2.1.1 Survey design.

The MENHBTS is a stratified random survey that targets inshore waters between the latitudes of 42.9 and 44.8 and longitudes of -70.8 and -66.9 within the 19.3-km (12-mile) limit. Samples were collected across a broad depth profile, from near-surface waters to deeper than 220 m. The survey area is divided into 20 strata based on depth (5–40 m, 41–70 m, 71–100 m, and >100 m) and five longitudinal regions based on oceanographic, geologic, and biological features [32] (Fig 1). The number of tows per stratum is determined by strata size, with larger strata containing proportionately more tows. This stratified sampling approach enhances the survey’s ability to detect and track changes in species distributions, abundances, and community composition over time. The number of tows for each season are provided in S1 Fig.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. A map of the Maine-New Hampshire Bottom Trawl Survey area.

The survey area is divided into five longitudinal regions (1-5) and four depth strata: 5–40 m (brown), 41–70 m (khaki), 71–100 m (light green), and depths greater than 100 m (dark green). The inset map shows the location of the survey area in the northeastern United States. The state and country boundaries were sourced from Natural Earth (public domain; http://www.naturalearthdata.com), and the detailed Maine coastline shapefile was sourced from the Maine Department of Marine Resources (https://dmr-maine.opendata.arcgis.com/datasets/a34df630902643f78cf6395de4f0c641_0/explore).

https://doi.org/10.1371/journal.pclm.0000843.g001

From fall 2000 to spring 2003, the survey was conducted using two nearly identical commercial fishing vessels, the F/V Tara Lynn and F/V Robert Michael, which were two Down East 54’ sisterships built from the same mold. Since spring 2004, the F/V Robert Michael has been the only vessel used for the survey. Further vessel details are available in Sherman et al. [32]. The survey employs a modified shrimp trawl net with a 5.1 cm mesh in the wings and a 2.5 cm mesh liner in the cod end. This gear configuration, coupled with 15.2 cm rubber cookies, is designed to minimize habitat disturbance while efficiently capturing a wide range of fish species [32]. Tows are conducted at a speed of 2.5 knots for approximately 20 minutes, covering a distance of about 1.48 km (0.8 nm). The net mensuration system is used to monitor the gear performance during each tow to make sure the gear is performing the same each tow. The net mensuration system measures door spread, wing spread, headline height, and bottom contact. This data are collected for each tow when the system/sensors work(s).

2.1.2 Consistency in temporal coverage.

The fall survey began in 2000, with sampling initiated in early to mid-October (2002–2011) or in late September (2012–2023) and completed in 4–5 weeks to avoid weather issues later in the season. The only exception occurred in 2000 when the survey took place during November. The 1–2 week variation in start dates was deemed reasonable to accommodate logistical changes (e.g., weather or vessel issues) and was not expected to introduce significant biases in species distribution estimates. However, the 2000 fall survey was excluded from analyses due to its anomalous sampling time relative to other years. Spring survey start dates occurred consistently during early May, with a 1-week variation across years, except during 2001, which started in late April. In addition, surveys in 2020 and 2022 deviated from the usual protocol due to COVID-19 disruptions; consequently, data from these two years were excluded from our analyses.

2.1.3 Consistency in spatial coverage.

The four depth strata were sampled in both seasons. However, the fourth stratum (>100m) was not sampled until 2003. Consequently, data prior to 2003 for both seasons were excluded from the analyses to ensure comparability across years. A few extreme outliers were identified in the depth and salinity data, which were either confirmed or corrected by ME DMR staff to ensure the data quality. The horizontal spatial coverage (latitude and longitude) of the survey remained consistent over the years for both seasons.

2.2.1 Survey data.

Temperature data were collected using a SeaBird CTD (19plus) at the surface and bottom during the MENHBTS [33]. These temperature measurements were taken after each trawl haul, following the survey’s stratified random sampling design. The stratified seasonal temperature for each year (temperature_{s, y}) was weighted by the area of each stratum:

where A is area, z is the index for vertical position (bottom or surface), y is year, s is season, d is day, t is stratum, k is kth sample in a stratum, n is number of samples.

2.2.2 OISST.

The NOAA 1/4° Daily Optimum Interpolation Sea Surface Temperature (OISST) data (https://psl.noaa.gov/data/gridded/data.noaa.oisst.v2.highres.html) were used to evaluate the quality of the MENHBTS temperature data [31]. OISST data within the survey area were extracted based on the survey sampling dates. The seasonal stratified SST averages were compared with the averages of the extracted OISST for each season. Pearson’s correlation coefficient was calculated to evaluate the correlation between the two datasets.

2.2.3 Change point analysis.

Change point analysis was performed on the seasonal stratified temperature to test for and identify the occurrence of thermal regime shifts across the trawl survey time series using the changepoint R package [34,35]. A regime shift is generally defined as a significant, abrupt, and persistent change in the structure and function of an ecosystem [36]. Change point analysis of the fall temperature data identified two distinct points of transition: 2010 for bottom temperature and 2012 for surface temperature (see Results). Although there is no consensus on the exact extent or spatiotemporal scale of data needed to identify a regime shift, most studies use at least 20 years of data [1,6,37]. Both surface and bottom temperatures of fall and spring were examined for regime shifts. Because these points were close in time and both indicated a shift to a warmer thermal regime, we selected 2012 as a single, conservative change point to define the two climate regimes for all subsequent biological analyses.

