Assessing Assemblage Composition of Reproductively Mature Resource Fishes at a Community Based Subsistence Fishing Area (CBSFA)

: Nearshore ﬁsheries in Hawai‘i have been steadily decreasing for over a century. Marine protected areas (MPAs) have been utilized as a method to both conserve biodiversity and enhance ﬁsheries. The composition of resource ﬁshes within and directly outside of the recently established H¯a‘ena Community Based Subsistence Fishing Area (CBSFA) on the island of Kaua‘i were assessed to determine temporal and spatial patterns in assemblage structure. In situ visual surveys of ﬁshes, invertebrates, and benthos were conducted using a stratiﬁed random sampling design to evaluate the efﬁcacy of the MPA between 2016 and 2020. L 50 values—deﬁned as the size at which half of the individuals in a population have reached reproductive maturity—were used as proxies for identifying reproductively mature resource ﬁshes both inside and outside the CBSFA. Surveys between 2016 and 2020 did not indicate strong temporal or spatial changes in overall resource ﬁsh assemblage structure; however, some species-speciﬁc changes were evident. Although overall resource species diversity and richness were signiﬁcantly higher by 2020 inside the MPA boundaries, there is currently no strong evidence for a reserve effect.


Introduction
Marine protected areas (MPAs) have been increasingly employed as an effective method for managing overfished nearshore coral reef ecosystems [1,2]. MPAs have been shown to increase fish biomass, diversity, and reproductive output within the protected area, as well as enhance adjacent areas via adult and larval spillover [1,3]. However, MPAs are not a panacea for overfished stocks, poor habitat quality, or ineffective management or enforcement [1,[3][4][5][6]. The effectiveness of a MPA is often reliant on careful consideration of key features, such as size, shape, configuration, larval connectivity and recruitment, life-history traits, habitat types, enforcement, and community support [3,[5][6][7].
Nearshore fisheries declines in the State of Hawai'i over the last century have paved the way for implementation of some level of protection in~17% of state waters, yet only 3.4% of nearshore waters are considered to be highly protected [2]. It is within these few, highly protected areas where resource fish biomass is substantially greater compared to lower or non-protected areas [8]. Additionally, with such a wide variety of MPA features, restrictions, and enforcement comes varying degrees of success. Contrasting examples include the successful recovery of herbivorous species in the Kahekili Herbivore Fisheries Management Area (KHFMA) on Maui [7] and the rapid depletion of reefs upon the opening of the Waikīkī-Diamond Head Shoreline Fisheries Management Area on O'ahu where restrictions to fishing are implemented on a yearly rotational basis [4]. In general, most of

Materials and Methods
Hā'ena is located on the north shore of Kaua'i Island within the Hawaiian Archipelago. Coral reef structures extend along the inner reef of the Hā'ena coast in shallow water, and along the forereef into deeper depths (~20 m). Its location on the north shore exposes Hā'ena's reefs to high wave energy and flushing year round, but particularly during winter months (November-March) when large North Pacific swells regularly generate waves in excess of 10 m. Several streams enter the ocean both within and adjacent to the CBSFA boundaries. These boundaries extend 1610 m (~1 mile) offshore and 5633 m (~3.5 miles) along the coastline [11,28] (Figure 1a). There are varying fishing restrictions within the CBSFA boundaries, including a designated area on the shallow backreef where all fishing is prohibited. This area is referred to as the Makua Pu'uhonua, which translates to "nursery area" in Hawaiian, and was set aside to protect juvenile fishes during this critical life-history phase.

Sample Design
Surveys were conducted both within and directly east of the CBSFA (hereafter referred to as "inside" and "outside") along the coast of Hā'ena. Due to varying benthic structure, the Nā Pali State Park on the west side of the CBSFA boundaries was intentionally not surveyed. A stratified random sampling design was used to pre-determine survey stations to allow for spatial representation by depth, habitat type, and location (inside and outside). Approximately 100 random points were generated and stratified by depth (shallow < 7 m, deep ≥ 7 m) using ArcGIS 10.6.1 (Figure 1b). Points were overlaid on National Oceanographic and Atmospheric Administration (NOAA) habitat base maps [28]. Surveys were conducted at the original pre-determined stations. If hazardous conditions were encoun-tered, depth estimates were in error, or <50% of the substrate was hard bottom, then divers swam the depth contour at a pre-determined compass heading (<100 m from the original site) until safe conditions, accurate depths, and >50% hard substrate were reached.

