The China Coastal Front from Himawari-8 AHI SST Data—Part 2: South China Sea

Abstract

High-resolution (2 km) high-frequency (hourly) SST data from 2015 to 2021 provided by the Advanced Himawari Imager (AHI) onboard the Japanese Himawari-8 geostationary satellite were used to study spatial and temporal variability of the China Coastal Front (CCF) in the South China Sea. The SST data were processed with the Belkin and O’Reilly (2009) algorithm to generate monthly maps of the CCF’s intensity (defined as SST gradient magnitude GM) and frontal frequency (FF). The horizontal structure of the CCF was investigated from cross-frontal distributions of SST along 11 fixed lines that allowed us to determine inshore and offshore boundaries of the CCF and calculate the CCF’s strength (defined as total cross-frontal step of SST). Combined with the results of Part 1 of this study, where the CCF was documented in the East China Sea, the new results reported in this paper allowed the CCF to be traced from the Yangtze Bank to Hainan Island. The CCF is continuous in winter, when its intensity peaks at 0.15 °C/km (based on monthly data). In summer, when the Guangdong Coastal Current reverses and flows eastward, the CCF’s intensity is reduced to 0.05 °C/km or less, especially off western Guangdong, where the CCF vanishes almost completely. Owing to its breadth (50–100 km, up to 200 km in the Taiwan Strait), the CCF is a very strong front, especially in winter, when the total SST step across the CCF peaks at 9 °C in the Taiwan Strait. The CCF’s strength decreases westward to 6 °C off eastern Guangdong, 5 °C off western Guangdong, and 2 °C off Hainan Island, all in mid-winter.

1. Introduction

The China Coastal Front (CCF) is a major ocean feature in the East China Sea (ESC) and northern South China Sea (NSCS). The CCF is a boundary between coastal waters off the mainland China and offshore waters. The CCF is physically associated and collocated with the China Coastal Current (CCC), which is also called the Zhejiang–Fujian Current or Zhe–Min Current in the ECS and Guangdong Coastal Current in the NSCS. Numerous studies of the NSCS elucidated many aspects of its circulation and frontal structure reviewed previously by Su (2004) [1] and recently by Shu et al. (2018) [2]. Nonetheless, a detailed climatology of the CCF is missing, probably because most efforts to date focused on establishing a general frontal pattern of the China Seas before zeroing in on individual fronts. The need to close this knowledge gap was the main incentive for our study of the CCF. In the first part of this investigation, we covered the ECS between 31°N and 24°N (Belkin et al., 2023) [3]. The present paper is dedicated to the NSCS between 24°N and 18°N. The Taiwan Strait (TS) is generally considered to be part of NSCS. The northern TS is covered in Part 1 (Belkin et al., 2023) [3], while the southern TS is covered in Part 2 (this paper).

The history of remote sensing studies of oceanic fronts in the NSCS is briefly reviewed below.

Table 1. Satellite studies of the China Coastal Front in the South China Sea from SST data. Place names: SCS, South China Sea; NSCS, Northern SCS; TS, Taiwan Strait. Algorithms and datasets: BOA2009, Belkin and O’Reilly (2009) [5]; CCA1992, Cayula and Cornillon (1992) [6]; CW2014, Castelao and Wang (2014) [7]; PH2010, Pi and Hu (2010) [8]; S2005, Shimada et al. (2005) [9]; W2020, Wang YT et al. (2020) [10] (combination of Canny (1986) [11], Castelao and Wang (2014) [7], and Wang YT et al. (2015) [12] algorithms); CCAIM, Xing et al. (2023) [13]; DOISST, Daily Optimum Interpolation SST; OSTIA, Operational SST and Sea Ice Analysis. Asterisk (*): AVHRR, MODIS, and AMSR-E data merged. Other acronyms: GM, gradient magnitude; NA, not available.

Reference	Sensor	Period	Algorithm	Region
Belkin and Cornillon 2003 [14]	AVHRR	1985–1996	CCA1992	SCS
Belkin and Cornillon 2007 [15]	AVHRR	1985–1996	CCA1992	SCS
Belkin et al., 2009 [16]	AVHRR	1985–1996	CCA1992	SCS
Belkin et al., 2024 (this study)	AHI	2015–2021	BOA2009	NSCS, including TS
Chang et al., 2006 [17]	AVHRR	1996–2005	S2005	TS
Chang et al., 2008 [18]	AVHRR	1996–2005	S2005	TS (north of 24°N)
Chang et al., 2010 [19]	AVHRR	2001–2007	S2005	NSCS, including TS
Chen JY and Hu ZF 2023 [20]	GHRSST	2002–2021	S2005	NSCS
Dong and Zhong 2020 [21]	AVHRR MODIS	2009–2012	GM	NSCS, including TS
Jing et al., 2015 [22]	OSTIA	2006–2013	GM	NSCS
Jing et al., 2016 [23]	GHRSST	2006–2014	GM	NSCS
Lan et al., 2009 [24]	AVHRR	1996–2005	S2005	TS
Lee et al., 2015 [25]	AVHRR	1996–2009	S2005	TS (north of 24°N)
Pi and Hu 2010 [8]	Misc. *	2002–2008	PH2010	NSCS, including TS
Ping et al., 2016 [26]	MODIS	2000–2013	CCA1992	TS (north of 22°N)
Ren et al., 2021 [27]	Model	2005–2018	Canny (1986)	SCS
Shi R et al., 2015 [28]	OSTIA	2006–2011	GM	NSCS
Wang DX et al., 2001 [29]	AVHRR	1991–1998	GM	NSCS, including TS
Wang YC et al., 2013 [30]	AVHRR	2006–2009	S2005	TS
Wang YC et al., 2018 [31]	AVHRR	2005–2015	S2005	TS (north of 24°N)
Wang YT et al., 2020 [10]	MODIS	2002–2017	W2020	NSCS, including TS
Xing QW et al., 2023 [13]	AVHRR	1982–2021	CCAIM	China Seas
Yang CY and Ye HB 2021 [32]	MODIS	2003–2017	NA	NSCS (114–117°E)
Yu et al., 2019 [33]	MODIS	2002–2017	GM	SCS
Zeng et al., 2014 [34]	MODIS	2002–2011	BOA2009	East Hainan
Zhang L and Dong J 2021 [35]	MODIS	2016–2017	GM	NSCS, 250 m L2 data
Zhang Y et al., 2021 [36]	OSTIA	2006–2015	GM	NSCS, 3D structure
Zhao et al., 2022 [37]	DOISST	1982–2021	CCA1992	China Seas

sums up relevant studies published in peer-reviewed international English-language journals. A few studies published in Chinese in domestic journals were recently reviewed by Shen and Belkin (2023) [4]. The review below is organized chronologically. Regional and thematical aspects are elucidated in subsequent sections of this paper.

