Fishing ground expansion and its consequences for estimates of sustainability in artisanal coral reef fisheries

Introduction

Achieving sustainability in fisheries requires balancing natural production with resource extraction, but this goal is complicated by conceptual, technical, and implementation challenges (1–3). Ecological production is constrained by energetic limits set by solar and physical inputs (4, 5), but measuring

sustainability is often undermined by incomplete data, poor or untested assumptions, and reliance on proxies when data are missing (6, 7). This study explores the potential for common fisheries data collection and projection methods and policies to produce recommendations that will lead to expansionist policies and potential overshoots between ecological and capture production.

Fishing ground expansion is frequently proposed to mitigate overfishing by dispersing effort and promoting nearshore recovery (8, 9). However, expansion requires new skills, capital, and adaptation to offshore conditions, making it costly and risky for many artisanal fishers (10–12). Moreover, expansion alters spatial effort distribution, the source of production and composition of the catch, which can complicate per-area estimates of effort, yield, and sustainability (13). Coral reef fisheries are particularly vulnerable to poor sustainability estimates because expansion shifts effort from productive nearshore habitats to lower-yield offshore deeper benthic and pelagic habitats (4, 14).

Traditional surplus production fisheries models assume constant stocks, a bounded fishing area, and a stable equilibrium between effort and yield (15). Yet, these assumptions can be inaccurate in complex, multispecies systems where ecological reorganization can influence production as stock biomass changes (16, 17). Technological drift, subsidies, gear competition, and behavioral responses further obscure effort–catch relationships (18, 19). Consequently, sustainability assessments may need to consider the dynamics of changes in effort, fishing area, gear usage, and ecological production. The outcomes of past static or single-variable models and assumptions of constant area and production are thereby tested using catch time series from nearshore areas of Kenya's fringing reef. In particular, the consequences of various models for changes in fishing area associated with national policies to expand fishing effort and capture area are discussed (20).

This study evaluates the long-term effects of fishing ground expansion on sustainability estimates in Kenyan artisanal coral reef fisheries. The study is based on the hypothesis that expansion dilutes per-area effort, capture efficiency, and in the long term the increased capture potentially fails to increase either employment or catch. Moreover, past effort–catch fisheries models overestimate ecological production, which promotes expansion policies and leads to excess effort and the collapse of the fishery. Using 23 years of Kenyan nearshore fisheries and ecological data, this study examines whether expansion achieves its intended recovery and sustainability goals or promotes overfishing and collapse. The consequences of fishing area changes for simple future effort and catch projection estimates using Bayesian and catchability assumptions are also examined. Therefore, the study tests the potential of commonly used historical fisheries research methods and associated expansion policies to promote capture rate recommendations not well-aligned with contemporary holistic social-ecological sustainability metrics. This thereby prompts a reconsideration of past methods and estimates used for estimating tropical nearshore fisheries sustainability.

Study context: data collection and regulations

This study assessed fishing area, access, and ecological and fisheries production in the nearshore coral reef fisheries of southern Kenya, which exhibit symptoms of overfishing and has been influenced by both protective and extractive subsidies (21, 22). Fishing behavior was influenced by national regulations and gear and boat subsidies (Figure 1b; Supplementary Table S1), with compliance varying among landing sites. Kenyan marine national fisheries catch data reported from 1950 to 2014 were used to estimate a Schaefer and Fox maximum sustainable yield (MSY) equilibrium effort of ~14.0–19.3 fisher/km²/y and yield of ~10.6–10.3 tons/km²/yr (24). The equilibrium effort and MSY were made prior to fish biomass production r and K being known and were used to constrain yield estimates from catch time series (7, 25, 26).

Catch data were collected through a long-term collaborative catch monitoring project by the Wildlife Conservation Society and Kenya's Fisheries Department. The project aimed to quantify catch production, share results with stakeholders, and promote agreements on restrictions to improve food security, reduce poverty, and conserve reef biodiversity (27). Community participation in data collection and annual meetings facilitated transparency, dialogue, and adaptive management. The primary goal was to evaluate the prevalence of fishing ground expansion and its implications for the recovery and sustainability of nearshore coral reef fisheries.

National and local regulations influenced fishing practices during the study period. The Fisheries Management and Development Act of 2016 introduced gear restrictions and bans on certain gears, while Beach Management Units implemented local bylaws (Supplementary Table S1). Enforcement was inconsistent across landing sites. For example, some rules were interpreted as distinguishing artisanal from commercially manufactured gears. Government subsidies frequently provided boats and gill nets (>45 mm mesh) to replace smaller, more competitive gears. However, these subsidies were often underutilized or sold to non-local stakeholders, and evidence suggests that benefits, including fisher incomes, declined shortly after subsidy implementation (21).

Methods

Study area and sites

The study focused on nearshore coral reef fisheries along ~200 km of Kenya's southern fringing reef, encompassing shallow lagoons and reef flats extending to the reef edge (~0.5–2 km²offshore). Thirteen landing sites with shallow-water fishing grounds (< 20 m depth; ~1–6 km² each) were included, with initial and final fishing area estimates made between 1995 and 2018 (Table 1; Supplementary Figure S1).

Table 1. Fishing landing sites and the initial (1995) and final area (2018) estimates (km²) and total and annual percentage change.

Mapping of fishing grounds (community co-production)

Fishing grounds were mapped via a collaborative process: stakeholder interviews and working groups produced hand-drawn maps, which were harmonized, digitized, ground-truthed with GPS, and converted into polygons (28). Final maps were validated with community members (Supplementary Figure S1). Stakeholder participation increased for the 2018 reassessment relative to initial mapping due to regular (i.e., annual) community meetings. The simple shore-to-reef geometry and midpoints between adjacent landing sites defined boundaries. The biggest changes in the area were to estimate the distance that fishers traveled from the reef edge (29). Some communities created small permanent closures (tengefu, < 1 km²), which could locally reduce fishing area (Table 1, Supplementary Table S1) (30).