Sample Design
Surveys were conducted both within and directly east of the CBSFA (hereafter referred to as "inside" and "outside") along the coast of Hāʻena. Due to varying benthic structure, the Nā Pali State Park on the west side of the CBSFA boundaries was intentionally not surveyed. A stratified random sampling design was used to predetermine survey stations to allow for spatial representation by depth, habitat type, and location (inside and outside). Approximately 100 random points were generated and stratified by depth (shallow < 7 m, deep ≥ 7 m) using ArcGIS 10.6.1 (Figure 1b). Points were overlaid on National Oceanographic and Atmospheric Administration (NOAA) habitat base maps [28]. Surveys were conducted at the original pre-determined stations. If hazardous conditions were encountered, depth estimates were in error, or <50% of the substrate was hard bottom, then divers swam the depth contour at a pre-determined compass heading (<100 m from the original site) until safe conditions, accurate depths, and >50% hard substrate were reached.

Fish and Benthic Surveys
Fish counts and sizes were estimated inside and outside the CBSFA using the Kauaʻi Assessments of Habitat Utilization (KAHU) rapid assessment technique over a five year period [28]. This method employed 25 × 5 m belt transects [29], designed by the UH FERL Fish Habitat Utilization Study (FHUS). The time frame of this study included surveys conducted in August 2016 (n = 55 inside, n = 43 outside), August 2017 (n = 59 inside, n = 49 outside), August 2018 (n = 71 inside, n = 32 outside), and June and August in 2019 (n = 58 inside, n = 40 outside), and June and August in 2020 (n = 79 inside, n = 44 outside). At each station, a 25 m transect was deployed in the direction of a pre-determined compass bearing (0°, 90°, 180°, or 270°). All fish species, counts, and sizes (total length [cm]) were recorded within a 5-m swath (125 m 2 total area per transect) for a minimum of 10 min [28]. Fish surveyors annually participated in calibration dives to account for observer variability.
A benthic surveyor followed the fish diver to quantify habitat types associated with each transect [28]. Photographs were taken of the substrate on a previously calibrated Cannon S100 camera at every 1 m mark at a 90° angle to the transect (n = 26). Invertebrate (urchins and sea cucumbers) abundances were also recorded. The surveyor also enumerated macroinvertebrates within a 2 m × 25 m swath (50 m 2 ) along the transect line. The

Fish and Benthic Surveys
Fish counts and sizes were estimated inside and outside the CBSFA using the Kaua'i Assessments of Habitat Utilization (KAHU) rapid assessment technique over a five year period [28]. This method employed 25 × 5 m belt transects [29], designed by the UH FERL Fish Habitat Utilization Study (FHUS). The time frame of this study included surveys conducted in August 2016 (n = 55 inside, n = 43 outside), August 2017 (n = 59 inside, n = 49 outside), August 2018 (n = 71 inside, n = 32 outside), and June and August in 2019 (n = 58 inside, n = 40 outside), and June and August in 2020 (n = 79 inside, n = 44 outside). At each station, a 25 m transect was deployed in the direction of a pre-determined compass bearing (0 • , 90 • , 180 • , or 270 • ). All fish species, counts, and sizes (total length [cm]) were recorded within a 5-m swath (125 m 2 total area per transect) for a minimum of 10 min [28]. Fish surveyors annually participated in calibration dives to account for observer variability.
A benthic surveyor followed the fish diver to quantify habitat types associated with each transect [28]. Photographs were taken of the substrate on a previously calibrated Cannon S100 camera at every 1 m mark at a 90 • angle to the transect (n = 26). Invertebrate (urchins and sea cucumbers) abundances were also recorded. The surveyor also enumerated macroinvertebrates within a 2 m × 25 m swath (50 m 2 ) along the transect line. The twenty-six benthic photographs collected at each transect were later processed by overlaying 30 points per photograph using the benthic image analysis program CoralNet [28].

