Habitat-Driven Population Parameters Insights for European Eel Anguilla anguilla (Linnaeus, 1758) in Croatian Waters

Abstract

Key parameters were estimated separately for the European eel, Anguilla anguilla (Linnaeus, 1758) subpopulations across freshwater and brackish environments within the Eastern Adriatic Eel Management Unit (EMU: EA). Between 2023 and 2024, European eel sampling was carried out at 23 locations along the Croatian coast (N = 678). Ages ranged from 1 to 13 years in freshwater and 1 to 8 years in brackish waters. The population structure was dominated by undifferentiated (42.8%) in freshwater and females (46.3%) in brackish habitats. Eels in freshwater exhibited a significantly higher b-coefficient in their length–weight relationship and better condition. Based on the otolith annuli patterns, age classes 3 to 5 predominated in both groups. A slightly longer asymptotic length and lower growth rate were found for the freshwater group compared to a shorter length and higher growth rate in the brackish habitat. Natural mortality was estimated at 0.174 ± 0.09 year⁻¹ and 0.191 ± 0.133 year⁻¹ for brackish and freshwater habitats, respectively. Total mortality was higher in freshwater (0.86 year⁻¹) in comparison with the brackish (0.83 year⁻¹) habitat. According to obtained results, more than 50% of eels aged three years are under exploitation. The maximum Yield per Recruit (Y/R) was 0.082 g/recruit in brackish at L_c = 44.88 cm, and a current L_c is 19.4 cm in the samples. In freshwater, Y/R peaked at 0.042 g/recruit at L_c = 55.49 and a current L_c 11.1 cm. It is recommended, following a precautionary approach to management, that the current fishing practices change in order to increase the minimum landing size (MLS), at least to 45 cm (above the current official MLS of 35 cm), to increase the fishing yield, and directly enhance population resilience. Findings emphasise the need for sustainable eel management policies considering different subpopulation parameters along the freshwater/brackish gradient at a small spatial scale when proposing and implementing stock management measures.

1. Introduction

Studies published in the last two decades highlight the critical conservation status of the European eel, Anguilla anguilla (Linnaeus, 1758) [1,2] and evaluate management strategies across Europe [3,4,5,6,7]. The European eel population has declined significantly since the 1970s, prompting conservation efforts across EU member states [3,6,7]. However, fewer than 50% of current management plans achieve their goals, indicating a need for more ecological and behavioural information [1,5].

European eel management strategies vary across EU member states, reflecting the complex nature of eel conservation. Key elements include improving connectivity and assisting elver migration across barriers, river restoration to increase habitat, and enhancing lacustrine environments for larger eels [8]. However, northern and southern regions face different challenges in eel population recovery [3]. Despite these regional differences in management approaches, genetic studies comparing maturing silver eels from southern and northern Europe have found no significant genetic differentiation, supporting the panmixia hypothesis for European eels [9,10]. This genetic homogeneity underscores the importance of coordinated conservation efforts across the species range. Also, there is a crucial need for developing effective management plans that balance conservation and fishery goals [11]. However, Ref. [1] advises a complete cessation of fishing, which is consistent with the fishery management objective of achieving Maximum Sustainable Yield (MSY).

Evaluating the influence of fisheries, predation and migration barriers on the population dynamics of the European eel requires a large set of data, including migration, restocking, natural and fishing mortality, predation, and hydropower impacts. Mathematical modelling has been proposed to assess potential spawning stocks and develop effective management plans, considering both conservation and fishery goals [3,5,12,13,14,15,16,17] covering the von Bertalanffy Growth Function (VBGF) applications, age-structured models, environmental influences and functional morphology. Additionally, region-specific strategies considering ecological, environmental, and socioeconomic factors are necessary for effective conservation [7]. To assess models used in management decisions, continuous system-specific time-series data are essential [16]. Unfortunately, data for most management units, especially in southern Europe, including the Eastern Adriatic coast, are unavailable or incomplete. Also, data for different habitats considering ontogenetic changes are often lacking, although there are indications of different strategies and population characteristics depending on habitat type. Survival rates vary between river systems, influenced by factors such as population density, water quality, and sex ratios [18].

Age as a key parameter for assessing natural and fishing mortality was identified by [19]. However, age estimation in European eels is complicated by phenotypic plasticity, longevity (>15 years; [20,21], stress-related rings in older individuals [22], and habitat-dependent annuli formation are influenced by environmental factors such as temperature, salinity, and food availability [23]. Also, recent studies revealed that eels in small estuaries had population dynamics parameters similar to those previously reported in larger habitats [24], highlighting that small estuaries contribute significantly to the eel population and, therefore, play an essential role in conservation strategies for European eels. Mediterranean coastal habitats form a significant part of the European eel range. The recovery of its global stock may rely heavily on efforts from Southern European and North African countries, making it crucial to understand eel biology and population dynamics to ensure the success of the EU conservation plan in these areas [25].

The Eastern Adriatic coast of Croatia has several characteristics which distinguish it from other Mediterranean coasts, including a dominant rocky nature, short canyon-type inflowing rivers, and only a few small estuaries. Eel passage in some rivers is blocked by large waterfalls, and all rivers are subjected to anthropogenic changes, notably dam construction, stream bank stabilisation, water reservoir establishment, and significant water diversion for communal and agricultural purposes, especially during summer [26]. Recent studies highlight the critical decline of European eel populations in several Croatian coastal and river ecosystems [27]. While some Montenegrin waters appear suitable for eels, Croatian rivers show declining eel conditions related to habitat loss and anthropogenic impacts [26,28,29,30,31]. The officially reported commercial silver eel catches in Croatia, mainly executed in the Neretva estuary, were usually less than 1 tonnes/year in the last decade in comparison with historical catches of 100 tonnes in the 1930s, while the yellow eel fishery was mainly executed by a recreational fishery with an estimated catch of up to 2 tonnes in 2022 [32]. Following this situation and recommendations by the General Fisheries Commission for the Mediterranean (GFCM), Croatia banned eel recreational fisheries since September 2023.

