Evaluating Leaf Area and Biomass Relationship in Posidonia oceanica (L.) Delile: A Tool for Non-Destructive Assessment

Abstract

Posidonia oceanica (L.) Delile is a seagrass endemic to the Mediterranean Sea and a significant contributor to human well-being through a variety of ecosystem functions, such as carbon cycling. Despite its ecological significance, most methods for estimating leaf biomass in this species are either destructive or expensive. In this study, 2500 individual leaves from 351 shoots of Posidonia oceanica were collected, across 16 sites in the Ligurian Sea, over two time periods (2016–2018 and 2024–2025), and analyzed for total leaf area and dry weight. An allometric equation following a power-law structure was derived using linear mixed-effects models and it was later validated via 10-fold cross-validation. Although some variations in the intercept term were observed, the allometric scaling structure remained consistent across space and time, providing the first robust species-specific allometric tool to estimate P. oceanica biomass, forming the basis of a proposed non-destructive protocol for meadow-scale biomass estimation.

1. Introduction

Posidonia oceanica (L.) Delile is a seagrass endemic to the Mediterranean Sea and one of the most ecologically significant components of the basin’s carbon cycle [1,2]. A defining feature of these meadows is their capacity for long-term carbon sequestration, achieved through the formation of the matte—a dense matrix of dead rhizomes, roots, and sediment that accumulates over centuries and stores substantial quantities of both organic and inorganic carbon [3,4,5,6,7,8]. The matte stocks vary between 40 and 237 kg C m⁻² across the basin [2,3,4,5,6,7,8,9], with the total organic carbon stock trapped in P. oceanica matte estimated at 711–1067 million Mg C [9,10].

The second major contribution of P. oceanica to the carbon cycle lies in its exceptional primary productivity. The leaf blades are the most productive part of the plant, accounting for 79% of the carbon fixed per shoot, followed by sheaths (17%) and rhizomes (4%) [2]. Gross carbon fixation capacity ranges between 33.5 g C m⁻² per year, in light-limited environments, to 426 g C m⁻² per year, in shallow and well-illuminated areas [2,6]. This high productivity contributes directly to carbon storage carbon storage, since it is estimated that roughly between 10% and 36% of biomass produced in meadows is permanently buried in the matte, constituting a long-term carbon sink [8,11,12].

Given the importance of P. oceanica in Mediterranean carbon budgets, a range of monitoring techniques have been developed to quantify its productivity and carbon fixation. One of the most widely applied direct methods is the leaf marking technique originally described by Zieman (1974), in which divers pierce all leaves at a fixed reference point and, after an interval of 15–30 days, collect shoots and measure new tissue growth from the ligule to the mark, deriving biomass production over the elapsed period of time [13,14,15,16]. An alternative reconstructive technique is lepidochronology, which requires the sampling of cuttings and the analysis of each sheath’s thickness to reconstruct decades of productivity [14,17,18]. More recently, non-destructive methods have been explored with the aim of reducing physical disturbance to the meadow. However, most of these techniques are expensive and require special materials, such as the aquatic eddy covariance, which measures oxygen flux above the canopy through an array of sensors [5].

Alongside these approaches, allometric relationships have been used as a means to derive P. oceanica leaf biomass and primary productivity without the need for destructive sampling [7,19]. Such techniques have been widespread for many years in the field of terrestrial botany [20,21,22,23,24]. Duarte (1991) made the first attempt at establishing an allometric relationship across seagrass species in the Mediterranean [25]. He demonstrated that a power-law relationship exists between morphological traits and biomass. Building on this work, subsequent studies have validated allometric power-law scaling of leaf biomass in terms of leaf area in species such as Zostera marina [26,27,28,29]. For P. oceanica, however, no reliable and validated allometric equation has been developed, with most existing productivity assessments still relying on destructive methods [30].

