Low optimal fisheries yield creates challenges for sustainability in a climate refugia

Reducing resource depletion and promoting ecosystem‐based management are considered key climate change adaptation policies. Therefore, the resource status of an identified climate refugia in a semi‐enclosed bay on the Kenya–Tanzania border was evaluated for sustainability. Both fisheries stock and catch assessment methods found low production and excess effort. Stock recovery in closures (up to 45 years) determined the best‐fit r and K values, which established a maximum sustainable production (MSY) of 2.98 ± 0.45 (SEM) tons/km2/year. Stock estimates in the bays' fishing grounds indicated that biomass was below the MSY and predicted to produce 1.8 ± 1.0 (SEM) or 1.1 ton/km2/year below the optimal MSY. However, landed fish at five studied fishing villages varied greatly from 0.22 to 2.9 tons/km2/year. MSY in the refugia was therefore considerably lower than estimates in nearby ocean‐exposed locations, which has been estimated at 5–7 tons/km2/year. Therefore, low to modest capture rates of fish will be required to allow the recovery needed to achieve sustainability and restore the refugia's ecology. The refugia's highest stocks and near‐MSY yields were captured in the national reserve. Therefore, broader implementation of the reserve's gear‐restriction policies should restore fisheries. High spatial variability in yield patterns indicate interactions between fisheries management, compliance, trade connections, and governance. In climate refugia, reducing cumulative impacts will require knowing and managing for lower fisheries production limits.


| INTRODUCTION
A priority for modern resource policy is to improve management in climate sanctuaries or refugia (Myers et al., 2000;Van Woesik et al., 2022).Climate refugia are locations having the properties of high biodiversity and the capacity to either avoid, resist, or quickly recover from climate change disturbances (McClanahan, D'Agata, et al., 2023;McClanahan, Darling, et al., 2023;West & Salm, 2003).Once refugia are identified, there is a need to evaluate their status and manage them to avoid losses of diversity and ecological services arising from local human impacts.Along the African coastline, this often entails reducing fishing pressure to promote stock recovery (McClanahan & Muthiga, 2016;Rehren et al., 2020;Robertson et al., 2018).Stock biomass is closely associated with the biodiversity of fish and other ecological resilience characteristics of fish communities (Graham et al., 2017;McClanahan, 2022;Zamborain-Mason et al., 2023).Large-scale studies suggest a series of ecological thresholds associated with fish biomass that will have consequences for ecological processes, such as fisheries production and reef growth (McClanahan, Graham, et al., 2011).Thresholds vary along fisheries production gradients but many including biodiversity are supported at or just above maximum sustained yield (MSY) fisheries levels (McClanahan, 2018a(McClanahan, , 2022)).What is less understood is the variability in fisheries production rates and the differences between existing and optimal catch rates.
A minimum goal for reef climate refugia should be to avoid overfishing.This is expected to sustain and balance the human needs for fish while supporting ecosystem properties and processes (McClanahan, 2022).A better understanding of the stocks, production, and yield potentials is therefore an important step for planning and management in refugia.Considerable progress has been made to better understand fisheries yield potentials (McClanahan & Azali, 2020;Samoilys et al., 2017).Yet, spatial variability and underlying factors that affect production are poorly known, particularly in avoidance refugia.The production function of the logistic population growth equation r is the most critical value for estimating production but also one of the least understood (McClanahan, 2018b).Therefore, most estimates of yields are based on landed catch and not stock and production, which has produced controversy over the accuracy of sustainability metric options (Pauly et al., 2013).Fish production is expected to vary with environments and yet environment-production relationships are poorly known (McClanahan, D'Agata, et al., 2023;McClanahan, Darling, et al., 2023).Large disparities in estimates of r for coral reefs are the result.For example, combining many sites over ocean-basin scales resulted in a mean r of $0.07 (MacNeil et al., 2015).However, a study of the western Indian Ocean reported higher provincial estimates of 0.23 ± 0.08 that reached 0.34 near fisheries closures (McClanahan, 2021;McClanahan & Graham, 2015).This variability undermines the ability to make accurate local estimates of production, which is increasingly important for knowing the limits of catch and fish recovery times.
A broad question for the climate change-coral reef fisheries production science is if stocks, production, and yields differ in climate refuges than other locations.If so, refuges will require special fisheries management policies.This requires knowing where climate refugia are located and then using stock recovery rates to estimate production.Several studies and metrics have identified a nearshore area on the Kenya-Tanzania border as an avoidance climate refugia (Beyer et al., 2018;McClanahan & Azali, 2021;McClanahan, Maina, et al., 2011).A proposed mechanism is that the deep water ($750 m) of the Pemba Channel creates cooler and stable temperatures that reduces the nearshore penetration of episodic warm oceanic water (McClanahan, 2020;Painter et al., 2021;Sekadende et al., 2020).Fishing is widespread but there are several modest sized fisheries closures of various ages that can provide space-for-time substitutions to estimate recovery rates (MacNeil et al., 2015).Here, recovery information in the local closures was compiled and compared with nearby fisheries catch assessment to evaluate different methods to estimate fisheries status and yield potential.Therefore, the goal was to determine the status and production potential of this refugia by comparing two yield estimation methods.