Evidently, the published data about eels for the Adriatic Sea is limited and non-comprehensive, while the implementation of general recommendations is complex due to socioeconomic and historical reasons and thus requires additional investigation. The aims of this study were to ascertain the structure of the endangered A. anguilla subpopulations in the Eastern Adriatic Sea by estimating key population parameters—including age, growth, condition factor, natural and total mortality, exploitation ratio and Yield per Recruit across two distinct habitat types in Croatian inland waters. These habitats include typical freshwater riverine environments and brackish/estuarine areas. Growth was analysed using various growth models, with the best-fit model selected according to Akaike’s Information Criterion (AIC) for each habitat type and sampling group. The main goal is to gather essential data to support the development of a management plan or to inform further management measures in Croatia.

2. Materials and Methods

2.1. Study Area, Sampling and Data Collection

Eel sampling was conducted seasonally (March, June, September, December) at 23 sampling locations from September 2023 to July 2024, along Croatia’s west coast (estuaries and rivers; Figure 1). The selected locations include bigger (Mirna, Raša, Zrmanja, Krka, Cetina and Neretva) and smaller (Dragonja, Dubračina, Ričina, Jadro, Žrnovnica, Mala Neretva and Ombla) rivers. The bigger rivers were sampled at two sites, typical freshwater and brackish ones. The smaller rivers had just freshwater sites for sampling. Additionally, eels were also investigated in Lake Vrana and Lake Prokljan, as two historically important habitats for eels and in two lagoons (Parila and Ston), as typical brackish habitats (Table S1).

Eel sampling in freshwater sites was executed by an electrofishing device (Susan 1030SMP, ~2.4 kW, Shenzhen Qinda Electronics Co., Ltd.; Shenzen, China) on a georeferenced 100 m length × 4 m width walking or boat pathway with single-pass per location. The sampling in brackish water was executed using simple modernised traditional eel traps without wings and with 4 mm mesh size of the final part. At each brackish site, five traps were deployed for the 48 h. The total weight (W) and total length (TL) were measured. After the field measurements, all eels were frozen for further analysis. After dissection, sex determination was executed according to the ICES manual [19].

2.2. Age Reading

Age classification of the specimens was based on sagittal otolith readings, which were extracted, ground, and polished, as described in the report of the ICES workshop on age reading of European and American eels [19]. Three readers (co-authors LG, SG, and BG) participated in otolith age reading using both a microscope and digital images. After individual readings, the ages were compared and finalised by group decision when initial readings differed. False rings were handled based on protocol proposed by ICES [19]. Age reading examples are presented in Figure 2.

2.3. Data Analysis

2.3.1. Length-Frequency Distribution and Length–Weight Relationship

Length–weight relationships for the eels were calculated based on the power model [33]: $\mathrm{W}\mathrm{e}\mathrm{i}\mathrm{g}\mathrm{h}\mathrm{t}=\mathrm{a}\cdot \left[\mathrm{T}\mathrm{o}\mathrm{t}\mathrm{a}\mathrm{l} \mathrm{L}\mathrm{e}\mathrm{n}\mathrm{g}\mathrm{t}{\mathrm{h}}^{\mathrm{b}}\right]$, for both habitats separately based on the raw measurements of the samples, where a, the intercept of the power function; b, allometric exponent of the regression. The length-frequency structure of the samples is illustrated in Figure 3 for the freshwater and brackish habitats, separately.

2.3.2. Growth Modelling

Growth was estimated based on the von Bertalanffy growth function (VBGF), the logistic and the Gompertz models [34,35,36,37,38] based on the average total length of the specimens at each age group:

$$\begin{array}{l}\mathrm{V}\mathrm{B}\mathrm{G}\mathrm{F}:{\mathrm{L}}_{\mathrm{t}}=\mathrm{L}\infty \left(1-{\mathrm{e}}^{-\mathrm{K}(\mathrm{a}\mathrm{g}\mathrm{e}-{\mathrm{t}}_0)}\right)\\ \mathrm{G}\mathrm{o}\mathrm{m}\mathrm{p}\mathrm{e}\mathrm{r}\mathrm{t}\mathrm{z}:\mathrm{L}\infty \cdot {\mathrm{e}}^{-{\mathrm{e}}^{\mathrm{m}-\mathrm{n}\cdot (\mathrm{a}\mathrm{g}\mathrm{e})}}\\ \mathrm{L}\mathrm{o}\mathrm{g}\mathrm{i}\mathrm{s}\mathrm{t}\mathrm{i}\mathrm{c}:{\displaystyle \frac{\mathrm{L}\infty }{1+\mathrm{m}\cdot {\mathrm{e}}^{-\mathrm{n}\cdot (\mathrm{a}\mathrm{g}\mathrm{e})}}}\end{array}$$

where Lt, the theoretical length at age t (cm); L∞, the asymptotic length (cm); K, the growth coefficient (year⁻¹); t, the age (years); t₀, the theoretical age at zero length (years); e, the base of the natural logarithm, m, the growth rate coefficients; and n, the age at the inflection point when the growth rate is maximal. The best model to describe eel growth based on the samples was selected using the Akaike criterion [39,40]. We selected to study eel growth using 3 typical growth models [24] since the specimens in the samples do not cover the whole life cycle of the species in the study area but only a part of the population (undifferentiated and yellow eels before silvering). Therefore, estimations of L∞, K and t₀ which will be used in the next stage of estimations (mortality, etc.) need to be calculated with the minimum possible confidence error.