This gap persists because the biomass-to-area relationship is dependent on a range of site-specific factors, including depth, light availability, hydrodynamics, and nutrient availability, which introduce variability across meadows and seasons [30,31,32]. Establishing such a relationship would represent a significant step toward the development of non-destructive methods for estimating leaf biomass and, subsequently, primary productivity across P. oceanica meadows [33,34]. This prospect carries particular relevance in the current conservation context, where several restoration programs are being implemented across the Mediterranean Sea to prevent the decline of this seagrass [35,36,37,38,39].

The objective of this study is to evaluate whether a robust allometric relationship between P. oceanica leaf area and dry weight can be established and serve as the basis for a non-destructive biomass estimation equation. The derived equation is integrated with current seagrass monitoring techniques to outline a protocol for meadow-scale biomass assessment.

2. Materials and Methods

A dataset of roughly 2500 individual Posidonia oceanica leaves was assembled, collected from 351 shoots across 16 different sites, scattered around the region of Liguria, Italy (Figure 1). The Ligurian Sea was chosen as a case study because of its ecological heterogeneity, encompassing meadows at various depths, conservation states and subjected to several anthropic pressures [40,41,42,43,44,45,46,47]. These include physical damage from anchoring, fishing activities or worsening water quality from maritime and fluvial works, such as dredging and beach nourishment [30,40,41]. The sites selected for this study host P. oceanica meadows at various depths, with some located within Marine Protected Areas (MPA) (

Table 1. Sites selected for the study, with coordinates, sampling depth, and conservation status, indicating sites located within a Marine Protected Area (MPA). Anthropogenic pressures are classified following Holon et al. (2015) [42] and their references are indicated. TU = touristic facilities; A = boat anchoring; ANT = coastline anthropization (intensive urbanization and high usage of coastal areas); CE = coastal erosion; WQ = water quality degradation due to urbanization and coastal runoff; FI = recreational fishing; n.a. = data not available.

Site	Longitude	Latitude	Depth (m)	Protection	Pressures
Punta Pedale	44°19′15.4956″ N	9°12′56.5662″ E	12	No	TU, A, ANT, FI [43,44]
Paraggi	44°18′39.4164″ N	9°12′34.9164″ E	8	Yes	TU [43,44]
Niasca	44°18′30.6612″ N	9°12′37.9038″ E	7	Yes	TU [43]
Sanremo Porto Vecchio	43°48′45.4602″ N	7°46′37.6206″ E	13	No	A, ANT [43]
Bergeggi	44°14′15.9072″ N	8°26′39.3252″ E	12	Yes	TU, WQ [43]
Punta Manara	44°15′11.037″ N	9°24′32.2056″ E	15	No	WQ, ANT [45]
Monterosso	44°8′26.901″ N	9°38′38.6946″ E	18	Yes	WQ [46]
Prelo	44°20′25.3746″ N	9°13′32.0766″ E	14	No	ANT, TU [41,46]
Framura	44°12′15.1302″ N	9°32′51.954″ E	13	No	TU, A, ANT [47]
Camogli	44°20′42.6402″ N	9°9′10.1088″ E	15	No	WQ, ANT, TU [44,47]
Diano Marina	43°54′22.6188″ N	8°5′6.1758″ E	12	No	n.a.
Ventimiglia	43°47′11.7492″ N	7°35′25.998″ E	15	No	A, ANT, WQ [41]
Noli	44°12′25.128″ N	8°25′17.9646″ E	11	No	WQ, ANT [47]
Spotorno	44°13′35.5974″ N	8°25′28.3218″ E	16	No	WQ, ANT [47]
Vado Ligure	44°17′20.079″ N	8°27′34.3368″ E	15	No	WQ, ANT [47]
Ospedaletti	43°47′49.8624″ N	7°42′35.0598″ E	10	No	n.a.