| ENVIRONMENTAL CONTEXT
The African border between Kenya and Tanzania runs east to west just south of an elbow-shaped coastline that creates a semi-enclosed bay (Figure 1).The depth is variable offshore and created by the interaction of geological rifting, historical sea level changes and riverine inputs (Kent, 1971).This creates high geomorphological and habitat diversity (Maina et al., 2015).South of the Kenyan coastline are a series of smaller Pleistocene reef islands (Wasini, Mpunguti, and Misali) and submerged coral reefs, some with emergent Pleistocene ($120,000 years before present) reef remnants (Kisite and Mpunguti).The deep Pemba Channel is related to Rift Valley faulting and runs longitudinally between the smaller coastal and larger islands of Pemba and Zanzibar $20-50 km off the coastal mainland (Kent, 1971).The Pemba channel is $750 meters deep and connected to deeper oceanic water of $1500 m off the eastern shore of Zanzibar Island (McClanahan, 2020).This deep and cool water enters close to shore from the southern Pemba Channel and localized upwelling is caused by the interaction of strong currents with depth gradients (Painter et al., 2021;Semba et al., 2019).
Reef fisheries are exposed to a mixing of cooler oceanic currents with the warmer lagoonal waters within the semi-enclosed bay and mangrove-fringed coastline.
Consequently, in situ measurements of sea surface temperatures on mid reefs indicate moderate temperature variability but little acute temperature stress that is often caused by penetration of warm oceanic waters during heat wave events, such as a positive El Niño or Indian Ocean Dipole anomaly (McClanahan, 2020).Coral reef organisms are therefore expected to be both acclimated to moderate temperature variability and protected from extreme warm thermal pulses (McClanahan, 2017(McClanahan, , 2020)).Tidal currents ($4-m maximum tidal range) are the main source of physical energy in the semi-enclosed bay.The island reefs have a diverse fish fauna that differ in species composition from ocean-exposed fringing reefs to the north (McClanahan & Arthur, 2001).The lower physical energy may have consequences for the production, community composition, and yields of reef fishes.The purpose of this study was to determine what the consequences of the semi-enclosed bay was for fisheries production.While the conditions may support high biodiversity and provide refuge from climate change, these same conditions might cause lower-than-expected fisheries production.Thereby lowering the optimal yield recommendations needed to sustain the refugia's ecology (McClanahan & Azali, 2021).Low production potential was addressed by evaluating biomass recovery rates in fisheries closures and trends in fishing effort and catch.

| Study design
Investigations were intended to evaluate yield estimates by accounting for several factors including effectiveness of the data recorders, area from which catch was landed, fisheries stock and catch-based approaches, and variable management systems.Stock assessment methods included visual census of fish stocks in a variety of management systems and over time since closure to fishing in four reefs closed to fishing (Figure 1).Space-for-time evaluations allowed estimates of r and K for the logistic fisheries production equation.Secondly, fisheries catch landed at five village landings sites were recorded for 16 months by employed and trained community members and fisheries officers.This allowed a comparison of fisheries observers, time, and their interaction on estimates of production.Thirdly, estimates of per area yields were possible by estimating the areas of the fishing grounds via a collaborative community mapping process.

| Field sampling
We studied five villages or landing sites (Mkwiro, Wasini, Kibuyuni, Vanga, and Jimbo) where fish were landed and weighted prior to entering the fish trade or taken home for consumption.Fish were weighed and priced by the dominant taxa or groupings that fishers used to sell them.Weighing and price groupings were goatfish, parrotfish, octopus, scavengers (Haemulidae, Lethrinidae, Lutjanidae), and a mixed group that included a diversity of coral reef fishes that had low market value (McClanahan & Kosgei, 2019).Fish were weighed and the number of boats, fishers, and gear used were recorded.Price data were collected once a month for each catch group during sampling visits and used to estimate incomes.Observers sampled fishers opportunistically with the instructions and goal of targeting all fishers that landed fish.However, the length of time that each observer spent at the landing site was not controlled for or recorded.