After Jean-François Cayula and Peter Cornillon at the University of Rhode Island (URI) developed a state-of-the-art edge detection algorithm (based on the histogram approach) known as the Cayula–Cornillon algorithm (CCA; Cayula and Cornillon 1992 [6]; Cayula and Cornillon 1995 [38]), the CCA was used in numerous studies, including a global survey of SST fronts funded by NASA (see Belkin et al. (2009) [16] and references herein).

Meanwhile, Dongxiao Wang and collaborators published a study of SST fronts in the NSCS (Wang DX et al., 2001 [29]), using a 9 km resolution, 8-year Pathfinder dataset (1991–1998, thus overlapping with the URI dataset) and defining SST fronts as pixels with gradient magnitude GM exceeding 0.5 °C/9 km. Wang DX et al. (2001) [29] identified six fronts and determined frontal envelopes (“corridors”) for each front, using a method originally developed by Hickox et al. (2000) [39] to calculate cross-frontal SST steps across 10 fronts in the East China Seas. Among the six fronts in the NSCS, Wang DX et al. (2001) [29] identified the Fujian–Guangdong Front and noted that “the coastal fronts off Fujian and Guangdong become a continuous front in wintertime” (ibid., p. 3966). The maps of long-term mean seasonal frontal probability (frequency) presented by Wang DX et al. (2001) [29] show the coastal Fujian–Guangdong Front extending westward until ~115°E, where this front abuts another front, the Pearl River (Zhujiang) Estuary coastal front, which in spring-summer protrudes offshore beyond the 50 m isobath. Farther west, beyond the Pearl River front, Wang DX et al. (2001) [29] identified the Hainan Island East Coast front. Thus, between the Taiwan Strait and Hainan Island East Coast, the CCF consists of three disconnected fronts of different origin. This conclusion that seems obvious after a quick inspection of seasonal frontal maps presented by Wang DX et al. (2001) [29] becomes less obvious in the light of most recent studies discussed below that point to the alongshore continuity of the CCF during the cold season, especially in winter.

Yi Chang and collaborators used the Shimada algorithm (Shimada et al., 2005) [9] to detect and map SST fronts in the NSCS, including the Taiwan Strait (Chang et al., 2006 [17]; Chang et al., 2008 [18]; Chang et al., 2010 [19]; Lee et al., 2015 [25]). In a study of winter fronts, Chang et al. (2006) [17] and Chang et al. (2008) [18] used high-resolution (0.01°) data from 1996 to 2005 to generate long-term mean monthly maps of GM that document the CCF (called the Mainland China Coastal Front, MCCF) and its seasonal variability. The monthly maps of GM presented by Chang et al. (2006) [17], Chang et al. (2008) [18], and Chang et al. (2010) [19] reveal a rather sharp front (MCCF) with GM up to 0.3 °C/km in January, when the MCCF’s intensity peaks. Using 1 km SST data from 2001 to 2007, Chang et al. (2010) [19] generated maps of frontal frequency FF for December through March. In December and January, the continuous CCF extends along the Fujian–Guangdong coast up to Hainan Island. During these months, the CCF shows no signs of a local break-up off the Pearl River Estuary. Perhaps, this fact can be explained by the drastically decreased discharge of the Pearl River (Zhujiang) in winter. Indeed, according to the most recent climatology by Liu ZZ et al. (2022) [40] based on the 1954–2020 data, the Pearl River discharge drops seven-fold from the summer maximum of ~50 × 10⁹ m³/month to the winter minimum of ~7 × 10⁹ m³/month (Liu ZZ et al. (2022) [40], their Figure 6). In late winter (February-March), the CCF in the southern Taiwan Strait does not continue along the Fujian–Guangdong coast. Instead, the CCF veers offshore to join the Taiwan Bank Front (TBF). The CCF-TBF merger was noted earlier by Wang DX et al. (2001) [29], Chang et al. (2006) [17], and Chang et al. (2008) [18]. Farther west along the Guangdong coast, the CCF breaks up into two parts (eastern and western) off the Pearl River Estuary in spring-summer when the Pearl River discharge increases seven-fold from ~7 × 10⁹ m³/month in February to ~50 × 10⁹ m³/month in June (Liu ZZ et al. (2022) [40], their Figure 6). The sharply increased discharge pushes the Pearl River plume offshore across the shelf to separate the eastern and western Guangdong’s segments of the CCF.

The spatial pattern of the CCF in the Taiwan Strait remains unclear. For example, the often-repeated claim that the CCF extends along the 50 m isobath is inconsistent with the existence of a broad (~50 km wide) shallow area northwest of the Taiwan Bank, with depths of <40 m and often <30 m. This shallow area (submarine isthmus connecting the bank with the continent) could, in principle, constrain the CCC and the associated CCF. Indeed, surface drifters used by Qiu et al. (2011) [41] mostly avoided this shallow area except for a few drifters that went straight across it. Other drifters went to or around the Taiwan Bank. The existence of two quasi-zonal topographic barriers (Taiwan Bank with the submarine isthmus to the west and Chang Yun Rise in the eastern Taiwan Strait) gave Huang et al. (2020) [42] a reason to call the Taiwan Strait a “quasi-cul-de-sac” during the winter northeast monsoon.

Shi R et al. (2015) [28] used the Operational SST and Sea Ice Analysis (OSTIA) daily data with a spatial resolution of 0.05° from 2006 to 2011 to calculate monthly mean GM in the NSCS. The monthly maps of GM presented by Shi R et al. (2015, their Figure 2) [28] reveal the CCF featuring rather low GM (<0.08 °C/km), even during winter, when the CCF is best developed. As we are about to see below, such low values of GM obtained in some other studies (e.g., Shi R et al., 2022 [43]) tend to underestimate the real intensity of the CCF.

Shu et al. (2018) [2] reviewed the NSCS circulation and paid special attention to (a) the Guangdong Coastal Current (GCC) and (b) the interaction between the Pearl River plume (PRP) and coastal currents in the NSCS. The PRP-GCC/CCF interaction is highly complicated and variable in time and space owing to a complex interplay of winds, tides, topography, river discharge, stratification, and circulation (e.g., Dong et al., 2004 [44]; Zu and Gan, 2015 [45]).

Yu et al. (2019) [33] used MODIS data with a spatial resolution of 0.04° × 0.04° from 2002 to 2017 to calculate SST gradients and SST trends in the SCS. This is probably the first study of long-term variability and trends of SST gradient magnitude (SST GM), the latter being a measure of front intensity (front sharpness), while front strength is measured by a total SST step (SST range or SST differential) across the front. Yu et al. (2019, their Figure 5a) [33] presented a map of climatological annual mean SST GM that revealed rather small maximum values of GM, just slightly over 0.04 °C/km. As mentioned above, such low values underestimate the real maximum values of GM obtained in other studies. For example, using high-resolution (1 km) satellite data from the NSCS shelf, Dong J and Zhong Y (2020) [21] observed sharp SST fronts in winter, with GM exceeding 1.3 °C/km. There are at least two factors that might help explain the unrealistically low values of GM reported in some studies: one is the relatively coarse spatial resolution of SST data used in such studies; another is the long-term averaging involved in the production of climatic maps of GM.