). While all sites are subject to differing degrees of anthropogenic pressure, site-specific measures of the nature and extent of these pressures fall outside the scope of this study, as comprehensive historical records are not readily available. Samples were always collected at the center of the meadow, in an area of high shoot density and cover. Shoots were collected across seasons in different proportions, with the highest amount in autumn (135), followed by summer (90), winter (63) and spring (63).

2.1. Sampling Design

At each site, the number of replicates was chosen according to the guidelines given for biomass monitoring by the PREI index (Posidonia Rapid Assessment Index) [14,47]. Accordingly, a total of nine P. oceanica shoots were collected randomly from three different stations by SCUBA divers at each site.

All morphological analyses were performed in the laboratory following standardized procedures [14,17,30,48,49]. For each shoot, juvenile and severely damaged leaves were removed, the remaining leaves were counted, and their length and width were measured, to obtain the following descriptors: the number of leaves per shoot (n_leaves·shoot⁻¹), leaf width (cm) and leaf length (cm). Leaf length was defined as the longest extension of the leaf from apex to base and the leaf width as the longest extension of any two points on the blade parallel to each other and perpendicular to the length axis [50]. After measurements, all the epiphytes were scraped off from the leaves and the clean leaves were dried at 60 °C for 48 h to measure dry biomass (g DW).

For each shoot, we calculated the total leaf area (cm²), mean leaf biomass per shoot (g DW shoot⁻¹), and the mean carbon content per shoot (gC shoot⁻¹). Because leaf area was estimated assuming a perfectly rectangular shape, a correction factor (CF) was calculated and applied using a subset of 160 randomly selected leaves, which were photographed using an OM System Tough TG-7 digital camera (OM Digital Solutions, Tokyo, Japan). The true leaf area was derived through the software Fiji (Version 1.54), National Institutes of Health, Bethesda, MD, USA, and divided by the corresponding rectangular estimate to obtain the CF [50,51,52,53]. Carbon content within shoots was then estimated using conversion coefficients gathered from the literature. The regional Mediterranean leaf tissue typically contains an average of between 37.8 ± 1.6% and 41.3 ± 0.4% carbon for leaf tissue [2,54], although earlier localized studies established lower values (27.4 ± 1.5%) [7]. Within this study, we used the values proposed by Pergent-Martini et al., (2021) [2], as they provide measurements of over a hundred leaves across various sites and depths, allowing the protocol the capacity to estimate carbon content across the whole Mediterranean basin. To finally derive an equation that estimates carbon content, without accounting for seasonal or depth-related variations, the Pergent-Martini et al. (2021) [2] conversion factor was inserted in the equation and the uncertainty of carbon content was calculated by combining the carbon fraction variability (CV = 4.23%) with the allometric absolute error, through propagation of relative errors in quadrature.

2.2. Data Analysis

To quantify the relationship between leaf dry weight (DW) and leaf area in Posidonia oceanica, a log–log allometric model (1) was created:

$$\log \left(D{W}_{ij}\right)=\alpha +\beta log\left(A_{ij}\right)+{\epsilon }_{ij}$$

(1)

where $D{W}_{ij}$ is the dry weight of leaves in a shoot i at site j, $Aij$ is the corresponding leaf area, and $\epsilon ij$ represents the residual error. For comparison, the linear model of leaf area and dry weight (2) was also considered, expressed as follows:

$$D{W}_{ij}=\alpha +\beta A_{ij}+{\epsilon }_{ij}$$

(2)

Diagnostics plots were subsequently performed to evaluate model adequacy, to verify assumptions of homoscedasticity and normality of residuals. To test differences in allometric scaling between the two sampling periods (i.e., 2016–2018 and 2024–2025), linear mixed-effect models (LMMs) with Gaussian error distribution were fitted on log-transformed data (Equations (3) and (4)).

$$log\left({DW}_{ijk}\right)={\beta }_0+{\beta }_{1}log\left(A_{ijk}\right)+{\beta }_{2}Datase{t}_{ikj}+u_{0j}+u_{1j}\mathrm{l}\mathrm{o}\mathrm{g}\left({\mathrm{A}}_{\mathrm{i}\mathrm{j}}\right)+{\epsilon }_{ij}$$

