In the Mediterranean Sea, Risso’s dolphin (Grampus griseus, Gg), long-finned pilot whale (Globicephala melas, Gm), and Cuvier’ beaked whale (Ziphius cavirostris, Zc), are considered low-density elusive species. Their assessment status under the IUCN Red list of threatened species recently changed from ‘Data Deficient’ to, respectively, ‘Endangered’ (Gg, Lanfredi et al., 2021), and ‘Vulnerable’ (Gm, Gauffier and Verborgh, 2021; Zc, Cañadas and Notarbartolo di Sciara, 2018). A distinct subpopulation of long-finned pilot whales, limited to the Strait of Gibraltar area, and listed as ‘Critically Endangered’, was also identified during the last assessment (Verborgh and Gauffier, 2021). The three species are listed in Annex IV of the EU Habitats Directive (HD, Directive 92/43/EEC) as species requiring a special protection regime across their natural range, both within and outside the Natura 2000 sites, to enable their Favourable Conservation Status (FCS) to be maintained or, where appropriate, restored, in their natural range. The core areas of their habitat must be identified, designated as Sites of Community Importance, included in the Natura 2000 network, and managed in accordance with their ecological needs. Moreover, Member States must regularly report to the EU on their conservation status. Cetaceans are also a target species of Descriptor 1 (Biodiversity) of the Marine Strategy Framework Directive (MSFD, Directive 2008/56/EC), which aims at achieving a Good Environmental Status (GES) of EU marine waters by establishing a common approach and objectives for the prevention, protection and conservation of the marine environment. Thus, information about the preferred habitats of cetacean species and the early detection of potential changes in their distribution is essential to identify needed conservation measures.

Despite the fact that the HD focuses on the conservation status of the species (i.e., the effects), and the MSFD on eliminating the causes (i.e., the threats) through mitigation measures that will restore the GES (Palialexis et al., 2019), the HD and MSFD have strong synergies. Under the MSFD, Member States are required to establish threshold values for each species through regional or sub-regional cooperation and, for species covered by the HD, these values shall be consistent with the Favourable Reference Values (FRV) established under the HD. Both HD and MSFD directives require reporting every six years equivalent parameters/criteria for the assessment of the species conservation status such as ‘Range’ (i.e., HD ‘The natural range of the species is neither being reduced nor is likely to be reduced for the foreseeable future’; MSFD D1C4 ‘the species distributional range and, where relevant, the pattern, is in line with prevailing physiographic, geographic and climatic conditions’) and ‘Habitat’ (i.e., HD ‘There is, and will probably continue to be, a sufficiently large habitat to maintain its populations on a long-term basis’; MSFD D1C5 ‘The habitat for the species has the necessary extent and condition to support the different stages in the life history of the species’). Similarly, the EO1 assessment within the Barcelona Regional Sea Convention (UNEP-MAP, EO1) is based on the Common Indicators (CI) 3 (‘Species distributional range’) and 1 (‘Habitat distributional range’). The IUCN Guidelines for the assessment of the conservation status of threatened species also foresee the assessment based on the criteria A2c (‘A decline in Area Of Occupancy-AOO, Extent Of Occurrence-EOO and/or habitat quality’) and B (‘Geographic range’). Specifically, the AOO is defined as ‘the area contained within the shortest continuous imaginary boundary that can be drawn to encompass all the known, inferred or projected sites of present occurrence of a taxon, excluding cases of vagrancy’ (IUCN, 2001), where ‘Projected sites’ are considered as the sites spatially predicted on the basis of habitat maps or models (area of potential habitat, also called Extent of Suitable Habitat, ESH). A suspected decline in the AOO could consequently be estimated based on the reduction of suitable habitat. In addition, also the Reporting Guidelines of the Habitats Directive (2017) suggest to evaluate the FRV as the AOO, or as the potential range in relation to available suitable habitat (‘Ecological potential’, the potential extent of range considering physical and ecological conditions).

Within such legal requirements, Species Distribution Modelling (SDM) is a promising approach to support the assessment of cetacean species. Indeed, as long as the amount/quality of input data is reasonably adequate, SDM can be used to support regulatory decision-making for conservation, i.e., by informing on spatial prioritisation through the identification of biodiversity hotspots, important areas for vulnerable species, or valuable habitats, overcoming the problems related to coarse or incomplete knowledge (Franklin, 2010; Maiorano et al., 2019). Time series of comparable data with sufficient statistical power, coupled with standardised SDM analyses, can help identify changes from a reference period. A significant reduction in the extent or a shift of species geographical distribution can then be related to environmental variability, habitat conditions or changes in population size, or to the effect of anthropogenic pressures. Moreover, the comparison of the suitable habitat predicted through SDM with the distributional range observed indicate potential suitable areas that are not used by the species.