| Fish stock biomass
Fish stocks were sampled using visual underwater counts to estimate biomass in fisheries closures over closure age to estimate stock weights in four different fisheries management systems (Figure 1).Management systems included a permanent national fisheries closure (Kisite Marine National Park), an associated gear restricted reef (Mpunguti Marine Reserve), four community managed closures (Wasini, Sii, Kibuyuni, and Mji wa Kale), and unregulated fishing grounds.Repeat-sampling was undertaken in Kisite, Chumbe Island and Misali Island closed to fishing in 1978, 1994, and 1998 respectively.Sampling was undertaken by snorkeling and scuba diving at depths between 1 and 15 m and mixed reef habitats of reef slopes, crests, and back reefs.Sites were all located on calcium carbonate bottoms colonized by hard and soft corals and various algae, with sand and seagrass being a smaller portion of the bottom cover.Fish were counted in 500 m 2 belt transects where individual encountered fish were identified to a family and sized by standard lengths in 10 cm bins to convert to individual and summed for community weights (McClanahan & Kaunda-Arara, 1996).No fish <3 cm were counted.Observed fishes that were not a member of the 24 preselected families were placed in an others group.When certain precautions are made, such as counting larger and diver-adverse fish while laying out transect lines, this method is accurate for the larger observable fish that are the focus of fishers catch and production estimates (Bernard et al., 2013;Emslie et al., 2018;McClanahan et al., 2007).
Biomass for estimating fisheries yield was pooled and evaluated into four groups that were total, fishable, target, and non-target biomass.The fishable biomass category was composed of all individuals >10 cm in 22 families with sharks and damselfish excluded due to their episodic variation (see McClanahan et al., 2019 for a list of taxa).Target biomasses estimates were sums of commercially exploited reef taxa >10-cm including Carangidae, Haemulidae, Holocentridae, Lethrinidae, Lutjanidae, Mullidae, Nemipheritidae, Scaridae, Serranidae, Siganidae, Sphyraenidae, and >20 cm Labridae (McClanahan, 2022).Target stock biomass contain economic or commercially valuable taxa whereas fishable biomass includes taxa consumed largely in households.
Recovery of biomass in fisheries closures evaluated the above fish groups but also resident, and migrant families separately.There was concern that migrant and resident species might have different recovery rates due to their different diets and foraging behaviors of using resources on and off the reef, which might influence their rates of production.These two groups were both a large part of the catch and therefore large differences in their production rates could lead to errors in estimating total fisheries production.Therefore, this division of the biomass was undertaken as opposed to other options, such as piscivores versus herbivores, that were expected to produce smaller differences in recovery and production estimates.Resident families were those known to largely feed on or very close to the reef and included the families Acanthuridae, Aulostomidae, Balistidae, Caesionidae, Carangidae, Chaetodontidae, Diodontidae, Fistularidae, Ginglymostomatidae, Holocentridae, Labridae, Lagocephalidae, Lethrinidae, Mullidae, Muraenidae, Pempheridae, Pinguipedidae, Pomacanthidae, Pomacentridae, Rhincodontidae, Scaridae, Scorpaenidae, Serranidae, Siganidae, Sphyraenidae, and Synodontidae.Migrants were in families known to school on the reef but largely to migrate daily or seasonally away from the reef to feed in surrounding seagrass, sand, and rubble habitats.These included the families Haemulidae, Lutjanidae, and Nemipheritidae.

| Fish catches
The five landing sites were monitored in two ways, by employed community members and fisheries department observers between February 2020 and July 2022.Prior to employing community members, a fisheries literacy exam of 28 questions was given to 114 people who expressed an interest in participating in the project.Based on the exam results, five people with the highest scores were selected and agreed to regularly measure landed fish.The employment selection decisions favored top-scoring females who lived locally but lacked fishing as an employment option.Kenya's fisheries department local office director was asked to identify two people to participate in the project and to receive some financial support.
Landing sites were visited 12.2 ± 0.3 (SEM) by employed community members and 3.0 ± 0.2 days per month by fisheries officers.Observers recorded the numbers of fishers and boats, landed fish were weighed in five different demersal finfish catch group categories of goatfish, rabbitfish, scavengers, parrotfish, and mixed catch were each measured to the nearest 0.1 kg.Mixed catch was a highly diverse group of mostly reef-associated fish that fishers did not like to separate because they received a similarly low price.Price data per kg were collected monthly for each catch group during sampling period and were used to estimate revenue of fishers based on their daily catch rates.Landing data was entered via cellphone using Atlan Collect software (http://collect.atlan.com/forms/), and data were downloaded into spreadsheets for analysis in R version 4.1.2and JMP16.0 statistics.Average prices were 198 ± 66 (SD) Kenya shillings (Ksh) and a 110 Ksh/US$ was used as the exchange rate in 2021 US dollars.The number of fishing days per year was 212 based on interviews and landing site reports of monthly fishing effort.

