Mackerel, a sleek and fast-moving fish, is commonly found in temperate and tropical seas globally and is valued as a food source and a sport fish (Davila, 1985; Schaefer, 1986). Belonging to the family Scombridae, mackerel are carnivorous fish that consume a diet of plankton, crustaceans, mollusks, fish eggs, and small fish (Hernández and Ortega, 2000; Yoon et al., 2008). The family Scombridae encompasses over 30 different species or varieties, which are commonly known as mackerel (Cabarello et al., 2003). Mackerels can be further categorized into true mackerels (Scombrin), Spanish mackerels or sister tribes (Scomberomorini), and other types of mackerels (Food and Agriculture Organization of the United Nations, 2022a).

In the Yellow Sea, East China Sea, and South Western East Sea, the mackerel has been a significant commercial fish (Limbong et al., 1991). From the late 2010s to 2019, the yearly catch was around 120,000 tons, but by 2019, fish catches had decreased to 100,000 tons during the previous ten years, which is attributed to climate change and its impact on the marine environment.(Food and Agriculture Organization of the United Nations, 2022b). The integration of satellite remote sensing with fishery data serves as a valuable tool for identifying environmental conditions conducive to fish aggregation, forecasting potential fishing zones, and monitoring habitat changes (Santos, 2000; Klemas, 2013). These days, understanding fishing ground information or estimating the habitat suitability index (HSI) is thought to be crucial for sustainable fisheries management. In marine fisheries, the HSI approach has been widely applied to examine historical and current fish population trends, and to forecast future changes in population dynamics (Chen et al., 2009; Li et al., 2014a; Galpalsoro et al., 2009; Morris and Ball, 2006). Previous research has assessed the fishing grounds for mackerel in the coastal waters of Japan and Korea. (Lee et al., 2018; Chen et al., 2009; Wang et al., 2021; Kunimatsu et al., 2023).

To promote sustainable exploitation and effective resource management, many studies have focused on understanding the effects of climate variability on habitat suitability and its correlation with fish distribution, abundance, and catch rates (Silva et al., 2016; Tanaka and Chen, 2016). Satellite images are utilized in this study due to their numerous advantages, particularly in oceanographic research, where monitoring large and dynamic marine environments is essential (Kavanaugh et al., 2021). Satellite remote sensing offers extensive spatial and temporal coverage, allowing researchers to consistently observe ocean conditions over vast areas, which would be both challenging and costly using traditional in situ methods (Blondeau-Patissier et al., 2014). By integrating satellite data, this study gains a comprehensive understanding of oceanic conditions, which is vital for HSI modeling a core objective that seeks to link satellite-derived environmental data with marine species distributions and catch data (Mondal et al., 2023).

This research aims to develop an HSI model for mackerel based on various marine environmental parameters. By utilizing satellite-derived data on these parameters within the study area, we can quantitatively analyze the relationship between mackerel catch data and calculate the corresponding HSI values. Over 10 years, we examined mackerel catch data to identify the key environmental factors that significantly impact mackerel HSI. Using this data, a quantitative statistical model was applied to determine the most influential parameters and identify the mackerel fishing grounds in Korean waters.

2.1. Study Area

Mackerel in the northwestern Pacific Ocean is divided into the Tsushima Warm Current (TWC) and Pacific stocks. The TWC stock is primarily found in the Yellow Sea, East China Sea, and East Sea, while the Pacific stock resides along the Pacific coast of Japan and the high seas of the Northwestern Pacific Ocean (Shiraishi et al., 2008). The mackerel caught by South Korea in its surrounding seas belong to TWC stock (Lee et al., 2018).