Fish biomass censuses

Fishable biomass was estimated from 239 underwater visual censuses across 47 sites (1987–2024) using 500 m² belt transects previously described (31). Censuses spanned the reef profile from shallow lagoonal habitats at neap low tide to ~20 m depth, quantifying fish >10 cm total length (≈50% target taxa—snappers, grunts, rabbitfish, parrotfish, larger surgeonfish—and ≈50% non-target taxa). These taxa were also captured at landing sites, enabling calibration between biomass and capture rates (Figure 2).

Ecological production rates were derived from long-term recovery in permanent fisheries closures: r = 0.22 (95% CI: 0.10–0.34), K ≈ 101 t·km⁻² (95% CI: 92.5–109.8), with an upper biomass limit ≈110 t·km⁻² (26). The recovery model estimates B_MSY ≈ 50 t·km⁻² and daily production at MSY ≈ 25.5 ± 4.1 kg·km⁻²·day⁻¹ assuming 220 fishing days per year is 5.6 tons/km²/year (Figure 2b). The model had an upper biomass limit of ~110 tons/km², and therefore no predictions were made above this value.

The ecological recovery model MSY of 5.6 tons/km²/year is nearly half the fisheries' dependent surplus production yield estimate of 10.5 tons/km²/year (24). Ecological fish production should differ from the selectivity of fisheries catch and production as the cessation of fishing allows species to recover without selective fish mortality (23). Moreover, the Fox and Schaefer surplus production based on total catch assumes an MSY at 37% and 50%, respectively, while the ecological recovery model is based on 50% of maximum biomass and uses empirically derived estimates of r and K values. Comparing production from natural ecological recovery vs. fisheries production is a useful because natural recovery is expected to reflect processes not influenced by strong gear and capture selectivity, thereby producing ecological sustainability metrics and more proportional taxonomic composition (16).

The coral reef profiles indicate a large expanse of shallow lagoonal, a shallower reef flat, and a smaller areal surface of the reef edge, which drops steeply to ~20 m before ending in a sand plain that extends further offshore. Therefore, most of the fishing grounds studied are shallow lagoons and reef flats from the low tide water mark onshore out to the reef edge. Fish, such as snappers, emperors, and grunts, school in deep water during the day and many of them feed on the detrital ecosystem of the deeper (>20 m) sand plain. Their schooling behavior can result in high but patchy distributions of fish biomass at depth. Therefore, biomass estimates can sometimes exceed the upper limits of the logistic biomass recovery model developed by McClanahan and Kosgei (26).

Catch monitoring

Observers sampled landed catches during the 18 lunar fishing days per month, averaging 2.8 ± 0.1 observer days·month⁻¹ (30.6 ± 2.0 days·yr⁻¹). Recorded variables included numbers of fishers, boats, gears, catch groupings, and weights (0.1 kg resolution). Price-based groups comprised goatfish, rabbitfish, scavengers, parrotfish, and “mixed catch.” Data entry transitioned from spreadsheets to Kobo Toolbox (https://eu.kobotoolbox.org/); analyses were performed in R 4.4.2 (Figure 3; Tables 2–4).

Table 2. Estimated fishing effort (mean ± 95%CI, fishers/km²/day) at the studied landing sites based on the three area estimates of the initial, final, and interpolated area calculations.

Percentage change from the final and initial sample presented as well as ANOVA or Kruskal–Wallis test of significance as appropriate for the three area estimates using post-hoc Tukey HSD or Dunn's pairwise comparisons test for paired area-estimate tests.

Table 3. Catch per unit area (CPUA; mean ± 95% CI, kg/km²/day) at the studied landing sites based on the three area estimates of the initial, interpolated, and final area calculation.

Percentage change from the final and initial sample presented as well as ANOVA or Kruskal–Wallis test of significance for the three area estimates using post-hoc Tukey HSD or Dunn's pairwise comparisons test for paired area-estimate tests.

^*Tradewinds estimates of area was reduced in fished area over time.

Table 4. Catch per unit effort (CPUE; mean ± 95% CI, kg/fisher/km²/day) at the studied landing sites based on the three area estimates of the initial, interpolated, and final area calculation.

Percentage change from the final and initial sample presented as well as ANOVA or Kruskal–Wallis test of significance for the three area estimates using post-hoc Tukey HSD or Dunn's pairwise comparisons test for paired area-estimate tests.

Metrics and area calculations

Daily metrics included per-area fishing effort (fishers·km⁻²·day⁻¹), total catch (kg·day⁻¹), catch per unit area (CPUA; kg·km⁻²·day⁻¹), and catch per unit effort (CPUE; kg· fisher⁻¹·km⁻²·day⁻¹). Daily measurements were pooled annually; both parametric and non-parametric tests (ANOVA, Kruskal–Wallis, Tukey HSD, Dunn's tests) were applied as appropriate to distributional assumptions (Tables 2–4).

Because area changed between initial and final estimates, per-area metrics were computed using three area baselines: initial, final (2018), and a time-weighted interpolated area. Interpolation linearly weighted the initial and final areas by the relative position of each year within the interval from the initial area start year to 2018.

Statistical modeling

Biomass model

Fishable biomass was modeled using a generalized linear mixed-effects model (GLMM) with Gaussian error distribution and identity link, with site as a random effect and fixed effects for year (1987–2024), expansion status (non-expanded vs. expanded), fished status (fished vs. unfished), observer, and belt-transect depth at low tide. Interaction terms included time × expansion status and depth × expansion status. Partial residual plots are reported by depth (2, 2.75, 5, 10, 15, 20 m); the area-weighted mean fishing depth was 2.75 m (Figure 2). A GLMM (gamma, log link) was also fitted for comparison, but the Gaussian generalized linear mixed effects model results are presented. This model is used to project production beyond the final sampling date in 2022 until 2100 in non-expanded and expanded fishing grounds (see “Bayesian and catchability future projections” section). Model goodness-of-fit metrics are presented in Supplementary Table S2.