However, relevant indicators or threshold values for assessing species range and habitat have not yet been developed (Palialexis et al., 2019), and some recommendations were only recently provided through an international scientific cooperation to define indicators, assessment methods, and data requirements for the assessment of marine turtles under the MSFD (Girard et al., 2022). Moreover, despite an increasing research effort, a limited number of studies attempted so far to infer temporal changes in cetacean distributional range or habitat use, and the ‘trend’ criterion for these parameters/criteria is still considered ‘unknown’ for almost all cetacean species in the Mediterranean Sea (last HD report 2013-2018), likely due to the lack of comparable data and standard methodological approaches.

The Fixed Line Transect monitoring Network (FLT Med Net) has been operating in the Mediterranean basin since 2007 collecting cetacean data along fixed trans-border transects regularly surveyed throughout the years. Using the dataset gathered across twelve years, this study aims to improve the knowledge on three low-density cetacean species of the Mediterranean basin Risso’s dolphin (Grampus griseus, Gg), long-finned pilot whale (Globicephala melas, Gm), and Cuvier’s beaked whale (Ziphius cavirostris, Zc), and evaluate potential approaches to support legislative requirements. In particular, using the dataset collected during the third HD six-years reporting cycle (2008-2012) as baseline, the study aims to assess potential changes in the range and habitat of the three species over the subsequent periods (short-term trend) testing four potential indicators: 1) Observed Distributional Range, ODR: changes in the extent of ODR detected within the area covered by monitoring effort; 2) Ecological Potential Range, EPR: change in the extent of Ecological Potential Range predicted by means of SDM; 3) Range Pattern: percentage of overlap, and shifts of ODR and EPR between the two time periods; 4) ODR vs EPR: changes in the proportion of observed distributional range vs the ecological potential range between the two periods. Overall, the study aims to test and evaluate such methodological approaches and indicators to contribute to the species assessment under the requirements of the main European nature legislative framework.

Cetacean monitoring was carried out from passenger ferries travelling along 11 trans-border transects, covering the Mediterranean Sea within the latitudes 43.6° N - 35.8° S and longitudes -5.5° E - 20.8° E, and connecting Italy, France, Spain, Greece, Tunisia and Morocco. These transects are included in the Fixed Line Transect Mediterranean Network (FLT Med Net, Arcangeli et al., 2019), and are representative of a large proportion of the Western-Mediterranean, the Adriatic Subregions, and two portion eastern and western of Ionian Sea in the Ionian-Central Mediterranean Subregion. Transects considered for the baseline period (2008-2012) covered the effort area shown in gridded grey in Figure 1 . In the second period (2013-2019) monitoring was also extended to the area in light grey along the east Spanish coasts and Gibraltar Strait on Western Mediterranean, and in the Adriatic-eastern Ionian Sea.

The monitoring activity was performed on a seasonal basis with at least three surveys per season along each sampling transect. Seasons were defined as winter (January to March), spring (April to June), summer (July to September) and autumn (October to December). Data on cetacean species were systematically collected following a standard protocol applied from large vessels (ISPRA, 2015) (FLT Net data, Supplementary Table 1 ). Ferries provided an observation point at 20−29 m above sea level and travelled at a mean speed in the range of 19−25 knots. Two experienced observers were positioned on the two sides of the command deck scanning both sides of the ship within an angle of 130° ahead in order to avoid re-counting the animals; observations were performed by naked eye and binoculars; binoculars and cameras were used to correctly identify the species and the number of animals. A dedicated GPS was used for automatically recording the survey track at the finest resolution, marking the beginning/ending points and the locations of cetacean sightings. Monitoring was carried out during daylight hours only in optimum weather conditions (≤3 on the Beaufort scale).

All the analyses performed for this study considered the sighting as the statistical unit, regardless of the number of animals within the sighted group. However, the mean group size was also examined to assess differences between the two periods. Data were analysed considering the different Mediterranean Subregions of the MSFD (https://www.eea.europa.eu/): Western Mediterranean (WMED), Ionian Sea and Central Mediterranean (Central MED), Adriatic, Aegean-Levantine Sea ( Figure 1 ). As data were homogeneously collected within the same set of conditions, detection probabilities were assumed the same across all surveys and between the two survey periods.