2.3.1 Species selection.

Data were subset to exclude all catch data not identified to the species level or without a full scientific name (29 groups were excluded). Although the sampling protocol for the MENHBTS has been generally consistent, this initial review recognized a suite of species that had undergone a change in identification level or grouping over time (see Species to watch in Section 2.5.2). Then for each of the remaining 163 species included in subsequent analyses, catch per unit effort (CPUE) was standardized to biomass (kg) caught per 20-minute tow, and only species with at least 30 tows with non-zero CPUE in a season for at least 16 years over the time series were considered further. This cutoff ensured the statistical robustness of the estimates for 19 species in fall and 16 species in spring in subsequent analyses (Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Summary of the 22 species (six species in fall, three species in spring, and thirteen in both seasons) selected in our analyses. Abbreviations of species common names were used in Table 2 and figures in this study.

https://doi.org/10.1371/journal.pclm.0000843.t001

The requirement for a species to be present in at least 16 years of the time series (approx. 80%) ensured that sufficient data were available for comparison in both regimes. Similarly, requiring at least 30 tows with non-zero CPUE per season was set to ensure adequate spatial coverage for reliable estimates of distribution metrics; however, this criterion can exclude species with sparser occurrences, such as Atlantic cod (Gadus morhua). A series of zero-CPUE tows could indicate that a species has moved out of the survey area or, alternatively, that its population has declined. While distinguishing between these scenarios was beyond the scope of this study, our criteria were designed to focus the analysis on species with a consistent presence for which between-regime comparisons would be most meaningful.

2.4.1 Length data.

Catch-at-length data were collected for a variety of fish and invertebrate species caught in the MENHBTS. All individuals were counted, weighed, and measured for length. Finfish total centerline length was measured to the nearest cm, except for species with heterocercal caudal fins (e.g., dogfish and sturgeon). Squid (Order Teuthoidea) were measured by mantle length (in cm), crabs were measured by carapace width (in cm) and lobsters were measured by carapace length (in mm). If a species exceeded 150 individuals within a tow, a random subsample of 100 individuals was collected for length measurements [32]. The total catch and the subsample were weighed separately to create an expansion factor (total catch weight/subsample weight). The length frequency distribution of the subsample was then expanded to the total catch by multiplying the number of individuals in each length bin by this weight-based expansion factor [32]. American lobster was the exception to this rule; sex and length data were recorded for every individual, even when the count exceeded 100, except in spring 2014, when a subsample was taken for a tow with abnormally high lobster catch (the expanded total catch for this tow was 11,203 individuals, weighing 2,231.28 kg). Notably, more than half of the American lobster collected in the MENHBTS were below the legal size (<83 mm).

2.4.2 Maturity data.

Sexual maturity data were collected for a subset of finfish species, following the MEDMR MENHBTS protocol [32]. Approximately 25 individuals per tow were examined, ensuring representation across all length classes and maturity stage [immature, developing, ripe, ripe and running, spent, and resting [38]], sex, and body length and weight information was recorded for each individual. In recent years, due to COVID-related disruptions and changes in survey staffing, the number of individuals examined has decreased to around 10 per tow. It should be noted that the maturity data were collected specifically for estimating size-at-maturity and body lengths within sub-samples were not random. Sexual maturity data were not available for invertebrates. For crustaceans such as American lobster, only female egg-stage data were recorded.

2.5.1 Structural changes.

2.5.1.1 Changes in species distribution.

Three indicators were developed to evaluate species spatial distributional changes within the MENHBTS footprint and time series: 1) Southwest to Northeast and poleward (alongshore), 2) Inshore-offshore (cross-shore), and 3) Depth. Because the survey was conducted in the nearshore area along the ME-NH coast (Fig 1), species distribution was first evaluated along the coastline. A linear regression was fitted to UTM coordinate data (northing and easting) of all tow locations to approximate the primary orientation of the coastline within the survey domain. The resulting regression line defined a standardized alongshore axis oriented from southwest (SW) to northeast (NE), with the southwestern-most tow designated as the origin (0 km). Each tow location was orthogonally projected onto this axis, and the Euclidean distance from the origin to the projected point was calculated. This distance was used as a continuous variable to represent the relative alongshore position of each observation, thereby facilitating spatial comparisons across the SW-NE gradient of the survey area. To evaluate cross-shore (inshore-offshore) distributional changes, the shortest distance between each original observation and the axis was used to describe the cross-shore location, perpendicular to the SW-NE axis (alongshore). Finally, depth was used to indicate changes in vertical distribution to deeper or shallower zones.

The center of gravity (CG) is a biomass weighted spatial indicator, which was used to evaluate species horizontal distribution (alongshore and cross-shore) and vertical distribution (depth) over years [39]:

where CPUE_{i, y} is catch per unit effort which was standardized as weight (kg) caught per 20-minute tow; location_i,y is the spatial coordinate (alongshore, cross-shore, or depth) for a given observation; y is year, and i represents the i^th observation.

To evaluate species distribution shifts under climate change, the year of 2012 was used to divide the time series into two regimes based on the results of the change point analysis described above. A Welch’s t-test was used to determine if the CG_y,i in the two regimes were significantly different.

The variation (CGSD_y) associated with the spatial indicator was estimated as [39]:

where n_y is the number of observations in year y.

To evaluate the consistency of the trawl survey in capturing target populations across the time series, first the spatial boundaries of the area sampled were defined. Lower boundaries were defined as SW, inshore, and deeper depth, while the upper boundaries were defined as NE, offshore, and shallower depth. The lower (B_{sd, lo, y}) and upper (B_{sd, up, y}) bounds of the 95% confidence interval were used to describe the species distribution boundaries within the overall spatial footprint of the trawl survey (note this differs from the overall geographic range of each species). Therefore, the species distribution boundaries were determined by the combination of CGy and CGSDy. The center of each species distribution was then evaluated over the time series to determine if it moved toward the lower or upper boundaries of one or more spatial indicators (alongshore, cross-shore, depth). Simultaneously, the CGSDy associated with the CGy was evaluated to determine if it decreased (indicating a distribution contraction) or increased (indicating a distribution expansion). The survey’s lower (B_{sv, lo}) and upper (B_{sv, up}) boundaries were defined as the 0.025 and 0.975 quantiles of the survey locations (alongshore, cross-shore, and depth) over the years. The “lower” and “upper” boundaries for each dimension correspond to the direction of decreasing and increasing values along each rotated coordinate axis, respectively.

To evaluate the consistency of the survey in capturing populations over its sampling history, the year 2012 was used again to divide the time series into two distinct regimes. Mean values of the species’ lower and upper distribution boundaries were then calculated for each regime:

where r1 is regime 1 (2003–2011) and r2 is regime 2 (2012–2023), and n_r1 and n_r2 are number of samples in r1 and r2, respectively. Lower and upper boundaries were denoted as lo and up.