Catch analyses

Effort, CPUA, and CPUE trends were evaluated annually across the initial, interpolated, and final area baselines (Tables 2–4, Figures 3, 4). Interpolated effort–catch relationships (effort vs. CPUA and effort vs. CPUE) were assessed using linear and linear-log models for CPUA (Figure 5a) and linear models for CPUE (Figure 5b). Model fit metrics are reported in the figures.

Future catch projections

We explored two simple forward-looking approaches that might be used to make future projections and fishing effort policies based on catch time series data.

• 1. Bayesian multilevel polynomial regression of CPUA over time. To project CPUA(y) forward from the final sampling date 2018 to 2100, a Bayesian hierarchical model was chosen. Model priors, summaries, fit (convergence and diagnostics), and equation terms were provided in Supplementary Figure S2 and Supplementary Table S3.
• The model encapsulated a second-degree orthogonal polynomial gamma family with log link, first-order autoregressive AR(1) process with covariance structure. To account for variation, landing sites were modeled as group-level (random) effects. I used R package brms version 2.23.0 interfaced with stan software. Since CPUA(y_{*i, t*}) had a Gamma(μ_{*i, t*}, ϕ) distribution we used the link log(μ_{*i, t*}) = η_{*i, t*} to make a linear predictor specified as

yi,t= Gamma(μi,t,ϕ)log(μi,t)=ηi,t, where ηi,t=β0+β1L1(t)+β2L2(t)+β3Ai+β4(L1(t)×Ai)+β5(L2(t)×Ai)+ui+εi,t,where εi,t=ρεi,t-1+δi,t

• This polynomial approach aligns with the autoregressive methods that uses polynomial models and past catch time series to predict future catches (32). Posterior predictions were summarized by the 5th, 50th (median), and 95th percentiles draws to express uncertainty.
• 2. Catchability based projections: These are grounded in the Schaefer catch concept: instantaneous catch C_*t* = q·B_t·E_t, with CPUE = qB. Catchability q was estimated from site level linear regressions of CPUE on biomass, computed both with empirical intercepts and under a zero intercept constraint. Effort was projected to 2100 via GLM (Gamma, log link) with time as the sole predictor (Figure 6). Projections capped biomass at B_MSY, with uncertainty bands shown for the lower and upper 95% biomass bounds (Figure 7).

Results

Changes in fishing area

Of the 13 fishing grounds evaluated, nine were measured and classified as expanded. Three did not change, and one decreased in size (Tradewinds), and therefore four sites were included in the non-expanded category (Table 1). The average increase in expanded fishing ground area was 71% over the measured time, or 5% per year. The one shrinking landing site decreased by 27%. For all 13 fishing grounds combined, there was a 48.2% area expansion, 3.3% per year, or from an average of ~2.6 km² to 3.6 km² per fishing ground.

Modeled biomass and ecological production from fish biomass censuses

The GLMM biomass model variables of time or year, expansion status, depth, fished status, and time x expansion status were significant (p < 0.001) and had moderate to high fit for marginal fixed effects (R² = 0.65) and the conditional full model with sites (R² = 0.78) (Supplementary Table S2). The interaction between depth and expansion status was not statistically significant. The partial effects of time, depth, and expansion status were strong, and fishable biomass was higher in the expanded compared to the non-expanded fishing grounds and increased with depth and time (Figures 2a, b). However, the expanded grounds' interaction with time was negative, and therefore the rate of rise of biomass was faster in the non-expanded than the expanded fishing grounds.

Ecological biomass recovery and production rates relative to MSY varied with depth. Only the deeper forereef edge sites had biomass at 15–20 m depths that was above the estimated B_MSY of ~50 tons/km²/y and ecological production rates of 25.5 ± 4.1 kg/km²/day for 220 fishing days (Figure 2b). Some deeper sites had biomasses greater than the recovery model's confidence intervals (~110 tons/km²), making it speculative to estimate their productivity. Depending on the non-expanded and expanded categories, deep sites displayed hump-shaped and declining net production over time, indicating reduced net production as their biomass exceeded B_MSY. Modeled biomass and production at 20 m were above MSY estimates from the beginning of the time series in expanded areas. The non-expanded areas reached MSY levels in 2001.

Among the more extensive shallower depths, recovery rates to MSY depended on depth and time. The biomass during the years 1995–2018 at the estimated average depth of 2.75m had a mean daily ecological productivity of 8.9 ± 2.0 kg/km²/day, which is 35% of the ecological production MSY of 25.5 ± 4.1 kg/km²/day. Modeled fishable biomass and production increased over time in both non-expanded and expanded fishing grounds. Also, the measured capture rate of production declined and matched ecological production rates at various times depending on the depth. This is discussed further in the concluding section of the results where modeled ecological production from biomass and empirical capture rates are directly compared.

Estimates of fishing effort and catch

Fishing effort changes

Area expansion was not linearly associated with an increase in total daily effort and catch (Figure 3). Rather, total effort increased early between 1995 and 2001 but then slowly declined below the initial levels (Figure 4). Specifically, the peak total effort for the expanded area landing sites in 2001 was 46.2 ± 17.2 fishers/day but declined to 18.0 ± 7.4 fishers/day by 2018. The peak total catch per landing site was also recorded in 2001 at 113.0±25.2 kg/day but declined rapidly after 2002 to 52.9 ± 17.3 kg/day in 2018.

Calculating fishing ground areas based on the initial, final, and interpolated areas affected estimates of per-area effort (Table 2), CPUA (Table 3), and CPUE (Table 4) for both area change categories. Effort, CPUA, and CPUE declined as the area estimates increased from the initial to the final calculations. For example, calculated per-area estimates of effort declined by 40% in expanded fishing grounds from 11.6 ± 1.02 to 6.7 ± 0.6 fishers/km²/day if the initial and final were used, respectively, in all area calculation (Figure 3a). The interpolated effort was 8.6 ± 0.9 fishers/km²/day. In the non-expanded landing sites, effort was more constant regardless of the method (8.3 ± 0.8, 8.4 ± 0.7, 8.5 ± 0.7 fishers/km²/day). For all sites combined, the average decline in fishing effort was 28% from 10.1 ± 0.7 to 7.3 ± 0.5 fishers/km²/day, with an interpolated value of 8.5 ± 0.6 fishers/km²/day. This suggests that area expansion was not associated with more fishers overall but with more dilution of effort into larger areas. Rather, per-area effort increased early between 1995 and 2001 but then declined below the initial levels by the end of the time series (Figure 4a). The expanded grounds' peak per-area effort in 2001 was 16.6 ± 7.4 fishers/km²/day but declined to 5.3 ± 2.4 fishers/km²/day by 2018.