Wang YT et al. (2020) [10] processed 4.5 km resolution MODIS data from 2002 to 2017 with their own algorithm that builds on Canny (1986) [11] modified by Castelao and Wang YT (2014) [7] and Wang YT et al. (2015) [12]. They used a small GM threshold for a front: 1.4 °C/100 km or 0.014 °C/km. Wang YT et al. (2015, their Figure 1b) [12] presented a map of long-term annual mean GM, in which the maximum GM scale value is a mere 0.04 °C/km. This rather low maximum scale value of GM is consistent with their previous study (Yu et al., 2019) [33]. Also, the maximum scale value of GM can be arbitrarily set to a low value to highlight weak fronts. The seasonal maps of frontal frequency presented by Wang YT et al. (2015, their Figure 2) [12] show the CCF as a robust feature in fall-winter (October-March) but a weak or absent feature in spring-summer (April-September), especially off western Guangdong.

Zhao et al. (2022) [37] used Daily Optimum Interpolation Sea Surface Temperature (DOISST) data from 1982 to 2021 with a spatial resolution of 0.25° × 0.25°. The data were processed with the Cayula and Cornillon (1992) [6] single-image edge detection algorithm (CCA-SIED) to detect SST fronts in the China Seas and study their spatial and temporal variability on a variety of scales. Alas, the massive study by Zhao et al. (2022) [37] was hampered by the rather coarse spatial resolution of the SST data used.

Chen JY and Hu ZF (2023) [20] used high-resolution (1 km) GHRSST data from 2002 to 2021 and Shimada et al. (2005)’s algorithm [9] to map SST fronts in the NSCS. The study area extended from 24°N (north of the Taiwan Bank) to 18°N (south coast of Hainan Island), and it was subdivided into three subregions. The detected SST fronts were classified into cold fronts, warm fronts, and cross-isobath fronts according to the fronts’ relationships with the underlying bathymetry following Ullman and Cornillon (1999) [46] and Ullman and Cornillon (2001) [47]. Seasonal maps of SST gradient magnitude GM revealed patterns consistent with similar maps published before. Monthly variability of several frontal parameters averaged over the three subregions was studied, including cross-frontal SST step, cross-frontal length scale, and cross-frontal GM, for the cold and warm fronts separately.

Xing et al. (2023) [13] developed a new front detection algorithm by modifying the Cayula and Cornillon (1992) algorithm [6] and applied the new algorithm to identify >50 fronts in the South China Sea and study their long-term variability over 40 years. This study resulted in the most complete detailed inventory of SST fronts in the SCS to date, which is consistent with the entire body of previous studies based on in situ and satellite data.

The above brief review of previous research reveals a few common threads. First, there were no dedicated studies of the entire CCF from satellite data, much less a comprehensive seasonal climatology of the CCF. Second, there is general agreement about the location (path) of the CCF in the NSCS. Third, there is evidence of the CCF’s spatial adjacency to the Taiwan Bank Front, albeit no clear picture of the frontal pattern in this area. Fourth, there are strong disagreements (exceeding an order of magnitude) regarding the maximum values of SST gradient magnitude GM in the SCS.

With the above issues in mind, we set out to produce an up-to-date monthly climatology of the CCF using the most recent satellite SST data of high spatial and temporal resolution. Our reliance on the most recent SST data mitigates possible effects of regional climate change that can be significant when recent data are compared with old data. Our emphasis on high-resolution data is meant to mitigate the detrimental effect of sparse data on gradient estimation. Finally, our approach was two-pronged: (1) elucidate general spatial and temporal patterns of the CCF and (2) provide numerical estimates of monthly SST ranges, maximum GM, and frontal frequency FF that can be used in a variety of academic and applied research and maritime activities.

The structure of this paper is as follows:

• Section 2: Data and Methods;
• Section 3: Results;
• Section 4: Discussion;
• Section 5: Conclusions.

2. Data and Methods

We used the same data and methods as in Part 1 of this study which focused on the ECS (Belkin et al., 2023) [3]. Nonetheless, to make Part 2 (this paper) self-contained, below we have provided sufficient descriptions of the data and methods that are similar to the respective section in Part 1 yet abridged.

2.1. Himawari-8/9 Satellites and Advanced Himawari Imager

The Japanese geostationary meteorological satellites Himawari 8/9 (Bessho et al., 2016) [48] carry the Advanced Himawari Imager (AHI) that has 16 spectral bands (visible light bands #1–3, near-infrared bands #4–6, and infrared bands #7–16), with a spatial resolution that varies from 2 km for bands #5–16 to 1 km for bands #1, 2, and 4 and 500 m for band #3. Every 10 min, the AHI provides full-disk images, while cloud-free full-disk composite images are available every four days. The satellites are stationed at 140.7°E and cover the 80°E–160°W, 60°N–60°S area. In this study, L3 level SST data with 1 h temporal resolution and 2 km spatial resolution from July 2015 through December 2021 that cover our study area (northern SCS, including Taiwan Strait: 18–26°N, 108–122°E) were downloaded from the website of the Japan Aerospace Exploration Agency, JAXA (https://www.eorc.jaxa.jp/ptree/ (accessed 5 August 2024). The hourly data are generated by the JAXA from the original 10 min data. The AHI data are processed with a novel cloud-masking algorithm (which uses 500 m resolution visual images that resolve individual cumulus clouds) and a novel cloud-tracking algorithm; these advanced algorithms combined with the high-frequency (every 10 min.) full-disk scanning enable the production of cloud-free imagery every four days, which is unprecedented (Bessho et al., 2016) [48].

2.2. Front Mapping

The Belkin and O’Reilly (2009) [5] algorithm (BOA) was used to map SST fronts. The BOA generates maps of gradient magnitude (GM) and gradient direction (GD). Frontal maps are generated by setting a threshold T for GM. Every pixel with GM > T is considered a frontal pixel. Once frontal maps showing the most intense fronts (with GM > T) are generated, maps of frontal frequency FF over any time period can be generated. Pixel-based FF is calculated as a ratio, FF = N1/N2, where N1 is the total number of times the given pixel was frontal, and N2 is the total number of times the given pixel was flagged as valid (i.e., cloud-free). The maps of GM and FF complement each other because gradient magnitude GM is a metric of front intensity (front sharpness), while frequency FF is a metric of front stability (front robustness).