Assessing Reproductive Maturity
L 50 values-defined as the size at which half of the individuals in a population have reached reproductive maturity-were obtained from previous studies within the main Hawaiian Islands (MHI) ( Table A1). Female L 50 values were used rather than male L 50 values owing to the disproportionate importance of large females in population reproductive output [17,19,[29][30][31]. Careful consideration was taken to use L 50 values that were most appropriate for Hā'ena because L 50 values have been shown to vary within and among islands in Hawai'i between individuals within a species [26]. Acanthurus triostegus (convict tang, manini), which is an endemic subspecies, was the only resource fish with L 50 values derived from gonad measurement on the northshore of Kaua'i. The majority of the remaining L 50 values were selected from similar reproductive studies conducted within the Hawaiian Archipelago [15][16][17][18][19][20]23,25,26,32]. A select few L 50 values were chosen from reproductive studies from the Northwestern Hawaiian Islands (NWHI) [32] or Papua New Guinea (Kyphosus spp., lowfin chub, nenue) [33]. Resource fish species with L 50 values from outside the Hawaiian Archipelago, with the exception of the one Papua New Guinea study, were excluded from our analyses.

Statistical Analysis
Changes in assemblage composition of resource fishes throughout the five years inside and outside the CBSFA were analyzed using both multivariate and univariate statistics in the R statistical software. Simpson's diversity, Menhinick's richness, and Pielou's evenness were also calculated using biomass metrics for the resource fish assemblage. Biomass was calculated using the following equation: where standard length in cm was converted from total length and the a and b parameters were obtained from the Hawai'i Cooperative Fishery Research Unit database. Non-metric multidimentional scaling (nMDS) is a rank-based analysis that was used to visualize trophic level patterns in resource fish assemblages. Fish taxa were categorized into trophic categories (corallivores, herbivores, mobile invertebrate feeders, sessile invertebrate feeders, piscivores, zooplanktivores, and detritivores) according to various published sources and FishBase (www.fishbase.org (accessed on 7 March 2021)). Species that occurred in <5% of stations were eliminated prior to analysis. Biomass was down-weighted to account for rare species and a Bray-Curtis dissimilarity matric was used. Permutation-based multivariate analysis of variance (PERMANOVA) was conducted using Type II sum of squares on trophic level (herbivore, piscivore, planktivore, and mobile invertivore) matrices to assess patterns of resource fish species assemblages through time (2016, 2017, 2018, 2019, 2020) and location (inside or outside). Homogeneity of variances were assessed for reliability of the PERMANOVA results using the PERMDISP2 procedure [34]. Pairwise comparisons with a Holm's correction was applied following significant (p < 0.05) PERMANOVA results. SIMPER (similarity percentages) were used to analyze resource species with the highest influence on the multivariate test; however, unequal sample sizes may result in unreliable SIMPER results [35]. Therefore, further univariate analyses were conducted to determine resource species patterns. Univariate analyses included ANOVA with Holm's correction on multiple pairwise comparisons on quarter-root or log 10 transformed data. However, the large number of zeros in most of the datasets meant zero-inflated models were most appropriate. Due to the inability to run biomass, a non-integer response variable, through zero-inflated models, data were manually split into two sections and generalized linear models (GLMs) were used. Non-normal data were first divided into presence/absence to account for the high number of zeros present. A binomial distribution with a complementary-log-log link function was used for presence/absence data to account for the unequal number of zerosto-ones in each matrix. Non-zero biomass data were analyzed using a Gamma distribution with a log-link function run on quarter-root transformed biomass. Least-squares means were applied to GLM models to examine pairwise comparisons.
Finally, distance-based RDAs (redundancy analyses) (dbRDAs) were used to compare biomass matrices with depth, habitat type, and percent cover of coral, calcareous coralline algae (CCA), turf, and macroalgae and abundances of invertebrates, to examine drivers of patterns in fish biomass over time and between management regimes. Turf was excluded from the analysis, as it was found to be highly correlated with CCA. An Akaike information criterion (AIC) forward and backwards step model selection was used to determine the critical variables to run in the dbRDA model using a Bray-Curtis index.