The primary spawning areas for mackerel are situated in the coastal waters of the East Asian marginal seas and along the Pacific coast of Japan (Li et al., 2014a; 2015a; 2015b; 2014b; Sogawa et al., 2019). The suitable spawning sea surface temperature (SST) for mackerel ranges from 15 to 22°C, with the optimal SST being 18°C. The primary spawning period occurs between March and June (Takasuka et al., 2008; Yukami et al., 2009). Over the past 55 years, from 1968 to 2022, the annual mean SST in Korean waters has increased by approximately 1.36°C, while the global mean SST rose by about 0.53°C during the same period. This indicates that the long-term rate of SST increase in Korean waters was roughly 2.6 times greater than the global average. Notably, since the 2010s, the rate of SST rise during the summer has been approximately 3.9 times higher than during the winter (Han et al., 2023). In this study, we focused on the coastal waters of South Korea, specifically the area outlined by the black box in Fig. 1, ranging from latitude 31° to 40° N and longitude 122° to 132° E. This has become the primary focus because most mackerel fishing in Korean waters occurs within this black box. This region includes the mackerel fishing grounds within South Korea’s EEZ. Our objective was also to analyze the ecological characteristics of mackerel about the relevant marine environmental parameters.

Figure 1. This figure illustrates the primary spawning areas, fishing areas, and migration routes of mackerel, as well as the oceanographic structure in the Northwest Pacific Ocean. The currents shown include the Yellow Sea Warm Current (YSWC), Tsushima Warm Current (TWC), and Kuroshio Current (KC). Red solid lines indicate warm currents, respectively. Yellow and green shaded areas denote the main spawning and fishing areas of mackerel, while dotted lines with arrows depict their migration routes (modified from Wang et al., 2022). The blue line marks the South Korean Exclusive Economic Zone (EEZ), and the study area is outlined by a black box.

2.2. Mackerel Catchment Information

The mackerel catch data utilized in this research includes information such as the date and location of capture (with corresponding latitude and longitude), fishing vessel tonnage, the volume and weight of the fish caught, and details about the fishing gear used, specifically purse seine nets in both coastal and open sea areas. This data was obtained from the National Federation of Fisheries Cooperatives, a government agency in South Korea. Additionally, for this research, we filtered the data to focus only on the reporting date, capture location (with latitude and longitude), vessel tonnage, and the weight or volume of the catch from 2011 to 2019. The primary reason for using data from 2011 to 2019 is that more recent data has not yet been fully obtained. Additionally, the global COVID-19 pandemic in 2020 disrupted many sectors, including the fishing industry, with restrictions on outdoor activities leading to a decline in mackerel fishing in Korea. This period also coincided with other factors, such as weather-related restrictions and spawning season prohibitions, further limiting fishing activities.

The data reflect the recorded catches for each group, with daily catch amounts and locations. To analyze changes in the marine environment and mackerel fishing grounds from a climatological viewpoint, daily data were grouped and aggregated into monthly figures from January 2011 to December 2019. Satellite spatial data were organized with a resolution of approximately 17 × 17 km, following the standards set by the Marine Ecological Map from South Korea’s Ministry of Oceans and Fisheries. In total, 8,751 catch records were collected in the study area between 2011 and 2019. An analysis of the catch distribution (Table 1) revealed that 92.64% of the recorded catches ranged between 0 and 300 metric tons (MT), while 644 records, or 7.36% of the total, exceeded 300 MT. In this study, any catches above 300 MT were classified as outliers and adjusted to 300 MT.

Table 1 . Frequency and percentage of reported mackerel catch data from 2011 to 2019.

Catch (MT)	Frequency	Percent (%)
0 ≤ × < 300	8,107	92.64
300 ≤ × < 600	426	4.87
600 ≤ × < 900	150	1.71
900 ≤ × < 1200	68	0.78
Total	8,751	100

2.3. Satellite Data on Marine Environmental Conditions

To select appropriate environmental parameters for mackerel fishing, we considered marine variables that influence fish migration patterns and the location of fishing grounds (Yu et al., 2018; Feng et al., 2021). The selected marine environmental parameters included SST, chlorophyll-a concentration (CHL), sea surface height (SSH), and primary production (PP).

