The fishing sector is a crucial source of nutrition, income, and livelihoods for hundreds of millions of people worldwide (Bejaoui et al. 2019). Global fish production has increased significantly, with per capita availability reaching a record high of 20.5 kg in 2018, up from less than 10 kg in the 1960s (FAO 2017). Projections suggest this trend will continue, with production rising from 171 million t (tonnes) in 2016 to 201 million t by 2030 (FAO 2017). However, the intensive use of aquatic resources has shown that, although they are renewable, they are not unlimited and require sustainable management to ensure their long-term contribution to global well-being (Harris and Roach 2021; Mili et al. 2023). This issue is especially urgent given the overcapacity and declining stock levels in many regions, emphasizing the need for better fisheries management, international cooperation, and advanced technologies (FAO 2024).

The management of transboundary fish stocks in the Mediterranean Sea is a complex issue, as the basin represents a critical biodiversity hotspot hosting about 7%–10% of the world’s marine species despite covering less than 1% of the global ocean surface (Klaoudatos and Theocharis 2026). This is particularly true of migratory species such as tuna, which require a regional approach to assessment. To facilitate management, the General Fisheries Commission for the Mediterranean (GFCM) has established geographical sub-areas (GSAs, Pons et al. 2019a). GSA 13, which comprises Malta, the Lampedusa and Linosa islands, is a key fishing zone in the central Mediterranean bordering the Tunisian coast. The Tunisian fleet operating in this sub-area includes purse seiners, trawlers, and small-scale vessels. In 2010, the national fleet comprised 11 082 vessels, of which 10 388 (94%) were small-scale boats, 321 were purse seiners, and 373 were trawlers (DGPA 2023). Within GSA 13, the fleet includes 215 vessels, with 171 purse seiners (10 of 6–12 m and 152 >12 m), 72 trawlers, and numerous polyvalent small-scale vessels. Tuna purse seiners typically measure between 15 and 38 m in length, with engine power ranging from 74 to 670 kw (DGPA 2023). Small tunas are primarily caught using purse seines, with additional gears including gillnets, traps (madragues), and handlines. The fishing season is highly seasonal, with peaks in landings recorded in May and June (Zarrad and Missaoui 2016). Tunisia, with its significant maritime domain and long fishing tradition, has recognized the potential of its pelagic resources, particularly large pelagics (Mili et al. 2025). These resources were considered under-exploited in the early 2000s (Gaamour et al. 2005; Hajjej et al. 2011; Hattour unpublished¹). This situation changed after the implementation of the National Strategy for the Development of Small-Scale Pelagic Fisheries (NSDSPF) (Mili et al. 2025). The family Scombridae, which includes tunas and bonitos, is of particular importance due to its high commercial and nutritional value (Hattour 2009; Karunarathna and Attygalle 2009). However, these species are characterized by their pelagic and migratory nature, which makes their stock assessment challenging at local and national scales, a limitation that is commonly acknowledged in the literature (Fromentin and Lopuszanski 2014; Dalaut et al. 2025; Klaoudatos and Theocharis 2026). To address this challenge, the presently reported study employs two complementary data-limited methods, LBB and CMSY++, that are specifically designed for data-poor and transboundary stocks (Froese et al. 2017; Froese et al. 2023; Winker et al. 2018). These methods do not require detailed migration data or stock boundary definitions; instead, they rely on length-frequency data (LBB) and catch time series (CMSY++) to estimate stock status indicators, providing robust baselines for regional assessments.

This study focuses on two small tuna species of significant commercial interest in the central Mediterranean: the little tunny, Euthynnus alletteratus (Rafinesque, 1810), and the bullet tuna, Auxis rochei (Risso, 1810). In Tunisia, small tuna landings averaged approximately 2813 t between 1995 and 2010, with A. rochei accounting for 38.9% and E. alletteratus for 27.1% of the landings (Zarrad and Missaoui 2016). Euthynnus alletteratus is a neritic, epipelagic predator found in the Atlantic and Mediterranean (Sant’Ana et al. 2021a). In the Mediterranean, it spawns mainly in summer (Hajjej et al. 2011) and has an opportunistic diet (Falautano et al. 2007). Although some biological studies exist (Hajjej et al. 2010; El-Haweet et al. 2013; Neghli et al. 2019), quantitative stock assessments are scarce and sometimes contradictory. Recent efforts to standardize maturity staging for this species across the Mediterranean highlight the ongoing need for consistent biological data to support robust assessments (Saber et al. 2019). Auxis rochei is one of the smallest tunas (Cayré et al. 1993; Sant’Ana et al. 2021b) and exhibits more coastal behavior and less migration than other tunas (Maguire et al. 2006). Our knowledge of this species biology is limited, with few studies conducted in the Mediterranean (De la Serna et al. 2005; Macías et al. 2008) and in Tunisia (Hattour 2009; Allaya et al. 2015). Furthermore, genetic studies have revealed a complex population structure in the western Mediterranean, with differentiation between Iberian and North African locations. This adds another layer of complexity to stock assessment and management (Allaya et al. 2017; Ollé-Vilanova et al. 2022). There are also no specific ICCAT management measures in place, and the lack of adequate data and stock assessments for these small tunas remains a recurring issue (Lucena-Frédou et al. 2017, 2021; ICCAT 2023). At the international level, small tunas fall under the mandate of ICCAT, although no specific management measures are currently in place for E. alletteratus and A. rochei (ICCAT 2023). The GFCM has established multiannual management plans for demersal species in the Strait of Sicily (GSAs 12–16), but these do not directly address small tunas. At the national level, Tunisian fisheries are governed by Law No. 94–13 on fishing activities. However, according to Tunisian legislation (JORT 1995), there are no specific regulations for small tunas regarding minimum landing sizes or catch limits. This regulatory gap underscores the urgency of the presently reported study to provide scientific baselines for future management.

Assessing the stock of these migratory species is difficult because of their extensive distribution and the multinational fisheries that exploit them. Traditional assessment methods range from global models using catch and effort data (Cadima 2002) to analytical models that include biological factors like growth and mortality (Beverton and Holt 1957; Ricker 1975). The Mediterranean Sea features small-scale, multi-species fisheries and limited long-term monitoring programs, leading to a persistent data-poor situation that hampers the use of standard assessment techniques such as Virtual Population Analysis, statistical catch-at-age models, and surplus production models (Klaoudatos and Theocharis 2026). Therefore, data-limited methods are becoming increasingly important for stock assessments when biological data series are unavailable (Al-Mamun et al. 2022; Sarr et al. 2023) and have been successfully used for various Mediterranean species (Kılıç et al. 2023; Mili et al. 2026).