Fishery's capture vs. ecological production

If the initial and final area estimates were used in the calculation, fisheries-dependent production or CPUA declined by 42% in expanded fishing grounds from 28.5 ± 2.0 to 16.6 ± 1.1 kg/km²/day (Figure 4b). The interpolated average annual CPUA was 21.6 ± 1.7 kg/km²/day, starting at 31.2 kg/km²/day and ending at 14.9 ± 5.2 kg/km²/day. In the non-expanded landing sites, average CPUA was more constant regardless of the three area-estimate methods at ~27 ± 2 kg/km²/day. For all sites combined, the average decline in CPUA was 26% from 28.0 ± 1.6 to 20.6 ± 1.2 kg/km²/day with an interpolated average value of 23.7 ± 1.4 kg/km²/day.

The fisheries independent production estimate from censused biomass gives an average ecological production estimate of 8.9 ± 2.0 kg/km²/day and an optimal or MSY of 25.5 ± 4.1 kg/km²/day. Therefore, capture exceeded ecological production rates except for the deepest reef edge locations (Figure 2b). Fish production estimates increased slightly over time (see blue and magenta production timelines in Figure 3b), but estimates of capture were above ecological production until the end of the time series when confidence intervals began to overlap.

Capture rates were 3.1 to 2.3 times above the ecological production depending on the year, but the capture production/ecological production gap declined over the time series. For example, peak interpolated CPUA in 2001 was 40.2 ± 13.0 kg/km²/day but declined to 14.9 ± 5.2 kg/km²/day by 2018. Modeled ecological production at 2.75 m depth increased from 10.2 ± 2.2 kg/km²/day to 13.2 ± 2.4 kg/km²/day over the same time (Figure 4b). Therefore, capture and ecological production rates were converging by the end of the time series.

The peak catch was higher in the expanded than the non-expanded fishing grounds, but the expanded area's CPUA declined faster and stayed permanently below the non-expanded values of 31.7 ± 14.7 kg/km²/day after 2002. Therefore, by either the total or per-area catch metrics, area expansion did not compensate for lost catch in the long term. Rather, catch rates declined faster after the overshoot approaching the ecological production baseline faster.

Catch per unit effort changes

If the initial and final area estimates were used in the calculation, CPUE rates declined by 42% in expanded fishing grounds from 1.4 ± 0.1 to 0.8 ±0.1 kg/fisher/km²/day (Figure 4c), whereas the interpolated CPUE was 1.0 ± 0.1 kg/fisher/km²/day. In the non-expanded landing sites, CPUE was more constant regardless of the method at ~1.4 ± 0.1 kg/fisher/km²/day. For all sites combined, the average decline in CPUE was 26% from 1.4 ± 0.10 to 1.0 ± 0.1 kg/fisher/km²/day with an interpolated value of 1.2 ± 0.1 kg/fisher/km²/day.

CPUE changed over time and showed some recovery after a low period from 1998 to 2008 toward a peak in 2009 of 1.6 ± 0.5 kg/fisher/km²/day for non-expanded and 1.4 ± 0.8 kg/fisher/km²/day for expanded grounds. Thereafter, there was more stability in CPUE until the final area estimate in 2018 (Figure 3c). This decline in CPUE arose from the effort overshoot and declining CPUA until 2005. However, the post-2005 recovery of effort and CPUE were not observed in the expanded fishing grounds, particularly if the final area in the expanded category was the basis of the calculation. Therefore, reducing effort in the expanded areas was associated with a recovery of CPUE, but the final CPUE rate was lower than in the non-expanded fishing areas.

Effort-capture rate relationships

The temporal patterns are simplified by plotting the site's fishing effort vs. CPUA for expanded and non-expanded fishing grounds and compared to the estimated ecological production (Figures 5a, b). The effort-CPUA relationships exhibited strong fits, and capture for all sites and years exceeded the estimated ecological production but nevertheless fell below the Fox MSY estimate. The non-expanded fisheries were ~1.6 times more efficient than the expanded area on a CPUA per fisher basis and data fit better to a linear than a logarithmic model. Expanded areas showed higher variation, low catch efficiency, and a more rapid decline in CPUA at high effort. The expanded areas had a greater range of effort and more variability when including the three area estimates. Some of the annual landings fell below the ecological MSY but were still above the estimated ecological production toward the end of the time series.

Effort-CPUE relationships displayed high between-site variability and weak linear fits even for site-level evaluations (Figures 5c, d). Among non-expanded fishing grounds, the site-level effort-CPUE relationships were negative apart from at one site (Kiruwitu). Among expanded sites, the individual sites' effort-CPUE relationship was mostly not significant, but three sites were significantly positive, namely Vipingo, Kanamai, and Mgwani. Neutral to positive effort-CPUE relationships suggest that high densities of fishers in expanded sites were occasionally but not consistently associated with greater CPUE, contrary to the standard fisheries model predictions (Figure 1a). A temporary overshoot of increased effort and high catch was the probable cause, but this was followed by rapid declines in effort, catch, and CPUE (Figure 4). Both effort-CPUA and effort-CPUE relationships were linear, which indicates that the linear model is more appropriate than the logarithmic model in these fisheries (Figure 1).

Catchability relationships and catch projections

Future catch projections using catchability models revealed the challenges of estimating the catchability coefficient q (Figures 6a, b). Using empirical catch and modeled biomass shows that there were differences in the slopes (q) between non-expanded and expanded areas. Additionally, the intercept differed from zero, particularly in the non-expanded sites, which indicates sources of compensation for capture production. Additionally, the non-expanded sites had a lesser slope than the expanded slope, which indicates a weaker reliance on the local biomass production.

