This study was conducted on the traditional and unceded territories of the Kwakwa̱ka̱’wakw peoples (Umista Cultural Society, n.d. ⁶ ) within the Broughton Archipelago. This region sits at the interface of two major bodies of water: the Johnstone Strait and Queen Charlotte Strait, near the northeast of Vancouver Island, BC, Canada ( Figure 1 ). The Broughton Archipelago features many islands in the west and glacially carved fjords in the east (Shugar et al., 2014; Davies et al., 2018). Due to its strong environmental gradient in seawater temperature and clear differences in bathymetry, this region can be distinctly divided into two subregions: the cooler, dynamic western archipelago, and the warmer, sheltered, eastern fjords (Foreman et al., 2009; Brewer-Dalton et al., 2014; Lin and Bianucci, 2023). This environmental gradient and spatial differences in bathymetry drive kelp abundance, with the larger and denser kelp beds that can be detected with satellite imagery only located in the dynamic subregion, whereas the smaller, fringing beds that are challenging to detect with satellite imagery are in the fjord subregion (Man et al., in prep). We focused on the dynamic subregion due to the limited availability of high-resolution satellite imagery in the fjord subregion ( Figure 1 ). This subregion is dominated by Nereocystis beds, except for on the north shore of Malcolm Island, which is lined with large, dense beds primarily composed of Macrocystis (Man et al., in prep; Sutherland, 1990). Macrocystis and Nereocystis are the only canopy-forming kelp species in the subregion, and “kelp” hereafter collectively refers to both species. Nereocystis, as an annual species, tends to display more interannual variability than Macrocystis, a perennial species (Dayton et al., 1984; Springer et al., 2010).

Local community members, including First Nations, have revealed that, generally, the kelp forests have declined in density and coverage in their territories (Broughton Aquaculture Transition Initiative (BATI), unpublished, 2021; Salmon Coast Field Station (SCFS), unpublished, 2023). Community members have also reported specific locations of kelp change in the region, including an increase around Malcolm Island, and decreases at an offshore kelp bed at the mouth of the Nimpkish River (“NR”), at a nearshore kelp bed off the shore of the Alert Bay Lighthouse (“ABL”), and along the salmon migration routes in the fjords near now-decommissioned open-net salmon farms (SCFS, 2023; Mountain, pers comm, 2023).

One Macrocystis site and eight smaller Nereocystis sites were selected to represent the temporal dynamics of kelp beds of different species and sizes exposed to different environmental conditions. The Macrocystis site encompasses the entire north shore of Malcolm Island (spanning 11.6 km²), and the eight smaller Nereocystis sites (ranging from 0.03-1.45 km² per site) represent the portion of the eastern shore of Malcolm Island with high kelp abundance and the smaller islands east of Malcolm Island ( Figure 1 ). Only one Macrocystis site was selected as this was the only area within the Broughton Archipelago where Macrocystis was present. The bottom substrate type at the Macrocystis site was primarily mixed rocky and sandy substrate, with kelp growing on the rocky areas (Haggarty et al., 2020; Man et al., in prep). The smaller site approach was chosen for the Nereocystis sites rather than mapping all of their coastlines due to the geomorphological complexity of the area, which imposed a challenge for the satellite remote sensing of kelp because of the increased land adjacency effects (Cavanaugh et al., 2021). The Nereocystis sites included the offshore kelp beds at the mouth of the Nimpkish River (NR), the nearshore kelp beds off the shore of the Alert Bay Lighthouse (ABL), the eastern coastline of Alert Bay (ABE), the eastern tip of Malcolm Island (MIE), Pearse Islands (PI), Bold Head (BH), Wedge Island (WI), and South Leading Islet (SLI) ( Figure 1 ). PI, BH, and WI had primarily shallow (<20 m depth) rocky bottom substrate, surrounded by deeper (>20 m) areas where kelp cannot grow (Haggarty et al., 2020). ABL, ABE, and MIE had mixed rocky and sandy bottom substrates, and NR and SLI had mixed rocky and sandy bottom substrates surrounded by pure sandy substrates which cannot support kelp growth (Haggarty et al., 2020).