The area covered by the effort was the largest in the WMED Subregion, while very limited in the Central MED during the first period (i.e., eastern Sicily), and increased during the second thanks to the inclusion of new Adriatic routes covering also the Northern Hellenic Trench ( Figure 1 ). No effort was performed in the Aegean-Levantine Subregion ( Table 1 ).

Between 10 to 37% of the effort area overlapped with the species observed range (ODR) in the WMED. In the Central MED instead, 99% of the effort area overlapped with Gg ODR during the first period (i.e., in the eastern Sicily), and a limited percentage with the ODR of Zc (2%) during the second period (i.e., in the Northern Hellenic Trench). In the Adriatic, 7% of the effort area intercepted the Gg ODR in the southern part.

ODR areas were mostly located in the northern part of the WMED Subregion for all the species ( Figure 3 ) with ODR for Gg also located in the westernmost MED, the Tyrrhenian-Sardinian channel and the southern Adriatic, Gm in the westernmost MED, and Zc in the eastern Ionian (i.e., Northern Hellenic Trench). In the northern area, the ODR generally overlapped between the two periods, with a tendency to shift towards offshore in the Sardinian-Balearic basin for all the three species, and in the Ligurian Sea for Gg ( Figure 3 , left).

Considering only the common area of effort between the two periods, the trend calculated over the ODR extents revealed an expansion in all the three species with a significant delta index >10% for Gg (+16%).

Based on AUCs, validation data, and well-known sites of species presence, model outputs showed strong predictive skill at the basin wide scale. The ROC plots exhibited high average AUCs for both training and test datasets and small Standard Deviation and overfitting values for all models ( Table 2 ), which indicates consistency and reliability. In general, performance of the prediction maps of the second period was higher compared to those of the first period when validated by the independent dataset collected during the same period. Performance was also higher for prediction maps for Gm2 (over 90% of correct prediction), while performance of Gg and Zc maps was fair-good in the WMED Subregion only (over 70% of correctly predicted sites).

In general, the areas of suitable habitats highlighted by the MaxEnt output maps were consistent with previous knowledge on the species ( Figure 4 ) with the highest incongruence noted for the Gm_2 prediction in the Aegean-Levantine Subregion. Standard Deviations were generally low (<0.4), especially for the unsuitable areas. However, uncertainty was highest in general in the Aegean-Levantine Subregion and in the central and southern areas of the Central MED Subregion for the Gg_1 and Zc_2 outputs.

The ‘Minimum training presence’ threshold produced binary maps restricted to the most suitable habitat only excluding a large number of presence sights. The values identified through the ‘Equal training sensitivity and specificity’ and ‘Maximum training sensitivity plus specificity’ thresholds resulted similar ( Table 2 ), but the first approach was chosen as being more conservative and was then used to define the EPR.

Some differences in the EPRs were found between the two periods ( Table 3 ) in the WMED, where the EPR of Gg decreased by almost -7%, while Gm increased by 57% and Zc by 4%. Results for the other Subregions were not reliable as they were based on very small probability of presence in those areas (<5000 km²).

In general, Distance from Canyon, Chl-a, and depth were the most important predictors for all the three species, followed by seamount distance and SST, but only for Gm and Zc ( Table 4 ). Chl-a was the most important parameter for the definition of Gg habitats, either as percent contribution or permutation importance, in both periods, followed by canyon distance during the first period and depth during the second. Distance from Canyon was the most relevant parameter for Gm during the first period, while Chl-a strongly contributed during the second period, followed by the distance from seamounts. Chl-a and distance from canyon were the most significant parameters also for Zc during the first period, while depth and distance from seamounts were the parameters that most affected the distribution of the species during the second period.

In addition to the investigated changes in the extent of range areas, the analysis of spatial pattern revealed some shifts in the location of the main range areas. Indeed, the percentage of overlapping spanned 40-70% for ODR for the three species reaching the maximum overlap for Zc, and 30-50% for EPR. The location of overlapping areas for ODR ( Figure 3 ) and EPR ( Figure 5 ) showed the permanence over the time of some well-known areas for the three species.

In particular for Gg, some well-known areas of the WMED were predicted in both periods (e.g., Alboran Sea, Balearic Sea, Corso-Ligurian-Provençal basin, several spots in Tyrrhenian Sea including the Pontine Archipelago, and eastern Sicily). The offshore waters of the Gulf of Lion were no longer identified as the most suitable during recent years, while some new areas emerged ( Figures 4 , 5 ). A general reduction of suitable habitat was identified in the Pontine Archipelago and around the Sicilian coasts. Other widespread spots of potential suitable habitat appeared dispersed in the WMED from the recent model. Outside the more reliable area of the WMED, some suitable areas with higher uncertainty emerged in the eastern Mediterranean basin such as the southern Türkish, the northern Aegean during the more recent period and the coasts between Lebanon and Egypt.