2.3. Front Strength and Cross-Frontal Ranges of Oceanic Variables

Another important aspect of every front is its strength, defined as the total cross-frontal range (or cross-frontal step, change, or differential) of the oceanic variable in question. Hickox et al. (2000) [39] estimated cross-frontal ranges by defining frontal boundaries and frontal envelopes (“corridors”) using maps of frontal frequency. Belkin et al. (2023) [3] used a different method by analyzing distributions of SST along a series of closely spaced cross-frontal sections. This approach allows both cross-frontal and along-frontal variability to be visualized, as demonstrated in Section 3.

2.4. Downstream Tracking of CCF

To investigate the CCF extension from the ECS into the NSCS, we used a simple but efficient technique of downstream tracking of ocean fronts. Most recently, this approach was used by Belkin et al. (2023) [3] to study the CCF in the ECS. The algorithm of downstream tracking can be summed up as follows: (1) make a provisional map of the front in question; (2) find a large number of cross-frontal oceanographic sections and arrange them along the front to minimize section-to-section spacing; (3) along each cross-frontal section, determine frontal parameters (e.g., SST) on both sides of the front; and (4) visualize and analyze the downstream distribution of the frontal parameters, especially SST on both sides of the front.

3. Results

3.1. Introduction

In this section, we present the most important results, namely, long-term mean monthly maps of SST, GM, and FF, and plots of SST along cross-frontal sections. These maps and plots visualize key frontal parameters: (1) intensity identified with gradient magnitude GM, (2) robustness quantified as pixel-based frontal frequency FF, and (3) strength measured as the total cross-frontal range of SST.

3.2. Sea Surface Temperature

The long-term (2015–2021) mean monthly maps of SST (Figure 1) show seasonal evolution and spatial variability of SST in the northern SCS. The large-scale pattern of SST in the northern SCS has two distinct seasons—winter (November–April) and summer (May–October)—in accordance with the monsoon-driven climate of this region.

3.2.1. Winter

After a summer-to-winter transition in October, the winter pattern of SST in the northern SCS is established in November and persists through April (Figure 1). This pattern can be exemplified by the month of January, when the coastal waters are relatively uniform along the entire Guangdong Coast between the Taiwan Strait in the east and Leizhou Peninsula/Qiongzhou Strait in the west. Across this 700 km distance, the coastal SST varies very little in the alongshore direction, remaining within a mere two-degree range between 17 and 19 °C. The cross-shelf gradient of SST is also uniform in the alongshore direction. The shelf-break is roughly demarcated by the 23 °C isotherm along the entire northern shelf of the SCS. Thus, across the shelf, the SST varies between 17 and 23 °C in the east (near Taiwan Strait) and 19–23 °C in the west (near Hainan). The alongshore uniformity of SST is suggestive of the dominant role of alongshore advection by the westward Guangdong Coastal Current. This explanation is consistent with the winter wind-driven circulation pattern forced by the monsoon-driven northeasterlies. The tongue of cold near-shore waters (15–17 °C) along the Fujian Coast manifests the southward China Coastal Current (Zhe–Min Current), which is driven by the monsoon’s northeasterlies.

3.2.2. Summer

After a winter-to-summer transition in May, the summer pattern of SST in the northern SCS is established in June and persists essentially unchanged through September (Figure 1). The summer pattern features upwellings in two areas: Taiwan Strait and the east of Hainan Island. The Taiwan Strait upwelling consists of two centers: one over the Taiwan Bank and another off the Fujian Coast. Both centers remain quite robust over the 4-month period in June-September. Compared with the Taiwan Bank upwelling, the East Hainan upwelling is relatively short-lived. It develops best in June-July, significantly weakens in August, and almost completely disappears in September. During summer, the coastal waters off the Guangdong Coast cannot be reliably distinguished from SST maps alone as the summertime warming reduces the horizontal gradients of SST. As a result, the entire shelf of the northern SCS is covered with uniformly warm waters (29–30 °C). Neither the CCF nor the shelf-slope front south of the Guangdong Shelf can be identified from the summertime SST maps.

3.3. SST Gradient Magnitude GM

Horizontal gradients of SST are typically at maximum in winter and at minimum in summer, when summertime warming all but obliterates the surface manifestations of most fronts (e.g., Hickox et al., 2000 [39]). Therefore, to elucidate the spatial patterns of SST gradients and facilitate visual detection of thermal fronts, Belkin et al. (2023) [3] used dynamic scaling of SST gradient magnitude GM. Specifically, monthly maps of GM were generated by adjusting color scales for GM based on monthly ranges of GM determined from monthly histograms of GM (Figure 2). These histograms reveal maximum values of GM in winter (up to 0.15 °C/km), while summertime GM are much smaller and do not exceed 0.05 °C/km in September-October.

The long-term mean monthly maps of SST gradient magnitude GM (Figure 3) reveal high-gradient zones (fronts) and their spatial and temporal variability. The most striking feature of the temporal variability is dramatic changes in the frontal pattern on a monthly time scale. In this respect, the northern SCS differs from the ECS, where the frontal pattern evolves on a seasonal time scale (Belkin et al., 2023) [3].

3.4. Statistics of SST Fronts: Frontal Frequency Maps

Maps of frontal frequency FF (Figure 4) show the most intense fronts with GM > 0.1 °C/km.

Pixel-based FF in Figure 4 is calculated as the number of times the given pixel’s GM exceeded the threshold of 0.1 °C/km divided by the total number of times the given pixel contained valid data during the given month. The large-scale patterns of fronts in FF maps (Figure 4) and GM maps (Figure 3) are mutually consistent. The main difference between the two is that the FF maps bring out the most robust (stable) fronts that retain their location during the given time period (in this case, a month). At the same time, the FF maps inevitably leave out myriads of migrating and short-lived (transient) mesoscale and submesoscale fronts that do not show up in the FF maps. Such dynamic small-scale fronts may play a significant role in marine ecology, owing to their ubiquity and sheer numbers. The statistics and ecological role of such dynamic small-scale fronts remain unexplored. Any statistical metrics of such fronts should include integration over a study area as opposed to pixel-based statistics. An example of such integral metrics (frontal index F1) can be found in Belkin et al. (2009) [16].

3.5. Cross-Frontal Distributions of SST along Fixed Lines across the Northern SCS

To explore the cross-frontal structure of SST, we plotted long-term mean monthly SST along 11 fixed lines across the northern SCS, including four parallels (23°N, 21°N, 20°N, and 19°N) and seven meridians between 112°E and 118°E (Figure 5).

The monthly plots of SST along the 11 fixed lines (Figure 6 and Figure 7) reveal mesoscale and submesoscale details of the cross-frontal structure of the SST field that are not apparent in the monthly maps of SST (Figure 1), GM (Figure 3), and FF (Figure 4). The monthly plots of SST vs. offshore distance also allow accurate estimations of cross-frontal SST steps (differentials dSST), defined as offshore SST minus inshore SST. Accurate estimations of dSST depend on accurate demarcation of the offshore and inshore boundaries of the front in question. For more methodological details, see Part 1 of this study by Belkin et al. (2023) [3].