2.3.3. Mortality Modelling and Y/R

Natural mortality-at-age (M) was estimated based on the allometric model using weight-at-age values (i.e., weights corresponding to each age class), which relates to mortality and specimen weight (M-W equation):

$$\mathrm{N}\mathrm{a}\mathrm{t}\mathrm{u}\mathrm{r}\mathrm{a}\mathrm{l} \mathrm{M}\mathrm{o}\mathrm{r}\mathrm{t}\mathrm{a}\mathrm{l}\mathrm{i}\mathrm{t}\mathrm{y} ({\mathrm{year}}^{-1})={\mathsf{\mu }}_1\left[\mathrm{Total} \mathrm{W}\mathrm{e}\mathrm{i}\mathrm{g}\mathrm{h}\mathrm{t} {(\mathrm{g})}^{-{\mathrm{b}}_1}\right],$$

where μ₁ is the natural mortality rate at 1 g weight: b₁, the allometric scaling factor and weight, the individual weight-at-age in g. To estimate the coefficients μ₁ and b₁ for the above equation, we used as natural mortality reference, the Jensen invariant MREF = 1.5 K (were K is the VBGF growth coefficient) and as weight reference, WREF, the maximum weight of the individuals in the samples per habitat [41,42]. With the MREF and WREF values known, the equations of natural mortality vs weight were solved using a linear programming algorithm [43,44] to estimate μ₁ and b₁ coefficients.

The M-W equation is nonlinear, and therefore to be analysed by linear programming, the raw data were log-transformed so the equation became

$$\ln \left({\mathrm{M}}_{\mathrm{R}\mathrm{E}\mathrm{F}}\right)=\ln \left({\mu }_1\right)+b_1\left(\ln {\mathrm{W}}_{\mathrm{R}\mathrm{E}\mathrm{F}}\right)$$

This equation was fed to the linear programming software Eureka (version 2.1; https://archive.org/details/eureka-solver-v2.1, accessed on 24 April 2025) with the constraints: −0.5 < b₁ < 0 and 0 < μ₁ < 4.1 year⁻¹ [45]. Optimum μ₁ and b₁ values are obtained through the minimization of the absolute errors in log space of the M-W equation (iterations within the software):

$$\underset{\mathsf{\alpha },\mathsf{\beta }}{\min }={\displaystyle \sum _{\mathrm{i}=1}^{\mathrm{n}}\left|z_{\mathrm{i}}-\mathsf{\alpha }-\mathsf{\beta }\left(u_1\right)\right|}, \mathrm{where} z_i{=\mathrm{lnM}}_{\mathrm{REF}}, u_1{=\ln (\mathrm{W}}_{\mathrm{REF}}), \alpha =\ln {\mu }_1, \beta =b_1$$

Applying the model to the M-W equation, the variables are the log intercept (lnμ₁), the slope (b₁) and the residual variables r_i⁺ and r_i⁻ which are set automatically during each iteration, to satisfy the constraints and minimise the objective. The constraints for the linear programming procedure then become

$$\begin{array}{l}{z}_i-\alpha -\beta (u_1)=r_i^{+}-r_i^{-}\\ r_i^{+}, r_i^{-}\ge 0\\ \alpha <\ln 4.1=1.411\\ -0.5<\beta <0\end{array}$$

Average mortality per habitat was estimated using the complete dataset from the samples, using the final model and the individual specimen weight to estimate natural mortality per individual.

Total mortality (Z) was assumed to be constant and was estimated based on the Length-Converted Catch Curve method [46]. Finally, Fishing mortality (F) per age class was estimated using formula: ${\mathrm{F}}_{\mathrm{a}\mathrm{t}-\mathrm{a}\mathrm{g}\mathrm{e}}=\mathrm{Z}-{\mathrm{M}}_{\mathrm{a}\mathrm{t}-\mathrm{a}\mathrm{g}\mathrm{e}}$ year⁻¹, while the Exploitation ratio (E) was estimated as [47]

$${\mathrm{E}}_{\mathrm{a}\mathrm{t}-\mathrm{a}\mathrm{g}\mathrm{e}}={\displaystyle \frac{{\mathrm{M}}_{\mathrm{a}\mathrm{t}-\mathrm{a}\mathrm{g}\mathrm{e}}}{\mathrm{Z}} {\mathrm{year}}^{-1}}$$

Yield per Recruit (Y/R) analysis was performed based on the Beverton–Holt model in order to examine potential scenarios for the management of the local eel population in relation to fishing pressure [48] for each habitat. The model was run several times varying L_c between the length at first catch, which equals to the smallest individual in the samples and the L∞. The maximum Y/R was estimated at each run. The values of Y/R were correlated with the L_c values and fitted to a simple 2-degree polynomial model:

$$\mathrm{Y}/\mathrm{R}=\mathrm{x}+\mathrm{y}\left[{\mathrm{L}}_{\mathrm{c}}\right]+\mathrm{z}\left[{{\mathrm{L}}_{\mathrm{c}}}^2\right],$$

where L_c is the length at first capture in cm and x, y, z, regression coefficients for the dome-shaped Y/R model.

The regression equation was then maximised using a simplex linear programming algorithm [43,44] to estimate its maximum (Y/Rmax) and the corresponding optimum L_c. The main management guideline that was derived from this analysis is that the eel fishing method (gear configuration and materials as well as the use of gears) so that the minimum landing size, which corresponds to these results, need to increase from the current (around 11 cm) so that the Y/R remains at maximum levels.

2.4. Statistical Analysis and Software

Comparison between samples was carried out using standard or non-parametric tests following normality testing based on the Shapiro–Wilk W test [49]. In the case of non-normality, the Kolmogorov–Smirnov test was used to compare 2 samples and the Kruskal–Wallis test with L multiple range testing for multiple sample comparisons was performed. In the other cases, a t-test was performed [50].