The SST data utilized in this study is sourced from The Group for High-Resolution SST (GHRSST), which supplies a range of global high-resolution SST products. These products are tailored for near real-time use, updated daily, and serve both professional operations and the broader scientific community (Minnett et al., 2019). The data is provided daily, with a spatial resolution of 10 km and an accuracy standard deviation globally exceeding 0.4 K (Donlon et al., 2012; Jha and Bhaskar, 2023; Martin et al., 2012). In this study, daily SST data was aggregated into monthly datasets, covering the period from January 2011 to December 2019.

We utilize CHL data from the Geostationary Ocean Color Imager (GOCI), one of the instruments aboard the Communication, Ocean, and Meteorological Satellite (COMS). The chlorophyll data is available through the Korea Ocean Satellite Center’s website (http://kosc.kiost.ac.kr), which provides hourly data ranging from 00:15 to 07:45 UTC with a spatial resolution of 500 meters (Jeon and Cho, 2022). After applying an atmospheric correction algorithm, the derived CHL values are obtained (Park et al., 2021; Yang et al., 2023). The hourly data is aggregated into daily and monthly datasets, based on the required time for the analysis.

The SSH data utilized in this study comes from satellite altimeter measurements, beginning with Topex/Poseidon in 1992 and later enhanced by Jason-1 in 2001 and Envisat in 2002. This data is provided by the Archiving, Validation, and Interpretation of Satellite Oceanographic (AVISO) service (Ballarotta et al., 2023; Niedzielski and Miziński, 2013). We aggregate the instantaneous sea surface height values, which capture short-term temporal variations in sea level, into monthly datasets according to the required time, with a spatial resolution of 0.25 degrees (approximately 25 km) (Taburet et al., 2021).

The PP refers to the formation of organic matter from inorganic compounds, such as carbon dioxide (CO₂) (Cullen, 2001), which is produced through photosynthesis and serves as an indicator of nutrient levels in the ocean surface layer (Lim and Jeong, 2022; Kulk et al., 2020). In this research, the Vertical General Production Model (VGPM) developed by Behrenfeld and Falkowski (1997) is used to estimate PP. VGPM has provided monthly data starting in 2003 that corresponds to the MODIS observation period, with a resolution of about 8 km (Westberry et al., 2023; Gall et al., 2024).

Due to the varying resolutions of marine environmental parameters, standardization is required to ensure that the resolution of the suitability index (SI) data matches that of the fish catch data. For satellite imagery, only successfully observed pixels are utilized. The spatial resolution of the marine environ mental variables differs, and to facilitate comparison with the catch data, the horizontal resolution of these variables was gridded to align with a uniform grid system. Using Python, we defined anchor reference lines spanning from 30 to 40°N latitude and 122 to 134°E longitude. Each grid cell has a resolution of 17 × 17 km, equivalent to 0.1667 degrees, with the assumption that there are 9 smaller grid cells within every 0.5-degree increment. This standardization ensures consistent spatial resolution across the study area for accurate analysis. Out of 8,751 total data points for mackerel catches in Korean waters, 7,853 were selected to develop the HSI model. The model is primarily based on data from 2011 to 2018, while data from 2019 is used for validation purposes.

2.4. Habitat Suitability Models

The HSI represents a habitat’s ability to support the survival of a species by incorporating key environmental parameters specific to that species. Originally developed by the US Fish and Wildlife Service (Terrell, 1985), this method produces a numerical index ranging from 0, indicating unsuitable habitat, to 1, representing optimal habitat. In this study, the satellite data was focused specifically on the EEZ of South Korea (Fig. 1) and we aim to develop models based on various marine environmental factors.

The variable X represents a marine environmental parameter, while the values a, b, c, and d are constants. There are four different combinations for calculating HSI values, utilizing various formulas and reference sources for the suitability index.

For the four models mentioned above, the likelihood and the Akaike Information Criterion (AIC) for each corresponding HSI set were calculated using Eq. (6) to decide the optimal model.