| Community mapping of fishing grounds
Estimates of fishing area were undertaken using a collaborative community mapping process in each of the study site (Figure 1).The first step was to define different resource user groups with respect to fishing gear and other marine-related occupations.This resulted in 15 stakeholder categories namely gleaners, seaweed farmers, speargun/sticks, hook and line, basket traps, ring net, shark net, long line, fence trap, beach seine, reef seine, drift net, set net, monofilament/scoop net, and tourist boat operators.Two representatives per category were chosen as informants, considering diversifying age and gender inclusion where applicable.The representatives formed small working groups of three to five persons each representing unique gear and presented with a map showing their village landing site.The groups were asked to hand-draw fishing grounds associated with each gear as well as landmarks used for navigation.The different maps were harmonized to a single main hand-drawn map through a subsequent group collaboration.The final consensus or master hand-drawn map was then digitized on Google Earth.Eventually, the digitized map was used in ground truthing exercise where global positioning system (GPS) coordinates of specific fishing grounds were recorded and used to create polygons to validate the finalized map.The ground-truthed map showing fishing border polygons, coordinates, and estimated fishing areas were then further discussed and validated by community members.At the end of the exercise, generated shapefiles of fishing grounds by gear were produced and used to calculate fishing areas.

| Fisheries stock recovery
Repeated visual underwater fish biomass samples were undertaken in the Wasini, Kisite, Misali, and Chumbe closures.Biomass at the specific recovery times created a biomass time series that was fit to the logistic or Schaefer model to determine r and K values for the above fish categories.Here, B t is the biomass in closure year t, B 0 is the initial biomass, K is the carrying capacity, and r is the recovery rate: MSY was then estimated for each fish category using the solution of the Verhulst equation (Bacaër, 2011).
Stock recovery data have been tested by various fisheries production equations or models and the logistic model performed best (McClanahan, 2022;McClanahan & Graham, 2015).The closure time-biomass relationship in the times series fit well to the logistic equation with strong significance of the r and K values (Table S1).Additionally, fisheries-catch models have been tested for their ability to predict long term (18-23 years) catch and yield data in Kenya's ocean-exposed fringing reefs located to the north of this study area (McClanahan & Azali, 2020).Among the commonly tested options, the logistic model had the lowest variation and fewest assumptions.For example, the Pella-Tomlinson model predicted a MSY between 4.9 and 6.7 tons/km 2 /year and required estimating an unknown recruitment parameter while the Schaefer model predicted yields with less variation between 5.5 and 5.7 tons/ km 2 /year and did not require the recruitment estimate.

| Fish catch analyses
Catch analyses focused on differences in (a) effort calculated as numbers of fishers who fished on that day divided by size of fishing area (fishers/km 2 ), (b) catch per unit effort (CPUE), as total catch divided by numbers of fishers who fished (kg/fisher), (c) catch per unit area (CPUA), as total catch divided by the size of fishing area (kg/km 2 ) and, (d) income as summed CPUE times price per fish category (US$/fisher) over time, between village landing sites, and fisheries management systems.Lost yields were derived by subtracting site yields from stock-established fishable maximum sustainable yield (MSY) (kg/km 2 ) while lost income was calculated as lost yields times the price (US$/km 2 ).Data pooled into months failed Shapiro-Wilk W normality test (<0.0001).Therefore, paired comparisons of yields and income by community and fisheries observers used Kruskal-Wallis tests.However, log transformation (log 10 ) of monthly pooled data passed normality test and enabled using ANCOVA statistics to test for effects of site, observer, time, and all interactions.We plotted the expected daily CPUE and incomes as a function of fishing effort for existing (1.8 tons/km 2 /year) and optimal yields (2.98 tons/km 2 /year) by dividing the annual catch (212 days/year) by the number of fishers.These were compared to the official daily poverty (2.9US$/day), personal (10.5US$/day), and family (17.1US$/day)living wage thresholds for Kenya in 2020.

| Stock biomass estimates
Recovery of fish biomass in the regional closures indicated changes in biomass over a 45-year closure period for total, fishable, target and non-target biomass (Figure 2a).The values of K declined along this gradient from 1278.0 ± 166.8 (SE) kg/ha for total, 1193.0 ± 178.4 kg/ha for fishable, 814.1 ± 135.7 kg/ha for target, and 372.1 ± 81.9 kg/ha for non-target biomass.Most of the biomass was composed of resident taxa, and this group had a K of 1075.0 ± 141.1 kg/ha whereas migratory taxa had a K of 213.5 ± 88.3 kg/ha.Pooling visual census data into fisheries management categories and changing units to tons for fisheries production purposes, indicated the highest total biomass in high compliance closures (117.8 ± 4.0 tons/km 2 ), followed by the national gear-restricted reserve (67.9 ± 5.2 tons/ km 2 ), community closures (37.9 ± 10.6 tons/km 2 ), and fishing grounds (27.8 ± 10.5 tons/km 2 ) (Table 1a).These values were reduced slightly for fishable biomass.Target biomass was, however, lower but still highest in high compliance closures (66.2 ± 2.5 tons/ km 2 ), followed by the national reserve (24.7 ± 1.3 tons/ km 2 ), community closures (14.0 ± 8.1 tons/km 2 ), and fishing grounds (9.0 ± 5.0 tons/km 2 ).