The presently reported study aims to evaluate the population properties and stock status of E. alletteratus and A. rochei in the central Mediterranean (GSA 13) using two complementary, data-limited approaches. The CMSY++ model (Froese et al. 2021) is a Bayesian state-space method that estimates fisheries reference points (MSY, F_MSY, B_MSY) and stock status (B/B_MSY) from catch time series. The Length-Based Bayesian Biomass (LBB) method (Froese et al. 2018; Wang et al. 2021) estimates stock status indicators and reference points using length-frequency data. By applying these robust, modern techniques to Tunisian fishery data for GSA 13, the study aims to provide valuable insights into the sustainability of exploiting these two key species and support the development of effective management strategies in the central Mediterranean. This study provides the first integrated stock assessment of these two small tuna species in the central Mediterranean using complementary data-limited methods.

Study area and data collection. This study focused on Geographical Sub-Area 13 (GSA 13) of the central Mediterranean, as shown in Fig. 1. The port of Teboulba (35°39'40.6"N, 10°57'33.4"E) was selected as it is of paramount importance as the main landing site for small tuna species in the region (Ben Salem and Franquesa 2005; Hattour 2009; Zarrad and Missaoui 2016). The harbor contributes with 15.43% to Tunisian marine fishery production and accounts for 71% of the total declared fishery production within GSA 13 (DGPA 2023), reflecting its status as the primary landing site for purse seine and small-scale fisheries in the region. The harbor is home to a fleet of 522 fishing vessels, including 74 purse seine vessels, and supports an estimated 2804 direct jobs, 1184 of which are in the purse seine sector (DGPA 2023). Previous studies have shown that over 33% of the total small tuna production across all Tunisian coasts is accounted for by landings at Teboulba Port (GSA 13), primarily due to offshore purse-seine activities (Ben Salem and Franquesa 2005; Hattour 2009; Zarrad and Missaoui 2016).

Fish sampling and data processing. Sampling was carried out from March to November 2022, coinciding with the peak landing period for these species. Specimens were collected directly from the commercial landings of purse seine vessels operating in GSA 13 at Teboulba harbor. A total of 48 surveys were conducted throughout the study period, targeting landings from multiple vessels to capture the variability in catch composition and size structure. The total number of measured individuals (916 specimens of Euthynnus alletteratus and Auxis rochei) is consistent with sample sizes reported in similar data-limited stock assessment studies in the Mediterranean (Demirel et al. 2020; Lucena-Frédou et al. 2021; Kılıç et al. 2023; Mili et al. 2025).

A total of 408 E. alletteratus and 508 A. rochei individuals were collected. Fork length (FL) was recorded for each specimen and is defined as the straight-line distance from the tip of the snout to the fork of the caudal fin (ICCAT 2016). Measurements were taken using a measuring tape for large specimens and an ichthyometer for smaller individuals, with a precision of ±1 cm. Measurements were taken at night during commercial landings to ensure specimen freshness.

Biological parameters. The length at first maturity (L₅₀) for each species was obtained from local studies conducted in Tunisian waters (GSA 13). For E. alletteratus, L₅₀ was set at 42.3 cm FL (Hajjej et al. 2010). For A. rochei, no published values of L₅₀ were available. These thresholds were applied to the length measurements to determine the proportion of immature individuals in the catches. Values from FishBase (Froese and Pauly 2025) indicate an L₅₀ of 41.8 cm for E. alletteratus and 35 cm for A. rochei, based on data from similar latitudes and ecosystems. Local estimates were prioritized to ensure the assessment reflected the specific biological characteristics of the stocks in GSA 13.

Catch data. Annual production data [kg] for E. alletteratus and A. rochei were obtained from landings recorded in the central Mediterranean (GSA 13) between 1996 and 2022. The dataset comprised 27 consecutive years of data for each species and was compiled from the official fishery statistics yearbooks of the Tunisian General Directorate of Fisheries and Aquaculture (DGPA 2023).

Statistical analyses. All statistical analyses were performed using R software (version 4.5.2, R Core Team 2025) within the RStudio integrated development environment (Posit team 2025). Figures were generated using the ggplot2 package in R. Descriptive statistics were calculated for both length-frequency and annual production data for each species. Normality was assessed using the Shapiro–Wilk test. The length distributions significantly deviated from normality for both species (E. alletteratus: p < 0.05; A. rochei: p < 0.05). Consequently, the non-parametric Mann–Whitney U test was used to compare the length distributions between the two species, as it is specifically suited for comparing the central tendency (median) of two independent groups with non-normal distributions, which aligns with our objective of testing size differences between species.

The Shapiro–Wilk test (W) confirmed normality for both species in the annual production time series (E. alletteratus: W = 0.95, p = 0.21; A. rochei: W = 0.94, p = 0.12). Temporal trends were analyzed using linear regression to estimate the average annual change (kg · yr^–1) and the strength of the linear relationship (Pearson’s). This analysis was complemented by the non-parametric Mann–Kendall trend test and Sen’s slope estimator to detect and quantify monotonic trends. Interannual variability was also assessed via the coefficient of variation and visual inspection of time-series plots.

Stock Assessment Methods. The stock assessment of E. alletteratus and A. rochei employed two complementary data-limited approaches, given the scarcity of comprehensive fishery data. The CMSY++ model uses catch time series, while the Length-Based Bayesian (LBB) model relies on length-frequency data. Both models are recognized as robust, low-input assessment methods (Pons et al. 2019a; Pons et al. 2019b; Chong et al. 2020; Lucena-Frédou et al. 2021).

CMSY++ model. CMSY++ is an advanced Bayesian state-space stock assessment method (Froese et al. 2023). It combines information on the stock’s productivity and exploitation history and can account for gaps and missing abundance data, which is its main advantage over other models (Froese et al. 2021; Ferragut-Perello et al. 2023).

This method uses catch (weight) data and, optionally, abundance data together with a prior for resilience (r) and broad priors for the ratio of biomass to unfished biomass (B/K). Using catch data alone, it generates refined estimates of stock status for key fishery reference points, such as MSY, F_MSY, and B_MSY (Froese et al. 2017; Ferragut-Perello et al. 2023). It provides assessments of the stock’s status, relative size (B/B_MSY) and the level of fishing pressure or exploitation (F/ F_MSY) at the beginning, an intermediate year, and the end of the time series, as described by Froese et al. (2021). The ranges for resilience and depletion priors are detailed in Table 1.