The locations of the Macrocystis and Nereocystis sites were selected opportunistically during a field visit and based on their importance to the Mamalilikulla First Nation, ‘Na̱mg̱is First Nation, and Kwikwasut’inuxw/Haxwa’mis First Nation. These three Nations formed the Broughton Aquaculture Transition Initiative (BATI), a coalition that emphasized the need to continue monitoring and protecting the Archipelago’s kelp forests due to their importance as nearshore salmon habitat (BATI, unpublished, 2021). The Macrocystis and the Nereocystis sites experience slightly different environmental and topographical conditions from each other, with the Macrocystis site having flatter slopes and lower tidal current speeds than the Nereocystis sites (Davies et al., 2018; Foreman et al., 2009).

The following data were compiled: (1) local, regional, and global-scale environmental conditions, (2) Landsat-derived canopy kelp area (1984 to 2023) at the Macrocystis site, and (3) Planetscope-derived kelp area (2016 to 2023) at the Nereocystis sites (both “kelp area” hereafter) ( Figure 2 ). In addition, very high-resolution Worldview-2 and GeoEye-1 imagery was used to validate kelp classifications derived from Landsat and Planetscope imagery ( Figure 2 ). The environmental variables were used to define climate periods (years with similar environmental conditions) ( Figure 2 ). Objective 1 (identify the temporal responses of Macrocystis pyrifera and Nereocystis luetkeana, respectively, to changes in temperature and climatic oscillations) was achieved by analyzing both long-term and short-term kelp time series alongside environmental changes at local, regional, and global scales ( Figure 2 ). Objective 2 (identifying spatial patterns of kelp persistence) was achieved by spatially combining yearly kelp areas ( Figure 2 ).

At the regional level, the mean spring SST ranged from 8.4°C to 11.1°C, and mean summer SST ranged from 9.6°C to 12.6°C ( Figure 3 , SST anomalies in Supplementary Figure S1 ), with no temporal trends in either regional spring or summer SST (modified Mann-Kendall’s test: spring: Z-statistic = 0.187, p=0.0911; summer: Z-statistic = 0.195, p=0.0785; Table 3 ). Based on the mean summer and spring regional SST, the changepoints analysis ( Supplementary Table S2 ) indicated three climate periods: Period 1 represents generally warmer ocean temperatures from 1984 to 1998, Period 2 represents cooler temperatures from 1999 to 2013, and Period 3 represents the highest temperatures of the time series from 2014 to 2023 ( Figure 3 ). Accordingly, mean spring and summer regional SST were different across all periods ( Figure 3 , Table 3 ). For both spring and summer, Period 1 (spring SST: 9.8 ± 0.41°C, summer SST: 10.7 ± 0.31°C) had significantly warmer SST than Period 2 (spring SST: 9.2 ± 0.47°C, summer SST: 10.1 ± 0.41°C, Figure 3 , Table 3 ). Furthermore, Period 3 (spring SST: 10.5 ± 0.37°C, summer SST: 11.6 ± 0.5°C, summer SST anomaly: 0.9 ± 0.53°C) had significantly warmer SST than both Periods 1 and 2 ( Figure 3 , Table 3 ).

At the local scale, the Landsat-derived SST showed some variability, with the Macrocystis site presenting warmer SST (11.5 ± 0.96°C) (climatological summer mean ± standard deviation) than the Nereocystis sites (10.1 ± 1.05°C). On average, the hottest Nereocystis sites were ABL, ABE, and SLI (~10.3°C), and the coolest Nereocystis site was BH (9.4 ± 0.72°C) ( Figure 4A , Supplementary Table S3 ). Across the three identified climate periods, the mean local SST significantly increased for the Macrocystis site and six Nereocystis sites (ABL, ABE, MIE, BH, WI, PI) ( Figure 4A , Table 3A ), with ~1.4°C ( Table 3A ) higher SST in Period 3 (Macrocystis site: 12.5 ± 0.42°C, Nereocystis sites: 10.6 ± 0.71°C) than in Period 1 (Macrocystis site: 10.3 ± 0.20°C, Nereocystis sites: 9.2 ± 0.54°C). However, neither Periods 1 nor 3 had significant differences with Period 2 regarding the mean local SST ( Figure 4B , Table 3A ). For the other two Nereocystis sites (NR, SLI), local SST measurements were not significantly different among the periods ( Figure 4B , Table 3A ), although local SST peaked in 2004 in sites NR and SLI, reaching mean summer temperatures of 14.2 ± 1.50°C and 12.0 ± 0.31°C. respectively.

