*3.2. Model Calibration*

Twelve species accounted for 98% of total assemblage biomass at the sampling sites selected for the model calibration phase, including 10 fish and two decapod taxa (Table 3).

The likelihood ratio tests among pairs of model formulations highlighted that the temporal factor significantly explained assemblage biomass (*p*-value = 0.002; test t1), and that the geographical factor (sub-basin) significantly explained assemblage biomass (*p*-value = 0.002) when added to a model already including temporal factor (t2). Among physico-chemical predictors, temperature significantly (*p*-value = 0.041) improved a model already including both temporal and geographical factors (t3.1), and salinity significantly (*p*-value = 0.012) improved a model already including temporal and geographical factors and temperature (t3.2). Dissolved oxygen (t3.3), turbidity (t3.4), sediment grain size (t3.5) and presence of macroalgae (t3.6) did not improve the model significantly (*p*-values > 0.05). Finally, the inclusion of the location factor to the model already taking into account temporal and geographical factors, temperature and salinity significantly (*p*-value = 0.045) explained assemblage biomass (t4). As a result, the GLM formulation including temporal factor, geographical factor, temperature, salinity and location was selected as the best model explaining the variability of nekton biomass.

Most of the species included in the analysis were predicted accurately (cross-validated average Spearman's coe fficients between 0.28 and 0.60), including big-scale sand smelt *Atherina boyeri* Risso, 1810, Mediterranean banded killifish *Aphanius fasciatus* Valenciennes, 1821, rockpool prawn *Palaemon elegans* Rathke, 1837, brown shrimp *Crangon crangon* Linnaeus, 1758, black-spotted goby *Pomatoschistus canestrinii* Ninni, 1883, marbled goby *P. marmoratus* Risso, 1810, Adriatic dwarf goby *Knipowitschia panizzae* Verga, 1841, thinlip grey mullet *Chelon ramada* Risso, 1827 and european flounder *Platichthys flesus* Linnaeus, 1758 (Table 3). Conversely, golden grey mullet *Chelon auratus* Risso, 1810, leaping mullet *C. saliens* Risso, 1810 and gilthead seabream *Sparus aurata* Linnaeus, 1758 were predicted with lower accuracy (cross-validated average coe fficients <0.25), and therefore they were excluded from the prediction of the expected assemblage. Overall, the whole assemblage was predicted with good accuracy (cross-validated average coe fficient = 0.46; Table 3).

**Table 3.** Average and standard error values of Spearman's coefficients calculated by cross-validation (k = 5) on species accounting for 98% of total biomass and on the whole assemblage. For each species the estuarine use functional guild (EUFG), the feeding mode functional guilds (FMFG) and the average biomass are reported. Guilds are abbreviated as follows. ES: estuarine resident species; ESs: solely estuarine resident species; MED: marine estuarine-dependent species; Bmi: microbenthivores; DV: detritivores; Bma: macrobenthivores; HZ: hyperbenthivores-zooplanctivores; HP: hyperbenthivores-piscivores; OV: omnivores.


Among the significant environmental factors, both site location within saltmarsh creeks and salinity showed a strong average influence on the whole assemblage biomass according to the selected model. By comparison, temperature had a weaker average effect (Figure 2). Nekton response to environmental and location factors varied markedly at species level (Figure 3). Overall, biomass of most of the species showed a clear response to the variables investigated, with only *A. boyeri* and *K. panizzae* exhibiting weak relationships with water temperature, salinity and site location compared to the other species. The marginal effect of temperature (i.e., with seasonal factor already taken into account) was positive for biomass of most of the species, while both salinity and location showed variable effects. In particular, salinity was negatively associated to biomass of *C. ramada*, *C. saliens*, *P. canestrinii* and *P. elegans*, and positively to biomass of *A. fasciatus* and *P. marmoratus*. The effect of sites located within saltmarsh creeks, as opposed to exposed saltmarsh edges, was positive for most of the species, with only *C. crangon*, *P. flesus* and *P. marmoratus* showing a negative relationship with creeks.

**Figure 2.** The average magnitude of the effect (and standard error) of relevant environmental and location factors estimated by the selected model. The values are calculated as the mean of the absolute (i.e., without sign) coefficients estimated for each species, weighted by average species biomass.

**Figure 3.** The estimated coefficients (and standard errors) of relevant environmental and location factors for species accounting for 98% of assemblage biomass.

### *3.3. Predicting the Expected Assemblages*

Model predictions applied to the selected scenarios of salinity reduction showed that the whole nekton assemblage (calculated as the cumulative biomass of species included in the analysis) would markedly respond to salinity variations, showing a progressive increase in biomass from the current to the low-salinity scenario. The predicted increase in biomass would occur similarly in all the seasons investigated, although this appears to be stronger and associated to a lower uncertainty in spring (Figure 4). Despite the overall trend observed in whole assemblage, some major differences could be found in the response at the species level (Figure 5). Only minor variations of species response among seasons were detected (see Figure A2 in Appendix B), hence average values were shown (Figure 5). Most of the species investigated showed an increase in biomass with decreasing salinity, including *A. boyeri*, *C. ramada*, *P. canestrinii*, *P. elegans* and *P. flesus*. Among them, *P. canestrinii* and *P. elegans* were associated to the most marked increase in biomass. In turn, *A. fasciatus*, *K. panizzae* and *P. marmoratus* exhibited an opposite pattern, with higher biomass densities under the current conditions and progressively lower values under reduced salinity scenarios. *C. crangon* did not show any relevant response to salinity.

**Figure 4.** Whole nekton assemblage biomass (g 100 m<sup>−</sup>2) predicted along the salinity gradient for each season investigated. The current salinity and the three scenarios of salinity reduction are highlighted.The dashed lines indicate the standard error of the prediction.

**Figure 5.** Species biomass (g 100 m<sup>−</sup>2) predicted along the salinity gradient. The seasonal values were averaged. The current salinity and the three scenarios of salinity reduction are highlighted. The dashed lines indicate the standard error of the prediction.

Aggregating species biomasses in ecological and trophic guilds highlighted the potential changes in assemblage functional attributes after the restoration (Figure 6). On the whole, biomass of estuarine resident and marine estuarine-dependent species are expected to increase accordingly with the decrease in salinity, although the latter guild showed a marked increase only under the low-salinity scenario. In contrast, solely estuarine residents exhibited a pattern of decrease from the current conditions to mid and low-salinity scenarios. In terms of trophic structure of the nekton assemblage, salinity reduction would drive an increase in biomass of detritivorous, macrobenthivorous, hyperbenthivorous-zooplanctivorous and hyperbentivorous-piscivorous species (Figure 6). Guild response did not exhibit major variations among seasons (see Figure A3 in Appendix B), hence the average values were shown (Figure 6).

**Figure 6.** Biomass (g 100 m<sup>−</sup>2) predicted along the salinity gradient for ecological and trophic guilds. The seasonal values were averaged. The current salinity and the three scenarios of salinity reduction are highlighted. The dashed lines indicate the standard error of the prediction. Guilds are abbreviated as follows. ES: estuarine resident species; ESs: solely estuarine resident species; MED: marine estuarine-dependent species; Bmi: microbenthivores; DV: detritivores; Bma: macrobenthivores; HZ: hyperbenthivores-zooplanctivores; HP: hyperbenthivores-piscivores; OV: omnivores.