Species distributions that were estimated to extend beyond the survey boundaries in either regime ( > B_{sv, up} or  > B_{sv, up} or  < B_{sv, lo} or  < B_{sv, lo}) were selected for further examination. Species distributions that remained fully within the survey boundaries across both regimes were considered to be sampled consistently over time and thus were not included in further analysis. Since this study only used MENHBTS data, it was assumed that a single CG occurred within the survey boundaries for each analyzed group. Therefore, the results apply only to the groups defined within the survey’s spatial and temporal footprints.

It is important to note that the estimated changes in distribution range represent the dispersion of a population around its CG, rather than a change in the total area occupied. The dispersion metric reflects how the population is arranged within the survey area, which means how tightly clustered or widely spread individuals are around their average location. In contrast, the area of occupancy describes the population’s total spatial footprint or external boundaries. A population’s occupied area could remain constant while its dispersion increases. For example, if a population shifts from a single, dense aggregation into two smaller, more distant aggregations at opposite ends of its range. In this scenario, the external boundaries have not changed, but the individuals are now, on average, much farther from the CG, resulting in a higher dispersion value. Methods are available to quantify the effective occupied area and assess distribution patchiness [40]; however, such analyses were not included in the scope of this study. For additional details on model sensitivities, please refer to Section 4.5.

Welch’s t-test was used to assess whether mean values of species distribution boundaries differed between the two thermal regimes at a significance level of 0.05. The alternative hypotheses for lower and upper boundaries are: and . If there were significant differences between a species’ lower and upper distribution boundaries, it was assumed these changes were likely driven by climate. Conversely, if significant differences were not observed, it remains unclear whether species were relatively less sensitive to thermal changes, adapting in place [41], or responding in a way that was undetectable by the methods used here.

For species whose distributions were estimated to extend beyond survey boundaries, the proportion of species distribution outside the survey boundary () was estimated as:

The potential impact of shifting species distributions on the MENHBTS was assessed by examining changes in the proportion of each species’ distribution that remained within the fixed boundaries of the survey footprint and over the course of the time series ().

Note that for depth changes, only the lower (deeper) survey boundary was considered () as most species are unable to shift beyond the upper (shallowest) depth boundary and onto land (i.e., the for depth is considered 0 in this study).

Given that the survey boundaries remained constant over time, changes in species distribution ranges were hypothesized to impact their availability to the survey (proportion of the species’ distribution contained within the survey boundaries []). Even if individual boundary shifts were not statistically significant, the combined effect of changes in both upper and lower boundaries could result in significant changes in the proportion of the species’ distribution that remained within the survey area over time (). This is because incorporates information from both the upper and lower boundaries of the species’ distribution. Species range expansions or contractions can both affect . Given the survey’s limited coverage area, it is likely that it did not encompass the entire distribution of each group. For each spatial dimension (alongshore, depth, cross-shore), values range between 0 and 1 and represent the proportion of a species’ distribution estimated within and outside the fixed boundaries of the survey footprint; for example, if  = 0.9 then 10% is estimated outside the survey boundaries and 90% of the distribution remains within the survey boundaries. Further, the difference in mean between the two regimes reflects how this proportion changed, either expanding or contracting as species distributions fluctuate over time, to result in an increase or decrease in P_in across the two regimes.

2.5.1.2 Changes in population length distributions.

To assess changes in population size structure, we analyzed length data collected in the MENHBTS from selected species (see 2.3.1 Species selection) with at least 30 tows with non-zero CPUE per season. The catch-at-length data for each stratum were weighted by the stratum’s area:

where l = length bin; i = tow; s = stratum; r = region; y = year; is stratum area-weighted catch-at-length (number of individuals) for length bin in year ; is catch-at-length observed in a subsample; is total catch weight (kg) of a tow; is subsample weight (kg) of a tow; CPUE was standardized as weight (kg) caught per 20-minute tow; is the number of tows in a stratum; and is area of a stratum.

To assess temporal changes in body length, we calculated the median, standard deviation, 0.025 quantile, and 0.975 quantile of length distributions for each population. These metrics were chosen to minimize the impact of outliers and provide a robust representation of the central tendency, variability, and range of size distributions. A Welch’s t-test compared metrics between regime 1 (pre-2012) and regime 2 (post-2012). Shrimp species (Northern shrimp, Dichelo shrimp, Dichelopandalus leptocerus, and Montagui shrimp, Pandalus montagui) were excluded from this analysis due to insufficient or lack of data. Results are presented as the rate of change in each metric between the two periods.

where m denotes the measure of median, 0.025 quantile, 0.975 quantile, or standard deviation.

2.5.1.3 Changes in length-at-maturity and composition of maturity stages.

A total of 45,857 observations across 20 species were collected from the MENHBTS, where each observation represents an individual for which both length and maturity status were recorded. After excluding unsexed and unidentified individuals, a subset of species with sufficient data (at least 30 tows with non-zero CPUE per season and sex for at least 16 years) was used for estimating length-at-maturity and analyzing spawning season trends.

Binomial generalized linear models [42,43] were applied to estimate the probability (P_M) of a female being sexually mature for each selected species and year, using a logit link function with individual length as the explanatory variable:

where and are model parameters, and is residual errors. Maturity stage of immature was classified as sexually immature, and the rest of maturity stages were classified as sexually mature. Length-at-maturity was defined as the length at which 50% of females were sexually mature. Differences in average length-at-maturity between the two regimes were tested using Welch’s t-test at a 0.05 significance level.

Since the maturity data were collected specifically to estimate size-at-maturity, variations in the size composition of sampled individuals could influence maturity stage proportions. Therefore, we compared the length distributions of maturity samples with those from random length samples that were used to assess population length distributions in the prior section. We also evaluated the proportions of females in each of the six maturity stages to explore potential shifts in spawning phenology. Only female maturity data were included in the analysis of spawning phenology, as the spawning season is typically determined primarily based on females [44], while males may provide a less precise estimate of the spawning season [45].While statistical tests were not feasible for evaluating changes in spawning phenology due to data limitations, visual inspection of the maturity stage composition data provides preliminary insights into potential trends over time.