The CMSY++ method uses an annual time series of catch data. Prior knowledge of population resilience is required to estimate the parameters of the Schaefer surplus production model: the intrinsic population growth rate (r) and the environmental carrying capacity (K). The following equation [1] governs the basic biomass dynamics:

B(t+1)=Bt+r(1–Bt⋅K–1)–Ct [1]

Where B_t is the biomass at time t, B_{(t + 1)} is the biomass at time (t+1), r is the intrinsic population growth rate, K is the environmental carrying capacity, C_t is the catch at time t, t is the time [year]

This method assumes that r, the initial relative biomass (B₀/K) and the final relative biomass (B_end/K), are known qualitatively (high, intermediate, or low). The value of K is constrained to vary between (maximum catch)/r and 4*(maximum catch)/r or between 2*(maximum catch)/r and 12*(maximum catch)/r, depending on the previous year’s biomass level.

As CMSY++ is based on the Schaefer model, it assumes a symmetric production function. This means that the biomass producing the maximum sustainable yield (MSY) is K/2, and MSY equals rK/4.

To account for reduced recruitment in severely depleted stocks, as predicted by standard stock-recruitment relationships (Beverton and Holt 1957; Ricker 1975; Barrowman and Myers 2000), a linear decrease in surplus production is incorporated when biomass falls below 0.25K, as described in Equation [2].

B(t+1)=Bt+(4Bt/K)×r(1–Bt/K)–Ct [2]

for B_t/K < 0.25

The term (4B_t/K) assumes a linear decline in recruitment below half the biomass capable of producing MSY (Schnute and Kronlund 1996).

For the CMSY++ analysis, prior ranges for the intrinsic population growth rate (r) were defined according to species resilience, in line with the classification of Froese et al. (2017). Both species were assigned a medium resilience prior (r: 0.2–0.8). The CMSY++ algorithm automatically derived prior ranges for carrying capacity (k) from the catch time series using the relationship between maximum catch and r (Froese et al. 2017, 2023). The model then identified viable r–k pairs through an iterative Monte Carlo approach, selecting the best estimates based on their fit to the catch data. All analyses were performed using the latest version of the CMSY++ R package (Froese et al. 2023), which offers enhancements such as parallel processing for quicker execution and improved management outputs, including Kobe plots.

LBB model. The Length-Based Bayesian (LBB) model, developed by Froese et al. (2018), is a method used to assess the exploitation status of fisheries with limited data. It is based on single or multi-year length-frequency (LF) data. It is suitable for species that grow continuously throughout their lives, including many economically important fish (Al-Mamun et al. 2022). The LBB model replaces absolute biomass and age with their relative values (Froese et al. 2018).

LF data can be used to estimate the following using standard fisheries equations: asymptotic length (L_∞); length at first capture (L_C); and fishing mortality (F) relative to each other. These calculations provide important metrics, such as relative natural mortality (M/K), fishing mortality relative to natural mortality (F/M), exploited biomass relative to unexploited biomass (B_MSY/B₀), and relative stock size (B/B_MSY). Additional parameters such as M/K, L_∞, and L₅₀ can be included for greater accuracy, although LBB primarily relies on LF data (Al-Mamun et al. 2022).

These estimates were used to determine the exploitation status of the E. alletteratus and A. rochei fisheries. A stock is considered healthy if B/B_MSY > 1, slightly overfished if 0.8 ≤ B/B_MSY ≤ 1, overfished if 0.5 ≤ B/B_MSY ≤ 0.8, severely overfished if 0.2 ≤ B/B_MSY ≤ 0.5, and collapsed if B/B_MSY ≤ 0.2 (Palomares et al. 2018). An F/M > 1 indicates an overfished stock. In addition, the L_mean/L_opt and L_C/L_Copt ratios were estimated. Values lower than one (<1) indicate growth overfishing, reflecting a truncated length structure and the capture of undersized individuals. A ratio of 95th percentile length to asymptotic length (L_95th/L_∞) close to unity (>0.9) indicates the presence of large fish species (Froese et al. 2018).

The LBB method operates on LF data from catches. It is based on the assumptions that recruitment remains constant over time and that growth follows the von Bertalanffy growth function (VBGF, von Bertalanffy 1938). This enables the analysis of the descending right-hand slope of the LF data (Wang et al. 2021). The model estimates the length at first capture (L_C) that would maximize yield, as well as the ratios of natural mortality to growth rate (M/K) and of fishing mortality to growth rate (F/K). The required biological parameters, including the asymptotic length (L_∞), were obtained from published studies (Hajjej et al. 2011) and FishBase (Froese and Pauly 2025), providing independent priors for the LBB analysis.

In the LBB model, length growth is assumed to follow the VBGF (3), as described by Beverton and Holt (1957) and expressed in Equation [3]:

Lt=L∞(1–e∧(–K(t–t0)) [3]

L_t is the length at age t, L_∞ is the asymptotic length, K is the growth coefficient, and t₀ is the theoretical age at zero length (Froese et al. 2018).

For fully selected fish, the number of survivors at length L (n_L) is expressed as follows in Equation [4]:

nL=nL start ×((L∞–L)/(L∞–Lstart ))∧(Z/K) [4]

where Z = F + M is the total mortality rate. The LBB model uses length-frequency data to estimate the ratios F/K and M/K, thus providing the relative fishing mortality F/M. Absolute M values for E. alletteratus were obtained from Pons et al. (2019b), whereas those for A. rochei were based on scenarios in the literature. Absolute F values were then derived using the formula [5]

F=(F/M)×M [5]

The selectivity of the fishing gear (purse seine in this study) is described by a sigmoid curve in Equation [6]:

SL=1/(1+e∧(–α(L–LC))) [6]

where S_L is the fraction retained at length L, and α describes the steepness of the selection curve (Froese et al. 2018).

The parameters of the selection curve, L_C, α, M/K, and F/K, are estimated simultaneously by applying the following equations [7] and [8]:

N ( L i ) = N ( L i − 1 ) × ( ( L ∞ − L i ) /

(L∞−Li−1))∧(M/K)×(F/K×S(Li)) [7]

C(Li)=N(Li)×S(Li) [8]

where C refers to the number of individuals vulnerable to the gear, N(L_i) refer to the number of survivors at length i.