On the regional scale, 57 MHWs were identified between January 1983 and August 2023. Among these, 27 were in Category I, 23 in Category II, 6 in Category III, and 1 in Category IV ( Figure 5A ). The most intense (Category IV) MHW occurred on 30 July 2020, lasting six days and reaching a maximum temperature intensity of 15.4°C (a +4.7°C anomaly), with a total cumulative intensity (calculated as the mean temperature anomaly of the MHW multiplied by the number of heatwave days) of 18.2°C days. The most cumulatively intense MHW occurred at the beginning of 2016, starting on 20 January and lasting 98 days, resulting in a cumulative total of 146.0°C days, and reaching a maximum intensity of 10.8°C (a +2.2°C anomaly). Note that there were several winter months with no SST data available, namely December 2018 - January 2019, December 2019 - January 2020, and November 2020 - Jan 2021 (see section 2.2.1), thus the lack of MHWs detected during those periods may be attributed to this reason. Furthermore, the pre-summer cumulative MHW intensity may be underestimated as a result. The pre-summer MHW time series did not display any temporal trends, but the summer MHW time series exhibited a statistically significant increase ( Table 3B ). Pre-summer MHW cumulative intensity was not significantly different across periods. However, summer MHW cumulative intensity was different across Periods 1 and 3, and 2 and 3, with Period 3 representing the highest MHW intensity (Period 1: 1.2 ± 54.32°C days, Period 2: 1.7 ± 6.21°C days, Period 3: 26.4 ± 106.84°C days) ( Supplementary Figure S2 , Table 3B ). Note that Period 3 had a high standard deviation for cumulative summer MHW intensity because most days did not have MHWs, but the MHWs that did occur had high cumulative intensity (°C days).

On the global scale, Period 1 (1984 to 1998) generally showed positive PDO (mean Z-score ± SD: 0.617± 0.855) and ONI (0.203 ± 1.13) values, reaching the highest PDO (2.09) and the second highest ONI (2.20) of the time series in 1997 ( Figure 5B ). Interestingly, the lowest ONI (-1.84) in the time series was also documented within Period 1 in 1988. Period 2 (1999 to 2023) displayed mostly negative PDO (-0.425 ± 0.817) and ONI (-0.292 ± 1.13), although a positive PDO and ONI were documented from 2002 to 2007. When comparing differences in oscillations between Periods 1 and 2, PDO was significantly more negative in Period 2 than in Period 1, whereas there were no significant differences in ONI ( Figure 5B , Table 3C ). Period 3 (2014 to 2023) started with positive ONI (0.180 ± 1.12) and PDO (-0.273 ± 1.09), with the highest ONI of the time series (2.21) documented in 2015. However, from 2020 onwards, both ONI and PDO shifted negative, with PDO reaching its lowest value of the time series (-1.91) in 2023, although ONI became positive again in 2023 (1.47). Ultimately, there were no significant differences in both ONI and PDO between Periods 1 and 3, and between Periods 2 and 3 ( Table 3C ).