| Stock growth estimates
The values of r among various fish categories were not highly variable with r = 0.12 ± 0.004 (SEM) for total, 0.10 ± 0.04 for fishable, and 0.14 ± 0.05 for target fish.However, an estimate of 0.10 ± 0.10 for non-target biomass had higher variation (Figure 2b,c; Table S1).The recovery of resident taxa was 0.12 ± 0.04 for both total and fishable, and 0.17 ± 0.06 for total target biomass.Recovery rates of the migrant taxa was more variable at 0.10 ± 0.10.This variation is attributable to their schooling behavior and subsequent patchy distribution in space (relative to the transect size) rather than variation over time.

| MSY yield estimates
Using the logistic model solution, the MSY estimates for each category suggested the highest MSY was 3.8 ± 0.5 tons/km 2 /year for total biomass but lower at 2.98 ± 0.45 and 2.85 ± 0.47 tons/km 2 /year for fishable and target biomass, respectively (Table 1b).MSY for migrant families target taxa was lowest at 0.53 ± 0.22 tons/km 2 /year.

| Expected stock-based yield and recovery estimates
The Mpunguti marine reserves that restricted fishing gears had the highest and most sustainable yields based on stock biomass estimates (Table 1c and Figure 3).Production estimates in the reserve were greater than community closures and lastly in the fishing grounds.The yield estimate in the marine reserve was close to the estimated MSY for total and fishable biomass but less than MSY for target biomass.In contrast, fishing grounds were estimated to be losing considerably more yield or an average of 1.10-1.75tons/km 2 /year of the fishable and target yields relative to MSY, respectively.Non-target fish yields were, however, closer to the MSY or losing 0.13 tons/ km 2 /year.This suggests overfishing of the migratory, fishable, and target but less overfishing for the non-target taxa.
Recovery times to MSY among the main fish and management categories suggest high variation from 3.4 to 16 years (Table 1d).Recovery times of target fish were longer at 5.7-16.5 years compared to 4.8-7.1 years for fishable biomass.Studied fish categories varied notably in their recovery times.
F I G U R E 2 Space-for-time estimated recovery of fish biomass in regional closures for total, fishable, and target biomass.Presented for all families combined, resident, and migratory families.See Table S1 for best-fit model coefficients.

| Fisheries catch patterns
Fishing effort, CPUE, CPUA yield, revenue and their interactions indicate significant observer, site, time, and interactions effects (Table 2).Time alone was only weakly significant for CPUE.Observer was significant for effort, per area yields, and revenue.Community observers recorded higher effort, yields, and revenue than fisheries officers.Community members were measuring larger catches than officers, but this varied by site, as indicated by the significant observer/site interaction.The largest observer disparity was observed in Mkwiro, where the community recorded a declining catch while officers recorded a rising temporal trend.Measurements at Kibuyuni indicated a downward trend in yield estimates for both observers and the difference between observers declined over time.Analysis of the distribution of the catch data found communities had higher median catch, lower skewness, and kurtosis than officers (Table 3).Thus, while all catch data had flat distributions and right skew, the community's measurements were closer to normally distributed than the officers.
Partial plots that account for multiple variables comparing observers indicated the complex relationships for sites and time among the metrics and landing sites (Figure 4).Per area spatial trends in yields were higher in the eastern sites of Mkwiro, Kibuyuni, and Wasini or nearer the deep Pemba Channel, the managed Kisite park and reserve, and open ocean (Figure 5).Vanga and Jimbo landing sites to the west had notably lower yields associated with a further distance away from the channel, the national park and reserve, but closeness to the international border and a paved road.Partial plots of temporal trends in yield, CPUE, and revenue largely reflected fishing effort trends.There was higher CPUE and revenue in sites with declining effort (Mkwiro, Kibuyuni, and Jimbo).Vanga and Wasini reported a small rise in effort that led to declining CPUE and revenue in Vanga but not in Wasini.Vanga and Jimbo revenues were below the Kenyan poverty line (2.9US$/day in 2020) while Mkwiro's income was above the individual livable wage threshold at the end of the sampling period (10US$/day) (Figure 6).
Given the estimated potential or MSY from the stock assessment, the lost yields and revenues were estimated (Table 3).The above spatial patterns indicate that community observers recorded fewer losses because of their higher yield estimates.Nevertheless, all communities were estimated to be losing potential yields with an average yield loss of 1.4-1.9tons/km 2 /year depending on the observer's estimates.These losses of yields were notably higher in Vanga and Jimbo where the lowest loss T A B L E 1 Fisheries stock-based estimates of (a) mean biomass categories (tons/km 2 /year ± SE) in each management system; (b) maximum sustainable yield estimates (MSY, tons/km 2 /year ± SE) for biomass categories and fish family group (see Figure 2) calculated as MSY = rK/4 where r and K are determined from logistic models fits presented in Table S1; (c) predicted yields (tons/km 2 /year ± SE) from the logistic model for biomass categories and management; (d) recovery time (years ± SE) for biomass categories and fish family group.