(3)

$$log\left(D{W}_{ij}\right)=\mathrm{E}\mathrm{q}\mathrm{u}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n} \left(3\right)+{\beta }_{3}log\left(A_{ij}\right)·Datase{t}_{ikj}$$

(4)

where $Datase{t}_{ikj}$ represents the sampling period k at the site j, and $u_{0j}$ and $u_{1j}$ are site-specific random intercept and slope, respectively. In the first LMM (3), the sampling period (Dataset_ikj) and leaf area (A_ij) were included as fixed effects. In the second model (4), their interaction was also included. Because multiple leaves were sampled close by or within the same sites, to account for spatial non-independence, the term site was included as a random factor in both LMMs. The significance of the model terms was assessed using the Wald χ² test analysis of variance. Different combinations of fixed and random factors were compared using the Akaike Information Criterion (AIC) and likelihood ratio tests (LRTs) using maximum likelihood estimation. Fixed effects were evaluated using Kenward–Roger approximated F-tests. To quantify the proportion of variance explained by fixed and random components, marginal and conditional R² were computed following the Nakagawa and Schielzeth framework. Marginal R² describes the proportion of variance explained by fixed factors, and conditional R² includes fixed and random factors [55].

Finally, the predictive performance of the model was assessed using 10-fold cross-validation, as well as the Root of the Mean of Square of Errors (RMSE) and the Mean of Absolute value of Errors (MAE). In addition to predictive performance parameters, a 95% confidence interval was constructed using a parametric bootstrap procedure (999 iterations). For each iteration, the dataset was resampled and the LMM was refitted to extract allometric parameters from the fixed effects, combining them with residual noise drawn from the RMSE. The resulting distribution was used to estimate the 2.5th and the 97.5th percentile as lower and upper bounds of the confidence interval (CI). A last validation of the predictive power of the final equation was achieved through the Concordance Correlation Coefficient (CCC), assessing the agreement between observed and predicted biomass values.

All statistical analyses were conducted in R (version 4.5.0), using R-studio version 4.5.0 and Claude 3.5 Sonnet (Anthropic) as an assistant to generate and de-bug code to (i) compare LMMs using the lme4 and lmerTest packages; (ii) compute performance metrics (RMSE, MAE and CCC); and (iii) produce publication-quality figures using ggplot2. All code generated with AI assistance was reviewed, tested and validated. Claude was not used to interpret results or draw conclusions. All analytical decisions were made independently by the authors.

3. Results

3.1. Shoot Canopy Morphology

A total of 351 shoots were analyzed. Leaf dry weight per shoot ranged from 0.097 g to 2.158 g, with a mean value of 0.719 g (±SD = 0.4221 g). Total shoot leaf area varied between 28.21 cm² and 663.36 cm², with a mean of 196.75 cm² (±SD = 94.46 cm²). The number of leaves per shoot ranged from 3 to 12, with an average of 5.69 leaves per shoot (±SD = 1.37 leaves). The site displaying the greatest total shoot area was Punta Pedale, with a mean area of 391 cm² (±SD = 155 cm²) and a considerable variation in leaf size over time, going from 101 cm² to 663 cm². Diano Marina displayed the lowest mean area of 118 cm² (±SD = 34.1 cm²), ranging from 54.2 cm² to 170 cm². When considering mean leaf area, Vado Ligure had the largest individual leaves, with a mean of 59.4 cm² per leaf (±SD = 12.8 cm²) and Diano Marina resulted as having the smallest with a mean of 20.1 cm² (±SD = 4.52 cm²). The correction factor for leaf area had a mean value of 0.996 (±SD = 0.086), indicating that manual measurements slightly underestimated the true leaf area on average.