2.5.2 Functional shift.

To provide a comprehensive assessment of community change, we analyzed abundance-based biodiversity (species richness, Hill-Shannon and Hill-Simpson indices) and biomass-based biodiversity (catch diversity) metrics.

2.5.2.1 Abundance-based biodiversity.

While species richness simply counts the total number of species present, it can be sensitive to rare species, which may lead to an overestimation of diversity in communities with uneven abundance distributions [46,47]. In contrast, the Hill-Simpson diversity index emphasizes dominant species and reflects the effective number of these dominant species within the community [48]. The Hill-Shannon diversity index, on the other hand, accounts for both common and rare species, making it responsive to changes in either group [47]. Although these diversity metrics are correlated, they capture different facets of community structure and are not fully interchangeable [46]. By employing multiple metrics, we can achieve a more comprehensive understanding of community diversity and its responses to environmental changes.

Hill numbers, are parameterized by the diversity order [49], and were employed to assess functional changes in the ecosystem. These indices consider both species richness and relative abundance, capturing different aspects of diversity, including species richness (), Shannon diversity (), and Simpson diversity () [48]. The Hill number for is defined as:

where is diversity, S is the number of species, is the relative abundance of species i.

The order determines the sensitivity of Hill numbers to species relative abundance. When , the Hill number () is equivalent to species richness, simply counting the number of species present. When (Hill-Shannon diversity), the Hill number is mathematically undefined, so an approximation is used to calculate it [48,50]. In this case, as approaches 1, the Hill number approximates the exponential of the Shannon entropy index [46,48]:

When , the Hill number (Hill-Simpson diversity) is equivalent to the inverse of the traditional Gini-Simpson index [51].

To account for sampling effort and ensure comparability across samples even when effort is variable, Hill numbers were standardized to a uniform target level of sample completeness (coverage) using the iNEXT R package [52], with species abundance data standardized to a 20-minute tow. Sample coverage measures how complete a sample is in representing the true diversity of a community [53]. It estimates the proportion of individuals in the sample that belong to the identified species, serving as an indirect estimate of the proportion of species that remain undetected within the community [46,47]. The iNEXT algorithm standardizes diversity to this target level using a two-sided approach with rarefaction and extrapolation [52]. For samples with coverage greater than the target, it uses rarefaction (interpolation) to calculate the diversity of a smaller subsample corresponding to the target coverage. For samples with coverage less than the target, it uses extrapolation to predict the diversity of a larger sample corresponding to the target coverage. For all analyses in this study, we standardized diversity estimates to a target coverage of 0.95 [47].

3.2.1 Changes in species distribution.

3.2.1.1 Center of gravity (CG).

For all spatial results, relative shifts are presented as changes within the MENHBTS footprint. A total of 22 species (six species in the fall, three species in the spring, and thirteen species in both seasons) were selected for distributional shift analyses based on data availability (S1 File and Table 2). Northeastward alongshore CG shifts were detected in six (American plaice, Hippoglossoides platessoides, Butterfish, Peprilus triacanthus, Monkfish, Lophius americanus, Red hake, Urophycis chuss, White hake, Urophycis tenuis, and Winter flounder, Pseudopleuronectes americanus) out of 19 fall species and four (Montagui shrimp, Red hake, White hake, and Winter flounder) out of 16 spring species. Fall alongshore CG shifts ranged from 17.4 km (American plaice) to 46.4 km (Winter flounder), while spring alongshore CG shifts ranged from 18.6 km (Montagui shrimp) to 62.1 km (Red hake) (S1 File and Table 2). Four species exhibited cross-shore CG shifts in fall with two species (American lobster and Jonah crab, Cancer borealis) shifting offshore, and two (Silver hake, Merluccius bilinearis, and Windowpane flounder, Scophthalmus aquosus) shifting inshore. In spring, four species (American lobster, Jonah crab, Longhorn sculpin, Myoxocephalus octodecemspinosus, and White hake) shifted offshore, while only one (Silver hake) shifted inshore. The magnitude of these cross-shore shifts ranged from 3.2 km (Silver hake) to 5.4 km (Windowpane flounder) in fall and from 3.6 km (Silver hake, American lobster, and Longhorn sculpin) to 4.7 km (White hake) in spring (S1 File and Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. A summary of distribution shifts of selected species. Only statistically significant shifts (p < 0.05) in the three-dimensional spatial indicators [alongshore (orange), cross-shore (blue), and depth (green) center of gravity (CG) and boundaries] and the proportion of species distributions within the survey boundaries (P_in) were included in the table. Values represent the absolute difference in the average values between the two regimes for each indicator. The cell next to the number under each variable indicates the direction of shifts. All distances are measured in kilometers, except for depths, which are in meters, and P_in, which is expressed as a proportion. Blank cells indicate non-significant shifts (p > 0.05). The ↔ denotes species boundary expanding, ↮  denotes species boundary contracting. ↑ denotes increased P_in, and ↓ denotes decreased P_in. NE = northeast, SW = southwest, Off=offshore, In=inshore, Dp = deep, Sh = shallow.

https://doi.org/10.1371/journal.pclm.0000843.t002

Five species (Silver hake, Red hake, Winter flounder, American lobster, and Jonah crab) exhibited similar shifts to deeper waters in both seasons. Conversely, Alewife (Alosa pseudoharengus) and Atlantic herring (Clupea harengus) showed no significant shifts in depth CG in either season (S1 File and Table 2). In fall, eight species (American lobster, American plaice, Jonah crab, Monkfish, Red hake, White hake, Winter flounder, and Witch flounder, Glyptocephalus cynoglossus) shifted towards deeper depths, with only one species (Windowpane flounder) shifting towards shallower waters. In spring, seven species (American lobster, American plaice, Dichelo shrimp, Jonah crab, Red hake, White hake, and Winter flounder) shifted towards deeper depths, with two species (Montagui shrimp and Windowpane flounder) shifting towards shallower depths. Fall depth CG shifts ranged from 12 m (Windowpane flounder) to 33.9 m (Jonah crab), while spring CG shifts ranged from 9.3 m (Dichelo shrimp) to 23.8 m (Jonah crab).

