190.6636	40.6636	21.33
9 May 2019–11 June 2019	East of Nanji Island to Beiji Island to Dongtou Island	PD	800	136.501	663.499	82.94
15–28 May 2019	Ningbo, Yushan	PD, Noctiluca scintillans	200	250.5753	50.5733	20.18
30 July 2019–2 August 2019	East of Putuo Mountain, Zhoushan	Chaetoceros brevis	100	204.2495	104.2495	51.04
28 April 2020–20 May 2020	East of Nanji to Dongtou to Wenling	PD	380	493.9996	113.9996	23.08
29 April 2020–27 May 2020	Shipu, Wenzhou to Yushan	PD	380	300.208	79.792	21.00
14–19 May 2020	Southeast of Jigushan, Wenling	PD	100	126.889	26.889	21.19

lists the red tide areas and composite relative errors for historical red tide events extracted via this algorithm in the East China Sea, including the data, location, dominant species, area in the bulletin, identified area of red tide, absolute error, and composite relative error between the area in the bulletin and the identified area. Specifically, the absolute error is calculated as the absolute value of the difference between the identified area and the area stated in the bulletin. The composite relative error is calculated by dividing the absolute value of the difference between the area in the bulletin and the area identified by this method by the maximum area of the two.

$$\mathrm{Absolute}\mathrm{error}=\left|\mathrm{i}\mathrm{d}\mathrm{e}\mathrm{n}\mathrm{t}\mathrm{i}\mathrm{f}\mathrm{i}\mathrm{e}\mathrm{d}\mathrm{a}\mathrm{r}\mathrm{e}\mathrm{a}-\mathrm{a}\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{i}\mathrm{n}\mathrm{t}\mathrm{h}\mathrm{e}\mathrm{b}\mathrm{u}\mathrm{l}\mathrm{l}\mathrm{e}\mathrm{t}\mathrm{i}\mathrm{n}\right|$$

(1)

$$\mathrm{C}\mathrm{o}\mathrm{m}\mathrm{p}\mathrm{o}\mathrm{s}\mathrm{i}\mathrm{t}\mathrm{e}\mathrm{r}\mathrm{e}\mathrm{l}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{v}\mathrm{e}\mathrm{e}\mathrm{r}\mathrm{r}\mathrm{o}\mathrm{r}={\displaystyle \frac{\left|\mathrm{i}\mathrm{d}\mathrm{e}\mathrm{n}\mathrm{t}\mathrm{i}\mathrm{f}\mathrm{i}\mathrm{e}\mathrm{d}\mathrm{a}\mathrm{r}\mathrm{e}\mathrm{a}-\mathrm{a}\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{i}\mathrm{n}\mathrm{t}\mathrm{h}\mathrm{e}\mathrm{b}\mathrm{u}\mathrm{l}\mathrm{l}\mathrm{e}\mathrm{t}\mathrm{i}\mathrm{n}\right|}{\max \left(\mathrm{i}\mathrm{d}\mathrm{e}\mathrm{n}\mathrm{t}\mathrm{i}\mathrm{f}\mathrm{i}\mathrm{e}\mathrm{d}\mathrm{a}\mathrm{r}\mathrm{e}\mathrm{a},\mathrm{a}\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{i}\mathrm{n}\mathrm{t}\mathrm{h}\mathrm{e}\mathrm{b}\mathrm{u}\mathrm{l}\mathrm{l}\mathrm{e}\mathrm{t}\mathrm{i}\mathrm{n}\right)}}\times 100\%$$

(2)

The statistical results indicate that the algorithm effectively identifies the locations of 46 red tide events from 2011 to 2020, with areas generally aligning well. However, some events exhibit discrepancies, where the identified area is either overestimated or underestimated. The difference between the algorithm’s identified area and the reported area ranges from 0.82167 km² to 663.499 km², with an average difference of 94.6109 km²; most differences are within 120 km². The composite relative error between the remote sensing identified area and the reported area varies from 0.32% to 94.25%, with an average composite relative error of 35.2%.

Overall, the algorithm proposed in this study can effectively determine red tide occurrence locations and extract relevant information about red tides, with the areas identified generally consistent with those reported. Some individual events show either overestimation or underestimation of areas. One major reason for the smaller identified areas is data loss due to cloud cover in certain marine regions. Remote sensing can cover a broader range and detect red tide information not reported in bulletins, which explains why the identified area from remote sensing may be larger than the reported area.

4. Discussion

4.1. Changes in Red Tide Distribution

Figure 13 shows the distribution of red tides from 16 August to 20 August 2016, which represents the largest red tide event occurring during the decade from 2011 to 2020. The figure clearly shows the distribution change in red tide at different times. This may be attributed to human activities that increase eutrophication, leading to the occurrence of red tides. Owing to the influence of wind and ocean currents, red tides change over several days.