3.6. Analysis of SST Distributions along 11 Fixed Lines across the Northern SCS

The monthly plots of SST along 11 fixed lines (Figure 6 and Figure 7) allow the identification and accurate estimation of various fronts in the northern SCS. Even though this study is squarely focused on the CCF, we also considered a few other fronts adjacent to the CCF, namely, the Taiwan Bank upwelling front, Fujian upwelling front, East Hainan upwelling front, Taiwan Warm Current front, and Pearl River plume front.

China Coastal Front (

Table 2. SST ranges (°C) across the CCF in the northern SCS determined from wintertime SST distributions along Lines 1–11 (Figure 6 and Figure 7). The locations of Lines 1–11 are shown over bathymetry in Figure 5. Lines 1 and 2 in the two rightmost columns are swapped since Line 2 crosses the CCF upstream of Line 1 (as evident from a detailed analysis of all the data). This swap allowed all the lines (columns) to be arranged along the CCF, going downstream from right to left, with Lines 1–8 arranged east to west and Lines 9–11 arranged north to south (Figure 5). The asterisks (*) mark unreliable values when the front’s boundary is poorly defined. The exclamation signs mark the most reliable values associated with well-defined boundaries of the front.

Line No.	11	10	9	8	7	6	5	4	3	1	2
Lon/Lat	19N	20N	21N	112E	113E	114E	115E	116E	117E	23N	118E
November	26.5–26.7 *	25.9–26.4	25.1–26.0	24.0–26.0	24.2–26.0	24.5–25.8	23.8–25.5	23.2–25.8	22.8–27.0	23.6–26.7	22.4–27.2
December	24.0–25.0	22.6–24.7	21.2–23.5 *	20.3–25.3 *	19.6–25.5 *	20.6–25.7 *	20.1–22.9 *	19.7–23.7	20.2–24.3	20.1–24.7	18.2–25.1
January	22.3–23.8 *	20.4–22.9	19.0–22.0 *	18.5–24.0 *	18.3–24.2 *	18.3–24.2 *	17.7–21.4 !	17.2–23.0 !	16.5–24.2 *	17.7–23.3 !	15.7–24.3 !
February	22.0–23.7	20.2–22.8 *	19.0–22.2 *	18.7–23.7 *	18.4–23.8 *	18.2–24.0 *	17.5–21.5	16.8–22.7 *	16.2–23.7	17.5–23.6	15.3–24.3
March	23.5–24.5	22.0–24.2	21.0–23.6 *	20.5–24.5 *	20.3–24.9 *	19.9–25.3 *	19.6–22.3	18.5–23.4 *	18.2–23.8	20.2–24.5	17.0–24.6
April	25.4–26.4 *	24.5–26.0	24.1–25.3 *	23.6–26.6 *	23.7–26.7 *	23.1–26.7 *	23.0–26.7 *	22.3–27.0	21.8–27.1 *	23.9 *–25.8	20.4–25.7

): This front develops in winter (November–April) when it stands out in SST plots along the 11 fixed lines (Figure 6 and Figure 7). During November through February/March, cross-frontal distributions of SST are remarkably similar between 118°E and 112°E. This qualitative similarity is suggestive of strong alongshore advection in winter by the westward Guangdong Coastal Current. Quantitatively, the wintertime cross-frontal ranges of SST decrease downstream (westward) along the front between 118°E and 112°E yet remain substantial even at 112°E, especially in mid-winter (Table 2).

The CCF is prominent in winter between 118°E and 112°E, that is, between the Taiwan Strait and the vicinity of Leizhou Peninsula. The cross-frontal SST range (offshore SST minus inshore SST) peaks at 9.0 °C in February at 118°E. Moving downstream (westward), the cross-frontal SST range decreases from 6 to 8 °C in the east down to 5−6 °C in the west. These high values are for mid-winter only (largely January and February). The respective values for transition months (early winter and late winter) are smaller yet significant, varying from 5 °C in the east down to 2−3 °C in the west. Overall, in winter, the SST steps across the CCF in the northern SCS are higher than those in the ECS, the latter determined by Belkin et al. (2023) [3] from the same data. This result appears striking and counter-intuitive because the maximum wintertime values of SST gradient magnitude GM in the ECS (determined by Belkin et al. (2023) [3]) are much higher than the respective values of GM in the northern SCS (Figure 2), 0.3–0.4 °C/km vs. 0.15 °C, respectively. These results can be summed up as follows: The CCF in the SCS is less intense (less sharp) but stronger than the CCF in the ECS. The lower intensity and greater strength of the CCF in the SCS are explained by the CCF being much wider in the SCS than in the ECS. These findings illustrate the importance and complementary nature of two frontal parameters, namely, front intensity and front strength.

Western terminus of the CCF in the northern SCS (Table 2): The CCF tapers off between Line 8 along 112°E and Line 9 along 21°N. The front’s abrupt weakening over such a short distance is remarkable. Indeed, in winter the CCF is well-defined along Line 8 but poorly defined along Line 9 (Table 2). Farther south, off Hainan’s east coast, Line 10 (20°N) and Line 11 (19°N) crossed a narrow coastal front with a mid-winter cross-frontal SST range of up to 2.5 °C. Is this front the real terminus of the CCF? A provisional answer is “no” because the coastal front off the east coast of Hainan is very narrow compared with the very wide CCF in the northern SCS as evidenced by cross-frontal distributions of SST in winter (Figure 6 and Figure 7). Also, the CCF is linked to the westward Guangdong Coastal Current, which extends to the Beibu Gulf via the Qiongzhou Strait between the Leizhou Peninsula and Hainan Island (Shi MC et al., 2002 [49]; Lao et al., 2022 [50]). Is there any evidence of the Guangdong Coastal Current branching south before entering the Qiongzhou Strait to flow west? Yes, the available evidence supports this view. According to schematics by Lao et al. (2022) [50], the West Guangdong Coastal Current bifurcates while approaching the Qiongzhou Strait, with the southward branch flowing past the north cape of Hainan Island. Thus, the southward extension of the Guangdong Coastal Current could be linked to the southward terminus of the CCF off eastern Hainan, especially in winter. The southward advection from the West Guangdong Shelf is invoked by Li JY et al. (2023) [51] to explain Chl-a variations off the east coast of Hainan Island.

East Hainan Front (

Table 3. SST ranges (°C) across the East Hainan Front determined from SST distributions along Lines 10 and 11 (Figure 6 and Figure 7). The locations of Lines 10 and 11 are shown over bathymetry in Figure 5.