Setting the intercepts to zero, the slopes of the non-expanded areas are higher than those in the expanded areas. This would suggest the non-expanded areas are either more efficient in capture, particularly at low biomass, or that catch at low biomass is being subsidized by migration of fish into the fishing grounds. In contrast, capture rates of biomass were more dependent but less efficient in the expanded than the non-expanded area. Therefore, if expanded areas were capturing fish further from shore with more biomass, capture rates were sensitive to minor changes in biomass. There were also notable differences among landing sites, which suggests catchability was influenced by local factors. Nevertheless, the mean responses were used to project catches in both area categories.

Modeled effort using the linear model based on historical trends indicates a predictable decline in effort for the non-expanded areas (Figure 6c). All area-effort estimates suggest a shallower decline in effort in the non-expanded than the expanded fishing grounds (Figure 6d). Moreover, in the expanded areas, the rates of decline differ among the three area estimates but approach a similar asymptote by the year 2100. Overall, the projected future effort or employment potential is lower in the expanded than non-expanded grounds.

Future catch models

The potential for simple future effort–catch projection models to make good policy decisions is explored using GLMM biomass/production and the Bayesian polynomial and catchability models described below.

Bayesian and catchability future projections

Forecast models of the capture (CPUA) and ecological production show that predictions varied with the model assumptions (Figure 7). In the non-expanded areas, the Bayesian multilevel polynomial model suggests a highly variable monotonic decline in CPUA (Figure 7a). The biomass/production GLMM model (blue-dashed lines) extended past 2018 indicates a recovery of optimal capture production by ~2040. After 2040, the net production declines as biomass exceeds optimal production levels. In expanded areas, CPUA also declines slowest when using the initial area and fastest when using the interpolated area. The recovery of production to MSY by the GLMM model is delayed until ~2075 (Figure 7b). Models have very high variability and confidence intervals suggesting many possible outcomes including rises in CPUA above what is ecologically sustainable.

The two catchability model intercepts and slopes also indicate the importance of model choices and assumptions (Figures 7c, d). In the non-expanded area, CPUA predictions indicate an undershoot for the empirical (Figure 7c) and overshoot for the zero intercept catchability equations (Figure 7e) relative to ecological production. The undershoot arises from the higher intercept and low slope derived from the empirical biomass-CPUE relationship (Figure 6a), which causes CPUA to remain below MSY levels and biomass to recover quickly. Therefore, compensatory production subsidizes the capture and underpredicts a final equilibrium. In contrast, the zero intercept and high slope catchability model (Figure 6c) produces an overshoot in catch, but an eventual decline of CPUA to eventually match production by the end of the time series (Figure 7e). In the case of expanded fishing grounds, CPUA is projected to decline and produce a slow recovery of ecological production for both catchability models (Figures 7d, f). This occurs because expansion causes a continuous decline in effort and CPUA by the model assumptions. Specific landing sites are expected to have different projections given the high site variation in catchability.

Discussion

Measuring ecological and fisheries capture rates together offered an energetic balance approach to evaluate sustainability. Seventy percent of landing sites expanded their fishing grounds by about 70% (≈5% annually) over 23 years. Despite this expansion, both total and per-area fishing effort and catch declined. Expansion diluted effort but did not sustain total catch or employment; both fell after the overshoot peak, and CPUE declined to near poverty thresholds before partially rebounding. Gear in expanded areas were less efficient in capture, produced lower yields, and reduced catch, while outcompeting other more effort-efficient fishing gear, and potentially delaying nearshore recovery by ~35 years. Therefore, offshore access did not promote nearshore recovery; and gains were concentrated at deeper reef edges rather than shallow lagoons. Over the long term, the fish biomass censuses and empirical modeling approach did not support the anticipated beneficial outcomes of the expansionist and subsidy policies. Despite Kenya's population growth, per-area effort declined after 2001, suggesting economic constraints set by the poverty threshold or unprofitable fisheries rather than fishing area and production limits (23).

The fish biomass-production model indicated slow recovery of biomass and CPUA in both non- and expanded area fisheries associated with declining effort (Figure 2). There was initially more biomass in the expanded areas at depth, and most recovery of biomass was in the deeper waters off the reef's edge. However, the rate of recovery was faster in the shallow non-expanded areas, which is contrary to near shore biomass recovery predictions for expanded areas. Despite the decline in effort and overall recovery from the overshoot, the balance between production and capture was not achieved for the larger expanses of shallow water. The recovery was slow, because effort and CPUA were maintained above the estimated ecological production MSY threshold. More sustained effort reductions are expected to recover natural ecological fish production and yields composed of the mixed life histories lost from heavy fishing (17).

The influences of immigration, changes in species composition, and optimal catch sizes were likely to differ between non- and expanded areas. In non-expanded areas, the high intercept and shallow slope for the biomass-CPUE catchability relationship indicates that capture was maintained above local ecological recovery production for many years. The effect was more evident in the non- than the expanded area fisheries, which suggests higher growth and immigration in the non- than the expanded areas. Expanded areas, in contrast, had a catchability relationship with a low intercept and high slope that suggests greater and direct dependency of CPUE on locally produced biomass. The potential importance of these processes and their causes will be briefly discussed below.

The effects of area change estimates

Continuous fishing area data are missing, but area was expected to expand the most during the early years, as effort increased and then declined after ~2001. During the early years, effort and CPUA rose together, but eventually CPUE declined to poverty levels (i.e., < 1.5 kg/fisher/km²/day) (26). Thereafter, effort declined steadily, and eventually CPUE partially rebounded and stabilized above the lowest catch overshoot values and between national poverty and minimum wage thresholds. These patterns are sensible from an ecological-economic overshoot theory (e.g., “bioeconomic equilibrium”), but the area-production elasticity would be more convincing if there were continuous area estimates. Measuring fishing area accurately over time is becoming more affordable as vessel monitoring systems (VMS) are adopted. In future, VMS should be able to determine the temporal and spatial elasticity of fishing area, capture production, and their likely control by the economic cost-benefit constraints of offshore fishing. Some previous reported catches in Kenya have exhibited rises and falls that may represent cyclical or economically driven effort and area of capture changes (32). A weakness of this study is that it is limited to only two area estimates but these still provide a reasonable long-term overview of stability and expansion in these fisheries.