Population dynamics analysis was performed using FISAT II (version 1.2.2 [47]). Statistical analysis was performed using the open-source software JAMOVI Desktop (version 2.7.13; https://www.jamovi.org, accessed on 24 April 2025), PAST (version 5.03; https://www.nhm.uio.no, accessed on 24 April 2025) and various routines in R language (version 4.2.2; https://cran.r-project.org). The difference in data distribution between the two groups was tested by a paired t-test, Wilcoxon Ranked sign test and 2-way ANOVA. Equation modelling was carried out using the CurveExpert software (version 2.6.57.3. www.curveexpert.net/, accessed on 24 April 2025). The exponents (slopes) of the length–weight relationships per habitat were compared using a t-test [51]:

$\mathrm{t}-\mathrm{t}\mathrm{e}\mathrm{s}\mathrm{t}={\displaystyle \frac{{\mathrm{b}}_{\mathrm{a}}-{\mathrm{b}}_{\mathrm{b}}}{\sqrt{\mathrm{S}{\mathrm{E}}_{{\mathrm{b}}_{\mathrm{a}}}^2+\mathrm{S}{\mathrm{E}}_{{\mathrm{b}}_{\mathrm{b}}}^2}}}$, where b_a and b_b, are the 2 slopes. Finally, all average values are reported with their respective standard deviation in the form of average ± SD.

3. Results

In total, 678 eels were collected over the sampling period from 23 sampling sites (304 at freshwater and 374 at brackish sites). The freshwater samples were composed of 130 undifferentiated (42.8%) individuals; 79 (26.0%) male and 95 (31.2%) female specimens. The brackish group samples were composed of 72 (19.2%) undifferentiated individuals; 129 (34.5%) males and 173 (46.3%) females.

3.1. Population Structure

The population structure in terms of length-frequency distribution between freshwater and brackish sites together with the average total length at age (red points) are presented in Figure 3a. The total length range for eels in the freshwater group was from 11.1 to 87.5 cm (mean 37.36 ± 13.56 cm) while eels in the brackish group ranged in total length from 19.4 to 67.5 cm (mean 42.81 ± 7.82 cm). Although age distribution for both habitats exhibits a main mode between 2 and 6 years of age (Figure 4). There is a difference in the abundance per length class in each group with mode of 35 cm in the freshwater samples and 41 cm in the brackish samples. According to the total length classes means, a total of 10 and 8 length groups were formed in freshwater and brackish environments, respectively. The distributions of the freshwater and brackish samples were found significantly different (Kolmogorov–Smirnov test D = 4.73, p < 0). In freshwater and brackish groups, the eels ranged in weight from 4.8 to 1750.4 g (mean 143.72 ± 197.15 g), and from 15.2 to 553.5 g (mean 165.99 ± 100.47 g), respectively. The population sexual structure shows that eel stock from freshwater and brackish groups are composed dominantly by undifferentiated and female individuals, respectively (Figure 3b). In the freshwater samples, undifferentiated specimens ranged in total length from 11.1 to 66.6 cm (mean 28.02 ± 9.39), males ranged from 25.3 to 45.0 cm (mean 36.2 ± 5.01) and females ranged from 27.2 to 86.2 cm (mean 48.85 ± 12.21). Total length distribution differs between sexes in the brackish group with undifferentiated specimens ranged from 19.4 to 52.1 cm (mean 34.45 ± 4.91), males ranged from 29.2 to 45.6 cm (mean 39.26 ± 3.24) and females ranged from 27.9 to 66.3 cm (mean 46.945 ± 7.38). LSD multiple range test showed that the total length distribution between males in freshwater and brackish habitat samples are homogenous groups. The same was obtained for females in both habitats. However, male and female total length distributions within habitats are statistically different (Kolmogorov–Smirnov 2 tailed test D = 0.355, p < 0.0001 for the freshwater habitat; Kolmogorov–Smirnov 2 tailed test D = 0.427, p < 0.0001 for the brackish habitat).

The total age range for eels in the freshwater sample was from 1 to 13 years, while eels in the brackish sample ranged in age from 1 to 8 years. A predominance of age classes 3 to 5 was observed in both freshwater (50.7%) and brackish (73.0%) habitats. Testing the total length per age distributions in the freshwater and brackish habitat samples, showed that there is no statistical difference and therefore it can be assumed that both eel groups belong to the same subpopulation (Kolmogorov–Smirnov test, D = 0.63, p = 0.82). Accordingly, no statistical evidence of differences in length–at–age distributions between habitats was found.

3.2. Length–Weight Relationship

The data on the length-weight relationships following the power model are summarised in

Table 1. Summary of the length–weight relationships for the eel per habitat.

Model	Habitat
	Freshwater	Brackish
Overall	a = 0.0004, b = 3.401, r2 = 0.956, std.error = ±39.82 g	a = 0.0016, b = 3.045, r2 = 0.946, std.error = ±23.48 g
Males	a = 0.0013, b = 2.985, r2 = 0.921, std.error = ±30.34 g	a = 0.0016, b = 3.038, r2 = 0.859, std.error = ±19.43 g
Females	a = 0.0005, b = 3.328, r2 = 0.974, std.error = ±59.82 g	a = 0.0015, b = 3.061, r2 = 0.955, std.error = ±28.25 g

. The equations are illustrated in Figure 5. The length–weight relationship for the brackish sample was described by following parameters a = 0.0016 ± 0.0002 and b = 3.045 ± 0.039 (N = 304; R2 = 0.945; Standard Error = ±23.48 g) and for the freshwater sample, a = 0.0004 ± 0.00007 and b = 3.401 ± 0.041 (N = 374; R2 = 0.962; Standard Error = ±39.82 g). A t-test revealed a significant difference between the exponent values for the freshwater and brackish habitats (t-test = 1149.2; d.f. = 678; p < 0.0001). Females exhibit a higher exponent in freshwater (3.328) than in brackish (3.061) habitat, while males exhibit a higher exponent in brackish (3.038) than in freshwater (2.985) habitat. Comparison of the coefficients a and b of the male equations between the brackish and freshwater habitats showed no statistically significant difference (t-test for a = −0.4205, p = 0.675; t-test for b = 0.317, p = 0.752). The comparison of the coefficients of the female equations between the brackish and freshwater habitats showed a statistically significant difference (t-test for a = −3.112, p = 0.0021; t-test for b = 3.425, p = 0.0007).