| Causes of low yields
Fisheries productivity estimates for MSY and realized catches were low by both stock recovery and fisheries yield assessments.The precise causes of this low production are not known but findings support our concern that the semi-enclosed bay has lower than the average Western Indian Ocean (WIO) production possibly via reduced physical energy in the forms of waves and currents.The low production findings contrast with previous production estimates using a similar space-for-time substitution method and more widely distributed WIO data from more fisheries closures.The study concluded that fishable and target fish production in the WIO province had an r of $0.23 (McClanahan, 2022).Many reef closures in the broadscale compilation were in more ocean-exposed environments compared to reefs leeward of the large Pemba and Zanzibar islands in this more site restricted compilation.We believe leeward locations have lower physical energy with consequences for benthic and fisheries production (Gove et al., 2015).When viewed from the literature on a global scale, however, low production may be common as indicated by the distribution of fish biomass recovery estimates in the absence of fishing along gradients of latitude, coral cover, sunlight and net primary production (McClanahan et al., 2019).A global compilation of coral reef biomass and recovery rates estimated an r of $0.07 and slow recovery, requiring 30 and 60 years (MacNeil et al., 2015).Knowing r is important because recovery rates have the largest effect on estimates of MSY, as per Equation ( 2) F I G U R E 3 Plots of differences between observed and fisheries stock estimates of maximum sustained yield estimates (tons/km 2 /year) for biomass categories (total, fishable, target and non-target) and different fish families (all, resident, and migratory taxa) by management type namely high compliance, marine reserve, community closure and fished areas.Estimates of maximum sustainable yields (MSY) are presented for each category as estimated in Table 1b.(McClanahan, 2018b).The current state of the literature suggests some difficulties with estimating and generalizing production in tropical reefs without knowledge of r, K, and fishing effort (i.e., Newton et al., 2007;Zamborain-Mason et al., 2023).The semi-enclosed bay located leeward of Pemba and Zanzibar islands is likely to have low physical energy of waves and currents, thereby causing the low fisheries production.Water flow is a strong predictor of benthic production and likely to vary with ocean exposure (Lowe & Falter, 2014).It is expected that coastal geomorphology and as well as ocean basin oceanography create differences in productivity at ocean basin scales of the Caribbean, Indian, and Pacific Oceans (Carpenter & Williams, 2007;Falter et al., 2013;Mumby et al., 2013).For example, fisheries yields in the Caribbean are reportedly low (i.e., <1.5 tons/km 2 /year) and associated with lower physical energy in terms of tidal range ($0.2 m), waves, and currents (Koslow et al., 1994).Influences on production appear to vary at many spatial scales according to environmental conditions that are currently only partially understood (Gove et al., 2015;McClanahan, D'Agata, et al., 2023;McClanahan, Darling, et al., 2023).Yield potentials will vary even when the variability in effort and capture rates are not considered.Consequently, the importance of knowing production and MSY variation among geographies and fisheries management areas.Kenya's ocean-exposed fringing reefs and the T A B L E 2 ANCOVA results for fisheries catch assessment of site, time, observer, site Â time, observer Â time, and observer Â site for (a) effort (fishers/km 2 /day), (b) catch-per-unit-effort (CPUE, kg/fisher/day), (c) yield (tons/km 2 /year), and (d) revenue (US$/fisher/day).studied ocean-protected bay being good examples of this geographic variability at the moderate scale of $100 km of coastline.Given the variability, further research will be needed to evaluate the causes and consequences of variable production.Estimating production has been difficult but provides critical information for improving subsequent yields, food security, and income estimates (McClanahan, D'Agata, et al., 2023;McClanahan, Darling, et al., 2023).Kenya's ocean-exposed fringing reef to the north of our studied bay has an estimated MSY between 5 and 7 tons/ km 2 /year, which suggests <4 years to recover biomass to MSY in the absence of fishing in these overfished fisheries (McClanahan et al., 2016).Similar or higher yield rates are commonly reported from tropical nearshore landings, but it is seldom known if they are sustainable (Dalzell, 1996).Historically, fisheries catch studies rarely undertake stock assessments and therefore stocks, production, and optimal yields are infrequently known (Dalzell, 1996;Samoilys et al., 2017).In these cases, MSY estimates were based on effort-catch equilibrium models, which typically overestimate yield potentials in the absence of realistic r and K estimates (McClanahan & Azali, 2020).For example, two global compilations found the mean landed fish had high variation of 4.8 ± 3.5 (SD) tons/km 2 /year from 67 studies (McClanahan, 2018a;Samoilys et al., 2017).A more individualized reef biomass modeling approach estimates a mean of $5.5 tons/km 2 /year but a larger range of 2.5-20.6 tons/km 2 /year (Zamborain-Mason et al., 2023).In catch landing studies, fishing effort and measures of ocean energy and productivity were seldom recorded, which makes it difficult to evaluate their influences on this high variability.Modeling studies suggest ocean temperature and productivity, fisheries management, distance to people, coral cover, and reef characteristics influence biomass and production (McClanahan et al., 2019;McClanahan, D'Agata, et al., 2023;McClanahan, Darling, et al., 2023;Zamborain-Mason et al., 2023).

