Introduction

Rising ocean temperatures are considered the primary driving factor behind widespread poleward shifts in the distribution of many marine fish species in the Northern Hemisphere (Pinksy et al. 2013; Fossheim et al. 2015; Cheung et al. 2010; Jørgensen et al. 2019; Fauchald et al. 2021). As poikilotherms, fish are sensitive to changes in ambient temperature, so range shifts may occur when they track movements in preferred thermal habitat (Sunday et al. 2012; Melbourne-Thomas et al. 2022). Among the fishes growing more abundant farther north in recent years are groundfish in the Bering Sea, including some with high commercial value, such as walleye pollock (Gadus chalcogrammus) and Pacific cod (G. macrocephalus) (Stevenson and Lauth 2019; Spies et al. 2020; Baker 2021; Rooper et al. 2021). Walleye pollock supports the largest U.S. fishery in terms of volume, and in 2021, 1.4 million metric tons of pollock worth over $371 million USD were caught in U.S. waters of the Bering Sea, Aleutian Islands, and Gulf of Alaska. In the same year, Pink salmon (Oncorhynchus gorbuscha) and Pacific cod accounted for the second- and third-largest fisheries in the region, with a combined catch of 355 thousand metric tons worth over $116 million USD (National Oceanic and Atmospheric Administration [NOAA] 2023a).

Most groundfish harvested in the Bering Sea are collected over the Bering Sea shelf in the Eastern Bering Sea (EBS) (54°N to 60°N). The shelf is a wide, relatively shallow (< 200 m) habitat, where oceanographic processes, such as tidal mixing and ice-induced stratification, support highly productive pelagic and demersal food webs and commercial fishing (Stabeno et al. 2001; Fig. 1). Recent studies indicate that walleye pollock, Pacific cod, and other commercial species often caught in the EBS have become more abundant in the Northern Bering Sea (NBS) (60°N to 66°N) in recent years, especially during periods when winter sea-ice coverage was low and bottom temperatures were high relative to long-term means (Stevenson and Lauth 2019; Siddon et al. 2020; Thoman et al. 2020). These conditions are likely to be amplified by long-term climate change, and poleward movements in some Bering Sea groundfish are projected to continue into the next century (Cheung et al. 2010; Fauchald et al. 2021). If so, commercial fishing efforts may undergo a dramatic redistribution in the Bering Sea as fishing vessels travel northward, in effect, following the fish. Such changes could have serious implications for those harvesting fish in the region and for those responsible for fisheries management, law enforcement, and emergency management, as well as for local communities with economies tied to the commercial fishing industry (Rogers et al. 2019; Hollowed et al. 2024).

Fig. 1
figure 1

A Study area, including the Bering Sea, Chukchi Sea, and Beaufort Sea (Esri 2022), B depth contours across the Bering Sea Shelf, and C the 50 × 50 km grid pattern defining cells used in the emerging hot spot and emerging cold spot analyses. Bathymetric data are from the U.S. Geological Survey and provided by the Alaska Ocean Observing System (2023)

Full size image

In this study, we investigated whether commercial fishing vessels in the Bering Sea have shifted their distribution poleward in recent years in accordance with previously documented shifts in heavily targeted stocks over the last decade (Stevenson and Lauth 2019; Fedewa et al. 2020; Spies et al. 2020; Rooper et al. 2021). We characterized temporal trends in the geographic distribution of commercial fishing vessels over the Bering Sea shelf from 2013 to 2022 by using emerging hot spot and cold spot analyses to identify areas where fishing vessel activity increased (emerging hot spots) and where it decreased (emerging cold spots) over the last ten years. Our hypothesis was that fishing vessels had shifted poleward, leading to emerging hot spots at the northern edge of the vessel operating range and emerging cold spots at the southern edge.

We recognize that factors in addition to fish distribution probably influence whether vessels track targeted stocks. For example, sea ice can limit a vessel’s northward movements during winter, and closed areas will restrict fishing activity in certain locations. Fuel costs could also make travel unprofitable, especially if harvests remain productive in the south, closer to ports and other coastal infrastructure. Global events may also play a role in the spatial dynamics of commercial fishing activities, with the global coronavirus pandemic and associated impacts on the workforce serving as one recent example (Love et al. 2021; March et al. 2021). This study does not attempt to tease apart these conflicting demands on the commercial fishing industry. Rather, it investigates whether poleward shifts in vessel distribution have taken place nonetheless.