3.2.1.2 Boundaries of species distributions.

While some species did not exhibit significant shifts in their CG, significant changes were observed in species’ CG confidence intervals, representing their upper and lower distribution boundaries (S2 File and Table 2). In the fall, six (Alewife, Butterfish, Longhorn sculpin, Monkfish, Red hake, and Witch flounder) out of 19 species expanded their NE boundary, while three species (Atlantic herring, White hake, and Winter flounder) contracted their SW boundary. Longhorn sculpin showed the largest expansion, extending its fall NE boundary by 64.2 km. In the spring, Jonah crab was the only species to expand its range both NE and SW, by 31.8 km and 71.4 km, respectively. Four (Montagui shrimp, Red hake, White hake, and Winter flounder) out of 16 spring species contracted their SW boundary, with Winter flounder showing the most significant contraction of 96.2 km. Additionally, Winter flounder and White hake contracted their range in both the NE and SW directions during the spring (S2 File and Table 2). Seven species (American plaice, White hake, and Red hake in fall and American plaice, Longhorn sculpin, White hake, and Winter flounder in spring) exhibited cross-shore shifts and contracted their inshore boundary, while only Butterfish expanded its inshore boundary in the fall by 10.7 km. Jonah crab and American lobster shifted their offshore boundaries by expanding their range in both seasons (6-6.9 km), while Windowpane flounder contracted its offshore boundary in both seasons (9.4 km in fall and 6.7 km in spring) (S2 File and Table 2). Eight (American lobster, American plaice, Dichelo shrimp, Jonah crab, Monkfish, White hake, Winter flounder, and Witch flounder) out of 19 fall species and six (American lobster, American plaice, Dichelo shrimp, Jonah crab, Red hake, and Winter flounder) out of 16 spring species expanded their depth range, with most shifting towards deeper waters. However, Windowpane flounder exhibited a contraction in the lower boundary in both seasons towards shallower depths. Jonah crab showed the largest depth expansion, with a 33.9 m shift in the lower depth boundary (towards deeper bottom) in the fall and a 23.8 m shift to deeper depths in the spring (S2 File and Table 2).

3.2.1.3 Consistency of the MENHBTS in determining species distribution.

Changes in are presented in the S3 File, with differences between the two regimes detailed in Table 2. Most significant changes in alongshore were relatively small, typically less than 0.1, with the exception of Longhorn sculpin in the fall, which experienced a decrease of 0.22. This difference indicates that the proportion of the species’ distribution remaining within the survey boundaries declined by 22% between the two regimes. The alongshore distribution expansion of American lobster in the fall and decreases in Longhorn sculpin in both seasons, led to a decrease in (differences ranged 0.05-0.22) between the regimes (Table 2 and S3 File), while the alongshore distribution contraction of Winter flounder, Red hake, and Silver hake in the spring and American plaice in both seasons led to an increase in (differences ranged 0.03-0.09) (Table 2 and S3 File). Similarly, cross-shore shifts influenced as was detected through a contraction of Winter flounder in fall, and Silver hake and Northern shrimp in spring, leading to an increase in (differences ranged 0.07-0.09), while the expansion of Jonah crab in both seasons led to a decrease in (differences ranged 0.07-0.08) (Table 2 and S3 File). Changes in cross-shore were also generally small, with most values below 0.1. While many species shifted towards deeper waters, only five species (American plaice, Monkfish, Acadian redfish, Sebastes fasciatus, White hake, and Witch flounder) in fall and three species (American plaice, Jonah crab, and Red hake) in spring experienced significant changes with decreases in depth (differences ranged 0.08-0.16) (Table 2 and S3 File). The average decrease in depth for these species was 0.11, with a maximum of 0.16 for Acadian redfish in the fall (Table 2 and S3 File).

3.2.2 Changes in population length distributions.

Of the 30 species examined, six (two in the fall, two in the spring, and two in both seasons) showed significant changes in median body length between the two regimes (S4 File and Table 3). Monkfish experienced a significant decrease in median length (33.8%) in the fall, while five other species (Shortfin squid, Illex illecebrosus, in fall, White hake and American shad (Alosa sapidissima) in spring, American plaice, and Longhorn sculpin in both seasons) showed significant increases in median length (S4 File and Table 3). Changes in median values could result from changes in length frequency of small or large size-groups. For Monkfish, the decrease in median length was associated with an increase in the proportion of smaller individuals (26.8% decrease in minimum length) (S4 File and Table 3). Conversely, for species like White hake and American plaice in spring, and Longhorn sculpin in both seasons, the increase in median length was associated with an increase in the proportions of larger individuals (increased maximum length) (S4 File and Table 3). American plaice exhibited particularly notable increases in median length, with a 27.1% increase in the fall and a 21.5% increase in the spring. Changes in the distribution of small and large individuals were also reflected in the standard deviation of length distributions. For example, American lobster experienced a decrease in standard deviation, indicating a reduction in variation and likely lower proportion of large individuals (S4 File and Table 3). In contrast, Longfin squid (Doryteuthis pealeii) in the fall and Jonah crab in the spring showed increased variation, with the former associated with increased proportion of large individuals (maximum size increased by 27%) and the latter with increased proportion of small individuals (minimum size decreased by 47.4%). It is important to note that changes in minimum or maximum length can occur without affecting the median or standard deviation. This was observed in Jonah crab, Witch flounder, and Red hake in the fall, and Alewife in the spring. Additionally, increased median lengths were observed for Shortfin squid in the fall and American shad in the spring without significant changes in maximum or minimum lengths.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Rates of change in median length, 0.025 quantile, 0.975 quantile, and standard deviation of the length distribution between the two regimes for analyzed species in fall and spring. Positive values indicate an increase in the metric, while negative values indicate a decrease. An asterisk (*) denotes a significant difference (p < 0.05) between the two regimes. Only species with significant changes in at least one metric are presented.

https://doi.org/10.1371/journal.pclm.0000843.t003

3.2.3 Changes in length-at-maturity and composition of maturity stages.

Only two species in each season (Witch flounder and White hake in fall, and Winter flounder and American plaice in spring) had sufficient data for analyzing changes in length-at-maturity and maturity stage composition.