The maximum kelp area at the Macrocystis site (1984 to 2023) covered 4.84 million m². Percent kelp area ranged from 24.0% (absolute area: 1.15 million m²) to 71.0% (3.44 million m²), with a mean of 50.0% (2.46 million m²) of the maximum kelp area. Further, we found a statistically significant increase in kelp area (Mann Kendall’s τ=0.247, p=0.0300, Figure 6A ) across all three climate periods. Among the climate periods, we observed statistically significant differences in kelp area between Periods 1 and 2 and between Periods 1 and 3, but not between Periods 2 and 3 ( Figure 6A ). Specifically, there was a slight increase in kelp area from 43.9 ± 9.34% in Period 1 to 53.8 ± 11.5% in Period 2, which remained high at 55.4 ± 10.4% in Period 3 ( Figure 6B ; Kruskal-Wallis χ2 = 7.56, p=0.0228; Dunn’s test: Period 1 vs Period 2: χ2 = 2.35, p=0.0380; Period 1 vs Period 3, χ2 = 2.35, p=0.0380, Period 2 vs Period 3, χ2 = 0.348, p=0.727). Considering the kelp time series in its entirety, without division between climate periods, the results of the linear model showed no effects of 1-, 2-, or 3-year averages of spring and summer SST, pre-summer and summer cumulative MHW intensities, PDO, or ONI on kelp area, regardless of the combination of predictor variables used ( Supplementary Table S4 ).

For the short-term time series, the maximum kelp areas across all Nereocystis sites ranged from 11,800 m² (BH) to 214,000 m² (MIE) ( Figure 7 ). On a site level, mean percent kelp areas across all surveyed years ranged from 19.0 ± 30.2% (4,010 ± 6,360 m²) at ABL to 58.5 ± 13.3% at PI (111,000 ± 25,100 m²), with a total mean percent kelp area across all sites of 43.1 ± 19.3% ( Figure 7 ) Most sites’ mean kelp areas exhibited parametric behavior, except ABL, which had a left-skewed pattern caused by kelp loss in the years 2018, 2019, 2022, and 2023 ( Figure 7 ). The eight Nereocystis sites also displayed variable temporal trends in kelp areas from 2016 to 2023. Six Nereocystis sites displayed no temporal trends (ABL, MIE, PI, BH, WI, SLI), one displayed a significantly decreasing trend (NR, modified Mann-Kendall’s Z-statistic = -4.26, p=0.000203), and one displayed a significantly increasing trend (ABE, modified Mann Kendall’s Z-statistic = 2.85, p=0.00443, Figure 7 ).

The spatial analysis indicated that the Macrocystis site had 77.0% persistence, i.e., 77.0% of the maximum kelp area was present for more than 19 years out of the 38-year time series ( Figure 8 ). The Macrocystis beds at the eastern part of the Macrocystis site were less spatially persistent than the western part ( Figure 8 ). Five Nereocystis sites (NR, ABE, MIE, WI, SLI) had 20.6-33.8% persistence, i.e., area that was present for more than 4 years out of the 8-year time series ( Figure 9 ). Sites PI and BH had larger proportions (53.1% and 53.6% respectively) of persistent kelp area, and ABL had no persistent kelp area (0.00% area). For both Macrocystis and Nereocystis sites, the persistent areas were mainly distributed in the center and inshore areas of each kelp bed ( Figures 8 - 9 ).

Overall, the kelp area in the Broughton Archipelago was mostly temporally persistent. Kelp percent area increased monotonically in the Macrocystis site from 1984 to 2023, specifically increasing by 9.90% in Period 2 and staying high throughout Period 3. The increase in kelp area at the Macrocystis site on the north shore of Malcolm Island corroborates reports from community members (SCFS, unpublished, 2023). This temporal increase indicated persistence in Macrocystis area from 1984 to 2023, which may be explained by the increase in local SST from ~10.0 to 13.0°C. This potentially represents a move towards more ideal environmental conditions for Macrocystis, which has an optimal thermal range from 12.0 to 17.0°C (Lüning and Neushul, 1978), although this range could vary among populations and life stages (e.g. spores, gametophytes, or sporophytes) (Muth et al., 2019; Hollarsmith et al., 2020; Le et al., 2022). For example, an increase from 9.5 to 12.9°C was experimentally associated with a ~5.00 μm increase in Macrocystis gametophyte germ-tube length (Le et al., 2022), and a 10.0 to 14.0°C increase to be associated with a ~2% increase in the relative growth rate of Macrocystis blades (Fernández et al., 2020). The local SST increases in the study area always remained below the upper thermal limits for Macrocystis, which lie around 18.0-25.0°C (Hay, 1990; Le et al., 2022; Ladah and Zertuche-González, 2007).