Because fishing grounds can change, it is important to revisit area estimations over time. Given enough time, small annual expansion rates of 1% to 5% will eventually produce capture/production inaccuracies and influence optimum catch estimates and sustainability status. For example, past per-area estimates of catch in Kenya have used a constant 800 km² at the national marine fishing grounds in the absence of good empirical estimates (24, 33). The findings here suggest that as area estimates change, so do current and future projections and recommendations for sustainable effort, employment, and catch. Therefore, repeatedly re-measuring area, or applying area interpolations for longer time series, is important to avoid inaccuracies arising from reliance on single initial or final area estimates. Moreover, in small-scale fisheries where humans, wind power, and reef locations drive effort, the patchiness and the scale of expansion require some consideration when making area calculations. The findings do suggest, however, that not all fishing communities responded the same. For example, the variability and changes in fishing gears, local adoption of community closures, economic dependency, influences on the cost of living, and other unmeasured variables may be as influential as an expanding human population. Therefore, expansion and effort–catch responses are not easily generalizable if this case study is typical.

Ecological and economic control

Despite potential area influences, the peak effort in 2001 was 16.6 ± 7.4 fishers/km²/day with a CPUA of 8.8 ± 2.8 tons/km²/year, which is high and variable effort, but within the confidence intervals of past national surplus production estimates (24). However, the rapid declines shortly after this peak indicate that these estimates were economically or ecologically unsustainable. Specifically, yields in 2001 were 4.8 tons/km²/year above the estimated ecological MSY and mostly composed of a few fast-growing and transient pelagic species (Figures 5a, b) (34). The Kenyan minimum wage in 2001 was at a low of US$62 (Ksh 4,800) per month and fishing incomes aligned closely with this minimum wage for 20 days of fishing (~2.4 kg/fisher/km²/day at ~US$1.0 per kg). Since 2001, the rising cost of living has exceeded the rising prices of fish by ~3.2% per year and provoked more fishers to leave the fisheries and search for more beneficial work (23).

Previous analyses have shown that CPUE equilibrates between the poverty and the national minimum and individual living wage levels regardless of the ecological productivity (26). This window of wage stability has been shown in both island and fringing reefs with low and high ecological production, respectively. Therefore, achieving incomes sufficient to meet living standards are drivers of fishers' effort, where minimum tolerable revenue establishes a threshold (23, 35). Human population numbers or ecological production are influential but not the key effort-control for effort-yield relationships. Rather, minimum acceptable incomes are a key negative feedback mechanism in many fisheries (36), which provokes effort reductions that restore and re-equilibrate effort and incomes near or above poverty wages (23, 37).

Sustainability is frequently defined as an energetic equilibrium, and this is more closely aligned with sustaining biomass, production, and biodiversity, metrics often included in sustainability evaluations (38). Here we see a social and ecological contribution to unsustainability. First, capture exceeded ecological production, and second, low catches and incomes drove fishers to reduce effort and presumably engage in more beneficial activities. Fishing should increasingly become an unattractive work option among fishers using gears that are outcompeted by low efficiency seine nets that capture a few small and fast-growing fish species (39). Therefore, as in many fisheries, the challenge is to better align stakeholder livelihood needs with ecological recovery rates and MSY that include species diverse fish assemblages (35, 37).

Temporal dynamics

Differential temporal dynamics were observed in effort-CPUE relationships between non-expanded and expanded areas during the data collection and future projection evaluations (Figures 4, 5, 7). When effort expanded, it went through a cycle of high catch, overexploitation, an eventual decline, and finally a small compensatory recovery in CPUE and CPUA. All of this occurred below the original surplus Fox and Schaefer MSY estimates (24). Thereafter, community biomass r and K values were known and used to lower optimal yield estimates (25). Therefore, overexploitation became evident when using ecological MSY estimates, but the minimum wage threshold that was the likely cause of fishers exiting the sector. Effort eventually declined below minimum wage but did not pass below the poverty threshold. The lower per-unit effort and area capture in expanded compared to non-expanded areas also created challenges for estimating MSY and income status. Effort was diluted with the fishing ground expansion, and CPUA was spread over a larger area. However, by the final measurements in 2018, expansion did not result in greater total effort or employment at the landing sites. Reduced effort after the peak in 2001 was likely to be more important in promoting the recovery of biomass than the dilution effect of expanding fishing grounds. These findings suggest the importance of including ecological production and economic limits when estimating future yield estimates.

Effort-CPUE/CPUA fits

The effort-CPUA relationship showed a strong rise during the early time series but eventually fell close to the national poverty threshold. The behaviors differed between the non-expanded and expanded areas. Specifically, the non-expanded areas did not exhibit the predicted downward inflection or density-dependent saturation effect predicted by surplus production models. Rather, effort rose and reversed and declined linearly once individual catches reached the poverty threshold. In contrast, expanded areas exhibited both poor catch efficiency and a downward deflection at high effort. Capture rates in some expanded sites also declined to the estimated ecological productivity after the peak catch in 2001. Therefore, effort in expanding areas seems less efficient in capture and responsive to declines in income.

Several factors likely created the high production and ecological resilience of these fisheries. These include weak or slow negative feedback between effort and catch, spill in or migration of fish from less fished locations, a fish community where faster are replacing slower life histories and exiting of less resistant fishers when catch incomes decline below their living needs. In total, these factors may produce social-ecological resilience but also potential lags between changes in effort and CPUA that are not well-represented in equilibrium surplus production models without estimates of biomass r and K (25). These ecological processes combined with fishers' persistence when receiving low incomes have likely created hyperstability in catch, particularly in the non-expanded areas.