Based on the measured shoot dry weights and carbon content values reported in the literature for Posidonia oceanica, leaf carbon stocks were estimated for each shoot. Using the most conservative estimate by Pergent et al. (1994), the mean carbon content of shoots was 0.197 ± 0.1162 g C shoot⁻¹, ranging between 0.0265 and 0.5913 g C shoot⁻¹ [7]. Based on the values provided by Pergent-Martini et al. (2021), the mean carbon content of shoots is 0.2717 ± 0.16 g C shoot⁻¹, with a range varying between 0.0366 and 0.8157 g C shoot⁻¹ [2]. As the latter values are derived from a comprehensive dataset compiled from various studies spanning the entire Mediterranean basin, they are considered more representative of regional variability and are therefore preferred for subsequent carbon stock estimates.

3.2. Leaf Area–Dry Weight Allometry

A strong positive relationship was observed between total shoot leaf area and total shoot leaf dry weight. A simple linear regression on raw-scale variables explained 74.4% of the variation in dry weight (R² = 0.744, p < 0.001). However, because plant size–biomass relationships typically follow power-law scaling [20,23,24,25,26,27,34,56], the relationship was further examined using a log–log transformation. Inspections of the marginal distributions of dry weight revealed a bimodal pattern, likely reflecting underlying a site- and season-specific structure in the data, which motivated the use of LMMs. Accounting for the hierarchical sampling structure, two random-effects structures were compared: a random-intercept model (Model 1,

Table 2. List of modeling equations employed in this study.

Model	Dependent Variable	Fixed Structure	Random Structure
1	Log(DW)	Log(Area)	1|Site
2	Log(DW)	Log(Area)	1 + log(Area)|Site
3	Log(DW)	Log(Area) + Dataset	1 + log(Area)|Site
4	Log(DW)	Equation (3) + Log(Area) · Dataset	1 + log(Area)|Site

) and a random-intercept-plus-slope model (Model 2, Table 2).

Within the first option, the factor site was inserted as a random intercept, allowing baseline dry weight to vary by site, while, in the second equation, it was inserted as a random slope, allowing the allometric scaling relationship to vary by site. The random-slope model provided a significantly better fit (χ² = 24.99, p < 0.001) and a lower AIC (−160.2 vs. −139.2) compared with the first model, indicating that the allometric scaling relationship varies significantly among sites. To understand whether the origin of site variation was related to the temporal origin of the data (2016–2018 vs. 2024–2025), mixed-effect models were run including dataset as a fixed effect (Model 3, Table 2), as well as the interaction with log-transformed leaf area (Model 4, Table 2).

ANOVA comparison between Model 3 and Model 4 resulted as marginally non-significant (χ² = 3.752, p > 0.05) with a very similar AIC (

Table 3. Comparison of model performances for LMM employed in this study, including Akaike Information Criterion (AIC), log-likelihood (LogLik), marginal R² and conditional R².

Model	AIC	LogLik	R2marginal	R2conditional
1	−139.21	73.606	0.727	0.919
2	−160.20	86.102	0.754	0.932
3	−171.82	92.909	0.835	0.921
4	−173.57	94.785	0.841	0.918

), leading to the retention of the simpler model without the interaction term. A final comparison, between Model 2 and Model 3, using LRT (χ² = 17.37, p < 0.001) and Kenward–Roger approximated F-test (F = 19.02, p < 0.001), confirmed that the dataset variable significantly improved model fit, justifying its inclusion as a fixed effect in the final model. Based on these results, the final predictive equation relating leaf area to dry weight is

$$DW=0.00248\times {\mathrm{A}\mathrm{r}\mathrm{e}\mathrm{a}}^{1.122}\times e^{ϵ}; \mathrm{with} ϵ\sim N(0,0.02867)$$

(5)

where 0.00248 (95% CI: 0.00141–0.00424) is the baseline coefficient and 1.122 (95% CI: 1.031–1.219) represents the estimated scaling exponent, indicating a slightly superlinear relationship between leaf area and dry weight and a clear deviation from isometric scaling (Figure 2).