Methods

This study was conducted using Vessel Management System (VMS) data generated by commercial fishing vessels inside the U.S. Exclusive Economic Zone (EEZ) within the Bering Sea over the Bering Sea Shelf (54–72°N and 140–179°W; Fig. 1). The VMS is a satellite-based communication system operated by the National Marine Fisheries Service of the National Oceanic and Atmospheric Administration (NOAA). Developed primarily to support law enforcement efforts associated with commercial fishing in the nation’s EEZ and treaty areas, VMS is required on fishing vessels operating in certain Alaska fisheries. The system involves on-board transceivers that transmit the vessel’s location (latitude–longitude) and other information (e.g., speed, heading) in a digital report to satellite sensors. Reports are then transmitted to ground stations where they are made available to the end user. The required frequency of transmission is one report per hour, increasing to one every 10 min when the vessel is near an environmentally sensitive area.

Study site

We focused this analysis on commercial fishing vessels operating from January 1, 2013, to December 31, 2022, within a 1 million-km2 region of U.S. federal waters over the Bering Sea Shelf in the EBS (54°–60°N) and NBS (60°–66°N). The study site also included the Chukchi Sea and Beaufort Sea as far north as the continental shelf break but not beyond (66°–72°N, Fig. 1).

The Bering Sea shelf includes three major oceanographic domains: the inner domain (depth < 50 m), middle domain (depth = 50–100 m), and outer domain (depth = 100–200 m). The heaviest fishing vessel traffic typically occurs in the outer domain, adjacent to the shelf break separating the shelf from the deep Aleutian Basin. In contrast, commercial fishing is prohibited in federal waters north of the Bering Strait, which includes the Arctic Management Area, spanning the Chukchi Sea west to the U.S.–Russia maritime boundary and the Beaufort Sea east to the U.S.-Canada maritime boundary (NOAA 2009; North Pacific Fishery Management Council 2009).

Commercial fishing vessels in the Bering Sea required to carry VMS transceivers include those harvesting walleye pollock, Pacific cod, Atka mackerel (Pleurogrammus monopterygius), sablefish (Anoplopoma fimbria), Pacific halibut (Hippoglossus stenolepis), and various crabs, including Paralithodes spp., Lithodes aequispinus, Chionoecetes bairdi, and C. opilio. We used the frequency of VMS report transmission as a proxy for commercial fishing vessel activity, assuming that decisions made by vessel operators about when, where, and how frequently to transmit reports in a particular location were similar from year to year.

Data analysis

Unprocessed VMS reports (hereafter, reports) were uploaded into ArcGIS Pro 3.0.3 (Esri 2023), and duplicate reports and reports from non-VMS transceivers (Automatic Identification System transceivers) were removed. Because our goal was to evaluate large-scale shifts in the distribution of the commercial fishing fleet as a whole, we included reports from all vessels at the study site, regardless of whether they were setting, trawling, or hauling gear, remaining stationary, or transiting between stations. In other words, we did not distinguish between reports generated by ships underway at certain speeds, using particular types of gear (e.g., trawling vs. long-lining), or targeting certain stocks (Lee et al. 2010). Rather, we investigated geographic shifts in the entire commercial fishing fleet, inclusive of all behaviors involved in the business of commercial fishing.

Most commercial fishing in the Bering Sea occurs during established open seasons, so we expected the number of reports to fluctuate seasonally. Moreover, sea-ice coverage in the Bering Sea is highly seasonal, which could limit vessel movements in the NBS during winter since sea ice is a navigational hazard for most fishing vessels. We tested the assumption that vessel activity was highly seasonal by plotting the number of vessels (V) operating and the number of reports they generated (R) each month, separately for each year of the study. We found substantial month-to-month variability in both V and R and a significant correlation between the two parameters (see Results). Therefore, we stratified the spatial analysis of interannual trends in vessel activity, i.e., emerging hot/cold spot analyses, by month and selected R as the dependent variable. We felt that R was a better indicator of fishing vessel activity in a given region than V because R is a function of both V and the amount of time that vessels spent in the region.