The findings here raise a concern that many fisheries-dependent yield estimates in the tropics are too high. For example, Dalzell and Adams (40) summarized coral reef yields in the Pacific and concluded that average multispecies MSY was 16 tons/km²/year (ranging from 6 to 20 tons/km²/year). Similarly, high catches have been reported in Kenya prior to 2001 (41), and MSY values estimated at ~10 tons/km²/year (24). However, these values are shown here to be both ecologically and socially unsustainable as they exceed natural multispecies recovery rates and reduce CPUEs to below living wages. Past studies in Kenya preceded recent catch time series methods that use catch-based r and K estimates to constrain yield potentials (7). How MSY catch estimates (CMSY) compared to fish ecological recovery r–K is still largely untested. Research using Kenyan biomass and catch data indicate lower MSY estimates when ecological r–K are used in surplus production models (25). CMSY are expected to overestimate ecologically sustainable estimates because catch selectivity extirpates species with slow life histories (42). Estimating MSY using fisheries and not ecological production is associated with losses of fish species, diverse gear use and livelihoods, and degradation of ecosystems (43). More realistic MSY values that consider ecological concerns are expected to be < 6 tons/km²/year but will vary with environmental, geographic, and species composition.

Fisheries work vs. employment

National fisheries statistics estimate fishing effort as the total number of people “employed” in the fishery and total yields landed. However, daily effort is lower than fishers enumerated by national employment fisher count procedures. For example, daily active fisher numbers declined after 2001 in our study while official Kenyan national statistics showed an exponential rise in total employed fishers (~14,000) until 2012 (24). Therefore, it is likely that people continued to enter the fishery for employment, but the subsequent poor incomes resulted in infrequent activity, or fishing considerably less than the traditional 220 days per fisher per year. Fishers in Kenya live in multiple-livelihood households and rely heavily on informal employment, and therefore fishing is one of their many work options. Formal fisheries employment typifies work in wealthy industrialized countries, but it is largely informal in tropical and agrarian countries.

Fishers' daily effort decisions change and will differ from calculated assumptions and therefore produce errors in both optimal effort and yield estimates. Unfortunately, assumptions of fishing day frequency used to compile national statistics will make it difficult to estimate effort-yield relationships that reflect the actual daily, lunar, seasonal, and annual cycles of fishing and other competing work options. Moreover, daily effort is less than the estimates of numbers of people employed annually, which will result in inaccuracies and frequent catch overestimates in catch reconstructions (44). These errors will influence the ability to correctly determine the direction of long-term trends when fishing employment trajectories diverge from active fishing. This divergence should be particularly problematic where fishers live among families with high dependence on informal work and multiple livelihoods. Consequently, I propose that catch overshoots and declines in effort and yield reported here are more common than calculated for official national, international, and reconstructed statistics i.e., (45) Sea Around Us Project; [https://www.seaaroundus.org]. This is particularly the case where fishing work and employee work calendars are not stable and formalized.

Benefits of a multivariate assessment approach

Isolated fisheries metrics commonly used to estimate sustainability were shown here to be problematic. For example, CPUE was not a good metric of effort, fishable biomass, or sustainability when comparing differences between sites and expansion categories. CPUE was variable and often decoupled from effort, biomass, and sustainable production. CPUE was also often stable and maintained for some time in the protracted capture/ecological production deficit period. Moreover, there was not a consistent negative effort-CPUE relationship when comparing landing sites (Figures 5c, d). Positive effort-CPUE relationships may represent the benefits of collective or concentrated fishing or a hyperstability effect. The slow reduction of fishing effort was responsible for the more stable CPUE, which hid the effort overshoot and the unsustainable nature of these fisheries.

In the same reef system, the biomass-CPUE relationship was flat in the non-expanded while rising steeply in the expanded areas (Figure 6a). When setting the intercept to zero, the relationship became more aligned with theory, but the slopes differed between landing sites and expansion categories (Figure 6b), thereby challenging any large-scale inferences about sustainability. Another example is the Bayesian polynomial model projections where a model unconstrained by ecological production could potentially make unrealistically variable and high CPUA estimates (Figures 7a, b). Making more bounded and realistic future models will require more stock, environmental, fish production, and socioeconomic data constraints. These findings further identify core problems when relying on catch time series data and models that do not include ecological production and human needs to predict sustainability.

Catch decoupled from production cannot produce realistic yields. Moreover, proxies of yields, such as total catch estimated by total numbers of employed fishers and their CPUE, cannot assess sustainability without estimating fishing area and daily active fishing effort. Common proxies of effort, including human population growth, total employed fishers, gear per landing site, restricted-sampling CPUEs, and unemployment rates, are therefore expected to produce yield estimates that eventually become decoupled from ecological production potential (46–48). These proxies, for example, include a constant ratio of fishers to the human population and total populations of fishers and gear rather than actual daily fishing effort. Estimates are not based on lunar and seasonal cycles of fishers with multiple livelihoods. Fishers make daily, lunar, and seasonal decisions about when, where, and how to fish based on work options and their knowledge of fishing grounds, potential production, associated costs, risks, and their personal and household needs. In poorly studied tropical fisheries, even these nuanced behaviors are being affected by slow social-ecological change, yet this is not well-understood because of the small investments in accurate fisheries monitoring and science.

Limited understanding and evaluations of fisheries variable metrics combined with expanding human needs are expected to promote policies that will exacerbate overfishing and poverty by encouraging overshoot behaviors (36). Large-scale fisheries decisions prioritizing expanding human needs while inaccurately measuring optimal production and the actual costs of capture have led to large-scale catch declines (37, 49). Total production goals combined with expansionist policies create false expectations. When coupled with weaknesses in knowledge, resource management, and decision-making, uninformed decisions can lead to rapid resource declines, collapses, and delayed recovery. Here, expanding fishing grounds produced a short-term positive effort-CPUA relationship that temporarily hid the expected negative effort-CPUE tradeoff relationships and delayed fish biomass recovery (Figures 5, 7). Yet these temporary behaviors produce deficits that were eventually paid back by losses in employment, a rising cost of capture, fish prices lagging costs and needs, changes in catch composition, reduced diversity of fishing gear, and a declining living standard (23, 41).