However, no maturity data were available for Witch flounder in 2003–2004 or for White hake in 2003, and only six samples were recorded for White hake in 2004. As a result, length-at-maturity estimates were not generated for these two species during 2003–2004.

A significant increase (1.75 cm) in the minimum length (0.025 quantile) of Witch flounder was observed during fall (S4 File and Table 3), indicating a reduced proportion of smaller individuals in the warmer regime (post-2012). However, no significant change in length-at-maturity was detected for fall Witch flounder between the two regimes (Fig 3). The length distribution of maturity samples closely mirrored that of the broader fall Witch flounder population, with a greater proportion of large females beginning in the late 2010s (Fig 4). Correspondingly, a decline in the proportion of immature females and an increase in pre-spawning (developing) females were observed in the 2020s (Fig 5). Notably, spawning females (ripe and ripe and running) began to appear in the samples starting in 2017 (Fig 5), suggesting a possible advancement (earlier onset) or protraction in the spawning season, especially given the relatively consistent survey timing since 2012. It was noted that an increase in the length-at-maturity for white hake and American plaice was observed after the late 2010s, which may indicate a delayed population-level response that needs further investigation.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Changes in length-at-maturity for four species.

Seasonal length-at-maturity estimates (black points) with 95% confidence intervals (error bars) for White hake and Witch flounder (fall; top panel), and American plaice and Winter flounder (spring; bottom panel). Horizontal lines represent the mean length-at-maturity for two regimes (pre- and post-2012), with the grey shaded area representing the 95% confidence interval of the mean. Solid red lines indicate a statistically significant difference (p < 0.05) in mean length-at-maturity between regimes, while the dashed black line indicates a non-significant (p > 0.05) difference.

https://doi.org/10.1371/journal.pclm.0000843.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Length distributions and female maturity samples for four species.

Boxplots of overall length distributions of combined sexes (F + M) (right panels) and lengths from female maturity samples [Length for maturity (F), left panels] for White hake and Witch flounder (fall; top two panels), and American plaice and Winter flounder (spring; bottom two panels). The box center line represents the median, with the lower and upper box edges marking the first (Q1) and third (Q3) quartiles. Whiskers extend to the smallest and largest values within 1.5 times the interquartile range from Q1 and Q3, respectively. Dots represent observations beyond these whisker bounds. (F = Female, M = Male).

https://doi.org/10.1371/journal.pclm.0000843.g004

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Maturity stage composition of four species.

Maturity stage composition (Right-hand side from top to bottom: immature, developing, ripe, ripe and running, spent, and resting) for White hake and Witch flounder (Witch fl) in fall (left panels), and American plaice (Am plaice) and Winter flounder (Winter fl) and in spring (right panels).

https://doi.org/10.1371/journal.pclm.0000843.g005

Mean length-at-maturity increased by 9.1 cm for White hake during fall between the two regimes (Fig 3). Although no statistically significant changes were found in the overall length distribution (S4 File and Table 3), a greater proportion of larger individuals was evident in the 2020s (Fig 4). Despite this, most female White hake remained immature, and spawning or post-spawning females were observed only rarely across the time series (Fig 5), indicating that the peak spawning period likely falls outside the survey’s sampling window.

Mean length-at-maturity increased by 1.4 cm between the two regimes (Fig 3) for Winter flounder during spring. However, no significant changes were observed in the overall length distribution (S4 File and Table 3). The consistently low proportions of spawning females (<2%) and moderate proportions of post-spawning females (5–20%) over time (Fig 5) suggest that the survey likely captured the latter part of the spawning season.

Mean length-at-maturity rose by 2.2 cm across the two regimes for American plaice during spring (Fig 3). Significant increases were detected in both median and maximum (0.975 quantile) lengths in the length distribution (S4 File and Table 3). As a result, a greater proportion of large females was included in the maturity samples during the second regime (Fig 4), accompanied by increased proportions of spawning females (ripe and ripe and running) observed during this period (Fig 5). These observations may suggest a potential shift (either advancement or protraction) in the spawning season, or could be influenced by other factors affecting changes in population length distribution (see Discussion).

3.3.1 Changes in abundance-based and biomass-based biodiversity.

3.3.1.1 Abundance-based biodiversity.

While species richness, which emphasizes rare species, did not change significantly between the two regimes (Fig 6), the Hill-Simpson diversity, increased significantly in the spring (Fig 6) suggesting the diversity of common species increased in the warmer regime. No trend was detected during fall (p > 0.05) using the Hill-Simpson index. In contrast, the Hill-Shannon diversity, which considers both rare and common species, increased significantly in the spring but decreased significantly in the fall (Fig 6).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Changes in species richness and diversity metrics.

Annual estimates (black points) of species richness, Hill-Shannon, Hill-Simpson, and catch diversity (listed from top to bottom on the right-hand side) in fall and spring. Horizontal lines represent the mean value for each metric during two regimes (pre- and post-2012), with the grey shaded area representing the 95% confidence interval of the mean. Solid red lines indicate a statistically significant difference (p < 0.05) in mean length-at-maturity between regimes, while the dashed black line indicates a non-significant (p > 0.05) difference. Note that the 95% confidence intervals of the annual mean for some metrics were too narrow to be clearly visible on the graph and are therefore provided in the supplementary material (S5 File). In contrast, confidence intervals were not calculated for catch diversity, as it is an empirical estimate derived directly from the data.

https://doi.org/10.1371/journal.pclm.0000843.g006

3.3.2 Species to watch.

A total of six species were identified as having undergone protocol changes related to identification level or grouping, and an additional 14 species had uncertain identification histories in the MENHBTS (S6 File). Among these, six were classified as anomalous species, as their native distribution does not include the GOM, while three had unknown native ranges.