Emerging hot spot/cold spot analyses were used to identify areas of the Bering Sea where, between 2013 and 2022, commercial fishing vessel activity significantly increased (emerging hot spots) or significantly decreased (emerging cold spots). This type of analysis requires data in the form of a space–time cube illustrated in Fig. 2. The dimensions of the cube were degrees longitude (x axis), degrees latitude (y axis), and time in years (z axis; 2013–2022). Because the analysis was stratified by month, we created 12 separate space-time cubes, one for each month of the year.

Fig. 2
figure 2

Conceptual diagram of a space-time cube. In this study, the response variable was the no. VMS reports converted to a relative frequency for monthly analysis

Full size image

Spatial bins were created in ArcGIS Pro, where we applied a 50 × 50 km grid pattern across the study site (n = 575 cells) and clipped it to the contours of the 200-m isobath. This spatial resolution was small enough to capture differences in vessel activity among cells in regions of high vessel density but large enough for differences between cells to represent meaningful shifts in vessel distribution over the ocean landscape. We then summed R in each grid cell (i) in each month (j) and each year (k), yielding Rijk as the dependent variable.

To account for unequal sample sizes each year due to interannual fluctuations in V and R, we converted Rijk to a relative frequency:

$$R_{ijk - rel} = \, \left( {R_{ijk} / \, R_{ijk - tot} } \right) \times 100,$$

where Rijk−rel = relative frequency of reports in cell i during month j in year k, Rijk = number of reports in cell i during month j in year k, and Rijk−tot = total number of reports in month j during year k.

Mann–Kendall trend tests were conducted in MATLAB (The MathWorks 2021) to detect significant upward or downward interannual trends in Rijk−rel over the 10-year time period. A separate test was run for each of the 575 columns in each space-time cube (Fig. 2). This trend test is a non-parametric tool widely used for detecting trends in time series data and measuring directional monotonic trends (positive or negative) in values in an ordered sequence (Mann 1945; Kendall 1975). Here, the sequence consisted of Rijk−rel values in chronological order from 2013 to 2022.

We visualized the results in ArcGIS Pro, where we uploaded the p values for each cell onto a map of the study site. Significant positive trends in vessel activity (p ≤ 0.05) were emerging hot spots and were coded dark red. Significant negative trends (p ≤ 0.05) were emerging cold spots and were coded dark blue. To visualize uncertainty around the significant trends, we also identified cells where 0.05 < p ≤ 0.10. In these cases, weaker positive trends were emerging warm spots and were coded light red; weaker negative trends were emerging cool spots and were coded light blue. These warm/cool trends are included in the map visualizations; however, because they were not statistically significant, they were not included in subsequent analyses.

We evaluated the extent to which emerging hot spots occurred to the north of emerging cold spots by plotting the latitude–longitude coordinates of the centroid of each column from the space-time cube that contained an emerging hot spot and, separately, an emerging cold spot. The coordinates were plotted on x–y axes, where x was longitude and y latitude, both in decimal degrees. In SAS® On Demand software (SAS Institute Inc. 2022), we calculated the 95% prediction ellipse around the coordinates of emerging hot spots and, separately, emerging cold spots for all months combined. The center of the prediction ellipse represents the mean latitude and mean longitude of the sample for which it was calculated. The shape of the ellipse is a function of variability around the bivariate mean.

Before comparing the geographic range of emerging hot spots with that of emerging cold spots, we used Shapiro–Wilk’s (W) tests to evaluate whether latitude and longitude were normally distributed in each group (Shapiro and Wilk 1965). They were not (cold: latitude W = 0.96, p = 0.001, n = 139; longitude W = 0.97, p = 0.002, n = 139; hot: latitude W = 0.97, p = 0.002, n = 131, longitude W = 0.95, p < 0.0001, n = 131), and log transformation did not resolve the distribution. Therefore, we used two-sample Wilcoxon Rank-Sum tests to evaluate differences between the latitude and, separately, the longitude, of emerging hot spots and emerging cold spots.