3.3. Growth

The growth equations are summarised in

Table 2. Estimated growth equations parameters for European eel, Anguilla anguilla Management Unit (EMU) Eastern Adriatic for freshwater and brackish groups.

	Habitat	Freshwater	Brackish
Model
VBGF	${\mathrm{L}}_{\mathrm{t}}=\mathrm{L}\infty \left[1-{\mathrm{e}}^{-\mathrm{K}(\mathrm{a}\mathrm{g}\mathrm{e}-{\mathrm{t}}_0)}\right]$	L∞ = 116.24 ± 11.96 cm, Κ = 0.107 ± 0.02, to = −0.171 ± 0.19, r2 = 0.995 Akaike = 25.48 (*)	L∞ = 90.755 ± 15.05 cm, K = 0.151 ± 0.04, to = −0.243 ± 0.21, r2 = 0.992 Akaike = 21.523 (*)
Gompertz	${\mathrm{L}}_{\mathrm{t}}=\mathrm{L}\infty \cdot {\mathrm{e}}^{-{\mathrm{e}}^{\mathrm{m}-\mathrm{n}\cdot (\mathrm{a}\mathrm{g}\mathrm{e})}}$	L∞ = 92.25 ± 5.89 cm, m = 0.905 ± 0.09, n = 0.270 ± 0.04, r2 = 0.992 Akaike = 30.92	L∞ = 72.96 ± 9.4 cm, m = 0.798 ± 0.14, n = 0.359 ± 0.09, r2 = 0.983 Akaike = 29.33
Logistic	${\mathrm{L}}_{\mathrm{t}}={\displaystyle \frac{\mathrm{L}\infty }{1+\mathrm{m}\cdot {\mathrm{e}}^{-\mathrm{n}\cdot (\mathrm{a}\mathrm{g}\mathrm{e})}}}$	L∞ = 86.52 ± 5.12 cm, m = 7.11 ± 1.44, n = 0.431 ± 0.06, r2 = 0.996 Akaike = 35.27	L∞ = 68.021 ± 7.64 cm, m = 5.674 ± 1.77, to = 0.557 ± 0.14, r2 = 0.976 Akaike = 32.42

(*) best growth equation per habitat based on Akaike criterion.

. The data show that the best equation to describe growth is the VBGF for both habitats based on the Akaike criterion (Table 2) following the exclusion of some problematic raw data. Correlation coefficients show a very high correlation for all models (r² > 0.9). The differences in the model variables between habitats are owed to the limited ages found in the brackish (up to eight age groups) in relation to the freshwater habitat (up to 13 age groups). The growth curves of the eel were presented in Figure 6. According to the VBGF curve data, the maximum length (L∞) in freshwater and brackish environments is estimated at 116.27 ± 11.96 cm and 90.76 ± 15.05 cm, respectively. Growth parameter k was 0.107 ± 0.02 for the freshwater and 0.151 ± 0.04 for the brackish group. Thus, a slightly longer asymptotic length and lower growth rate were found for the freshwater group compared to a shorter length and higher growth rate in the brackish group.

3.4. Natural Mortality

The results on natural mortality-at-age are summarised in

Table 3. Summary of the results on mortality-at-age data for the European eel Anguilla anguilla (Z: total mortality; M: natural mortality; F: fishing mortality; E: exploitation rate).

Age	Freshwater				Brackish
	Z	M	F	E	Z	M	F	E
1	0.860	0.467	0.393	47.40%	0.830	0.374	0.456	54.9%
2	0.860	0.294	0.566	68.18%	0.830	0.231	0.599	72.1%
3	0.860	0.212	0.648	78.03%	0.830	0.194	0.636	76.7%
4	0.860	0.163	0.697	84.02%	0.830	0.159	0.671	80.8%
5	0.860	0.124	0.736	88.64%	0.830	0.132	0.698	84.1%
6	0.860	0.107	0.753	90.71%	0.830	0.114	0.716	86.3%
7	0.860	0.086	0.774	93.29%	0.830	0.099	0.731	88.0%
8	0.860	0.071	0.789	95.02%	0.830	0.091	0.739	89.0%
9	0.860	0.071	0.789	95.01%
10
11
12

. Weights correspond to each age class. The natural mortality equations for each habitat are

$$\begin{array}{l}{\mathrm{M}}_{\mathrm{F}\mathrm{R}\mathrm{E}\mathrm{S}\mathrm{H}}=0.996\left[\mathrm{W}\mathrm{e}\mathrm{i}\mathrm{g}\mathrm{h}{\mathrm{t}}^{-0.378}\right]\\ {\mathrm{M}}_{\mathrm{B}\mathrm{R}\mathrm{A}\mathrm{C}\mathrm{K}}=0.993\left[\mathrm{W}\mathrm{e}\mathrm{i}\mathrm{g}\mathrm{h}{\mathrm{t}}^{-0.392}\right]\end{array}$$

The similarity of the coefficients (constant: two tailed t-test t = 0.17, d.f. = 676, p = 0.86; exponent: two tailed t-test t = 0.14, d.f. = 676, p = 0.88) allows the elaboration of the overall population natural mortality as

$${\mathrm{M}}_{\mathrm{O}\mathrm{V}\mathrm{E}\mathrm{R}\mathrm{A}\mathrm{L}\mathrm{L}}=0.994\left[\mathrm{W}\mathrm{e}\mathrm{i}\mathrm{g}\mathrm{h}{\mathrm{t}}^{-0.385}\right]$$

The values of the allometric exponent, b₁, in the above equations are between −0.392 and −0.378, while the values of μ₁ are in range from 0.993 to 0.996.