Policies aiming to maximize both employment and total catch ignore energetic limits and economic trade-offs, yet they are often promoted by fisheries scientists [e.g., (50)]. Pro-poverty and cornucopian policies do not properly recognize the energetic limits of fisheries, trade-offs between total catch and employment, and long-term social-ecological recovery costs. Most of the related capture accounting methods do not account well for the effort reductions and marginal employment that affects fishery capture globally (37). Thus, there is a significant difference in the policies when considering the high numbers of employed fisheries vs. the recognition of their marginal or part-time employment and subsequent widespread overfishing and poverty (36, 51, 52). The use and value of access restrictions are quite different, where the former sees success in high employment while the latter recognizes the widespread poverty of partial employment that does not provide living wages. Expansionist employment policies need to first strengthen ecologically based fisheries science so that it fully acknowledges the area requirements, energetic limitations, and ecological recovery rates required to supply human needs.

Caveats and conclusions

The study was fortunate to measure production and capture rates simultaneously in small and confined coral reef fringing reef fishery over many years. The small model system and modern analytical methods provided reasonable estimates of production and capture. However, like most large and complex-scale social-ecological systems, not all possible influential variables were measured or confirmed for their accuracy or impact. In many cases, potentially important variables were overlooked or were unaffordable to measure over large scales of space and time. The common short-term focus of fisheries science on effort and capture rates will therefore frequently miss slower ecological and fisheries processes, such as hyperstability, that were revealed in this longer time series. Specifically, the findings revealed a collapse of effort and yield when relying on prior surplus production MSY estimate, which was driven by economic limits more than production failures.

Catch time series are often the basis for evaluating fishery sustainability (53). However, without a knowledge of the balance of production and capture, these time series sustainability assessments are contentious (54). In contrast, per-area and production estimates can also be criticized for the limited coverage of the samples and species (55). Here, the time and spatial coverage is large and area interpolation methods applied to test for biases and errors inherent in these estimation methods. Nevertheless, one weakness of this study was that assessments of fishing areas were only made twice. However, the fishing areas could have vacillated over time with the costs, opportunities, hazards, and livelihood pressures on fishers. Therefore, fishing ground areas may be more dynamic than portrayed here. Regardless, a simple fundamental rule of ecological science is to measure and recognize area and production limits to natural resource management (4, 56). Fisheries science can improve data collection to include capture area and therefore production to better measure and evaluate sustainability. Equilibrium models that use or assume a fixed fishing area, fishing days per fisher, poor knowledge of unfished biomass and production, and an effort–catch equilibrium may often fail to accurately estimate what is more likely a shifting community of fish, fisher behaviors, and gear, and thereby optimal per-area yields (42, 49).

During community meetings held with fishers, they recognized and communicated how the cost of living and minimum wages motivated their decisions to fish and their ability to contribute to multiple-livelihood households. In contrast, fishers were unaware of the technical measures and causes of production and yield limits. The regular collaborative informational forums communicated these technical findings to increase appreciation of production potential. Therefore, the fishers' costs and ecologists' estimates of production potential provided a more holistic systems view and dialogue concerning the social-ecological fisheries system. Yet stakeholder information gatherings and dialogues are frequently overlooked or simplified in fisheries science and management. For example, access restrictions are frequently evoked as a limit to fish catch without an appreciation of the complexities of variability in production, decision-making, household and national needs, long-term consequences of short-term decisions, the importance of unfished locations, and other social-ecological contexts. For resources common to meet stakeholder needs, access restrictions are required to negotiate and balance needs and limits (57). Dialogue and informational forums used here can facilitate this important sharing of stakeholder knowledge and negotiations.

Historically, the process of area expansion and associated subsidies has been difficult to evaluate as per its pervasiveness, drivers, and consequences for fisheries yield and sustainability estimates (19). There is a wider recognition of the growing human populations and declining nearshore catches in coral reefs (8, 58, 59), but this study provides a rare empirical case study of the consequences of area expansion. The observations here show the complex and contextual behaviors around expansion and overfishing and how limitations are set by both production, ecological interactions, and fishers' economic constraints. The subsidy's expectation of shifting fishing effort offshore to recover nearshore fish populations failed and will protract the needed recovery.

The post-overshoot slow recovery of stocks was more likely provoked by eventual constraints in ecological productivity and the economic squeeze arising from the cost of living (23). Reduced effort was followed by the slow and broad decline in effort and slow recovery of biomass. Nevertheless, ecological recovery trajectories may present a future opportunity to achieve a balance between natural production, biodiversity, and human needs. Coupling area expansion with nearshore closures might better achieve the proposed nearshore recovery. However, scientific evaluations of their efficacy have been limited by low adoption rates, compliance, and replication requirements (60).

Some key factors and lessons are provided here to assist navigating options to supply human needs. However, as the cost of living rapidly rises throughout tropical and other nations, a new social-ecological context is developing (23, 36). The results here should temper the expectations for increased food and nutrition through increased fisheries access (61). The future will require decisions that are potentially less ecologically tolerant of the poorly informed exuberance of the past. There is a great need to accurately estimate ecological and fish production rates, to avoid single-variable proxies of sustainability, and to set realistic ecological limits to fishing effort and catch potential in the world's fisheries ecosystems. Policies prioritizing employment and total catch too often ignore energetic limits and economic trade-offs, and therefore amplify overshoot and poverty. Future strategies should integrate ecological production, area dynamics, and livelihood thresholds to set realistic expectations and avoid social-ecological collapses (62).

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/frish.2026.1716016/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Michele Luca Geraci, CNR/IRBIM, Istituto per le Risorse Biologiche e le Biotecnologie Marine, Italy

Reviewed by: Khaled Youssef AboueFadl, Aswan University, Egypt

Pamela M. Chiroque-Solano, University of Trás-os-Montes and Alto Douro, Portugal