To characterize spatial trends in emerging hot spots and emerging cold spots during the year, we calculated the geographic centroid of each group separately for each month. We then combined data from consecutive months with similar spatial patterns. This yielded four time periods of interest: October–February, April–May, June–August, and September (see Results). We used two-sample Wilcoxon Rank-Sum tests to evaluate differences between the latitude and, separately, the longitude, of emerging hot spots and emerging cold spots in each of these four time periods. This part of the analysis did not account for non-linear relationships between spatial coordinates due to Earth’s curvature. Instead, these tests were used to compare values of latitude and longitude of the two-sample populations plotted on a comparable linear scale.

While meaningful shifts in vessel distribution were possible in any compass direction, we defined poleward range shifts as those where emerging hot spots occurred at higher latitudes (°N) than emerging cold spots. In addition to the statistical analyses previously described, we summarized these differences by calculating the distance (km) and bearing (°) by which the center of the range of emerging hot spots was offset from that of emerging cold spots, separately for each month.

Results

We analyzed over 13.8 million VMS reports generated by 447 commercial fishing vessels operating in the Bering Sea between January 1, 2013, and December 31, 2022 (Table 1). Reports were available in continuous 1-min intervals during this time, except for two time periods when no reports were available: September 1–October 9, 2017, and March 10–May 19, 2019 (Table 1). Commercial fishing vessels were present in the study site during all months of the year, mainly in waters over the Bering Sea shelf south of Bering Strait. Five commercial fishing vessels were underway in the Chukchi Sea and Beaufort Sea during the months of July–October. They appeared to be transiting between coastal communities or supporting survey work as opposed to actively fishing.

Table 1 Monthly total number of fishing vessels observed by the vessel monitoring system over the Bering Sea shelf from January 1, 2013, to December 31, 2022
Full size table

Seasonal (monthly) trends in the number of vessels (V) were generally consistent from year to year (Fig. 3a). Vessels were most numerous in January–March, becoming less numerous in April–May. Numbers rose to a second, smaller peak in June–August before declining in September–December (Fig. 3a). The number of reports generated by these vessels (R) showed similar seasonal trends (Fig. 3b), and the two parameters were significantly correlated (Pearson r = 0.81, p < 0.0001, n = 115). Despite consistent month-to-month trends, V was highly variable from year to year (Fig. 3a). For example, monthly values of V ranged outside the 95% C.I. around the monthly mean in June 2014, October–December 2014, March 2015, November–December 2015, February 2016, August 2017, November 2017, and January 2018 (Fig. 3a).

Fig. 3 
figure 3

a Number of fishing vessels per month and b number of reports per month between January 1, 2013, and December 31, 2022, except for September 2017, October 2017, and March–May 2019 because of incomplete data during those months. Gray shading is the 95% C.I. around the monthly mean

Full size image

Figure 4 shows maps of the spatial distribution of the total number of reports produced by fishing vessels each month, all years combined. It also shows the relevant monthly median sea-ice extent in each year (2013–2022). Sea-ice data are available from the National Snow and Ice Data Center (https://noaadata.apps.nsidc.org/NOAA/). Reports occurred in the highest number during January–March, when vessels were only present in the southern portion of the EBS. During this time, sea ice covered much of the NBS. Fishing vessels were observed in both the EBS and NBS in June–October, when sea ice was largely absent from the NBS (Fig. 4).

Fig. 4
figure 4

Maps of the total number of reports generated across all years for the months of a January, b February, c March, d April, e May, f June, g July, h August i September, j October, k November, and l December. In all months, n = 10 years, except April (n = 9 years) and September (n = 9 years)

Full size image

Emerging hot/cold spots

Overall, we found 131 emerging hot spots of vessel activity and 139 emerging cold spots (Table 2). Figure 5 shows an example of the trends that characterized emerging hot spots, warm spots, cool spots, and cold spots in the Bering Sea. Emerging hot spots and emerging cold spots were distributed across the Bering Sea shelf, south of the Bering Strait. No significant trends in vessel activity were observed in the Chukchi Sea or Beaufort Sea.