Results
A total of 156 fish species were surveyed over the 5 year period across all locations. Mobile invertivores and herbivores occurred at over 90% of the total stations, while planktivores, corallivores, and piscivores occurred at >50% of the stations surveyed. Of the 156 species, 32 (20.5%) were classified as resource fishes by the Hā'ena community, and were composed mainly of herbivores (82.8%) and piscivores (48.5%). After eliminating species that occurred in <5% of the stations, 65 fish species overall and 19 resource fish species remained for analyses.

Overall Fish Assemblage
Fish assemblage structure showed a high degree of overlap among years (Figure 2a). Assemblage structure outside the CBSFA was more concordant and was a subset of the assemblage inside the CBSFA (Figure 2b). Clear spatial patterns in overall fish asseblages were evident among sub-locations in the nMDS plot (Figure 2c). Fish assemblages inside and outside shallow (<7 m) and inside and outside deep (≥7 m) strata had similar assemblage structures, while the Makua pu'uhonua had a distinct assemblage (Figure 2c).

Trophic Level Assemblage
Reproductively mature trophic level resource fishes had distinct assemblages in 2016 and 2017, while no distinctions were evident between 2017, 2018, 2019, and 2020. There were small differences between the fish assemblages at deep water stations inside the CBSFA compared to deep water stations outside the boundaries. Shallow water stations inside and outside the boundaries have similar assemblages, yet were distinct from deeper water stations. Trophic level resource fish biomass revealed significant differences among years (PERMANOVA, Table 1). Specifically, pairwise comparisons detected higher biomass in 2017 and 2019 compared to 2016 (Pairwise, Table 2). Reproductively mature herbivores and piscivores had higher abundances in 2020 compared to 2016 (χ 2 = 7.15, p = 0.041 and χ 2 = 7.68, p = 0.025, respectively). There was a higher number of herbivores inside the CBSFA (χ 2 = 18.05, p < 0.001), while planktivores were more abundant outside the boundary (χ 2 = 4.09, p = 0.045). Of the species present, mobile invertivores and planktivores had a higher biomass outside (χ 2 = 11.83, p < 0.001 and χ 2 = 4.41, p = 0.039, respectively), while herbivore and piscivore biomass was not significantly different by year or location (both p > 0.05).
Despite similar assemblage compositions through time and locations, specific species level changes were analyzed to examine species-level differences. Ten herbivore species, six piscivore species, and four mobile invertivore species were further analyzed. Spatial ordination plots did not reveal any distinct species assemblages through time or location for piscivores or mobile invertivores; however, herbivore species may be responsible for the distinction in assemblages between 2016 and 2017. Herbivore biomass in 2016 was significantly lower compared to 2020 in pairwise comparisons (pairwise, F 4,438 = 1.77, p = 0.002).