For species with consistent identification over time, a total of 36 species were present in only one of the two regimes (S7 File). Of these, seven were classified anomalous, and five had unknown native distributions. Two species in the fall were identified as potentially disappearing species: Bigeye scad (Selar crumenophthalmus) (native) and Moon snail (Lunatia heros) (unknown); both were observed in at least five years of the first regime but were absent in the second regime. Potentially emerging species included Northern searobin (Prionotus carolinus) (native) in the fall and Sculptured shrimp (Sclerocrangon boreas) and Friendly spine shrimp (Spirontocaris liljeborgii) (both unknown) in the spring, each with at least five years of occurrence in the second regime and absent in the first regime. Cusk (Brosme brosme) was an interesting case as they were only detected during a single year (2006) in spring in the pre-2012 regime and then were not observed again until the fall of 2022 in the post-2012 regime.

The majority (70%) of species characterized as either having undergone a protocol change (S6 File), found to be an anomalous occurrence, or listed as a species to watch in future years for emergence or disappearance (S7 File) were recorded in only one or two years per season, and with fewer than 10 observations per season. Based on their low frequency, no trends in total or standardized abundance were observed, and all were considered rare in the MENHBTS dataset (S8 File).

4.2.1 Changes in species distribution.

Similar to previous studies documenting poleward shifts in species distributions in the Northeast U.S. Continental Shelf Large Marine Ecosystem (NEUSCS LME) [1,8,9], many inshore species in the MENHBTS exhibited northeastward shifts in their alongshore CG. These northeastward shifts were typically linked to either a significant expansion of the northeastern boundary, a contraction of the southwestern boundary, or both. However, responses were variable and species-specific in magnitude and directionality. For example, the alongshore CG of the American plaice in the fall shifted northeastward without expansion or contraction in their alongshore distribution boundaries. Likewise, shifts in cross-shore and depth CGs were often associated with boundary expansions or contractions in one or both directions. This suggests that species distribution changes were not uniform across spatial dimensions, with boundaries shifting inconsistently, and sometimes in opposite directions [65]. Such complex patterns may also be influenced by the survey’s limited spatial footprint, which can mask boundary shifts that occur outside the survey area.

Our findings are consistent with Kleisner et al. [62], who observed that most GOM species showed stronger shifts in depth than latitude. While most species in our study moved northeastward, offshore, or deeper, some species like Windowpane flounder exhibited a unique pattern, shifting inshore and shallower during the fall. Additionally, Silver hake shifted inshore in both seasons, and Montagui shrimp moved shallower in the spring. Although population expansion, as seen in Silver hake [1,54,66], could potentially explain such shifts, other factors such as fishing and predator pressure may also contribute. Our findings suggest these shifts in direction that are opposite of what is expected may be the result of range contractions. For instance, Windowpane flounder experienced contractions at its offshore and deeper boundaries, resulting in an overall shift toward inshore and shallower waters in the fall, a finding consistent with Mills et al. [1].

Though the MENHBTS survey area is smaller than that of the NEFSCBTS, species distribution shifts observed in the MENHBTS were generally consistent with broader trends of northeastward, offshore, and deeper movements seen with data from the NEFSCBTS [1]. However, many shifts were not statistically significant, possibly due to the more limited spatial and temporal coverage of the MENHBTS. In contrast, the NEFSCBTS, with its broader geographic and temporal scope, is more likely to detect significant shifts in species distributions. Still, the NEFSCBTS’s limited access to the inshore GOM region reduces its ability to capture changes in these inshore populations; therefore, our results provide new insights into changes occurring in inshore regional habitats. Furthermore, species selection in this study was based on data availability, which may exclude species with patchy distributions, low abundance, or issues related to gear selectivity. Although not analyzed here, species such as Atlantic cod is an example of a species that merits further investigation as additional years of data become available in the MENHBTS time series.

4.2.2 Changes in population length distributions.

A common response to warming waters among many marine organisms is a decrease in body size [25] but this is not always the case [26]. This study found an increase in median body length for seven out of eight species that exhibited significant size changes, a result that contrasts with broader trends of decreasing body size in previous studies of the NEUSCS LME [1]. Variable changes in minimum, median, and maximum sizes indicate non-uniform responses to changing conditions from climate-related and other stressors and require further examination on a case-by-case basis. For example, observed changes in minimum size may be affected by recruitment failures, while declines in maximum body size could stem from fishing pressure, which can truncate population size structure [67,68]. If the timing of the survey remains relatively consistent, changes in size structure may also signal phenological shifts reflecting size-specific redistribution, differential seasonal growth rates or migration patterns of different size groups or life phases [69].

Catchability in the survey is also important context for interpretation. For instance, sublegal lobsters (<83 mm carapace length), which comprise the majority of the MENHBTS catch, did not show a change in median size over the time series, rather a decline in maximum size was observed. This pattern could be the result of larger individuals shifting offshore to deeper waters beyond the survey area or into preferred habitats in untrawlable areas such as rocky, cobble bottom and ledges [70]. The presence of fixed gear in the environment and occupation of traps by larger lobsters could also be affecting their availability to the MENHBTS. Furthermore, environmental conditions can directly influence catchability, as changes in water temperature or dissolved oxygen may trigger behavioral responses that alter a species’ susceptibility to the fishing gear [70]. Future studies using data from other MEDMR surveys (e.g., ventless trap and sea sampling) could provide additional insights into patterns of size-distribution in the region.

Regardless of the mechanism, changes in size structure are important indicators of ecosystem health with implications for fisheries management [68]. Changes in body size have been associated with decreased body condition, productivity levels, and energetic flows [71]. For example, declines in the size structure of Atlantic herring in the GOM have been associated with diminished condition of predatory species of conservation and management concern including Bluefin tuna (Thunnus thynnus) [72]. Overall, size-based shifts impact availability to the MENHBTS and its ability to accurately assess abundance of juvenile and early adult life stages in species such as Red and White hake that use inshore waters as essential ontogenic habitats [33].

4.2.3 Changes in length-at-maturity and composition of maturity stages.

While seasonal changes in life history have been detected in some marine species in the NEUSCS LME, phenological shifts are generally less well understood compared to spatial shifts [9–11,28]. This is largely due to the fact that capturing subtle changes in phenology requires high-resolution biological data spanning over multiple decades, which can be time-intensive to collect and remains limited for many species [73]. In this study, only four species had sufficient maturity data for phenological analyses. An increase in length-at-maturity across the two regimes was observed for three of the four species analyzed. Additionally, higher proportions of large individuals were evident in the length distributions of fall Witch flounder and spring American plaice, resulting in significant increases in body length over time. Although White hake in fall also showed a greater proportion of large individuals in the 2020s, this did not translate into a statistically significant shift in overall length. These patterns were mirrored in the maturity stage composition for American plaice, with higher proportions of spawning females recorded during the warmer regime.