Table 2 Coordinates (decimal degrees) of the centroid of the 95% prediction ellipse for emerging hot spots and emerging cold spots of vessel activity
Full size table
Fig. 5
figure 5

Relative number of reports generated in cells identified as emerging hot spots (p ≤ 0.05), emerging warm spots (0.05 < p ≤ 0.10), emerging cool spots (p ≤ 0.05), and emerging cold spots (0.05 < p ≤ 0.10) during the month of January (2013–2022)

Full size image

Overall, emerging hot spots (n = 131) occurred at higher latitudes in the Bering Sea than did emerging cold spots (n = 139; Wilcoxon Rank-Sum Z = 6.97, p < 0.0001). Specifically, the center of the geographic range of emerging hot spots was located 230 km directly north (001°) of emerging cold spots (Fig. 6, Table 2). The only significant trends in vessel activity observed at latitudes > 60°N were emerging hots spots (Fig. 6). Emerging cold spots, on the other hand, were found only in the southern portion of the operating range in the EBS (< 60°N; Fig. 6). Emerging hot spots also extended into these lower latitudes (Fig. 6).

Fig. 6
figure 6

The 95% prediction ellipse around emerging hots spots (n = 131) and emerging cold spots (n = 139), all months combined

Full size image

When we considered each month separately, we found a strong seasonal component to the pattern of emerging hot spots occurring to the north of emerging cold spots. This pattern is shown graphically in Fig. 7. White cells are locations where reports were never recorded over the 10-year study period. Dark gray cells are locations where at least one report was generated but no significant temporal trend in vessel activity was detected. Dark gray cells show that the operating range of commercial fishing vessels was restricted to the EBS in January–April but expanded into the NBS in May–December. Cells in dark red and dark blue are locations where reports not only occurred but also where the relative number of reports significantly increased or decreased between 2013 and 2022.

Fig. 7
figure 7

Maps of emerging hot spots, emerging warm spots, emerging cool spots, and emerging cold spots in the months of a January, b February, c March, d April, e May, f June, g July, h August i September, j October, k November, and l December

Full size image

In October, November, December, January, and February, emerging hot spots of vessel activity occurred at significantly higher latitudes and lower longitudes than did emerging cold spots (Fig. 7a, b, j–l; Table 3). The center of the geographic range of emerging hot spots was 230–387 km to the NNW (315°–002°) of that of emerging cold spots (Figs. 8a, 9). Emerging hot spots typically occurred in the middle domain of the EBS, while emerging cold spots occurred in the outer domain, adjacent to the shelf break. Their geographical ranges were not spatially distinct and overlapped in the middle latitudes of the vessel operating range. Nevertheless, patterns suggest that vessel activity in the southern edge of the EBS during these months generally declined over the years, while vessel activity in the northern and western part of the EBS increased (Figs. 7a and b, 8a).

Table 3 Results of two-sample Wilcoxon Rank-Sum test comparing the geographic coordinates (decimal degrees) of the 95% prediction ellipse for emerging hot spots and emerging cold spots of vessel activity
Full size table
Fig. 8
figure 8

The 95% prediction ellipse around emerging hots spots and emerging cold spots observed in a October–February, b April–May, c June–August, and d September

Full size image
Fig. 9
figure 9

Distance (km) and bearing of the center of the prediction ellipse for emerging hot spots (coordinates variable) relative to the center of the prediction ellipse for emerging cold spots (coordinates 0,0), separately for each month

Full size image

In March, the only significant trend was a single region of declining vessel activity (Fig. 7c). In the subsequent two months, April (Fig. 7d) and May (Fig. 7e), emerging hot spots and emerging cold spots occurred across the shelf in both the middle and outer domains of the EBS. Apart from a 150 km2 area of increasing vessel activity over the western region of the EBS in April, there was no strong spatial distinction between the geographical ranges of emerging hot spots and emerging cold spots (Fig. 8b).

Emerging hot spots were again prevalent north of emerging cold spots in June (Fig. 7f), July (Fig. 7g), and August (Fig. 7h; Table 3). Emerging hot spots were especially widespread in the NBS in July and August, ranging up to St. Lawrence Island and Norton Sound in July (Fig. 7g) and as far north as the Bering Strait and the U.S.–Russia maritime boundary in August (Fig. 7h). During August, the geographical range of emerging hot spots was centered 515 km to the N (346°) of the geographical range of emerging cold spots (Fig. 9). In contrast, emerging cold spots were mainly located in the central EBS in July and near the southern shelf break in August (Fig. 7g, h). These patterns represented significant differences in the range of emerging hot spots and cold spots in June–August (Table 3), even though some emerging hot spots were present in the south, in Bristol Bay, during all three months (Fig. 7g, h).