The AIC can obtain the information value of the model by using maximum likelihood estimation and the number of independent variables in several models. The k value is the number of independent environmental parameters and L is the log-likelihood estimate. Information scores from various models are calculated from the AIC test. The smallest value of AIC is shown to be better in model fit. (Bevans, 2023).

3.1. Distribution of Mackerel Catch and Analysis of Marine Environment

From Fig. 2(a), we can see a picture of mackerel fishing in Korean waters which is very close to the Korean mainland and rarely approaches the EEZ edge line except for the southern part which borders Japan. The average annual mackerel catch in the 2010s was approximately 120,000 tons (Fig. 2b), which is less than half the average mackerel catch recorded in the 1990s of 150,000 to 220,000 tons (Lee et al., 2016; Kim and Kim, 2023). Monthly average catch data showed a significant decrease from March to May (Fig. 2c). In accordance with the Decree on the Implementation of the Fishery Resources Management Law of 2016, a ban was implemented from April to June, limiting the catch of fish with a total length (TL) smaller than 21 cm, while the large purse seine fishing industry also implemented a ban fishing period from March 14 to April 14 (Kim et al., 2023). In addition, large purse seine fisheries voluntarily cease mackerel fishing for approximately 40 days during spring, when the proportion of juvenile mackerel increases (Lee et al., 2016). Fig. 3 illustrates the relationship between marine environmental parameters like SST, SSH, CHL, and PP, all of which can impact mackerel catch areas, presented through histogram visuals. The data shows the distribution of values for each parameter, particularly highlighting the range between the 10th and 90th percentiles as follows: SST ranged from 14.45 to 26.72°C, SSH from 0.61 to 0.84 m, CHL from 0.32 to 1.6 mg m^–3, and PP from 654.94 to 1.731.3 mg C m^–2 day^–1. Within these ranges, about 92% of the total mackerel catch was observed.

Figure 2. Amount caught by (a) spatial, (b) yearly, and (c) monthly distribution of the mackerel fish from 2011 to 2019 (unit: MT).

Figure 3. Histogram distributions of (a) SST ranged from 14.45 to 26.72°C, (b) PP from 654.94 to 1.731.3 mg C m^–2 day^–1, (c) SSH from 0.61 to 0.84 m, and (d) CHL from 0.32 to 1.6 mg m^–3, on the mackerel catch (x 102 MT) from 2011 to 2018.

Additionally, SST showed two significant catch peaks at 14–15°C and 24–25°C (Fig. 3a). SST is characterized by distinct seasonal variations, especially in the diverse geographical regions of Korea’s coastal waters. Therefore, instead of developing a single HSI model for mackerel, it is more effective to create separate models for each season. Supporting this approach, a previous study by Lee et al. (2018) developed HSI models for winter/spring (December–May) and summer/fall (June–November) for mackerel in Korea’s coastal waters. In this study, we aimed to enhance model performance by incorporating seasonal trends in marine environmental changes. The mackerel HSI model was developed for each season based on the spatial distribution of fishing grounds (Fig. 2a), dividing the year into winter (December –February), spring (March–May), summer (June–August), and fall (September–November).

3.2. Formulation of Seasonal HSI Model

In analyzing the HSI model for each season in South Korean waters, we categorized the SI values into four seasons based on marine environmental parameters, as illustrated in Fig. 4. The distribution of SI values for each season reveals several peaks corresponding to various environmental parameter variables, as well as differing mackerel catches across the seasons. As previously noted, mackerel are primarily caught in the winter and summer months. In winter, higher mackerel catches were linked to SST around 11 to 17°C, SSH of approximately 0.65 m, CHL of 0.86 mg m^–3, and PP of 915 mg C m^–2 day^–1. Conversely, in summer, substantial fish catches were observed at SST levels around 28°C, SSH of approximately 0.81 m, CHL of 0.49 mg m^–3, and PP of 940 mg C m^–2 day^–1.