On the contrary, regional and global environmental conditions were not linked to significant changes in kelp area (based on the linear model results, Supplementary Table S4 ). This is likely as regional and global conditions occasionally diverged from the conditions kelps were facing locally in the nearshore environment (Brewer-Dalton et al., 2014; Lin and Bianucci, 2023). For example, the colder SST in Period 2 was only observed in the regional SST and inferred from the negative PDO, but not the local SST time series, which increased continuously throughout all three climate periods. Furthermore, occasional intense MHWs and higher local SST were observed during and after 2020, despite a shift to more negative ONI and PDO during the same time. A similar disparity was found between local SST and global climatic oscillations in other parts of BC. For instance, local SST continuously increased after 2020 in the Salish Sea, despite climatic oscillations transitioning to negative phases (e.g. Amos et al., 2015; Mora-Soto et al., 2024a), suggesting that the local SST increase in the Broughton Archipelago may be linked to local oceanographic conditions and variation in coastal geomorphology (Brewer-Dalton et al., 2014; Lin and Bianucci, 2023). Despite mismatches in environmental conditions of different scales during certain years of the time series, generally speaking, Period 3 had higher local and regional SST, a higher frequency and magnitude of MHWs, and more positive PDO and ONI. Similar conditions observed in the Strait of Georgia and Barkley Sound resulted in a subsequent decrease in kelp area (Mora-Soto et al., 2024a; Starko et al, 2022). However, in the dynamic subregion of the Broughton Archipelago, both local and regional measurements of SST remained within Macrocystis’ optimal thermal range during Period 3, and consequently, we did not observe decreases in the kelp area. This reinforces the observation that local temperature gradients can greatly influence kelp responses in the face of regional events such as MHWs (Starko et al., 2024b). It is unknown whether the climatic oscillations affected other environmental conditions not investigated in this study (e.g. nutrient availability) (Whitney, 2015; Bond et al., 2015), which may have impacted kelp physiology (Hollarsmith et al., 2022). Regardless, climatic oscillations did not directly relate to significant changes in the kelp area at the Macrocystis site.

Aside from environmental changes, there may be biotic conditions at play, such as a low abundance of sea urchins (Eisaguirre et al., 2020). Although not investigated in this study, Man et al. (in prep) found no sea urchins during a one-time sampling effort during the summer of 2023 at the Macrocystis site, suggesting that the absence or low abundance of sea urchins may have been conducive to kelp persistence at the Macrocystis site. Continuous time-series data about other environmental and biotic conditions could elucidate what specifically drove the increase in Macrocystis area in the study area.

Our observations of Macrocystis persistence in cooler waters corroborated patterns observed in cooler areas throughout the Northeast Pacific coast. For instance, centennial increases in the Macrocystis area were also observed in Southeast Alaska by Hollarsmith et al. (2024), which the authors attributed to various factors including temperature increases and the reintroduction of the sea otter, a keystone predator. Persistence in Macrocystis area was also observed in Cumshewa Inlet (Gendall et al., submitted), Ella Beach, BC (Mora-Soto et al., 2024a), and the outer coast of Washington (Pfister et al., 2018), where local summer SST increases (~10.0 to 14.0°C) over the past few decades were similar to those observed in the Broughton Archipelago and did not reach the upper thermal limits of Macrocystis. Similarly, these cooler, northern regions (BC and Washington) displayed Macrocystis persistence in response to the Blob of 2016. Conversely, the warmer southern regions (Central to Southern California, and Baja California Norte and Sur), which reached summer temperatures of ~17.0-24.0°C, experienced areal decreases to ~2-57% of their pre-Blob baseline (Mora-Soto et al., 2024b; Pfister et al., 2018; Cavanaugh et al., 2019; Bell et al., 2023).