Line	November	December	January	February	March	April	May	June	July	August	September	October
10 (20°N)	25.9–26.4	22.6–24.5	20.4–22.8	20.2–22.7	22.0–24.0	24.5–26.0	27.3–28.8	28.4–29.9	28.2–30.2	28.7–30.2	29.0–29.7	27.9–27.9
11 (19°N)	26.5–26.6	24.0–25.0	22.3–23.8	22.0–23.7	23.5–24.5	25.5–26.3	27.5–29.2	27.6–30.0	27.8–30.2	28.5–30.0	29.4–29.7	28.0–28.1

): This front persists year-round except October, which appears as a transition month between summer and winter (Figure 6 and Figure 7). In summer, this front is maintained by wind-driven coastal upwelling generated by the monsoon-related southwesterlies (e.g., Jing ZY et al., 2015 [22]; Jing ZY et al., 2016 [23]). The strength of the front (identified with the total SST range across the front) in summer exceeds 2 °C, peaking at 2.4 °C in June-July along Line 11 at 19°N (Table 3). In winter, this front’s dynamics is different since the front is largely maintained by the southward advection from the north. The strength of the front in winter peaks at 2.5 °C in February along Line 10 at 20°N, thus rivaling the maximum strength of this front in summer (Table 3).

Taiwan Strait upwelling fronts (

Table 4. SST ranges (°C) across the Taiwan Strait upwelling fronts in May-September determined from SST distributions along Line 1 and Line 2 (Figure 6 and Figure 7). For each upwelling’s crossing, three values are given in accordance with the V-shape (“valley”) model described in Part 1 of this study by Belkin et al. (2023) [3]: inshore SST, minimum SST in the upwelling center, and offshore SST. The Taiwan Bank upwelling front is crossed by Line 1 at 23°N. The Fujian coastal upwelling front is crossed by Line 2 (118°E). The locations of Lines 1 and 2 are shown over bathymetry in Figure 5.

Line	May	June	July	August	September
1 (23°N)	26.1–25.5–28.3	28.3–26.6–29.8	29.1–27.1–30.3	29.2–27.0–30.1	28.5–27.4–29.5
2 (118°E)	No upwelling	27.3–26.5–29.5	27.7–26.5–30.0	28.0–27.0–29.8	28.2–27.5–29.5

): During the warm season (May-October), the persistent upwelling-favorable southwesterlies result in the formation of upwelling fronts in the Taiwan Strait. The upwelling fronts can be easily recognized in cross-frontal distributions of SST by their signature V-shape as illustrated and discussed in Part 1 of this study by Belkin et al. (2023) [3]. The seasonal development of the upwelling front east of the Taiwan Bank is well documented by monthly distributions of SST along Line 1 at 23°N (Figure 6 and Figure 7). The Taiwan Bank upwelling front emerges in May and develops best from June through September (Figure 6 and Figure 7). West of this upwelling, the SST distribution along 23°N (Line 1) has a distinct M-like shape in May-September, being best developed in June-August. The M-shaped distribution of SST across the Taiwan Bank at 23°N can be indicative of a minor upwelling between two “hot spots” of SST. The seasonal persistence of this feature is likely associated with the rugged bathymetry of this area. Another major upwelling center is located off the Fujian Coast. The seasonal development of upwelling off Fujian is well documented by monthly distributions of SST along Line 2 at 118°E (Figure 6 and Figure 7). The Fujian coastal upwelling front is developed from June through September (Figure 6 and Figure 7).

Inshore CCF front in the western Taiwan Strait (

Table 5. SST ranges (°C) across the inshore CCF in the western Taiwan Strait in November-April determined from SST distributions along Line 1 at 23°N (Figure 6 and Figure 7). The location of Line 1 is shown over bathymetry in Figure 5.

Line	November	December	January	February	March	April
1 (23°N)	23.1–23.6	19.4–20.0	17.0–17.8	16.5–17.4	18.7–20.5	22.0–23.2

): A careful examination of SST distribution along Line 1 (23°N) in winter (Figure 6 and Figure 7) reveals a minor but persistent near-shore front with a cross-front SST range between 0.5 and 1.8 °C. This front is present from November through April. During the warm season (May–October), this inshore front appears as the westernmost part of the M-shaped pattern along 23°N (Line 1) (Figure 6 and Figure 7) described above.

4. Discussion

4.1. China Coastal Front as a Major Link between the ECS and SCS

In Part 1 of this study (Belkin et al., 2023) [3], we focused on the CCF in the ECS, without addressing an important issue of the CCF extension into the SCS. In the present paper, we followed the CCF from the Taiwan Strait to Hainan Island. In winter, the CCF extends via the Taiwan Strait to the SCS, where the CCF is associated with the Guangdong Coastal Current. Thus, in winter, the CCF continues uninterrupted from the northern ECS (Yangtze Bank) all the way along the mainland China coast up to Hainan. The wintertime continuity of the CCF between the Yangtze Bank and Hainan documented in this study confirmed the results of previous studies (e.g., Wang DX et al., 2001 [29]; Chang et al., 2010 [19]; Wang YT et al., 2020 [10]; Tan et al., 2023 [52]).

4.2. China Coastal Front as a Chain of Regional Coastal Fronts

The CCF in winter appears as a single continuous front from the Yangtze Bank to the east coast of Hainan Island. Nonetheless, it is often convenient to consider individual segments of this very long front, especially if a study focuses on a small area. Therefore, historically, the CCF was sometimes considered as a chain of regional coastal fronts, with each regional front being a link of the chain. Despite a multitude of different names assigned to various regional fronts that are parts of the CCF, there is a general consensus regarding the CCF’s regional segmentation. From north to south, the CCF consists of the Zhejiang–Fujian Front, Taiwan Bank Front, East Guangdong Front, West Guangdong Front, and East Hainan Front. These are major links of the chain that we call the CCF. There are also minor branches of the CCF identified by earlier researchers and documented in detail in this study. For example, in the western Taiwan Strait (between the Taiwan Bank and mainland China), there is a near-shore shallow branch of the CCF which was clearly revealed, particularly in December and January, by high-resolution maps of frontal frequency in Chang et al. (2010, their Figure 2) [19]. The same near-shore branch is also revealed by our high-resolution maps of SST gradient magnitude GM (Figure 3), being best defined in December.

4.3. China Coastal Front in the Northern SCS in Summer

Driven largely by the summer monsoon winds, the Guangdong Coastal Current reverses and flows northeast. During several summer months (typically, July through September), the CCF-associated SST gradient (Figure 3) is barely noticeable off eastern Guangdong and it nearly vanishes off western Guangdong. The western and eastern Guangdong segments of the CCF appear disconnected. The most obvious reason for the disconnect is the Pearl River plume, which in summer extends across the Guangdong Shelf up to the 50 m isobath and beyond (Zhi et al., 2022, their Figure 14) [53]. In the western Taiwan Strait, the Taiwan Warm Current flows northeast along the mainland China coast. Here, the CCF is an upwelling front maintained by the southwesterlies. The SST distribution along Line 2 (118°E) in June-September (Figure 6 and Figure 7) has a characteristic V-shape as this line crosses the Fujian upwelling. The V-shape distribution (valley model) of SST is typical for the CCF during summer in the East China Sea and Taiwan Strait (Belkin et al., 2023) [3].