Suitable Gm habitats were predicted in the WMED Subregion, in the Alboran Sea and along the continental shelf of Balearic, Gulf of Lion and the Corso-Ligurian-Provençal basin. A small area was highlighted in the Pontine Archipelago, and other patch areas were predicted around Sardinia Island. During the second period, new ODR areas were identified over the Alboran Sea and the Strait of Gibraltar due to the added effort in this region which intercepted the known important areas for the species identified by the large EPR. Outside the WMED, the large prediction stretching from the Aegean to Libya seems unreliable given the current knowledge on the species distribution.

Some well-known suitable areas were highlighted in both periods for Zc in the WMED such as the Alboran Sea, Ligurian Sea, northern Tyrrhenian Sea, and Balearic Sea. In the Central MED and Adriatic Subregions, the Hellenic Trench, northern Ionian Sea, and southern Adriatic Sea were predicted during the second period only with higher uncertainty.

A shift of centroids’ core areas between the two periods was detected for the ODR and the EPR predicted over the WMED Subregion ( Figure 6 ). The shift on EPR for the other Subregions or at all MED scale was not considered as based on a very limited predicted area in one or both periods ( Table 3 ).

Results showed that all the species regularly occur in almost the same areas or in a smaller proportion of their ecological potential habitat during both periods (ODR equal or smaller than EPR), with the only exception of Gg, whose ODR in the second period was larger than the EPR ( Table 5 , SM Figure 1 ). In the WMED, the proportion of suitable habitat effectively occupied by the species ranged between 62% for Gm_1 and 158% of Gg_2. No significant changes were detected in the proportion of occupied vs potential habitat over the two periods for the Zc (-1%), while for Gg and Gm increased this proportion by 59% and 46% respectively. Limited area was predicted for Gg and Zc in the Central MED, effectively occupied by the Zc by 50%, while the Gg was recorded largely outside the predicted potential area. Gm was never detected either in the surveyed areas of the Central MED or in the Adriatic Subregions. The spatial pattern of observed and predicted potential areas showed large overlap, but with some local differences ( SM Figure 1 ). Both the areas of observed and predicted range of Gg in the northern part of the WMED expanded mainly towards offshore waters and stretched in patchy suitable areas in the centre. However, the shift in ODR detected in the more recent years in the western portion of the Corso-Ligurian-Provençal basin brought Gg outside predicted suitable areas. A contraction in suitable areas was instead detected in the south Tyrrhenian, where the species was no longer present, while new areas emerged in the Sardinian channel. A suitable area was confirmed in eastern Sicily in both periods. Gm observed range was almost similar across periods in the northern WMED, except for an enlargement towards offshore waters in the Sardinia-Balearic basin, which almost corresponded with the predicted potential range despite the latter being more scattered and fragmented during the more recent years. On the other side, a relevant area potentially suitable for Gm was revealed in both periods not overlapping any ODR in the central Tyrrhenian Sea. No noteworthy changes in observed and predicted range were detected for Zc in the northern part of WMED, while a new area emerged in the Sardinian channel both for the observed and predicted range.

The sampling strategy of the FLT Med Net was set in order to homogeneously cover large portions of the Mediterranean basin, with regular monitoring of the sampled areas during all the seasons (Arcangeli et al., 2019). A recent study revealed that sampling designed along multiple fixed ferry routes detected more species and were able to recover known patterns in species richness and distribution at smaller sample sizes better than unconstrained sampling points (Boyse et al., 2023). Results of this study confirm that the sampling design of the FLT Med Net proved adequate for catching the known distribution of the species, providing high modelling performance, and allowing trends analysis even for rare or elusive cetacean species such as Risso’s dolphin, long-finned pilot whale and Cuvier’s beaked whale. This was particularly the case for the WMED Subregion, and especially during recent years when new monitored transects also covered the westernmost portion of the basin, the Alboran sea and the Strait of Gibraltar area (roughly 80% of WMED covered by the effort). In the Adriatic Subregion, the effort strategy resulted in coverage of almost the whole region although with still some uncertainty in the northernmost area, as also assessed by Zampollo et al. (2022). The Central MED was instead only represented by the effort in the eastern Sicilian coast and the Greek Ionian portion, and no effort was performed in the Aegean-Levantine Subregion, which leaves open opportunities for improvement. Indeed, an adequate proportion of the effort area intercepted the main distributional range and suitable habitats of Gg, Gm and Zc in the WMED Subregion (between 10-37% for the observed distributional range, over 46% of the predicted ecological range), and a more limited proportion in the Central MED and Adriatic Subregions, in correspondence with some known important areas for Gg (i.e., eastern Sicily. e.g., ACCOBAMS, 2021) and Zc (i.e., Northern Hellenic Trench, e.g., Frantzis et al., 2003). Therefore, in the WMED the sampling design of FLT Net proved to be adequate to intercept the ecological variability of the area, producing reliable results also outside the area of effort, whereas more transects are instead required to improve reliability in understudied Subregion (e.g., Central and Aegean-Levantine Subregions). Moreover, as the distributional range and habitat use of species varies seasonally, the seasonal based temporal resolution of sampling strategy allowed including the potential seasonal displacement of the species and thus the entire species range. The approach was also effective in terms of monitoring costs vs. acquired information, and these methods and indicators are suitable to be replicated across all seas.