ANCOVA test terms
MSY determined from effort-catch rates and equilibrium assumptions in East Africa have predicted MSY at >10 tons/km 2 /year (Mkenda & Folmer, 2001;Tuda & Wolff, 2015).These results might suggest underfishing and a need for more effort to achieve MSY.However, a Kenyan study monitoring fishery where stocks were below MSY (B msy ) and exhibiting high initial yields of $10 tons/km 2 / year found a long-term $3% per annum decay in yields in the absence of fisheries closures (McClanahan, 2021).Between 1995 and 2020, per area yields declined from $10 to $4 tons/km 2 /year where stocks were $20-25 tons/km 2 T A B L E 3 Site summary mean (±SE) of yields (tons/km 2 /year), revenue (US$/km 2 /year), lost yields (tons/km 2 /year) and lost revenue (US$/km 2 /year) estimates plus respective Kruskal-Wallis tests of significance between community and fisheries observers.or half the B msy (McClanahan, 2018a(McClanahan, , 2021)).Therefore, fisheries models using non-equilibrium assumptions overestimate optimal effort and sustainable yield potentials in the absence of reasonable r and K estimates (McClanahan & Azali, 2020).Effort-CPUE relationships are therefore not fixed or at equilibrium but expected to change slowly when overshooting the capture limits set by environmental production.These studies suggest that stock-based surplus production models are better than catch for estimating sustainable production.

| Comparison of stock and catch methods
Our study was comprehensive in evaluating and comparing production by both fisheries stock and catch assessment methods.The stock recovery information needed to make comparisons was provided by replicate fisheries closures of different ages and repeated sampling.The two approaches produced estimates that were largely in agreement.Measured stocks were <B msy and catches were proportionally lower and related to stock levels in the studied management categories.Moreover, effort was above the levels needed to achieve MSY at the measured biomass and, as expected, declining.Declining effort was associated with rising CPUE but not catch per unit area, which is also an indicator of overfishing.Effort would need to decline further for stocks to recover the lost per area yields and therefore exhibit a considerable time-delayed response given the reported slow recovery rates (Kerwath et al., 2013).Finally, overfishing was evident in all locations except the Mpunguti marine reserve where higher fish stocks were associated with gear restriction policies and enforcement.
Here, some loss of optimal biomass of the higher priced migratory taxa was observed but not the lower priced resident taxa.These multiple sources of evidence indicated that the stock and catch methods aligned, especially when measured by the more thorough community data recorders.

| Biodiversity refugia and productivity tradeoffs
The study indicates potentially complex interactions between resilience (resistance and recovery), biological diversity, and productivity.The implication of this study is that the semi-enclosed bay provides resistance to acute thermal stress but at a cost of high production and fast recovery of fish biomass.Thus, resistance may be either unrelated or possibly inversely related to recovery (McClanahan et al., 2012).This makes it challenging to evaluate climate resilience as a single compositive variable but rather as having elements of avoidance, resistance, and recovery (McClanahan, D'Agata, et al., 2023;McClanahan, Darling, et al., 2023).Similarly, high diversity appears to be associated with low productivity contrary to the positive diversity-productivity hypothesis (Hooper et al., 2005).The likely reason is that production is determined by primary producers and associated environmental factors and not the consumers studied here.Therefore, many species may benefit from the reduced physical ocean energies and lower acute stress rather than high primary production.Consequently, the studied area is best classified as either an avoidance or resistance rather than a recovery refugia (West & Salm, 2003).The study was not intended to test these ecological refugia hypotheses, but this understanding may contribute to help better planning and managing fisheries in avoidance and resistance refugia.Practically, the failure to acknowledge environmental gradients and low production can easily lead to overfishing and an impoverished fishery.Where inhabitants rely on fisheries resources, the long-estimated recovery periods create hardships but also opportunities for economic diversification.Clearly, if the conditions in rural Kenya are representative, extreme overfishing and the loss of yields and detrimental changes to the reef ecosystem are likely to be common in populated and resource-demanding locations.Moreover, the international border and associated conflicts appears to have heightened challenges of management compliance (Tuda et al., 2019).For example, destructive effects of dynamite fishing on corals and reef structure were observed in the Vanga and Jimbo fishing grounds near the international border.The Kenyan and Tanzanian fisheries management and enforcement systems differed, and this can lead to challenges with cooperation, and potentially short-term complianceavoidance and economic opportunism among stakeholders (Nguyen et al., 2018).These reef fisheries are unproductive and these villages are therefore highly reliant on the pelagic fisheries.