Figure 4. Seasonal histogram distributions of suitability index (SI) for the (a) SST (°C), (b) SSH (m), (c) CHL (mg m^–3), and (d) PP (mg C m^–2 day^–1) during spring (March, April, May: red), summer (June, July, August: orange), fall (September, October, November: green), and winter (December, January, February: blue).

The SI curves were adjusted for each marine environmental variable (see Fig. 5). Table 2 shows the seasonal equations for the integrated SI model along with their statistical validation outcomes. The integrated SI model exhibits robust statistical performance, indicated by low root mean square errors and high coefficients of determination.

Figure 5. Seasonal fitted SI curves inferred from the relationship between the mackerel catch and SST (first row), SSH (second row), PP (third row), and CHL (fourth row).

Table 2 . Models for seasonal Suitability Index (SI) developed based on marine environmental factors.

Season	SI model	RMSE	R2
Spring	SI PP = 0.0049 + 0.981 * (– (PP – 1392.9505)2 / ((200.250094 * 2)2))	0.3043	0.88
	SI CHL = 0.0136 + 0.98 * (– (CHL – 0.78)2 / ((0.272 * 2)2))	0.1135	0.87
	SI SSH = 0.0048 + 0.98 * (– (SSH – 0.664)2 / ((0.081 * 2)2))	0.3154	0.73
	SI SST = 0.0081 + 0.98 * (– (SST – 15.76)2 / ((1.59 * 2)2))	0.1346	0.84
Summer	SI PP = 0.0049 + 0.991 * (– (PP – 1039.9505)2 / ((200.250094 * 2)2))	0.2954	0.87
	SI CHL = 0.00361 + 0.99 * (– (CHL – 0.47)2 / ((0.18 * 2)2))	0.2311	0.91
	SI SSH = 0.0031 + 0.97 * (– (SSH – 0.807)2 / ((0.056 * 2)2))	0.2988	0.75
	SI SST = 0.0082 + 0.98 * (– (SST – 27.73)2 / ((1.42 * 2)2))	0.1125	0.85
Fall	SI PP = 0.0049 + 0.992 * (– (PP – 1467.9505)2 / ((280.250094 * 2)2)) + 0.0005 + 0.59 * (– (PP – 2367.9505)2 / ((150.250094 * 2)2))	0.3363	0.89
	SI CHL = 0.00089 + 0.98 * (– (CHL – 2.36)2 / ((0.3 * 2)2)) + 0.00089 + 0.55 *(– (CHL – 1.05)2 / ((0.38 * 2)2))	0.2445	0.83
	SI SSH = 0.0087 + 0.98 * (– (SSH – 0.75)2 / ((0.0641 * 2)2))	0.3256	0.79
	SI SST = 0.0083 + 0.98 * (– (SST – 17.27)2 / ((2.22 * 2)2))	0.1134	0.83
Winter	SI PP = 0.0049 + 0.991 * (– (PP – 889.9505)2 / ((200.250094 * 2)2))	0.2536	0.85
	SI CHL = 0.0015 + 0.98 * (– (CHL – 0.87)2 / ((0.319 * 2)2))	0.1453	0.81
	SI SSH = 0.0065 + 0.98 * (– (SSH – 0.69)2 / ((0.042 * 2)2))	0.1893	0.82
	SI SST = 0.0005 + 0.98 * (– (SST – 11.79)2 / ((1.23 * 2)2)) + 0.0006 + 0.98 * (– (SST – 17.23)2 / ((1.51 * 2)2))	0.1145	0.80

The HSI value is derived by aggregating the SI for each environmental parameter over the entire season. This study identifies CHL as the most influential factor for mackerel fish, with an AIC value of 71,373. This is followed by SSH at 78,603, SST at 89,152, and finally, PP with the highest AIC value of 95,902. When considering multiple models that incorporate all variables, the Continued Product Model (CPM) has an AIC value of 68,490, while the MINM scores 126,895, the AMM achieves 57,894, and the GMM has an AIC value of 111,691. These results indicate that the AMM is the most suitable for calculating HSI for mackerel in the South Korean region. This finding aligns with prior research with different species by Lee et al. (2023), which demonstrated that the AMM effectively calculates mean values, in contrast to CPM, MINM, and GMM, which are significantly influenced by outliers (Chen et al., 2009).