4.2. Algorithm for Identifying Dominant Algal Species in Red Tide

As indicated in Table 1, the primary algal species responsible for red tides in the East China Sea are diatoms and dinoflagellates. Figure 14 shows the differences in the phytoplankton absorption coefficient spectral curves for red tides dominated by diatoms and dinoflagellates. The figure shows that the spectral curves of the two algal types are quite similar, with peaks at approximately 440 to 490 nm, a minimum value at 660 nm, and another peak near 680 nm. However, the peak for diatom red tides at approximately 440–490 nm is relatively flat, whereas the peak for dinoflagellate red tides in the same range is more pronounced.

On the basis of the spectral analysis mentioned above, this study employed the ratio a_ph490/a_ph443 to distinguish between red tide algal species. Multiple red tide events over the past decade were collected, and the spectral curves for red tide waters were extracted, resulting in 218 spectral curves. These spectral curves were then analyzed and displayed via scatter plots to observe their distributions. Figure 15 shows that the a_ph490/a_ph443 values for dinoflagellate red tides are relatively low, whereas those for diatom red tides are relatively high. According to the statistical data, this study defined a_ph490/a_ph443 > 1.1 as indicative of dinoflagellate red tides and a_ph490/a_ph443 < 1.1 as indicative of diatom red tides.

Figure 16 shows the algal species identification results of the two red tides with the largest areas caused by different algal species during the period from 2011 to 2020, and both occurred in the Yangtze River Estuary waters. The figure shows that the extraction result of the dinoflagellate red tide on 17 May 2016, was very good. Except for misjudgments in a few locations, all other locations were identified as dinoflagellate red tides. For the identification of the diatom red tide on 16 August 2016, some were identified as diatom red tides, and some were identified as dinoflagellate red tides. More diatom red tides than dinoflagellate red tides were identified, so it was judged as a diatom red tide. Most of the inshore areas were identified as having diatom red tides, whereas some offshore areas were misjudged. Future research will further analyze the subtle differences in the phytoplankton absorption coefficient spectral curves between dinoflagellate and diatom red tides to distinguish between these two types of red tides.

4.3. Advantages of the Algorithm

Current commonly used remote sensing algorithms for determining the red tide water color focus on remote sensing reflectance, which is the overall performance of water spectral data obtained from red tide waters resulting from the combined effects of various factors, including phytoplankton, colored dissolved organic matter, suspended particles, and even bottom substrates [19,20,21], which limits the algorithms to local biochemical, physical, and hydrodynamic conditions at the time of observation. This results in significant regional applicability limitations and weaker transferability [22,23]. The absorption coefficient of phytoplankton does not change with variations in other substances. Since red tides are caused by phytoplankton, the use of the absorption coefficient of phytoplankton to identify red tides can exclude the interference of substances irrelevant to red tides, increasing the accuracy of the algorithm and improving its applicability.

The algorithm mentioned in this paper was applied to red tide events in the Bohai and Yellow Seas, and it was found to be effective in identifying red tides, showing good consistency with high chlorophyll concentrations. Figure 17 shows the red tide near Tianjin Central Fishery Port in the Bohai Sea on 27 March 2017, and Figure 18 shows the red tide in the Yellow Sea near the coastal area from the mouth of the Paidan River to the Fuzi River on 17 May 2017. Validation in the South China Sea was not possible, as the coverage of the GOCI imagery did not include this region. In conclusion, the algorithm proposed in this paper is effective. It successfully extracted red tide information in the East China Sea from 2011 to 2020. Moreover, the algorithm was also proven to be effective in extracting red tide information in the Bohai and Yellow Seas.

5. Conclusions

Red tides are caused by the massive proliferation or accumulation of phytoplankton, resulting in changes in water color. Phytoplankton play a crucial role in the optical characteristics of red tide waters, and the absorption coefficient of phytoplankton is an important component of these optical properties. Therefore, this study started with the spectral curves of the phytoplankton absorption coefficients and used GOCI data to invert these coefficients via the QAA algorithm. The differences in the phytoplankton absorption coefficient spectral curves among red tide waters, nearshore non-red tide waters, and offshore non-red tide waters during red tide events were analyzed and compared. The findings reveal that offshore non-red tide waters exhibit a peak near 660 nm, which is absent in red tide waters. Conversely, red tide waters show a peak at approximately 680 nm, which is not present in nearshore non-red tide waters. Consequently, this study established a red tide extraction algorithm using the criteria a_ph660-a_ph555 < 0 and a_ph680−a_ph660 > 0. The algorithm was applied to extract information on red tide waters from East China Sea events, demonstrating its effectiveness in accurately determining the locations of red tides and extracting relevant information.