3.3. Model Validation

The predictive performance of the model was assessed using 10-fold cross-validation. Predictions were generated using fixed effects only, excluding random site effects to evaluate general predictive capacity. Folds were generated by random partitioning of observations. The model showed good predictive performance, with RMSE = 0.241 (log scale) and MAE = 0.189 (log scale). This corresponds to a multiplicative error of $\times 1.27$, back-transformed from RMSE, and an absolute error of ±20.8% on the original scale, back-transformed from MAE. Model bias was negligible (bias = −0.0019). Model agreement was further evaluated using the CCC (0.915), indicating a strong reproducibility of biomass estimates derived from the allometric model (Figure 3). Residual diagnostics revealed no major deviations from normality or homoscedasticity, supporting model assumptions (Appendix A).

Combining the allometric equation for leaf biomass and area with the leaf carbon fraction reported by Pergent-Martini et al. (2021) [2] (37.8 ± 1.6%) as a fixed multiplier, an equation estimating carbon content (C) based on leaf area can be derived (Equation (6)). The combined prediction uncertainty of such equation is ±21.2%, propagating from the absolute error (±20.8%) and the variability of the carbon conversion factor (4.23%) [2].

$$\mathrm{C}=0.000952\times {\mathrm{A}\mathrm{r}\mathrm{e}\mathrm{a}}^{1.122}\times e^{ϵ}; \mathrm{with} ϵ\sim N(0,0.02867)$$

(6)

4. Discussion

This study provides one of the first robust allometric characterizations of the relationship between P. oceanica leaf area and dry weight. This relationship is described by an allometric power law equation, where the resulting exponent (β = 1.122) indicates a slightly superlinear relationship, meaning that leaf mass increases proportionally faster than area [31,56,57,58,59]. The validity of this scaling exponent across the two sampling periods was confirmed by the non-significant interaction between leaf area and dataset (p > 0.05), indicating that the departure from isometry is a consistent property of the species rather than a result of sampling period. The allometric factor’s departure from perfect isometry suggests that larger leaves become progressively denser or thicker as they grow, likely reflecting progressive investment in structural tissue with increasing leaf size [25,60]. To account for the not-rectangular shape of P. oceanica leaves, a correction factor was applied (0.996 ± 0.086 SD), following Schrader et al. (2021) [50]. The correction factor indicates a slight underestimation of leaf area probably because P. oceanica leaves present natural irregularities along their blades or small herbivore bite marks, preventing the leaf outline from conforming perfectly to a rectangular approximation [4,14]. The robustness and uncertainty range of the allometric equation across 16 sites and two sampling periods confirms its potential as a reliable predictive tool for estimating P. oceanica leaf biomass from non-destructive area measurements.

The use of power-law allometric models to predict leaf biomass from morphological measurements has precedent across several terrestrial plant taxa [23,25,56,61,62,63,64]. Panayotidis and Simboura (1989) [65] were among the first to study the relationship between leaf surface and biomass in Mediterranean seagrasses, suggesting that the allometric relationship investigated has general value for many species [65]. Other studies, such as Echavarria-Heras et al. (2011), followed up on this approach, using the length ratio as a tool to estimate leaf biomass in Zostera marina through an allometric model [58]. Giovannetti et al. (2008) [30], whose work initially inspired this study, were among the first to test this on P. oceanica leaves and also found a significant relationship between leaf area and weight in the Ligurian Sea using a linear regression. They acknowledged the limits of a linear approach in its capacity to account for seasonal and environmentally influenced biomass peaks, suggesting further mathematical modeling was needed [30].