Kelp persisted in most of the Nereocystis sites from 2016 to 2023, although some sites showed different trends in the kelp area. Specifically, ABL did not display a significant temporal trend, but exhibited kelp losses in four of the eight studied years, corroborating community anecdotes of kelp decrease in the area (SCFS, unpublished, 2023). Similarly, kelp also declined significantly at NR, confirming community reports of kelp decrease near Alder Bay and Green Island, which comprise the eastern areas of NR ( Figure 9 ; SCFS, unpublished, 2023). An examination of Figure 9 revealed that the eastern areas of NR were indeed less spatially persistent. The observed increase in kelp area at ABE also reinforced local reports of a recent increase in kelp area after past decreases (SCFS, unpublished, 2023). Some of our results, however, contrasted with some other community reports of kelp trends: e.g. a kelp increase at MIE after urchin harvesting began in that area, although the start date of urchin harvesting was not reported (Mountain, pers comm, 2023). Such discrepancies between the local community members’ observations and our results may be linked to different study periods or time scales since they did not report the years where the changes in the kelp area were observed (SCFS, unpublished, 2023; Mountain, pers comm, 2023). Thus, it is possible that the changes identified by the local community were observed outside of the timeframe covered in this study between 2016 to 2023. Furthermore, short-term studies (<20 years) are less likely than long-term studies (>20 years) to capture patterns of kelp decrease due to kelp’s high interannual variability and environmental conditions’ multi-year or decadal dynamics (Wernberg et al., 2019). Therefore, it is also possible that our short-term time series was simply not long enough to detect the changes reported by local communities, or that Nereocystis’ high interannual variability in canopy-forming area may have skewed community observations based on when the observations were made.

The observed overall kelp persistence in the Nereocystis sites may be partially explained by the local SST (~9.0-12.0°C), similar to the patterns observed in the Macrocystis site, Local SST increased at some of the Nereocystis sites but mostly remained below the upper thermal limits of Nereocystis (around 12.0-16.0°C for sporophytes, and 16.0-18.0°C for gametophytes, depending on the population) (Pontier et al., 2024; Korabik et al., 2023; Weigel et al., 2023). However, local SST may not be the only variable affecting kelp area changes, as kelp area only decreased at ABL and NR, despite local SST not increasing beyond 12.0°C between 2016 and 2023 at both sites. While examining the spatial patterns of kelp loss, we noted that the sites ABL and NR were close to each other and near the mouth of the Nimpkish River, which brings pulses of warmer water above 20.0°C into the nearshore environment. This suggests the possibility that warmer freshwater pulses not captured from satellite-derived local SST and long-term regional SST averages may have affected kelp abundance ( Figure 1 , Barbosa & Man, personal communication, March 7, 2023). ABE, a site where the kelp area increased, did not have significantly different environmental conditions than the other six sites, which displayed no significant temporal change. Thus, the different temporal kelp trends in ABE, NR, and ABL may also be attributed to other environmental and/or biotic factors not researched in this study. One potentially important environmental factor not addressed here may be seasonal sediment outflows from the Nimpkish River, which may have increased the turbidity of the water, decreasing light availability at NR and ABL for kelps and affecting their growth. The phenomenon of river outflows limiting kelp growth has been observed in Southern Chilean Patagonia (Huovinen et al., 2020), however, further study is needed to test this hypothesis for these two sites. A potentially important biotic factor may be the understory kelps which can outcompete the more ruderal Nereocystis (Dayton et al., 1984; Springer et al., 2010), since a one-time underwater field inspection conducted by Man et al. (in prep) uncovered high understory kelp abundance at NR and ABL, which cannot easily be detected from satellite imagery (Cavanaugh et al., 2021). Similar to the Macrocystis site, regional and global environmental conditions did not appear to affect the overall kelp persistence in the Nereocystis sites, likely because these conditions differed from local SST conditions.