Given estimates of L_∞ and M/K, the length at which cohort biomass is maximized (L_opt) is calculated as follows using equation [9]:

Lopt =L∞×(3/(3+M/K)) [9]

The optimal length at first capture (L_Copt) can then be derived based on the fishing pressure (F/M) using equation [10]:

LCopt =L∞×(2+3F/M)/[(1+F/M)(3+M/K)] [10]

The relative biomass (B/B₀) and the estimated L_C from LBB provide direct management indicators. If B/B₀ is less than B_MSY/B₀, catches should be reduced. Conversely, if L_C is less than L_Copt, fisheries should target larger individuals (Froese et al. 2018). This method was implemented using the Bayesian Gibbs sampler JAGS (Plummer 2003), run in R (R Core Team 2025) to fit the observed length proportions. The analysis was performed in R using the LBB_33a.R script (Froese et al. 2018) with the default settings from the TropFishR/bsmart package (Mildenberger et al. 2017). The LBB results can also provide priors for the current biomass in the MSY analysis (Froese et al. 2017).

Annual production. Figure 2 shows the annual landings of both species (Euthynnus. alletteratus and Auxis rochei) from 1996 to 2022. The time series reveal markedly different production scales and variability patterns between the two species.

For E. alletteratus (Fig. 2, Table 2), annual production averaged 963 t (median: 960 t), with a standard deviation of 433 t and a coefficient of variation (CV) of 45.0% (Table 2). Landings fluctuated considerably over the study period, ranging from 109 t in 2010 and a peak of 2142 t in 2017. Relatively lower catches occurred in the late 1990s, followed by an increase with peaks in 2000 (1453 t), 2008 (1246 t), and, notably, in 2016–2017 (1592 t and 2142 t, respectively). A period of low catches occurred between 2010 and 2012 after which the stock recovered strongly. Linear regression suggested a modest positive trend of 19.5 t · yr^–1, but this was not statistically significant (R² = 0.112, p > 0.05). The Mann–Kendall test confirmed the absence of a significant monotonic trend (τ = 0.189, p = 0.207), with Sen’s slope estimating a median annual change of 9.2 t · yr^–1 (95% CI: –12.4 to 32.8 t · yr^–1).

For A. rochei (Fig. 2, Table 2), production was lower and more variable, averaging 414 t with a coefficient of variation of 77.3% (Table 2). Landings ranged from 2 t in 1996 to a historic maximum of 989 t in 2003. The most striking feature of this series is the peak observed from 2001 to 2003, with landings reaching 630 t, 933 t, and 989 t, respectively. This was followed by a collapse in 2004–2005, with landings reaching just 176 t and 2 t. Subsequent years showed a moderate recovery, with secondary peaks in 2007 (842 t), 2015 (760 t), and 2017 (910 t); however, these remained below the levels observed in the early 2000s. The series also includes numerous years with very low landings, such as the period from 1996 to 2000 (14–93 t) and in 2009–2010 (61–88 t). The Mann–Kendall test confirmed the absence of a significant monotonic trend (τ = 0.035, p = 0.082), with Sen’s slope estimating a median annual change of 8.0 t · yr^–1 (95% CI: –6.8 to 21.6 t · yr^–1).

Overall, E. alletteratus had an average landing rate approximately 2.5 times higher than A. rochei and exhibited substantially lower inter-annual variability (CV: 45.0%; 77.3%). Neither species showed a statistically significant directional trend over the 27-year study period, although both displayed pronounced multi-annual fluctuations. This contrasting variability suggests that A. rochei may be more susceptible to environmental drivers and fishing pressure dynamics. Meanwhile, the recent post-2010 recovery in E. alletteratus landings warrants continued monitoring.

Size structure of the catches. A total of 508 A. rochei and 408 E. alletteratus were sampled at Teboulba harbor during the 2022 fishing season. Descriptive statistics of lengths are presented in Table 3. The normality of the distributions was tested using the Shapiro–Wilk test. For A. rochei (W = 0.94, p < 0.001) and E. alletteratus (W = 0.91, p < 0.001), the normality hypothesis was rejected, justifying the use of non-parametric statistics (median and quartiles) alongside classical parameters.

The lengths of the 508 A. rochei individuals ranged from 18 to 44 cm, with mean and median values of 35.1 cm and 36 cm, respectively. The distribution was bimodal, with the first peak occurring at 33 cm and the second at 40 cm (Fig. 3). The Shapiro–Wilk test confirmed a non-normal distribution (p < 0.001). Around 42% of the captured individuals were smaller than the size at first maturity (L₅₀ = 32.6 cm), suggesting that a significant proportion of the catch were juveniles. The maximum observed length was 44 cm, and no individuals exceeding this size were recorded in the samples.

The 408 E. alletteratus specimens’ lengths ranged from 38 to 90 cm, with a mean of 61.1 cm and a median of 57 cm. The distribution showed a main mode in the 55–59 cm class and a secondary mode around the 80–84 cm class (Fig. 3). This reflects the coexistence of recruiting adults and large spawners. The distribution deviated significantly from normality (Shapiro–Wilk test, p < 0.001). Only 2% of individuals measured less than 42.3 cm (L₅₀), indicating that the fishery primarily targets mature fish. The interquartile range (IQR = 11 cm) reflected moderate size variability, with individuals exceeding 80 cm regularly present, which favors the conservation of reproductive potential.

A comparison of size distributions using the non-parametric Mann–Whitney test (U) revealed a highly significant difference between the two species (U = 18 507, p < 0.001). Euthynnus. alletteratus attained considerably larger sizes than A. rochei, with medians of 57 cm and 36 cm, respectively (Table 3; Fig. 3). This contrast is noteworthy given that both species are exploited by the same purse-seine fleet, using the same gear, and operating in the same fishing grounds (GSA 13). Despite this identical fishing pressure, E. alletteratus is primarily captured as adults, whereas A. rochei catches include a substantial proportion of juveniles. These contrasting demographic structures support the application of Length-Based Bayesian Biomass method (LBB) to evaluate their respective stock statuses.

Stocks assessment (LBB). The LBB method provided further insights into the exploitation status of the two species (E. alletteratus and A. rochei), revealing markedly different levels of fishing pressure and stock depletion (Table 4).