These limitations are addressed in the present study through the application of LMMs, which explicitly account for site-level variability as a random factor, enabling a more robust allometric characterization. Furthermore, the use of LMMs allowed for the comparison between sampling periods throughout this study. The absence of a significant interaction between the sampling period and the area indicates that the allometric relationship is not dependent on time. However, the significant effect of dataset as a fixed term indicates that for a given leaf area, biomass differs between sampling periods. This pattern suggests that environmental or temporal factors can affect the baseline biomass level between datasets without altering the underlying scaling mechanism. Considering the high concordance score (CCC = 0.915), the low prediction error observed during cross-validation and these results, the allometric model describes the relationship between P. oceanica leaf area and dry weight in a manner that accounts for spatial or temporal constraints, strengthening its value as a broadly applicable predictive tool.

The spatial and temporal independence of the allometric equation positions it as a practical foundation for non-destructive protocols to estimate P. oceanica biomass. Such protocols would involve the non-destructive measurement of leaf area in the field, using underwater measurements of leaf length and width to derive individual leaf area. A minimum of nine randomly selected shoots per station is recommended, consistent with standard P. oceanica monitoring guidelines [14]. The allometric equation derived here is then applied to each shoot to estimate shoot dry weight, which may be scaled to meadow level using shoot density and percentage cover to derive the leaf standing crop (LSC) [14]. Both shoot density and percentage cover can be measured with non-destructive protocols, using standard quadrat-based methods [14,66]. Within the application of allometric equation, the error term must be considered since the multiplicative error structure of the allometric model results in biomass predictions typically within ±20.8% of observed values, a margin that should be accounted for when interpreting meadow-scale estimates.

For large-scale or temporally repeated assessments, spatial heterogeneity in meadow structure must also be considered. Variations in shoot density and percentage cover—both of which influence the scaling from shoot-level to meadow-level biomass—should be monitored at pre-defined stations stratified according to the meadow’s structural heterogeneity [2,14,41,66,67,68]. This stratified approach ensures that spatial gradients in meadow condition are adequately captured rather than averaged out.

Despite these limitations, the promising results observed within this study depict the allometric estimate as an attractive alternative to traditional sampling methods. Lepidochronology, for instance, requires the physical uprooting of shoots or the imposition of physical stress on the plant, potentially harming the meadow. In contrast, the allometric equation derived here provides a non-destructive method to estimate leaf biomass with predictive accuracy, requiring only the measure of leaf dimensions in the field. It must be acknowledged that the non-destructive nature of this approach carries inherent trade-offs. Physical sampling methods such as lepidochronology allow for the potential study of tissue carbon content and the reconstruction of growth patterns through the examination of rhizome scales, information that can be obtained only by physical sampling. The allometric approach is therefore best understood not as a wholesale replacement for destructive methods but as a tool that is especially valuable when the primary objective is biomass estimation and when meadow integrity must be preserved.

The need for such non-invasive monitoring is particularly relevant in restoration contexts. Pansini et al. (2024) and Bacci et al. (2024) both emphasize that restoration projects require monitoring metrics that do not damage transplanted patches, especially given the long timescales over which restoration must be assessed, often exceeding 10 years [16,69]. The allometric equation therefore opens up new possibilities for non-harmful monitoring of biomass, both at a local and seascape scale.

Estimating carbon content within leaves represents a further valuable application of this protocol in restoration and monitoring contexts. In this study, the Mediterranean average carbon content of 37.8% in P. oceanica leaves, reported by Pergent-Martini et al. (2021), is used to derive an equation estimating carbon content from leaf area [2]. It must be acknowledged that this is a general estimate that does not account for seasonal, geographical or ecological variations in leaf carbon content. Consequently, the carbon equation is most reliable when integrating data across multiple seasons and sites, rather than individual shoots sampled at a specific time and depth [32]. Furthermore, the conversion of dry weight to carbon content introduces a layer of uncertainty on top of the allometric confidence interval, as it combines the carbon ratio variability (±1.5–1.6 percentage points) with the equation’s error term, resulting in a combined prediction uncertainty of ±21.2%. The difference in the conversion factors of Pergent-Martini et al. (2021) and Pergent et al. (1994) suggests that carbon content may vary with meadow ecological condition, season, and geographical location, which prompts caution when applying the fixed conversion factors across the dataset [2,7]. Nevertheless, the combination of literature data regarding carbon content and the allometric equation presented here represents a promising first step toward further methodological development of a reliable tool for estimating meadow carbon content in P. oceanica leaves. Such estimates could contribute to the growing body of work on Blue Carbon, providing a basis for carbon credit quantification to support funding of conservation and restoration projects [64], a factor that would greatly benefit restoration programs, in line with the objectives to achieve in the European Union Environmental Action Programme (EAP) [41,64,65,66,67,68,69,70,71]. However, such applications remain limited at the current stage of methodological development, as further applications of the allometric equation derived here are required for a standardized industrial scale of application.