From the results observed, the optimal HSI model for mackerel habitat in Korean waters is the AMM, which integrates the Suitability Index for environmental parameters SST, SSH, CHL, and PP. Using the HSI from the AMM model, we calculated the monthly HSI for 2019 and compared it with mackerel catch data (Figs. 6a, b). The AMM HSI model defines spring from March to May, summer from June to August, fall from September to November, and winter from December to February. A higher HSI value denotes a more favorable habitat for mackerel, indicating the potential to predict enhanced catch outcomes.

Figure 6. Distribution maps of (a) monthly mackerel catch and (b) monthly HSI in 2019, derived from the arithmetic mean model using SST, SSH, CHL, and PP in the South Korean EEZ.

Fig. 6(b) shows areas with HSI values above 0.6, corresponding to the highest catches, particularly in the Yellow Sea during the summer months (June, July, and August) and extending into early fall in September. The HSI values drop in October, likely due to mackerel migration, with dominant values falling below 0.4. The HSI then increases again in August and remains steady through the winter months (December, January, and February) until early spring in March, with some fishing grounds in the southern region displaying HSI values between 0.4 and 0.6.

Despite the AMM HSI results indicating a low value of 0.4, mackerel catches were still recorded in the East Sea during April, May, and October. During these months, the marine environment remains similar due to limited ocean surface heating from solar radiation. However, the large purse seine fishery in Korea faces a closed season, with a ban on mackerel landings for one month, usually between April and June (Owiredu et al., 2024), which likely accounts for the low HSI in May. Additionally, Lee et al.(2018) report that mackerel migrate northward in October, further contributing to the decline in HSI values in Korean waters, dropping below 0.4.

Mackerel form spawning grounds along the coasts of the Yellow Sea and South Sea in the spring, then migrate to the East Sea and surrounding South Korean waters by November. During the winter, they return to the South Sea. The relationship between marine environmental factors (SST, SSH, CHL, PP) and mackerel catches was analyzed, identifying optimal conditions: SST between 14.45 and 26.72°C, SSH from 0.61 to 0.84 m, CHL ranging from 0.32 to 1.6 mg m^–3, and PP between 654.94 and 1,731.3 mg C m^–2 day^–1. Considering the seasonal variation of these factors, especially SST, we developed an HSI model for each season, categorizing the year into winter (December–February), spring (March–May), summer (June–August), and fall (September–November), using data from 2011 to 2018.

The HSI was calculated using four methods (CPM, MINM, AMM, GMM) and evaluated for performance. The AMM method, utilizing all four marine environmental variables, showed optimal performance. A comparative analysis using the AMM-based HSI and the monthly catch data for 2019 confirmed that mackerel fishing grounds varied by season and that high HSI values (> 0.6) corresponded to high catch areas. However, discrepancies were noted in April, May, and October, where the HSI was low (< 0.4) despite significant catches. In summary, while this study primarily focuses on physical marine environmental variables (SST, CHL, SSH, PP) for assessing mackerel habitat suitability, future research could benefit from incorporating the ecological characteristics of mackerel, such as their feeding patterns and spawning behavior, to refine habitat models. Additionally, investigating how climate change impacts habitat suitability and distribution over time would provide valuable insights for sustainable fisheries management.

This study received funding from the Korea Institute of Marine Science & Technology (KIMST) promotion, supported by the Ministry of Oceans and Fisheries (RS-2022-KS221660).

No potential conflict of interest relevant to this article was reported.