Average natural mortality, M, per habitat was estimated at 0.174 ± 0.09 per year (weight range 40–560 g) for the brackish water group and = 0.191 ± 0.133 per year (weight range 20–1750 g) for the freshwater group. Overall, M was found 0.182 ± 0.01 per year. Natural mortality relations with age per habitat are illustrated in Figure 7.

3.5. Total Mortality, Exploitation Ratio and Yield per Recruit (Y/R)

Total mortality for the brackish habitat was found 0.83 ± 0.21 year⁻¹ and for the freshwater habitat, 0.86 ± 0.34 year⁻¹. The average values of fishing mortality were estimated 0.656 ± 0.09 per year and 0.669 ± 0.12 per year, for the freshwater and brackish water groups, respectively. The average exploitation rate for freshwater and brackish habitats was found and 78.9% and 80.7%, respectively.

The values of fishing mortality-at-age and exploitation ratio-at-age are summarised in Table 3. According to obtained results, more than 50% of eels aged three years are under exploitation.

The models for the correlation between length at first capture, L_c, and yield per recruit, Y/R, for each habitat are (Figure 8).

$$\begin{array}{c}\mathrm{Y}/{\mathrm{R}}_{\mathrm{FRESHWATER}}=0.008+0.001\left[{\mathrm{L}}_{\mathrm{c}}\right]-0.00001\left[{{\mathrm{L}}_{\mathrm{c}}}^2\right],{\mathrm{r}}^2=0.967,\mathrm{std}.\mathrm{error}=\pm 0.006 \mathrm{g}/\mathrm{recruit}\hfill \\ \mathrm{Y}/{\mathrm{R}}_{\mathrm{BRACKISH}}=0.016+0.003\left[{\mathrm{L}}_{\mathrm{c}}\right]-0.00003\left[{{\mathrm{L}}_{\mathrm{c}}}^2\right],{\mathrm{r}}^2=0.871,\mathrm{std}.\mathrm{error}=\pm 0.019 \mathrm{g}/\mathrm{recruit}\hfill \end{array}$$

The maximum Y/R for the freshwater habitat was found to be 0.042 g/recruit at L_c = 55.49 cm when the current L_c from the samples is 11.1 cm. The maximum Y/R for the brackish habitat was found 0.082 g/recruit at Lc = 44.88 cm when the current L_c from the samples is 19.48 cm. It is obvious that the current fishing method for the fish samples used in the target area should change to increase MLS above the current official level of 35 cm [26]. Precautionary advice would be to increase MLS at levels equal to the optimum L_c estimated for a maximum Y/R, i.e., to 45 cm to increase the yield and directly contribute to the local population’s resilience.

4. Discussion

Following recruitment declines of the European eel since the 1980s, various recovery measures have been implemented, including Council Regulation (EC) No 1100/2007, banning eel fishing in EU countries like Ireland, Slovenia and Malta, but also in Norway as a non-EU country, including alongside seasonal restrictions across European waters. Commercial and traditional fisheries have collapsed by over 90% compared to pristine biomass levels across Europe, with the Eastern Adriatic coast reflecting similar declines [26,52]. While recommendations and measures are given at a general level, considering the panmixia hypothesis for European eels [9], they are mostly based on local population studies for eels from rivers in Northern or Central Europe, while population data are insufficient or non-existent for different habitats in southern and southeastern European regions, including Croatia. The question arises as to how realistic it is to expect the implementation of all recommendations and the effectiveness of measures in EU waters if not all ecological components of eel or fishing characteristics within different regions are known and therefore not considered when creating policies.

The size distribution of European eels varies significantly across a broad latitudinal range and diverse water environments [17]. Our results revealed that in brackish, estuarine habitats, eels were within a narrower length and weight range but were generally longer and heavier, while eels of a wider length and weight range, but lower mean values, were present in freshwater habitats. These results are in line with previously reported observations that larger eels, in terms of length and weight, tend to occupy deeper areas with stronger river flow, while smaller ones may be found in shallower zones associated with wider stretches with abundant cover [53], since freshwater habitat features strongly influence the size distribution of eels inhabiting them [54]. Similarly, longer and heavier individuals of uniform range have been identified in estuarine waters [24,29]. However, length–weight differences in catch structure can arise due to sampling selectivity without clear biological meaning. Electrofishing, as the primary method for freshwater eel sampling, is more selective for larger eels due to depth limitation and thus can affect eel sample structure. Moreover, [54] reports that sampling methods used in coastal and estuarine waters are commonly less selective for larger eels since “drop-trap” sampling shows that small eels are also present in coastal waters. In our study, larger individuals were not more frequently present in samples from freshwater environments, which means that they certainly did not have a higher catch coefficient compared to brackish ones. Further on, the population structure in the Eastern Adriatic EMU was dominated by undifferentiated individuals and larger females in both brackish and freshwater habitats. Females of A. anguilla are usually larger than males of a similar age and attain a greater body size [12,55,56], so presented results confirm previous results for the population consisting mostly of resident females and yellow- and silvering individuals in the Mediterranean lagoon [25]. However, there were no differences between either females or males within the investigated environments, indicating their homogeneity at the wider level regardless of the type of environment.