Fish and Benthic Community Relationships
Depth and invertebrate abundance only explained ~5% (adj R 2 ) of the variance within reproductively mature resource fishes at the trophic assemblage level. Variations in the biomass of planktivores, piscivores, and mobile invertivores (5% adj R 2 ; Figure 6a,d) were explained by depth, while the variance in herbivore biomass was mainly explained by the abundance of invertebrates (6% adj R 2 , Figure 6a,c). Within the piscivore assemblage, CCA, macroalgae (Mac), and depth explained ~ 6% (adj R 2 ) of the variability in C. Although spatial patterns were not evident in the mobile invertivore nMDS plots, the PERMANOVA identified a significant location term where reproductively mature mobile invertivore biomass was higher outside the boundaries (p = 0.009). This was likely driven by the higher presence of Mulloidichthys flavolineatus (yellowstripe goatfish, weke) inside the boundaries, as well as higher biomass of the introduced Lutjanus kasmira (bluestripe snapper, ta'ape) outside the boundaries (χ 2 = 6.47, p = 0.018 and χ 2 = 9.41, p = 0.003, respectively; Figure 5).  Although spatial patterns were not evident in the mobile invertivore nMDS plots, the PERMANOVA identified a significant location term where reproductively mature mobile invertivore biomass was higher outside the boundaries (p = 0.009). This was likely driven by the higher presence of Mulloidichthys flavolineatus (yellowstripe goatfish, weke) inside the boundaries, as well as higher biomass of the introduced Lutjanus kasmira (bluestripe snapper, taʻape) outside the boundaries ( 2 = 6.47, p = 0.018 and 2 = 9.41, p = 0.003, respectively; Figure 5).

Fish and Benthic Community Relationships
Depth and invertebrate abundance only explained ~5% (adj R 2 ) of the variance within reproductively mature resource fishes at the trophic assemblage level. Variations in the biomass of planktivores, piscivores, and mobile invertivores (5% adj R 2 ; Figure 6a,d) were explained by depth, while the variance in herbivore biomass was mainly explained by the abundance of invertebrates (6% adj R 2 , Figure 6a,c). Within the piscivore assemblage, CCA, macroalgae (Mac), and depth explained ~ 6% (adj R 2 ) of the variability in C.

Fish and Benthic Community Relationships
Depth and invertebrate abundance only explained~5% (adj R 2 ) of the variance within reproductively mature resource fishes at the trophic assemblage level. Variations in the biomass of planktivores, piscivores, and mobile invertivores (5% adj R 2 ; Figure 6a,d) were explained by depth, while the variance in herbivore biomass was mainly explained by the abundance of invertebrates (6% adj R 2 , Figure 6a,c). Within the piscivore assemblage, CCA, macroalgae (Mac), and depth explained~6% (adj R 2 ) of the variability in C. melampygus and A. virescens (Figure 6b). The majority of variance in each of the resource fish matrices remains unexplained by the variables included. melampygus and A. virescens (Figure 6b). The majority of variance in each of the resource fish matrices remains unexplained by the variables included.

Discussion
Examining assemblage composition shifts and distributions through time is important in assessing the effectiveness of protected areas to implement adaptive management strategies [37,38]. Our overall results found no major shifts in fish assemblage composition in space or time inside or outside the Hāʻena CBSFA. Furthermore, the dbRDA models comparing resource fish biomass to depth and benthic community composition estimates explained very little of the variance in resource fish above their L50 values (Figure 6). While overall, reproductively mature resource fish assemblages remained fairly constant over time, species level shifts were evident. The significantly increasing biomass values of the piscivorous C. melanpygus outside the CBSFA boundaries in later years may suggest early signs of spillover. Positive increases in reproductively mature resource fish diversity and richness through time and between management regimes, as well as positive trends of the species that show significant relationships through time and location suggest continual monitoring can be beneficial.