Kelp area in the Nereocystis sites remained mostly persistent during and after the Blob ( Figure 7 ). It is important to note that the time series for the Nereocystis sites started at the end of the Blob (2016), lacking a pre-Blob baseline for the Nereocystis sites. Therefore, our results could mean that Nereocystis was either not affected by the Blob or was affected by the Blob and did not recover to possible higher pre-Blob abundances. This echoes findings of Nereocystis persistence after the Blob in Oregon (Hamilton et al., 2020) and in the southern Salish Sea (Mora-Soto et al., 2024a, 2024b), where SST remained cooler (~12.0-15.0°C). On the other hand, this contrasts with findings of Nereocystis trends in Northern California (McPherson et al., 2021; Cavanaugh et al., 2023; Bell et al., 2023), Southern Puget Sound (Berry et al., 2021), and the northern and central Salish Sea (Mora-Soto et al., 2024b; Starko et al., 2024a), where Nereocystis area declined after the Blob and showed limited recovery. The reasons for these kelp declines vary from higher SST (summer temperatures: ~13.0 to 20.0°C) to an increase in sea urchins after the loss of a keystone predator (sunflower sea stars) after the Blob (Hamilton et al., 2021), neither of which have been documented in the Broughton Archipelago.

Overall, we identified primarily persistent kelp forests, including Macrocystis and Nereocystis kelp areas, in the dynamic subregion of the Broughton Archipelago. As the documented SST trends remained within the favorable range for kelps throughout both time series, we cannot conclude if the kelps would be resilient to further SST increases (up to 20.0°C) in the study area, such as those observed SST increases in the northern Salish Sea, Southern Puget Sound, the sheltered parts of Barkley Sound, and in California. Regardless, the kelp forests have likely persisted for a long time in the study area, as all but two sites (BH and WI) were historically documented to have kelp present in the 1850s to 1950s based on records in the British Admiralty nautical charts ( Figure 10 ; Costa et al., 2020). Note that this historical information only serves as a record of kelp presence, not kelp absence (Costa et al., 2020), thus, the lack of historical kelp documentation at BH does not necessarily mean that kelp was absent from the 1850s to the 1950s. The likely centennial persistence of kelp suggests that the dynamic subregion of the Broughton Archipelago represents a climate refuge for kelp. Similar patterns of persistence can be expected and have been observed in other similar temperature regimes such as the more exposed areas of the Strait of Juan de Fuca (Mora-Soto et al., 2024a; Pfister et al., 2018) and Oregon (Hamilton et al., 2020), in the absence of other stressors such as high water turbidity and sea urchin abundance (Huovinen et al., 2020; Eisaguirre et al., 2020). It is important to consider that this observation of persistence in the dynamic subregion may not apply to other parts of the Broughton Archipelago, such as the fjord subregion, which have smaller, fringing kelp beds that are subject to significantly different environmental conditions (Man et al., in prep). Indeed, local community members have anecdotally noted decreases in kelp distribution and area in the fjord subregion, including near now-decommissioned open-net salmon farms (Mountain, pers comm, 2023). Future research can investigate kelp area changes in the fjord subregion using very-high-resolution satellite imagery (e.g. Worldview-2 at 0.46 m spatial resolution), which may be capable of accurately detecting the smaller, fringing kelp beds.

Kelp beds of both species had similar spatial patterns of persistence, with the center and inshore areas of kelp beds being more persistent than the edges. This reinforces spatial patterns found in Macrocystis forests in Southern California (Young et al., 2016), and Nereocystis forests in Northern California (Arafeh-Dalmau et al., 2023), Oregon (Hamilton et al., 2020; Arafeh-Dalmau et al., 2023), and the outer coast of Washington (Arafeh-Dalmau et al., 2023). The spatial pattern of persistence may be associated with local variations in environmental and biotic factors, including current conditions, kelp dispersal, and sea urchin abundance (Jackson and Winant, 1983; Graham, 2003; Reeves et al., 2022). For instance, water current velocities are higher at the edges of the kelp bed than on the inside due to kelp plants’ ability to buffer water currents (Jackson and Winant, 1983). Therefore, physical disturbances to kelps are more prone to happen at the kelp bed edges (Bekkby et al., 2019), potentially affecting its spatial persistence (e.g. Young et al., 2016). Currents may also play a role in kelp spore dispersal, with currents typically traveling further away at the kelp bed edge than in the interior, carrying zoospores away from the bed (Graham, 2003). In contrast, in the kelp forest interior, the drag from the high density of kelp sporophytes modifies current flow to primarily oscillate within the kelp bed, maintaining high levels of spore supply (Graham, 2003), and potentially contributing to the higher spatial persistence in the kelp bed interior. Biotic factors, such as increased urchin grazing at the edges of a kelp bed compared to the inside, may also cause lower kelp persistence at the edges (Reeves et al., 2022). However, it is unlikely that this is the situation at most of our study sites, as only some were documented to have abundant urchins (Man et al., in prep).