Length–weight relationships in European eels are crucial for understanding their growth patterns, biomass estimation and population health, making them essential for fishery management and conservation. The condition factor (CF) varies across habitats and seasons, reflecting different ecological conditions [53]. The allometric growth exponent (b-coefficients), which can indicate fish growth trends across habitats, is available for a large number of different European eel subpopulations [57]. In the present study, eels, both females and males, in brackish water show a positive b-coefficient of 3.045, but are significantly lower when compared to 3.401 obtained for eels in freshwater. Females show higher b-coefficients in freshwater (3.328 vs. 3.061), while males in brackish water show higher b-coefficients (3.038 vs. 2.985). However, females have higher b-coefficient in comparison with males. Several authors provide evidence of life-history variability across sexes in eels, including differences in migration timing and condition. Namely, male silver eels typically migrate earlier in the season and at a smaller size than females, which can affect their energy reserves and overall condition [52,58,59]. In the present study, sex-specific differences in b-coefficients were significant in freshwater eels but not in brackish, likely reflecting habitat-dependent sexual segregation, with larger and faster-growing females predominating in freshwater. In contrast, brackish habitats were dominated by smaller males with a narrower size range, reducing both biological divergence and statistical power to detect sex effects. Generally, these b results are similar to those described for European eels in brackish and freshwater habitats further south in Europe [18,25,60], and higher than in high latitudes [61] showing the effect of decreasing optimal conditions for eel growth, with decreasing water temperature range and time of the growing season in higher latitudes [62,63,64]. However, usually across Mediterranean brackish ecosystems, mostly lagoons, b-coefficients for eels range from 3.223 to 3.470 [24,65,66,67], suggesting good habitat quality (density, food availability or pollution) positively influences growth rates. Higher values of b-coefficient in freshwater environments suggest that, in our case, prevail superior trophic conditions inland compared to the coast. Poorer trophic conditions in Parila lagoon (brackish site of Neretva River) are possibly linked to habitat degradation and advanced competition of eel and invasive blue crab (Callinectes sapidus Rathbun, 1896) [29,68] that has significantly impacted the composition and structure of estuarine communities in recent years [69]. Interestingly, in the brackish group, there were no differences in the b-coefficient between males and females, while in the freshwater group the differences between sexes were significant, with females having a clearly higher value.

Understanding the age distribution of eels in various habitats is crucial for stock assessments, but this data is often lacking or fragmented. Efforts like the ICES Working Group on Eels (WGEEL) aim to address these gaps by compiling and analysing data across the species’ range. In the present study, we observed a fairly narrow age distribution for Adriatic Sea European eels (predominance of 3–5 age classes in both environments), similar to [25], which could be explained by exploitation effects (age truncation and growth compensation)), high underlying habitat productivity [69,70], or both [70]. European eels display a wide range of ages that vary by habitat type and regional context [71]. In freshwater, several studies report age ranges beginning near 1 year and extending to 15 years [72,73,74]. In brackish and estuarine habitats, the findings are more variable. Some papers describe narrow age spans of 0.5–4.5 years in the North African lagoons context [71], whereas [24] reported ages between 4 and 21 years in small French estuaries, while [75] noted ages from 2 years to beyond 25 years. Several studies also note that eels in warmer or estuarine conditions exhibit earlier silvering and faster growth relative to populations in freshwater [62,76]. The traditional view of long-lived eels with low growth rates does not fit the picture of Mediterranean eel populations [25]. Certainly, differences in methodologies, from otolith increment counting with validation through mark-recapture to length-based inference, help explain part of this heterogeneity in reported results.

The study of growth based on three different models showed that the age data are best described with the VBGF model, which is a common growth model used to explain eel growth [24,76,77]. Recently, different approaches, such as otolith biochronology, direct age-length regressions, or habitat-specific comparative analyses, have been used to fit better the different habitats that the species uses during its life cycle and the different environmental conditions [78,79]. However, the preference for the von Bertalanffy Growth Function (VBGF) over Gompertz or logistic models in eel management is primarily due to its mathematical compatibility with standardised Yield per Recruit and mortality models, rather than a lack of biological fit, while otolith-based chronologies and other approaches provide a better ecological context.

Most of the available studies on eel age and growth (see Table S2) have been conducted in freshwater habitats at higher latitudes [56,80,81] (papers cited in Table S2), and less is known about the population dynamics in brackish systems (estuaries and coastal lagoons) in southern areas [55]. The present study of the growth analysis of eels from brackish and freshwater habitats provides the first results for the Eastern Adriatic. The values of the L∞ show that the eels can grow larger in the freshwater habitat than in the brackish habitat. In addition, growth coefficients (K, year⁻¹) indicate that they grow faster in the brackish than in freshwater habitats (ANOVA test, p < 0.001), confirming previous reports that estuaries and lagoons along the Mediterranean coasts support higher growth rates compared to river habitats and probably offer more favourable conditions to support eel growth [24,57,61,73,82,83,84]. This is possibly owed to the higher temperatures (in this study, we measured a temperature difference of almost 1.6 OC for the freshwater habitat, which is also statistically significant; ANOVA test, p < 0.05) and the higher productivity of estuaries and coastal lagoons and the lower osmoregulation costs [85] in relation to the freshwater habitats. Growth coefficients estimated in this study belong to the higher range values of studies (Table S2). According to Table S2, the values of the K coefficient range between 0.013 and 0.123 year⁻¹ for freshwater habitats and between 0.058 and 1.095 year⁻¹ for brackish habitats. Reported values below 0.04 year⁻¹ are not considered without further research, since based on those, the species’ longevity exceeds 100 years (based on the 3/K rule [86]) and, in some cases, reaches 230 years (K values of 0.013 year⁻¹). However, we believe that our findings require further investigation of this specific population as they are based on samples containing individuals aged up to 13 years, while the growth coefficient (K) from the VBGF growth equation, suggests a life span of up to 28 years for the freshwater habitat group, and up to 20 years for the brackish habitat group. Data on European eel longevity show that an upper acceptable age limit for the longevity of the European eel is 30–32 years, which is in accordance with our data [87]. Combining these facts, i.e., the small number of age groups in the samples and the similarity between the longevity estimates based on our data with those estimated based on radiocarbon techniques [87], we can conclude that the growth curve of this eel population reaches its senescence phase early during its life cycle and therefore, samples with individuals up to 13–15 years of age are enough to allow the estimation of the growth equation coefficients with increased accuracy.