For A. rochei, the analysis indicated an extremely depleted stock, with current biomass estimated at 8% of the unexploited level (B/B₀ = 0.08, CI: 0.05–0.10). The biomass relative to B_MSY was also low (B/B_MSY = 0.21, CI: 0.14–0.28), indicating that the stock was below the level required to sustain maximum yields. This severe depletion is consistent with a high fishing mortality rate (F/M = 15, CI: 12–19), indicating that fishing mortality is higher than natural mortality. The length structure revealed that around 58% of the captured individuals were mature (≥ L₅₀ = 32.6 cm), suggesting that most of the catch comprises adults. However, a significant proportion (42%) of juveniles were also present (Table 4; Fig. 4A). The length at first capture used in the analysis (L_C= 27 cm) corresponds to the observed prior, reflecting the size at which 50% of individuals become vulnerable to the gear. Although the model estimated a higher L_C (43.5 cm), we retained the observed value for management recommendations to align with current fishing practices. With a L_mean/L_opt ratio of approximately 1.0, the average size of the captured individuals is close to the optimal length, suggesting that the fishery is not severely truncating the larger size classes. However, the combination of a low L_C/L_Copt ratio (0.79) and high fishing mortality (F/M = 15) indicates severe overexploitation, reducing the stock to 8% of its unexploited biomass.

The situation was less severe for E. alletteratus, though still concerning. Current biomass was estimated at 25% of the unexploited level (B/B₀ = 0.25, CI: 0.19–0.32) and B/B_MSY was 0.67 (CI: 0.52–0.87), indicating an overfished stock. Fishing mortality exceeded natural mortality (F/M = 1.4, CI: 1.1–1.7) but remained within the range commonly observed in exploited fisheries. The length structure was healthier, with 98% of the captured individuals being mature. The L_mean/L_opt ratio of 0.85 suggests that fishing selects individuals slightly below the optimal length (Table 4; Fig. 4B).

The contrasting stock status of the two species was highlighted by LBB indicators (Table 4). Although both species were classified as overfished, A. rochei exhibited critical levels of depletion, with current biomass at just 8% of the unexploited level (B/B₀ = 0.08), and a fishing mortality rate higher than the natural mortality rate (F/M = 15). By contrast, E. alletteratus had a higher relative biomass (B/B₀ = 0.25) and more balanced fishing pressure (F/M = 1.4), with almost all of the individuals captured being mature (98%).

These length-based indicators were complemented by a catch-only CMSY++ stock assessment, which provided fisheries-dependent reference points and a temporal trajectory of stock status relative to management targets.

Stocks assessment (CMSY++). The CMSY++ analysis revealed contrasting stock status for the two species (Table 5). For E. alletteratus, the analysis estimated an intrinsic population growth rate (r) of 0.229 (95% CI: 0.15–0.295) and a carrying capacity (K) of 20 600 t (95% CI: 14 600–35 800). The MSY was estimated at 1180 t · yr^–1 (95% CI: 926–1620). In 2022, the stock biomass was 8650 t, corresponding to a B/B_MSY ratio of 0.75 (95% CI: 0.45–1.11), indicating that the stock was below the biomass capable of producing MSY. The fishing mortality rate (F = 0.142) relative to the MSY fishing mortality rate (F_MSY = 0.114) gave an F/F_MSY ratio of 1.39 (95% CI: 0.73–3.03), indicating that overfishing is occurring (Fig. 5; Table 5). According to the Kobe plot classification, E. alletteratus currently indicates that overfishing is occurring, but the biomass is near MSY, though the wide confidence intervals suggest some uncertainty regarding the exact stock status (Fig. 6A; Table 5).

The CMSY++ analysis for A. rochei estimated an intrinsic growth rate (r) of 0.299 (95% CI: 0.163–0.409) and a carrying capacity of 6250 t (95% CI: 3930–12600) (Fig. 7; Table 5). The MSY was 467 t · yr^–1 (95% CI: 326–728). In 2022, the stock biomass was 2020 t, giving a B/B_MSY ratio of 0.52 (95% CI: 0.23–0.86). This indicates that the stock is below the biomass required to produce the MSY, meaning it is overfished. The fishing mortality rate (F = 0.145) relative to the MSY fishing mortality rate (F_MSY = 0.149) gave an F/F_MSY ratio of 1.30 (95% CI: 0.55–6.75), indicating overfishing. The Kobe plot trajectory places A. rochei in an overfished state, with overfishing occurring, confirming that its stock status is more critical than that of E. alletteratus (Fig. 6B; Table 5).

Both species are experiencing overfishing, but A. rochei is in a more depleted state, with a biomass of approximately half that of the MSY level, compared to three-quarters for E. alletteratus. The wide confidence intervals, particularly for F/F_MSY, reflect the uncertainty of catch-only assessments. To assess the stability and reliability of the CMSY++ stock status classifications, retrospective analyses were performed by progressively truncating the time series.

Retrospective analysis. Retrospective analyses were performed to evaluate the robustness of the CMSY++ estimates by sequentially removing the most recent years of data. For E. alletteratus, the B/B_MSY and F/F_MSY estimates for the final years remained relatively stable across all four retrospective runs (B/B_MSY range: 0.60–0.74; F/F_MSY range: 1.14–1.52) (Fig. 8A). This indicates that the model outputs are not overly sensitive to the inclusion or exclusion of recent data points.

However, the retrospective analysis revealed greater sensitivity for A. rochei. Consistent estimates were produced by runs including the full time series (1996–2022) and runs ending in 2021 (B/B_MSY: 0.52–0.58; F/F_MSY: 1.20–1.35). By contrast, runs ending in 2020 and 2019 produced markedly different results, with B/B_MSY exceeding 1.0 and F/F_MSY falling below 0.64 (Fig. 8B). This suggests that recent catch dynamics (2021–2022) have substantially influenced the classification of the stock status. Although estimates based on the full-time series are considered the most reliable, this retrospective uncertainty should be acknowledged.

This study provides the first quantitative stock assessment of the small tuna species Euthynnus alletteratus and Auxis rochei in the central Mediterranean Sea (GSA 13). By combining two complementary, data-limited methods (LBB and CMSY++), we reveal different stock statuses for these two species, which are exploited by the same fishery. This has major implications for their management.