The increases in observed length-at-maturity and changes in size distribution in the MENHBTS differ from patterns reported in regional stock assessments. The stock assessment report for Witch flounder noted a slight expansion in age structure based on landings and survey catch-at-age data [74]; however, analyses of the MENHBTS data found no significant changes in length-at-maturity and a non-significant decline in the standard deviation of length. In contrast, the regional stock assessment for the GOM Winter flounder indicated little change in size structure over time [20] while the analysis of the MENHBTS found a slight increase. Because regional stock assessments for these species do not yet incorporate maturity data from the MENHBTS, these contrasting findings potentially reflect localized changes in inshore waters that are not representative of stock-wide patterns.

The underlying drivers of length composition changes of the populations remain unclear, though it is hypothesized they may relate to shifts in spawning phenology or changes in spawning locations (also see Section 4.2.2). The reported spawning season for GOM Witch flounder spans from April to August, with peak activity occurring in May and June [38]. Notably, spawning female Witch flounder were not observed in fall samples until 2017, suggesting a possible advancement or protraction of the spawning period. Continued monitoring of this species and additional sampling outside the survey window could help verify if a shift in spawning timing is occurring in inshore waters. Phenological shifts may also be indirectly reflected in distributional changes, particularly for migratory species [27,73]. Although the MENHBTS design, which follows a consistent SW-to-NE sampling pattern, did not facilitate a detailed analysis of migratory phenology, it is essential to recognize that distributional shifts can be linked to temporal range expansions or contractions [73]. Furthermore, shifts to offshore and deeper waters could increase migration distances for some species to affect their phenology [75]. Alternatively, the increase in the proportion of spawning females in Witch flounder and American plaice may be linked to changes in their spawning locations as species adjust their distributions to follow optimal thermal conditions as waters warm [76–78].

Larger-sized mature females of American plaice were found in the second regime (Table 3 and S4 File), indicating a population shift in inshore waters. Because alterations in spawning location and timing can affect the survival of early life stages [9], continued monitoring and further studies are needed to fully grasp these emerging patterns. It is important to note that the MENHBTS is limited in duration to only five weeks during the spring and fall seasons, and not designed to cover the full spawning seasons of most species. While the MENHBTS data did not permit precise quantification of phenological shifts, it distinguished species whose critical life history phases are in and out of sync with the survey sampling window and identified several species to monitor closely in the coming years to verify potential early signals of phenological change. Future considerations for the MENHBTS include identifying priority species where additional phenological data could improve management goals while balancing the inherent challenges of expanding survey coverage under highly variable weather conditions during winter and the risk of interactions with fixed gear fisheries, particularly in summer (Waller and Staudinger, in review).

4.3.1 Changes in abundance-based and biomass-based biodiversity.

While global species richness is on the decline, the NEUSCS LME has seen an increase over recent decades [30,79]. Regional changes in species richness have been linked to increased water temperature [79] and shifts in species ranges and biomass [8]. Although species richness (abundance-based) remained relatively stable in the MENHBTS throughout both seasons, species diversity (abundance-based)—measured using the Hill-Shannon and Hill-Simpson indices—showed contrasting seasonal trends. In the spring, species diversity (abundance-based) increased, indicating a more even distribution of abundance among species. In contrast, species diversity (abundance-based) decreased in the fall, suggesting that a smaller number of species dominated the overall community. Catch diversity (biomass-based), which reflects the number of high-biomass species, declined in both seasons. This indicates that the portfolio of dominant, fishery-relevant species is shrinking. The divergence of these two metrics is informative: even as the broader community became more even in the spring, the total community biomass became increasingly dominated by a smaller number of species. A reduction in catch diversity implies a decrease in the number of dominant species, which may increase vulnerability in GOM fishing communities as the portfolio of fishing opportunities declines [65].

Our analysis of functional shifts revealed contrasting trends in species richness between inshore habitats of the GOM and the broader NEUSCS LME [79], suggesting that populations exhibit different responses at regional and local scales. Contrasting trends in diversity metrics between the MENHBTS and NEUSCS LME may result from several factors. The inshore area is constrained by the coastline, therefore the directionality of species distributional shifts may disproportionately trend towards offshore and deeper waters and exceed the MENHBTS footprint. This could be compounded by changes in seasonal inshore-offshore migrations to affect local species composition and diversity. Declines of cold-water species (e.g., Northern shrimp) in the GOM during recent years has been attributed to warmer ocean temperatures and subsequent changes in trophic relationships [80]. A lack of response to management actions (e.g., seasonal fishery closures) suggests that some species may be unable to recover in the warmer conditions of the current regime (post 2012). In contrast, species that are more adaptable to warmer conditions may become more dominant and impact the regional community through competitive and predatory interactions. Continued monitoring and assessment of species diversity metrics will be crucial to identifying species that are relatively vulnerable or resilient to changing conditions and need management or conservation interventions.

4.3.2 Species to watch.

Anomalous observations in biodiversity surveys provide important records of cryptogenic and rare species as well as first observations of species invasions from range expansions or human introductions [81]. The assessment of anomalous species in MENHBTS resulted in two lists being produced. The first is an account of protocol variations and unknown identifications, thus establishing reliability measures in the future for monitoring and tracking purposes. The other list is described as the ‘watch list’ and is based on species recorded only in one regime, in spite of it being challenging to distinguish expansion trends and random events [81].

Applying a detection threshold, we have identified two potentially disappearing species (Bigeye scad, Moon snail) and three potentially emerging species (Northern searobin, Sculptured and Friendly spine shrimps). However, this approach is not perfect. For example, there was a potential underestimate of Black Sea Bass (Centropristis striata) range expansion [54,82] into MENHBTS due to trawl bias against structured habitats. These shortcomings included potential aquaculture introductions (European flat oyster) and near threatened species (Sea pen, [83]).