September was an unusual month for long-term trends in vessel activity. Emerging hot spots appeared in the southernmost portion of the EBS near the Aleutian Islands (Fig. 7i), at lower latitudes and lower longitudes than emerging cold spots (Table 3). The center of the geographical range of emerging hot spots in September was 443 km to the ESE (108°) of the center of the range of emerging cold spots (Figs. 8d, 9).

Discussion

This study investigated whether commercial fishing vessels in the Bering Sea shifted to higher latitudes over the last 10 years, concurrent with documented poleward range shifts in highly valuable groundfish, including adult walleye pollock and Pacific cod (Spies et al. 2020; Stevenson and Lauth 2019). Recent studies suggest that juvenile walleye pollock and other demersal species in the region, such as northern rock sole (Lepidopsetta polyxystra) and adult male snow crabs, have also become more abundant at higher latitudes (Stevenson and Lauth 2019; Fedewa et al. 2020; Levine et al. 2023).

Geographic range shifts in Bering Sea groundfish and some other demersal species are often linked to regional warming trends that affect weather, currents, and, ultimately, sea-ice formation (Eisner et al. 2020; Siddon et al. 2020; Clement Kinney et al. 2022). When sea ice forms, a series of physical processes produce a persistent pool of cold (< 2 °C) water near the ocean floor, sometimes reaching into the EBS as far south as the Alaska Peninsula. This “cold pool” serves as a thermal barrier to temperate fish, including adult walleye pollock and Pacific cod. However, when the water column is relatively warm, sea-ice coverage at lower latitudes is reduced and the cold pool recedes north, allowing species to move poleward without leaving their thermal niche (Wyllie-Echeverria and Wooster 1998; Clement Kinney et al. 2022).

Since 2014, the Bering/Chukchi Sea region has been relatively warm, even when compared to other multiyear periods of warm ocean temperatures in the long-term record (1979–2023). Bottom temperatures were especially warm from 2017 to 2020, with the winter of 2017/2018 having the lowest maximum sea ice extent in nearly 40 years (Duffy-Anderson et al. 2017, 2019; Stabeno et al. 2019; Huntington et al. 2020; Siddon et al. 2020). That year, the cold pool was “almost non-existent” in the Bering Sea (Stabeno and Bell 2019), with cold (< 2 °C) bottom water only present in a small region of the NBS, west of St. Lawrence Island (Baker 2021). According to previous studies, adult pollock biomass in the NBS during summer/fall of 2017 was 65× greater than it was in 2010, when water temperatures were relatively cold (Stevenson and Lauth 2019). Adult Pacific cod biomass in the NBS was 10× greater in 2017 than in 2010. Densities of adult walleye pollock and Pacific cod in the NBS remained relatively high through at least the next 2 years (2018 and 2019; Eisner et al. 2020; Baker 2021). Other authors also report that conditions in the Bering Sea remained relatively warm until 2021, when bottom water cooled and groundfish shifted south and/or to deeper waters. In 2022, sea surface temperatures and the spatial extent of the cold pool in the EBS resembled average conditions in general, although temperatures over the shelf were above average in summer months (Siddon 2022).

In this study, we found evidence of a poleward shift in commercial fishing vessel activity in the Bering Sea over the last 10 years, but only during certain times of year. Vessel activity increased in the higher latitudes, including those of the NBS, in October–February and July–August and decreased in the south. In both periods, the geographical range of emerging hot spots, where vessel activity increased, was centered 204–515 km to the N, NE, or NW (depending upon month) of the center of the geographical range of emerging cold spots, where vessel activity decreased. Moreover, during both periods, the northern portion of the operating range, which included the NBS, contained only emerging hot spots and no emerging cold spots. Emerging hot spots were especially widespread in the NBS during August, appearing as far north as Norton Sound, St. Matthew Island, St. Lawrence Island, and the Bering Strait, as well as west to the U.S.–Russia maritime boundary. When we visually compared emerging hot spots in August to the location of the cold pool in June–August in 2019 and 2021 as reported by Rohan et al. (2022), we found that recent increases in vessel activity coincided with the outer edge of the cold pool, particularly just south of the coldest (< 0 °C) bottom temperatures. These areas were located southwest of St. Lawrence Island, contiguous with the U.S.–Russia maritime boundary. In prior years (2010, 2017), the cold pool covered a greater area of the Bering Sea, extending as far south as the Alaska Peninsula (Rohan et al. 2022; NOAA 2023b).