Stock status and landing. The LBB analysis indicates that A. rochei is in a critical state, with an estimated biomass of only 8% of the unexploited biomass and a high fishing mortality ratio. The CMSY++ results confirm this diagnosis, indicating that the stock is overfished and that overfishing is occurring. This situation is alarming, given that our study is the first to provide a quantitative assessment of this species in the central Mediterranean. Previous work has been limited to semi-quantitative risk analyses (Lucena-Frédou et al. 2017) and descriptive studies of size structure and biology (Hattour 2009; Zarrad and Missaoui 2016).

By contrast, E. alletteratus has a less severe yet still concerning status. Both methods indicate an overfished stock and ongoing overfishing. Our results are consistent with previous assessments (Pons et al. 2019a, 2019b; Lucena-Frédou et al. 2021) and provide the first detailed estimate in the central Mediterranean (GSA 13). The convergence of the LBB and CMSY++ methods bolster confidence in this diagnosis, aligning with the ICCAT Small Tunas Working Group’s recommendation (ICCAT 2018, 2023) to use multiple analytical approaches and reduce uncertainty.

The time series from 1996 to 2022 reveals contrasting dynamics between the two species. For E. alletteratus, the recent increase in landings, with notable peaks in 2016–2017, coincides with regional trends observed in the Mediterranean (ICCAT 2025; Sant’Ana et al. 2021a) and likely reflects a displacement of fishing effort following the introduction of bluefin tuna quotas (Failler et al. 2025). The strengthening of data collection frameworks and the spatial delineation of stock units by ICCAT have likely improved the reliability of reported catch statistics (ICCAT 2023). However, the absence of a statistically significant long-term trend indicates that these production peaks are episodic rather than indicative of sustained stock recovery. Our findings are consistent with national fishery statistics for Tunisia (Hattour 2009; Zarrad and Missaoui 2016), which documented increases in small tuna landings following the implementation of the National Strategy for the Development of Small-Scale Pelagic Fisheries (NSDSPF; Mili et al. 2025).

Landings of A. rochei exhibited greater interannual variability compared to E. alletteratus, with a pronounced peak in the early 2000s followed by a sharp decline. This instability, also observed in Mediterranean-scale data (ICCAT 2025), suggests a strong dependence on environmental fluctuations and seasonal concentrations (Valeiras et al. 2008), rendering the species particularly vulnerable to fishing pressure (Lucena-Frédou et al. 2017; Sant’Ana et al. 2021b). In Tunisia, the significant contribution of this species to small tuna landings (Zarrad and Missaoui 2016) implies that the local fishery is highly susceptible to this natural variability.

Mechanisms driving vulnerability. Despite being exploited by the same purse-seine fleet in GSA 13, the divergent stock status observed in these two sympatric species calls for an examination of the mechanisms underlying this disparity.

Firstly, the demographic structure of the catches differs. For E. alletteratus, the low proportion of juveniles (2% below L₅₀ = 42.3 cm) suggests relatively good selectivity. The specimens measured between 38 and 90 cm (FL), with a mean of 61.1 cm and a primary mode in the 55–59 cm class. This is consistent with historical observations in Tunisia by Hattour (2009), who reported a similar modal class of 56–60 cm. However, the rarity of very large individuals (>90 cm) in our samples, despite a growth potential reaching 100–122 cm in the Mediterranean (Claro 1994; Sant’Ana et al. 2021a), suggests a truncation of the population size structure (Table 6). Several studies have reported the increasing rarity of individuals exceeding 100 cm in the Mediterranean (Hattour 2009; Hajjej et al. 2010; El-Haweet et al. 2013), a trend that is consistent with our findings. The elimination of the oldest age classes is concerning because it deprives the stock of Big Old Fat Fecund Female Fish (BOFFFs), whose disproportionate contribution to reproductive success is vital for stock resilience (Hixon et al. 2014; Barneche et al. 2018). Taken together with the rarity of large individuals, this suggests that the fishery is exploiting medium-sized adult fish without allowing them to reach full maturity and contribute optimally to stock renewal (Vasilakopoulos et al. 2011; Vasilakopoulos et al. 2014; Konoyima et al. 2024). Our results contrast with those of Zarrad and Missaoui (2016) and Macías et al. (2008), who reported size ranges of 33–63 cm and 36–46 cm FL, respectively, from the Tunisian coast in 2014 and in the Spanish Mediterranean basin. Both studies point to a greater prevalence of smaller individuals (Table 6). This discrepancy may be due to interannual variability in recruitment dynamics in Tunisian waters and potential differences of the fishing gear selectivity in the Spanish Mediterranean.

For A. rochei, the warning signal is the significant capture of immature individuals. Around 42% of individuals caught are immature, which puts recruitment directly under pressure from fishing. This proportion is consistent with the observations of Zarrad and Missaoui (2016), who reported an abundance of small individuals (>32 cm) during the autumn and winter months in Tunisian waters. The size distribution is characterized by two peaks at 33 cm and 40 cm, spanning 18–44 cm. The observed modal class (33–41 cm) is consistent with the findings of Hattour (2009) and Kahraman et al. (2011) in the central and eastern Mediterranean, respectively (Table 6). This range represented over 90% of the individuals studied by these authors. Our mean size of 35.1 cm is very close to the values reported by Lucena-Frédou et al. (2021) for the Mediterranean region during the period 2009–2019 (38 cm). Compared to published data from other Mediterranean areas (Kahraman et al. 2011; Saber et al. 2018), the observed sizes of A. rochei were similar. The presence of these juveniles, which grow rapidly during their first months (Hattour 2009), suggests that the recruitment peak is subject to fishing pressure (Valeiras et al. 2008; Sant’Ana et al. 2021b). The range of sizes at first maturity reported in the Mediterranean (32.5–36.5 cm; Bök and Oray 2001; Kahraman et al. 2011; Plandri et al. 2009; Saber et al. 2017) confirms that a significant proportion of our catches consists of immature and sub-adult (Table 6). Contrary to the apparent stability observed at the Mediterranean scale (Lucena-Frédou et al. 2021), our results, considered alongside the national and regional literature, indicate significant temporal variability in size structure.

Secondly, the variability in landings between years is more pronounced for A. rochei than for E. alletteratus. This instability, characterized by peaks followed by sharp declines, suggests that Auxis rochei depends on sporadic recruitment events and localized seasonal aggregations (Valeiras et al. 2008; Sant’Ana et al. 2021b). This makes it particularly vulnerable to fishing gear such as purse seines. This is consistent with the marked seasonality of landings (Sabatés and Recasens 2001) observed by Zarrad and Missaoui (2016) and Hattour (2009). These authors indicate that, in Tunisia, catches are concentrated in spring and summer, with a significant proportion landed at Teboulba port. Such spatial and temporal concentration exacerbates vulnerability. The lower variability of E. alletteratus could indicate a stock structure that renders it more resilient to increased fishing pressure (Adams et al. 2018).