Discussion
Examining assemblage composition shifts and distributions through time is important in assessing the effectiveness of protected areas to implement adaptive management strategies [37,38]. Our overall results found no major shifts in fish assemblage composition in space or time inside or outside the Hā'ena CBSFA. Furthermore, the dbRDA models comparing resource fish biomass to depth and benthic community composition estimates explained very little of the variance in resource fish above their L 50 values ( Figure 6). While overall, reproductively mature resource fish assemblages remained fairly constant over time, species level shifts were evident. The significantly increasing biomass values of the piscivorous C. melanpygus outside the CBSFA boundaries in later years may suggest early signs of spillover. Positive increases in reproductively mature resource fish diversity and richness through time and between management regimes, as well as positive trends of the species that show significant relationships through time and location suggest continual monitoring can be beneficial.
While few species level trends may be evident, the lack of major shifts in overall biomass at a temporal or spatial scale in reproductively mature resource fish assemblages may indicate several possible outcomes: (1) the CBSFA is having no effect, (2) five years is not sufficient time to see an effect, (3) habitats within and outside the CBSFA may be dissimilar, (4) the CBSFA boundaries are too small to show an effect, and/or (5) low sample sizes resulted in low statistical power. It is likely that the five year period may not have been sufficient time to begin seeing reserve effects. The Kahekili Herbivore Management Area on Maui took > 6 years before effects were witnessed [7]. It is also possible that the 2018 record breaking freshwater flood event may have delayed the effects of protection. This freshwater event caused major landslides and flooding that resulted in significantly lower total fish biomass on shallow Hā'ena reefs [39]. Furthermore, the nMDS plots of the overall assemblage, trophic level, herbivore, and piscivore matrices demonstrate overlap on a spatial scale between inside and outside assemblages at the shallow stations and overlapping inside and outside assemblages at the deep stations. These patterns are consistent with the study design in comparing similar habitats within and outside the boundaries and are consistent with a lack of influx of resource fishes moving into the CBSFA if individuals are acquiring equal resources both inside and outside the boundaries.
The size and mosaic of habitats of an MPA can greatly affect the success of that area and the Hā'ena CBSFA may be too small for the mobilities of resource fishes that we examined [40]. Within such a diverse group of organisms with varying functional roles and demands for energy [41], it is necessary for some species of fishes to travel longer distances to fulfill their energy, reproductive, and social requirements [42][43][44]. Although the majority of the Hā'ena CBSFA is not fully protected against fishing, it is one of the larger MPAs in the state at roughly 8 km 2 , with is larger than the median size (1.2 km 2 ) of MPAs in Hawai'i [2]. Yet, considering the effective size for MPAs (10 to 100 km 2 [2], the Hā'ena CBSFA remains below the lower threshold.
Although sample sizes were >30 for all stations by year and location, limitations in sample sizes of individuals were evident for select species. Therefore, uncommon species with low sample sizes were excluded by eliminating species that occurred in <5% of stations for the purpose of improved detection of an MPA effect. Even in unfished regions, fish population distributions are naturally skewed to smaller individuals that are more abundant in size and as they become larger, their abundance decreases.
L 50 values can vary not only by location and water temperature but also by season and year [45,46]. Although the L 50 value for Acanthurus triostegus came directly from gonad measurements of individuals located around the Hā'ena area, L 50 values of most other species were derived from measurements of individuals from the MHI, some from the NWHI, and one from Papua New Guinea (Table A1). L 50 values are the best estimate that can be used to assess reproductive maturity using in situ observations. Hence, conducting further research to acquire L 50 values for all resource fish species from the Hā'ena region could be beneficial in ensuring precise analyses of that specific location.

Conclusions
Overall resource fish assemblage composition did not change temporally or spatially over five years despite the implementation of fishing regulations. Results of this study suggest continuing annual surveys to evaluate long-term trends in order to better predict how resource fish assemblages may be changing within the current management regime. These monitoring data are essential if future adaptive changes in rules and regulations are to be implemented and effective. Furthermore, determining the habitat types and benthic structures that are benefical to specific resource fishes at the Hā'ena CBSFA is crucial in assessing any emerging patterns of assemblage composition. Such contemporary research and management practices serve the purpose of providing data that allows local stakeholders and communities to adjust their rules and regulations as needed, thus, allowing an adaptive management strategy to be implemented [47]. CBSFAs place local and traditional knowledge and practices at the forefront of fisheries management, allowing accountability for fisheries by local community members [37]. One of the social side effects of most MPAs in Hawai'i is the prevention of local fishers from practicing traditional fishing methods or incorporating adaptive management strategies due to temporary or permanent closures [48]. The integration of local and traditional knowledge and practices into the CBSFA design gives local people who use the protected area on a frequent basis accountability and a voice that can be beneficial in creating rules and regulations, especially as there are many communities that still depend on marine resources for subsistence.  Institutional Review Board Statement: Ethical review and approval were waived for this study, due to methodologies that do not disturb or alter the marine biota.