4.4. Guangdong Coastal Current and China Coastal Front

The China Coastal Current (CCC) commonly associated with the China Coastal Front (CCF) originates in the ECS, where it is often called the Zhejiang–Fujian Current or Zhe–Min Current. In winter, the CCC extends to the northern SCS, where it is called the Guangdong Coastal Current (GDCC). Some authors distinguish the Eastern and Western GDCC. The inshore edge of GDCC extends over shallow waters with rugged bathymetry, which makes ship-based oceanographic surveys problematic. Therefore, direct current measurements with surface drifters appear as a viable alternative to ship surveys (Lin HY et al., 2020 [54]; Yang LQ et al., 2021 [55]; Yang LQ et al., 2023 [56]). Using such drifters, Yang LQ et al. (2021) [55] demonstrated that in winter the GDCC flows largely southwest along the Guangdong Coast in the western Taiwan Strait, at times achieving a speed of 1 m/s. Yang LQ et al. (2021) [55] reported a bifurcation of the GDCC off eastern Guangdong (south of Shanwei). The separation between the two current jets attains its maximum (>100 km) off central Guangdong. Both current jets eventually merge before the GDCC approaches Hainan Island. None of the drifters reported by Yang LQ et al. (2021) [55] got caught into the inshore GDCC, which flows toward the Leizhou Peninsula and Qiongzhou Strait. Also, the aforementioned bifurcation of the GDCC did not result in a bifurcation of the CCF (Figure 6 and Figure 7).

4.5. Westernmost Extension of the China Coastal Current

The importance of the connection between the Guangdong Shelf and Beibu Gulf was recognized long ago (e.g., Shi MC et al., 2002 [49]). Recently, these earlier findings were supported via a water mass analysis by Lao et al. (2022) [50] who demonstrated that waters carried by the GDCC via the Qiongzhou Strait play an important role in the water mass composition of the Beibu Gulf, especially in winter when the GDCC waters dominate the Beibu Gulf. However, notwithstanding the westward extension of the China Coastal Current via the Qiongzhou Strait observed by Shi MC et al. (2002) [49] and Lao et al. (2022) [50], we do not see any sign of the China Coastal Front (associated with the China Coastal Current) extending westward into the Beibu Gulf. It is possible that the 2 km resolution of the AHI SST data is not fine enough to resolve the CCF’s branch in the Qiongzhou Strait. It is also possible that the CCC’s branch in the Qiongzhou Strait is not manifested in SST. High-resolution data (e.g., Landsat imagery) can be used to resolve this issue.

4.6. East Hainan Fronts

The summer cooling off the northeast coast of Hainan is dynamically different from the summer cooling off the southeast coast of Hainan (Lin PG et al., 2016 [57]; Li YN et al., 2020 [58]; Bai et al., 2020 [59]). The local bathymetry off the northeast cape of Hainan Island (along our Line 10 at 20°N; “NE Cape” in Jing et al. (2015) [22]) features a zonally oriented elongated bank (shoal) protruding eastward. Tides interact with this feature, resulting in topographic upwelling and surface cooling along 20°N. The crucial role of bathymetry and tides in amplifying the effects of wind-induced upwelling was investigated by Bai et al. (2020) [59], who found the western edge of the upwelled water (“cold spot”) to have a peculiar, jagged morphology and to be repeatedly observed at the same location, apparently being steered by bathymetry. Advection from the north by the southward branch of the West Guangdong Coastal Current is another factor that affects SST at 20°N, and this factor plays a crucial role as shown by Li JY et al. (2023) [51]. Our results (Table 3) show that the front’s strength (cross-frontal SST step) peaks at 2.4–2.5 °C in winter and summer, although the front’s dynamics is fundamentally different during these opposite seasons (Jing ZY et al., 2015 [22]; Jing ZY et al., 2016 [23]; Lin PG et al., 2016 [57]; Li YN et al., 2020 [58]; Li JY et al., 2023 [51]). The magnitude of nearshore surface cooling (identified with the front strength) determined in this study (up to 2.5 °C) coincides with the results by Shi WA et al. (2021) [60] based on daily Himawari-8 AHI 2 km SST data, while we used monthly data.

4.7. Fujian and Guangdong Coastal Upwelling Fronts

In summer, monsoon-driven southwesterlies blow along the mainland China coast and drive upwelling off Fujian and Eastern Guangdong (Hu JY and Wang XH, 2016 [61]; Hu JY et al., 2018 [62]; Shi WA et al., 2021 [60]). The Fujian upwelling stands out in our monthly SST maps in June-September (Figure 1), when the upwelling center remains stable in the western Taiwan Strait south of 24°N, with minimum SST around 27 °C during the entire four-month period. The monthly SST distributions along our Line 1 (23°N) document the seasonal evolution of an SST front between the upwelled and offshore waters (Figure 6 and Figure 7). As shown in Part 1 of this study (Belkin et al., 2023) [3], the CCF dynamics fundamentally changes twice a year. In winter, the CCF is a water mass front. In summer, this CCF becomes an upwelling front maintained by the monsoon-driven southwesterlies.

4.8. Taiwan Bank Fronts

Lan KW et al. (2009) [24] used the Shimada algorithm [9] to detect upwelling fronts around the Taiwan Bank in summer, when the SST gradient magnitude GM reaches 0.2 °C/km. Their results (presented as maps of GM) are qualitatively consistent with our results (Figure 3), although their estimates of peak values of GM are higher than ours: 0.20 vs. 0.06 °C/km, respectively. This quantitative discrepancy can be explained, at least partly, by the higher resolution of SST data used by Lan KW et al. (2009) [24] vs. this study (1 km vs. 2 km, respectively). Zhang F et al. (2014) [63] conducted a thorough analysis of satellite data and cruise surveys of the Taiwan Bank fronts and reported the fronts’ locations and cross-frontal structure in temperature and salinity.