Even though tag–recapture studies have not been conducted in the study area, it is difficult to assess whether the local Croatian population can inhabit equally freshwater and brackish habitats as in other cases where populations of eels have been found to inhabit estuarine for their whole life cycle [24,79,88,89], instead of mixed strategies [90]. We may speculate that, in our case, both habitats can be inhabited by the specific population due to their similarities in the growth-performance indicators, which we calculated at 3.16 and 3.09 for freshwater and brackish habitats, even though we cannot discard the possibility that the brackish habitats are intermediate for this population during its migration towards freshwater habitats. In any case, this further underlines the importance of estuaries for European eels in the Southern European and Mediterranean areas [24,54,73,79,91].

Studies estimating natural mortality in eels reveal that demographic models integrating body mass, temperature, and stock density principles are widely used [12]. According to this concept, the average natural mortality was found to be lower in the brackish (0.174 year⁻¹) than the freshwater (0.191 year⁻¹) habitat. Estimated mortality rates (−0.392 and −0.378) fall within the values (from −0.25 to −0.46 per year; [45]) for the European eel [92,93]. In particular, [45] concluded that an appropriate value for the European eel, as derived from 15 different stock studies, is −0.46. Other studies provided estimates of 0.31–0.78 year⁻¹ for river or lake populations, with mortality varying by factors such as age, size, density and migration risks [12]. The values of μ₁ (0.993–0.996) also fall within the value range of 0.2–4.1 per year [45]. In addition, based on the nomographs presented by [45], and the average temperature of 15.9 °C, the eel stock under study can be considered as a low-density stock for both sexes. The negative effect of eutrophication processes in coastal lagoons and estuaries to the survival and growth of eels in terms of density was highlighted by [25] because while fishing pressure keeps density low, high eutrophication can significantly increase the density and consequently affect natural mortality.

While natural mortality, linked to habitat loss and contamination, plays a role, fishing pressure remains the dominant threat [94]. In the present study, mean total mortality was higher in freshwater (0.86 year⁻¹) in comparison with brackish (0.83 year⁻¹) habitats. However, the calculated exploitation ratio of 78.9% and 80.7% estimated for freshwater and brackish, respectively, reveals high fishing pressure in both habitats, with more than 50% of eels aged three years under exploitation. European eel populations face severe fishing-induced mortality, contributing significantly to their decline across European ecosystems [95,96]. Silver eel migration mortality reaches 22–26% in the Netherlands and up to 60% in a Danish fjord [97]. Across various ecosystems, fishing mortality often exceeds sustainable limits, particularly during migration [98], whereas regulated areas like Baltic coastal waters pre-2009 saw more controlled impacts [4].

The Y/R analysis revealed that the current L_c values for both brackish (19.40 cm) and freshwater (11.10 cm) habitats in Croatia are much lower than the estimated minimum landing size (44.880 cm versus 55.489 cm). It is obvious that the current fishing method in the fishing area should change in order to increase the official MLS from 35 cm to equal or above 45 cm. All Y/R analyses performed by ICES groups following analysis by [3] show exploitation rates far above sustainable levels, leading to the recommendation of zero catch across all life stages in all habitats and fisheries to meet the EU’s goal of 40% spawning stock biomass escapement [17]. Obviously, continued overfishing may lead to long-term depletion of eel stocks, and full recovery may take centuries [99]. Data on recreational and illegal fisheries remain inconsistent despite declining commercial fisheries pressure [100]. Habitat restoration efforts aim to improve coastal lagoons and water quality, particularly in the Mediterranean, while ICES continue to recommend zero catch for all life stages and habitats to reduce human-induced mortality.

5. Conclusions

It is clear from the present study that there are differences in population parameters between eels living in brackish/freshwater environments in the EMU: EA, and that they fall within the range obtained for other similar areas, especially lagoons and estuaries of the Mediterranean Sea. High values of natural- and fishing mortality and exploitation rates were determined, and the Y/R analysis indicates that an increase in the minimum landing size should be discussed to enhance the local population’s resilience and, indirectly, increase the fishing yields for fishers. These results underscore the importance of balancing ecological sustainability with economic considerations in the near future when formulating eel management policies. The Croatian government banned recreational eel fishing across coastal and riverine areas in 2023, while long-term management measures in the EMU: EA, particularly in the Neretva Estuary and Prokljan Lake, as key traditional fishing grounds, have to be further discussed and measures based on stakeholder consensus implemented. Integrated assessments, such as those in the Baltic Sea, could improve stock management, while freshwater protected areas have shown increased silver eel production compared to exploited zones. Our research confirms and highlights again the need for catchment-scale approaches, considering habitat connectivity and diversity. For sure, long-term recovery relies on continued monitoring and adaptive management across diverse ecosystems, from coastal marshes to inland waterways. However, given the European eel’s critically endangered status with no signs of recovery, if no measurable improvement occurs within the next three–five years, urgent management measures may be necessary, including a full moratorium on eel fishing in Croatia. Our results can definitely serve as a base for the development of a management plan or to inform further management measures.