Thirdly, the recent displacement of fishing effort linked to the bluefin tuna management has likely benefited the two species asymmetrically. The introduction of bluefin tuna quotas (Failler et al. 2025) has led to a shift in fishing efforts towards smaller tuna species, establishing E. alletteratus as a major substitute resource on a Mediterranean scale (Sant’Ana et al. 2021a; ICCAT 2025). Our data fit into this regional dynamic, with peaks in Tunisian production in 2016–2017 coinciding with maxima in the Mediterranean. In contrast, A. rochei does not appear to have benefited sustainably from this shift, as evidenced by the low catch levels between 2017 and 2021, which are consistent with ICCAT data (2025). This difference could reflect lower resilience (Ollé et al. 2019), a different ecology (Valeiras et al. 2008; Kahraman et al. 2011) involving more coastal areas that are less accessible to fleets that have shifted.

Fourthly, there are differences in life history traits and biological parameters (Sant’Ana et al. 2021a; Sant’Ana et al. 2021b). Comparing our estimates with values from the literature provides essential context. For E. alletteratus, our LBB estimate of L_max (118 cm; 95% CI: 116–120 cm) is consistent with the Mediterranean reference value (117 cm; Hattour 2009; Pons et al. 2019b). A synthesis of estimates reveals significant biogeographic variability (Table 6). In the Mediterranean, asymptotic lengths tend to be high (105–136 cm; Kahraman and Oray 2001; Hattour 2009; Hajjej et al. 2010; El-Haweet et al. 2013), reflecting the species’ significant growth potential (Table 6). However, K-values tend to be low (0.106–0.322 yr^–1), suggesting slower growth and greater longevity (Sant’Ana et al. 2021a). In the Atlantic, however, reported asymptotic lengths are generally lower (77.5–112 cm; Adams and Kerstetter 2014; Vieira et al. 2021), with higher K-values (0.690 yr^–1) indicating faster growth and reduced asymptotic size (Table 6). In theory, slower growth and greater longevity make E. alletteratus more susceptible to overexploitation, but potentially more resilient if spawners are protected. For A. rochei, our LBB estimate of L_∞ (48.6 cm; 95% CI: 47.7–49.4 cm) falls within the range of existing Mediterranean values: 46.7 cm in Tunisia (Hattour 2009); 45.1–57.4 cm in Turkey (Bök and Oray 2001; Kahraman et al. 2011); and 43.8–45.5 cm in Spain (Valeiras et al. 2008). Auxis rochei exhibits faster growth (K = 0.860 yr^–1), but it matures early and has a short lifespan (Sant’Ana et al. 2021b). This makes it dependent on stable annual recruitment. The strong variability of this recruitment, coupled with high fishing pressure on juveniles, can explain its rapid collapse.

Robustness and sensitivity. The convergence of LBB and CMSY++ results for E. alletteratus is remarkable. Our LBB estimate of L_∞ (118 cm) is similar to the value of 117 cm proposed by Pons et al. (2019b) in their Mediterranean synthesis and is consistent with previous Tunisian studies (105–117 cm, Hattour 2009; 130.8 cm for females, Hajjej et al. 2011). This cross-validation validates the quality of our length-frequency data and the reliability of other estimated parameters, notably the F/M ratio of 1.4. Regarding natural mortality, estimates for E. alletteratus in the Mediterranean vary: El-Haweet et al. (2013) obtained contrasting values of 0.392 and 0.167 yr^–1 using different methods (Table 6). More recently, Pons et al. (2019a, 2019b) applied various empirical methods and proposed a median value of 0.43 yr^–1 for the Mediterranean. Using this reference, we derive a fishing mortality of F = 0.60 yr^–1, which is higher than the F_MSY value of 0.114 yr^–1 estimated by CMSY++, confirming the diagnosis of overfishing (Table 6).

For A. rochei, L_∞ estimated in GSA13 fall within the range of existing Mediterranean values, thus validating our data. However, the absence of a Mediterranean-specific estimate of natural mortality is the main gap in our knowledge. The only available references come from the Indian Ocean (M = 1.18–1.9 yr^–1; Jasmine et al. 2013; Surya et al. 2023). If we apply a hypothetical value of M = 1.2–1.5 yr^–1, the F/M ratio of 15 leads to a fishing mortality of F = 18–22.5 yr^–1, which is unsustainable (Table 6). Even with a lower M, the diagnosis of severe overexploitation would remain unchanged. Nevertheless, uncertainty in M affects the precision of reference points and justifies dedicated research (Punt et al. 2021). Regardless of which biologically plausible M estimate is used, the conclusion of overexploitation remains unchanged due to the magnitude of the F/M ratio.

A retrospective CMSY++ analysis of A. rochei reveals sensitivity to recent data (2021–2022). Truncated analyses ending in 2019–2020 produce more optimistic diagnoses (B/B_MSY > 1). This sensitivity reflects the difficulty of assessing highly depleted stocks with limited time series, and highlights the importance of continuous monitoring to stabilize estimates (Ji et al. 2019; Al-Mamun et al. 2022; Mili et al. 2026). Runs including the full time series (1996–2022) and runs ending in 2021 produced consistent estimates (B/B_MSY: 0.52–0.58; F/F_MSY: 1.20–1.35). In contrast, runs ending in 2020 and 2019 produced different results, with B/B_MSY exceeding 1.0 and F/F_MSY below 0.64. This suggests that recent catch dynamics (2021–2022) have substantially influenced the classification of the stock status. While estimates based on the full-time series are considered the most reliable, this retrospective uncertainty must be acknowledged.

The sampling design employed in this study (48 surveys over nine months at the primary landing site, Teboulba) was specifically designed to capture the seasonal and inter-vessel variability in catch composition. However, two limitations should be acknowledged. First, sampling was restricted to a single year (2022), which may not capture interannual variability in recruitment and size structure. Second, sampling was conducted only at the landing site, precluding direct observation of at-sea and size selectivity. Nevertheless, the sample sizes are consistent with those reported in comparable data-limited studies in the Mediterranean (Demirel et al. 2020; Kılıç et al. 2023; Mili et al. 2025), and the narrow confidence intervals obtained for key parameters (Table 4) support the adequacy of the sampling effort.