Despite the promising applications of this allometric equation, several limitations must be acknowledged. The most significant one is the failure to account for environmental covariates such as depth or water quality, which can severely influence the morphology of P. oceanica leaves [5,11,30,31,72,73,74]. Instances of this can be found in Jiménez-Casero et al. (2023), where it is shown that the structure of P. oceanica leaves is affected by nutrient discharges through photo-physiological adjustments or the reduction in shoot architectural complexity [75]. Both these features severely affect the plant’s capacity to allocate leaf tissue, thereby impacting the length and width of the leaf, which in turn can alter the allometric relation [76,77,78,79]. This limitation is further exacerbated by the constraints of the dataset, where the size range of leaves in shoots ranged between 28.21 cm² and 663.36 cm². This range excludes juvenile leaves or extremely large leaves, as most allometric models rely on mature leaves [2,14,34,64]. Future works should aim to expand the current work and cover a broader size range, including juvenile shoots, providing a greater relevance when applying the non-destructive protocol to restoration sites. Finally, although the Ligurian dataset is robust, further validation across meadows throughout the Mediterranean would help confirm the broader applicability of the derived equation, particularly for meadows subjected to a wider range of stressors that may alter leaf morphology and carbon content [16]. Such future operational field validation of this proposed protocol would include the study of meadow biomass through traditional sampling methods and the comparison of these results to the predictions of the allometric model. Once small-scale implementations have been tested and the protocol is fine-tuned, the large-scale implementation of biomass estimation could be tested. At larger scales, implementation would further require the use of advanced technologies such as remote sensing. For instance, the implementation of a side-scan sonar (SSS) or multibeam echosounder to measure a meadow’s extent and delineate its boundaries provides key information in the creation of a sampling plan and in the accurate estimate of whole-meadow canopy biomass [9,80].

5. Conclusions

The allometric equation developed in this study represents a noteworthy tool for researchers and managers studying Posidonia oceanica leaf biomass from non-destructive measurements of leaf area. Validated across 16 sites and two sampling periods, the model demonstrates a stable allometric structure, with a consistent scaling relationship between leaf area and dry weight despite environmental and spatial variability. While differences in baseline biomass between datasets and sites were observed, the underlying allometric exponent remained constant, preserving a fundamental scaling relationship in P. oceanica leaves across conditions.

Thus, the allometric relationship described here represents an important step towards the development of a non-destructive approach for assessing seagrass biomass. By linking easily measurable morphological traits to dry weight, the model enables the estimation of shoot-level biomass and its subsequent scaling to meadow level using standardized measurements of shoot density and cover. Hence, this work provides one of the first frameworks for integrating individual-level measurements into seascape-scale biomass assessments. Importantly, this approach has not yet been tested in the field at full operational scale. However, this work initiates a body of research that establishes the basis for a non-destructive monitoring protocol, incorporating species-specific growth dynamics and extending this framework towards estimates of primary productivity.

Overall, the results highlight the strength of allometric scaling as a predictive tool in seagrass ecology and support the development of a standardized, low-impact method for long-term monitoring of Posidonia oceanica meadows.