4.9. Pearl River Plume Front

The front location is determined by the complex dynamics of the plume, with major factors being river discharge, winds, tides, bathymetry, Coriolis force, and ambient coastal currents (Su, 2004 [1]; Dong LX et al., 2004 [44]; Ou SY et al., 2009 [64]; Zu TT et al., 2014 [65]; Zu TT and Gan JP, 2015 [45]; Hu ZF et al., 2022 [66]; Zhi et al., 2022 [53]). The Pearl River discharge is often approximated by a sum of discharges of three main tributaries (Xijiang, Beijiang, and Dongjiang); long-term mean estimates of the sum vary slightly (285.2 × 10⁹ m³/a by Zhang SR et al. (2008) [67] or 283.4 × 10⁹ m³/a by Liu ZZ et al. (2022) [40]; either way, the long-term mean annual discharge is ~9000 m³/s). The Pearl River discharge varies seasonally by a factor of 6, from ~3000 m³/s in winter to 18,000 m³/s in summer according to Zhi et al. (2022, their Figure 8) [53]. The discharge variations strongly affect the Pearl River plume’s extent and location. In winter (November–April), the Pearl River plume is usually deflected to the west by the combined effect of the Coriolis force, monsoon-driven northeasterlies, and westward Guangdong Coastal Current, extending along the 10 m isobath off western Guangdong (Zhi et al., 2022, their Figure 14) [53]. In summer (May–October), the Pearl River plume extends far offshore, reaching the 50 m isobath and beyond (Zhi et al., 2022, their Figure 14) [53], and it can be deflected eastward by the eastward Guangdong Coastal Current and monsoon-driven southwesterlies. We observed the Pearl River plume front in winter, particularly in monthly maps of SST gradient magnitude GM (Figure 3) and frontal frequency FF (Figure 4), despite the plume’s relatively small size, owing to the sharply decreased discharge in winter (Liu ZZ et al., 2022) [40].

4.10. Intensity and Strength of the China Coastal Front

Cross-frontal SST gradients, especially their peak values, are commonly used as a measure of intensity (sharpness) of fronts. In this study, we provided monthly histograms of SST gradient magnitude GM (Figure 2) and monthly maps of GM (Figure 3). The histograms were used to determine maximum values of GM and use them for dynamic mapping of GM. Thus, the maximum values of GM in our maps (0.10 °C/km; Figure 3) are reasonably close to the actual maximum values of GM generated by the BOA algorithm (0.15 °C/km; Figure 2). Another important frontal parameter is strength, defined as a total cross-frontal SST step (differential dSST). As demonstrated in this study, the CCF is very strong in winter, when dSST peaks at 9 °C, while the CCF intensity (sharpness) is moderate, peaking at 0.15 °C/km. The contrast between the extreme strength and moderate intensity is explained by the much greater width of the CCF in the northern SCS vs. the CCF width in the ECS. Thus, both frontal parameters—intensity and strength—are important and complementary as both parameters are needed for complete characterization of fronts. Also, the significant breadth of the CCF in the SCS (50–100 km) enables this front to exert a substantial influence on the marine atmospheric boundary layer: As observed by Shi R et al. (2017) [68], when air parcels are slowly moving across a wide SST front, these air parcels are bound to be affected by the SST differential.

4.11. Comparison of Cross-Frontal SST Gradients (Figure 2 and Figure 3) with Published Data

Numerous studies of the CCF—both in situ and satellite studies—provided estimates of the front intensity (sharpness), defined as the maximum cross-frontal SST gradient magnitude GM. These estimates vary widely, spanning more than an order of magnitude. The two main reasons for such variations are (1) differences in resolution of observed SST data and (2) data processing techniques involving spatial and temporal interpolation, smoothing, and averaging of observed data that all result in gradient reduction. Satellite SST data resolution of different radiometers varies by one to two orders of magnitude, whereas the resolution of in-situ SST data varies by two orders of magnitude and more. Typically, a higher data resolution results in an increase in estimated GM across the front in question. Even though the CCF in the SCS is quite strong, with a cross-frontal dSST peaking at 9 °C in winter, this differential extends over the wide (50–100 km) front. The maximum GM reported across the CCF in the northern SCS exceeded 1 °C/km, based on high-resolution satellite data (Dong and Zhong, 2020) [21]. Other estimates are much smaller, varying from 0.3 °C/km (Chang et al., 2006 [17]; Chang et al., 2008 [18]) to 0.2 °C/km (Chang et al., 2010 [19]; Zhang Y et al., 2021 [36]) to 0.15 °C/km (this study) to 0.08 °C/km (Shi R et al., 2015 [28]) to 0.07 °C/km (Chen JY and Hu ZF, 2023 [20]) to 0.06 °C/km (Ren et al., 2021 [27]) to 0.04 °C/km (Yu Y et al., 2019 [33]; Wang YT et al., 2020 [10]) to 0.03–0.04 °C/km (Xing et al., 2023 [13]). Nonetheless, even moderate-intensity SST fronts significantly affect winds in the atmospheric marine boundary layer, thus being important to marine meteorology and the climate of the South China Sea (Shi R et al., 2015 [28]; Shi R et al., 2017 [68]; Shi R et al., 2022 [43]).

5. Conclusions

The high-resolution (2 km) high-frequency (hourly) SST data from 2015 to 2021 provided by the Advanced Himawari Imager (AHI) flown on the Japanese geostationary satellite Himawari-8 were used to study the China Coastal Front (CCF) in the northern South China Sea. The Belkin and O’Reilly (2009) algorithm (BOA) was used to generate monthly maps of SST gradient magnitude GM. Using 0.1 °C/km as a GM threshold, monthly maps of frontal frequency (FF) were generated. Monthly maps of SST, GM, and FF document the spatial patterns and seasonal evolution of the CCF from the Taiwan Strait to the east coast of Hainan Island. The horizontal structure of the CCF was investigated from cross-frontal distributions of SST along 11 lines (seven meridians and four parallels) that cross the northern South China Sea. The monthly distributions of SST along these lines were used to determine inshore and offshore boundaries of the CCF and calculate the CCF strength, defined as the total cross-frontal step dSST, a difference between offshore and inshore SST. The CCF strength is at its maximum in the Taiwan Strait, where wintertime dSST peaks at 9 °C. During winter, the CCF off Guangdong is a strong front, with dSST gradually decreasing westward, from 6 to 8 °C off eastern Guangdong, to 5–6 °C off western Guangdong. Farther southwest, off the east coast of Hainan Island, the CCF is much weaker, with dSST down to 2 °C even in mid-winter. The CCF in the northern South China Sea is much wider than the CCF in the East China Sea. Combined with the results of Part 1 of this study, where the CCF was documented in the East China Sea, the new results reported in this paper allowed us to trace the CCF from the East China Sea via Taiwan Strait into the northern South China Sea and farther west to the east coast of Hainan Island. In winter, the CCF is a continuous water mass front from the Yangtze Bank to Hainan Island. In summer, the CCF is discontinuous off Guangdong, where the eastern segment of the front is maintained by the monsoon-driven coastal upwelling, while the western segment of the front vanishes almost completely. The discontinuity—the break between the eastern and western segments of the front—is amplified by the Pearl River plume, which in summer extends southward across the shelf up to the 50 m isobath and beyond.