Collectively, these observations align with the hypothesis that changes over time in bottom water temperatures may impact the distribution of commercial fishing vessels in the Bering Sea by influencing the distribution of groundfish, at least during summer. However, more information is needed to evaluate this relationship. After all, reasons for poleward shifts being limited to October–February and July–August are likely complex, since not only are biological responses to climate change and other environmental stressors species-specific, but numerous factors in addition to groundfish distribution can influence decisions that commercial operators make about when and where to focus their fishing effort. Such constraints may include sea conditions, fuel costs, vessel capabilities, crew schedules, catch regulations, and area closures (see Haynie and Pfeiffer 2013; Watson and Haynie 2018). We did not attempt to analyze factors playing a role in the poleward shifts we observed, but we do note that the timing of poleward shifts in vessel distribution included the busiest fishing seasons in the Bering Sea. According to the VMS dataset, these were January–March and July–August, when both the number of vessels operating in the study site and the number of reports generated peaked each year.

These months also coincided with the two annual open seasons for walleye pollock: “A” Season and “B” Season. The “A” Season takes place in late January–April, the latter part of which includes the beginning of the spawning season for walleye pollock. Typically, spawning over the shelf begins in March–April, near the eastern Aleutian Islands (Unimak Island) and the Alaska Peninsula, later shifting north to the Pribilof Islands in late spring/summer (Bachelor et al. 2010). We observed greater vessel activity > 800 km NW of the eastern Aleutian Islands in January and 300 km ENE of the Pribilof Islands in May, making it seem unlikely that those vessels were associated with the winter pollock fishery. On the other hand, Eisner et al (2020) investigated spatial relationships between adult walleye pollock and environmental conditions in the Bering Sea in 2010 and 2017–2019 and hypothesized that warm bottom temperatures in 2017 and 2018 caused pollock to overwinter farther NE and NW of traditional spawning sites.

The “B” Season for walleye pollock usually takes place in June–October, when adult pollock disperse across the shelf to feed on other fish (Ianelli et al. 2021; Stevenson et al. 2022). As noted above, we observed the strongest evidence for a poleward shift in commercial fishing vessel activity during July–August, with emerging hot spots distributed across the northern EBS and NBS. Because sea ice is absent from the Bering Sea during July–August, it is possible that vessels were able to track poleward-shifting groundfish distributions more closely during this time of year, whereas poleward movements during other times of year would be limited by sea ice. During July–August, some fishing vessels were underway as far north as the Chukchi Sea, although commercial fishing is prohibited in waters north of the Bering Strait, setting a northern limit to commercial harvest.

In contrast to the idea that observed patterns were driven by the walleye pollock fishery alone, we also found northward shifts outside the open seasons for pollock, during October–December. Field surveys and genetic studies indicate that Pacific cod are among the species whose range has shifted poleward in recent years. Pacific cod in the Bering Sea are caught from late November to January, so it is possible that vessels targeting cod may have also moved north in recent years (Stevenson and Lauth 2019; Spies et al. 2020). Fisheries for other demersal species, such as sablefish, snow crab,Footnote 1 and tanner crab (Chionoecetes bairdi), are also open during the time when vessel activity increased in northern regions, underscoring the need for more detailed information about the correlation between fishing vessels and targeted fish populations to better understand the relative contribution of shifting fish distributions on commercial fishing effort in the Bering Sea.

Such information might also help to explain observations in September, when fishing vessel distribution showed a pattern opposite that of other months. In September, emerging hot spots occurred in the southern range of the study site, near the eastern Aleutian Islands, while emerging cold spots occurred to the N and NW. The VMS dataset does not contain information on fish species targeted by vessels, but September generally marks the beginning of the open season for king crab (Paralithodes spp.), sablefish, and Pacific cod caught with pot gear. September is also the beginning of the open season for Atka mackerel (NOAA 2024). If these vessels were targeting species at the start of an open season, our findings underscore the potential for management actions to influence the distribution of fishing vessels in the Bering Sea, even while climate change drives large-scale shifts in fish distribution.