| Benefits of community data recorders
Comparing recordings of community members and fisheries officers exposed frequently unconsidered sources of variability in fisheries landings.Based on the results, communities should be able to produce high quality landing data.Recording higher effort, more days, and higher yields suggest that community employees had better coverage and representation of their fishery.Hiring based on exam results may have helped to identify the most capable people.Chosen people did, however, also require training and supervision to correct some errors identified during the early periods of data collection.Among the early problems were fishers not wanting their catch weighed, double counting fishing effort when more than one trader purchased fish, and data recorders not reporting zero catches.These problems were all rectified in the first several months by supervision and explaining the process to recorders, fishers, and traders.Thus, engaging and training communities can help establish cooperative and community-based fisheries management that reduces travel costs and other limitations associated with employed fisheries officers.
A community approach to catch recording may not always work.For example, two landing sites in the region, Majoreni and Shimoni, did not participate.The lack of participation likely underestimated the yields, but these villages mostly captured fish in other unmapped fishing grounds.There was, however, some overlap.For example, a small number of fishers from Shimoni were observed landing fish caught in the Mpunguti marine reserve before dawn.Night fishing and landings are seldom officially recorded, which leads to universal underestimates of actual yields.It is, however, more likely that community recorders would potentially measure these catches by knowing and responding to local fisher's behaviors.Syncing fisher and recorders behavior can be a common problem of fisheries assessment.The participatory community approach appeared successful, which may eventually lead to broader participation and co-management.

| Sustainability needs of climate refugia
The study was an early effort to evaluate fisheries in a climate refugia and to expose challenges created by low production.Given that low production may be widespread (MacNeil et al., 2015), reducing biomass through excess effort and the associated loss of biodiversity is expected elsewhere (McClanahan, 2022).For example, the frequency of islands, rifts, and deep water canyons are less common in the western Indian Ocean than the Central Indo-Pacific province.Yet, this geographic heterogeneity in combination with sea level changes have played pivotal roles in promoting high biodiversity (Barber & Meyer, 2015;Gove et al., 2015).For example, we expect more climate refugia in the more geographically variable Central Indo-Pacific Archipelago province.Studies support this prediction of less acute thermal exposure, coral sensitivity, and abundant climate refugia in the Central Indo-Pacific Archipelago (Beyer et al., 2018;McClanahan et al., 2020).Therefore, there is the potential for low fisheries production where geographic and environmental forces promote avoidance and resistance refugia.
Maintaining high fish stocks is seen as a primary way to reduce the detrimental impacts of climate change.For example, a modeling study suggested that fisheries restrictions implemented in the WIO province could increase coral cover by $15% in 2050 relative to a business-as-usual scenario without restrictions (McClanahan & Azali, 2021).This effect is expected if the resident benthic-attached fish populations are maintained, many of which are not targeted or high-priced (McClanahan & Muthiga, 2016).Therefore, some combinations of gear restrictions, as observed in the Mpunguti reserve, combined with appropriate species restrictions could generate the needed resilience.Recovery of the currently overfished reefs is clearly a high priority to secure the future of this refugia and to increase food and incomes.MSY estimates recommend modest sustainable fishing effort when fish stocks have recovered to B msy (Figure 6).While reducing fishing effort and gear is a challenging prospect, past efforts have shown that co-management and gear restrictions have improved fish populations in nearby Tanzanian reefs (McClanahan et al., 2015).Creating community support, alternate livelihoods, and economies that reduce the pressures to fish should encourage participation and compliance.Regardless, fisheries effort and production policies and expectations in climate refugia need to be modest to promote climate resileince and maximize yields and revenues.

F
I G U R E 4 Partial plots of modeled (a) effort (fishers/km 2 /day), (b) CPUE (kg/fisher/day), (c) yield (kg/km 2 /day) and (d) revenue (US $/fisher/day) by the two observers (community and fisheries officers) in each landing sites as a function of sampling time.ANCOVA results presented in Table 2. Plot of the fish catch (mean ± SE) at the five landings sites relative to the stock-based estimate of fishable MSY.Presented as estimated by the two observer groupscommunity recorders and fisheries department personnel.
Additionally, median, skewness, and kurtosis by observer type for all sites pooled is presented.Sites arranged from northwest to southeast of distance from the international Tanzania-Kenya border.Abbreviation: NS, no significance. Note: Plots of predicted daily yields and revenue of individual fishers as a function of fishing effort.Presented for stock and yields at the time of the study (1.8 tons/ km 2 /year) and the fisheries stock-based estimate of MSY (2.98 tons/km 2 /year).Shown are Kenyan poverty, individual, and family livable wages thresholds in 2020.
F I G U R E 6