Furthermore, the absence of a significant directional trend in the 27-year landings series for both species should not be interpreted as a sign of stability. In fact, it masks large multi-annual fluctuations (Dimarchopoulou et al. 2023), which, combined with the state of overexploitation highlighted, justify increased vigilance and the implementation of precautionary management measures.

Implications for management. Our results highlight the need for management measures specific to the biological characteristics and stock status of E. alletteratus and A. rochei. Despite being exploited by the same purse-seine fleet, the two species exhibit markedly different demographic structures, exploitation patterns, and resilience, underscoring the need for species-specific management strategies (Sant’Ana et al. 2021b).

According to the ICCAT recommendations (ICCAT 2018, 2023), the situation for A. rochei appears critical and justifies strong and immediate measures. Fishing mortality exceeds sustainable levels, indicating the need for a reduction in fishing effort (Hall et al. 2017; Tidd et al. 2026) targeting small tuna in GSA13, potentially combined with explicit catch limits. Another priority is increasing the length at first capture from the current 27 cm to the optimal value (L_Copt = 34 cm). It is essential to protect the immature fraction, which currently represents around 42% of catches, as this segment of the stock is the most vulnerable. A seasonal closure during periods of high juvenile abundance, particularly in autumn and winter, as reported by Zarrad and Missaoui (2016), would protect incoming cohorts before they contribute to reproduction. Finally, enhanced and systematic size monitoring at landing sites is necessary to detect any further deterioration.

By contrast, the management priorities for E. alletteratus are different. The main concerns are the reduction in size and the rarity of large spawners. Protective measures targeting large individuals should be considered, such as implementing a maximum landing size (prohibiting the landing of specimens measuring more than 80 cm) and temporarily protecting spawning aggregation areas. These locations can be identified through the knowledge of fishers. Preserving large, highly fecund females is critical for stock resilience (Barneche et al. 2018). Increasing the length at first capture from 54 cm to the optimal value, L_Copt = 66 cm, would enable individuals to reach full maturity and contribute to stock renewal. This objective could be achieved by implementing temporal and spatial restrictions during key growth periods. It is important to note that the current low proportion of juveniles in catches is a relative strength and must be maintained by preventing any increase in fishing effort.

Additionally, according to Tunisian legislation (JORT 1995), there are currently no specific regulations in place for these species. We therefore recommend establishing a minimum legal catch size corresponding to the optimal length estimated in this study: 34 cm for A. rochei and 66 cm for E. alletteratus.

Beyond species-specific measures, the pronounced interannual variability observed, particularly for A. rochei (see Sant’Ana et al. 2021b), highlights the necessity of continuous monitoring and adaptive management (Lyons et al. 2008). The absence of a clear long-term directional trend in landings should not be interpreted as stability; rather, it masks significant multi-annual fluctuations that justify a precautionary approach (Hixon et al. 2014; Barneche et al. 2018). The pronounced seasonality documented in Tunisia and the Mediterranean (Sant’Ana et al. 2021a; Sant’Ana et al. 2021b), particularly the concentration of landings at the port of Teboulba in spring and summer (Hattour 2009; Zarrad and Missaoui 2016), could exacerbate vulnerability if fishing effort becomes concentrated in space and time. Such dynamics reinforce the need for flexible, evidence-based management that can respond rapidly to variability in recruitment and stock fluctuations.

Perspectives. Several priority research areas have emerged to improve scientific knowledge and the effectiveness of management for E. alletteratus and A. rochei stocks.

The first issue is the need to estimate natural mortality for A. rochei in the Mediterranean. The absence of a region-specific estimate is a significant source of uncertainty in stock assessments. Secondly, expanding the size-frequency monitoring over time and space is essential. Annual sampling should be extended to additional landing sites to capture spatial heterogeneity in exploitation patterns and recruitment pulses. Thirdly, integrating standardized fishing effort data is a priority. Systematically collecting catch per unit effort data from the purse seine fleet would enable a relative abundance index to be constructed, substantially improving the robustness of CMSY++ outputs.

Further methodological development should include applying integrated data-limited models such as LIME (Rudd and Thorson 2018), which combine length-frequency data with catch time series. These approaches can reduce the bias inherent in models based on a single data source and provide more stable estimates of stock status.

Finally, given the transboundary nature of these small tuna stocks, regional coordination is indispensable. Within the frameworks of the International Commission for the Conservation of Atlantic Tunas (ICCAT) and the General Fisheries Commission for the Mediterranean (GFCM), harmonization of assessment methods and management measures would enhance coherence and effectiveness. Coordinated monitoring and shared reference points are essential to ensure sustainable exploitation on a Mediterranean scale.

This study provides the first quantitative stock assessment of Euthynnus alletteratus and Auxis rochei in the central Mediterranean (GSA 13), employing complementary data-limited methods (LBB and CMSY++). Our results reveal contrasting stock statuses for these sympatric species, which are exploited by the same purse seine fleet.

Auxis rochei is in a critical state: it is severely overfished (B/B_MSY = 0.21; 0.52), experiencing unsustainable fishing mortality (F/M = 15), and 42% of the catches are juveniles, which requires urgent intervention. By contrast, E. alletteratus is overfished (B/B_MSY = 0.67; 0.75) and subject to overfishing (F/F_MSY = 1.39); however, it has a healthier demographic structure, with low juvenile catch rates (2%), although large spawners (>90 cm) are sporadic.

Based on these findings, we recommend the following: for A. rochei: increasing the length at first capture from 27 cm to the optimal 34 cm; reducing fishing effort; and for E. alletteratus: protecting large spawners through maximum landing sizes (>80 cm); and increasing the length at first capture from 54 cm to 66 cm. In the absence of specific regulations in Tunisian legislation (JORT 1995), we strongly advocate establishing minimum legal catch sizes corresponding to these optimal lengths.

The pronounced interannual variability, particularly for A. rochei, highlights the need for continuous monitoring and adaptive management. Future research should prioritize estimating natural mortality for A. rochei in the Mediterranean and fostering regional coordination within the ICCAT and GFCM frameworks. By providing robust scientific baselines for GSA 13, this work supports regional initiatives for the sustainable management of these vulnerable yet vital transboundary resources.