Poleward shifts in vessel distribution like those we observed in October–February and July–August, regardless of their cause, may have implications for individuals working in the fishing industry and for those managing regional fisheries. Elevated harvests in northern regions will require fisheries managers to adjust population models to account for higher catch rates in the NBS and lower catch rates in the south, especially during summer (e.g., July–August) according to our results. Novel fishing pressure in the NBS may also prompt the need for new place-based protections or seasonal regulations to protect harvested populations and subarctic ecosystems. Poleward shifts in fishing effort during the busiest fishing seasons in the Bering Sea (January–March, July–August) can also have implications for maritime safety. Commercial fishing is one of the most hazardous occupations in the nation, with vessel disasters (sinking and capsizing) accounting for most fishing-related fatalities in U.S. waters (Lucas and Case 2017; Case and Lucas 2020). If fishing vessels continue to move northward in the Bering Sea, farther from permanently stationed emergency response assets, work may become even more risky. In this case, emergency managers in surrounding coastal communities may need to reevaluate emergency plans with respect to search and rescue capabilities and marine pollution response operations, especially in the summer months of July and August, when poleward shifts in fishing vessel traffic were most dramatic.

Climate-based species distribution models for marine fishes in the Bering Sea and elsewhere suggest that borealization of subarctic and Arctic fish communities is likely to continue into the next century (Cheung et al. 2010; Rooper et al. 2021). How far northward will commercial fishing vessels travel in response to changing environmental conditions? Current physical, economic, or legal limits on commercial fishing in the ocean’s highest latitudes may continue to define the northern extent of fishing vessel distribution; however, they, too, may change in time. For example, sea-ice coverage prevents most fishing vessels from accessing the NBS during winter, yet shelf areas of the Pacific Arctic  are forecast to be without ice in summer months within the next 25–30 years (Årthun et al. 2021), potentially extending the navigable season in the NBS into later autumn. Logistical and economic constraints, such as travel time and fuel costs, may limit profitability of fishing in the northernmost regions of the Bering Sea, but the industry may adapt to new conditions by increasing the use of catcher/processor vessels in the polar regions (Watson and Haynie 2018). Policy-based restrictions also play an important role. For example, bottom trawling is currently banned in the Northern Bering Sea Climate Resilience Area, and all commercial fishing is prohibited in federal waters north of the Bering Strait (North Pacific Fishery Management Council 2009). These protections would reduce the likelihood that poleward shifts in vessel distribution will extend into the Chukchi Sea as the climate continues to warm, although such policies, too, can change over time.

In conclusion, this study shows that the commercial fishing fleet has increased its presence in northern latitudes of the Bering Sea during October–February and July–August over the last decade. Because this region has been undergoing periods of warming and cooling for decades (and probably longer) due to large-scale atmospheric-oceanic processes, such as those associated with the Pacific Decadal Oscillation (Hunt et al. 2021), it is still unclear whether the patterns we observed are part of a longer-term warming trend associated with long-term global climate change. Climate models suggest that rising ocean temperatures including bottom water temperatures, and recent reductions in sea ice coverage in the Bering Sea are indicative of conditions accompanying post-industrial climate change and that many groundfish will continue to move poleward (Thoman et al. 2020; Fauchald et al. 2021; Rooper et al. 2021). If so, this study may serve as a window into the early stages of a long-term redistribution of commercial fishing effort in the Bering Sea, a region that supports one of the largest fishing industries in the world. To better prepare for future conditions, additional research is needed in the short term to (a) understand which sectors of the commercial fishing industry are currently experiencing the most dramatic shifts and (b) to evaluate the extent to which fishing vessel shifts are correlated with target stocks versus other factors that influence decisions about where to focus fishing effort. Fishery-specific information will be especially important in determining our collective ability to mitigate future climate change impacts by ensuring that effective protections are in place in the Bering Sea, while continuing to promote a sustainable and profitable fishing industry in the region.