Beyond the potential influence of environmental and biotic drivers on the spatial patterns of persistence, the variable environmental conditions during the satellite imagery acquisition could have also affected the observed kelp area. Potential environmental conditions affecting the observed kelp area in the satellite images include the tidal height, which can submerge the edges of the kelp bed (short-term dataset tidal height: 0.716-2.50 m), and the currents that move kelps in different directions (Timmer et al., 2024). Timmer et al. (2024) found that kelp bed area can decrease by an average of 22.5% around the edges of the kelp bed per meter of tidal increase during low current speeds (<0.100 m/s) and 35.5% at high current speeds (>10.0 m/s). However, that study was conducted using drone imagery, thus the specific impacts of a tidal height increase on kelp bed edge submersion as detected from satellite imagery may slightly differ due to the difference in spatial resolution. We have reduced the influence of tidal height on detected kelp area by 1) aggregating images acquired under the different tidal heights for the long-term dataset and 2) testing and confirming with a linear mixed model the lack of a significant effect of tidal height on kelp area for the short-term dataset. No information on current speeds during the time of satellite imagery acquisition was available, however, according to the regional model by Foreman et al. (2009), tidal current speeds are generally high (~0.390 m/s) at the Nereocystis sites, thus it is possible that some kelp was submerged by the tidal currents. Nonetheless, the non-persistent areas in the short-term time series were all larger than 35.5% of the maximum kelp areas, thus it is unlikely that the spatial pattern of persistence can be entirely attributed to tidal height and tidal current speeds.

The proportion of persistent kelp area varied within the Macrocystis site. A visual assessment of the Macrocystis site showed that the western parts have more persistent kelp areas, whereas the eastern part had a smaller percentage of persistent kelp areas. Fieldwork conducted for Man et al. (in prep) and historical data from the BC Shorezone Mapping System ¹¹ revealed submerged eelgrass beds interspersed between the Macrocystis at the eastern end of the north shore and homogenous Macrocystis patches at the western end. Thus, interspecific competition between Macrocystis and eelgrass may lead to smaller areas of persistence, a phenomenon that has been documented between other seaweed and seagrass species (Alexandre et al., 2017).

The kelp beds in the Nereocystis sites generally had lower proportions of spatially persistent area than in the Macrocystis site. ABL, the only Nereocystis site with years of kelp loss, was not spatially persistent, even in the center of its maximum kelp area. NR, the Nereocystis site with significantly decreasing kelp area, had 22.8% persistent area, with no persistent areas in the eastern part of the site. The sites that displayed no significant temporal changes, i.e., temporal persistence, had 20.6-53.6% spatially persistent areas; and the site where kelp was increasing (ABE) had a 33.8% spatially persistent area. It is important to note that for the Nereocystis sites, the threshold for kelp persistence was only 4 years out of the 8-year time series, much lower than that of the longer time series (19 out of 38 years), which may have led to the different percentages of persistent area between the long-term and short-term datasets. However, the lower persistence levels in the Nereocystis sites, as compared to the Macrocystis sites, may also be partially explained by Nereocystis’ ruderal quality and annual life history, which leads to higher interannual variability than other perennial kelp species, such as Macrocystis (Dayton et al., 1984; Springer et al., 2010). The phenomenon of Nereocystis having higher spatial variability than Macrocystis was also observed on the Washington coast (Pfister et al., 2018), which the study authors also attributed to Nereocystis’ ruderal nature.