Most of the Observed Distributional Range (ODR) of the species highlighted by the Kernel analysis and the Ecological Potential Range (EPR) predicted on the basis of suitable habitat modelling were consistent with previous knowledge on the species, especially for the WMED Subregion, further confirming the importance of the north-western Mediterranean for Gg, Gm and Zc (ACCOBAMS, 2021). Consistency in these areas was also found across periods, with a general enlargement in the areas of distribution, and a shift towards more offshore areas in the Sardinian-Balearic basin for the three species, and in the Ligurian Sea for Gg. Outside the WMED, some known important areas for Zc such as the Ionian Sea and the deep Hellenic Trench were predicted, even if for a limited extent, during the second period only, when monitoring effort was added in the Adriatic-eastern Ionian region. Higher uncertainties or unreliable areas were revealed, as expected, in unsurveyed areas of the Central or the Aegean-Levantine Subregion.

Findings of this study on both ODR and EPR of Risso’s dolphin (Gg) confirmed the permanence across the two investigated periods of some well-known important areas for the species in the WMED Subregion. The species is mostly found in the Western-Mediterranean Sea from the Alborán Sea, including deep offshore waters (Cañadas et al., 2002; Cañadas et al., 2005), to the south of the Provençal basin, with high values along the Algerian coast and the Balearic Islands (ACCOBAMS, 2021; Lanfredi et al., 2021). However, findings of this study no longer identified the offshore areas of the Gulf of Lion as most suitable during recent years, while highlighting new distributional areas in the offshore waters of the Sardinian-Balearic basin and Ligurian Sea. The species was considered favoured by the proximity of the continental slope, primarily in the north-western basin (Bearzi et al., 2011), with a very specialised niche and a habitat spatially restricted on the upper part of the continental slope (Praca and Gannier, 2008). A high fidelity for the Provencal continental slope, without strong seasonal pattern in abundance (Laran et al., 2010; Laran et al., 2017), and a transient use of the offshore area was also confirmed on a long-term basis between 1989-2012 by Labach et al. (2015). Nonetheless, during recent years Gg was sighted in more offshore environments than previously reported in literature (ACCOBAMS, 2021). This is also in line with the trend observed by Azzellino et al. (2016), who reported a significant decrease in Gg abundance between the early ‘90s and 2014 in coastal and continental slope areas of the Ligurian Sea, with stable occurrence in pelagic areas. The result was assumed as a loss of coastal group or a shift in animal distribution (Azzellino et al., 2016). Moreover, apart from the more defined sites, widespread spots of potential suitable habitats appeared dispersed in the WMED in the current study. A general reduction of suitable areas was also detected in the Pontine Archipelago, and around the Sicilian coasts and Ionian Sea, where only a portion of suitable habitat persisted eastern of Sicily and Taranto Gulfs where strong side fidelity was found by other studies (e.g., Monaco et al., 2016; Carlucci et al., 2020a; Cipriano et al., 2022). Relatively large groups of Risso’s dolphins were reported further east in the southern Adriatic and Ionian Seas and the deep Hellenic Trench from ASI visual surveys, but no sightings were reported from acoustic surveys (ACCOBAMS, 2021) in line with the uneven prediction produced by this study. During the first period, some suitable areas emerged in correspondence of the Türkish Mediterranean, Palestinian and Israeli coasts consistent with the few contemporary reports (Öztürk et al., 2011; Kerem et al., 2012). The absence of effort in this area prevents any conclusion on whether or not the predicted reduction reflects a true species negative trend. The few encounters of Gg in mixed-species groups with striped dolphins and short-beaked common dolphins in the deep waters of the semi-closed Gulf of Corinth (e.g., Frantzis and Herzing, 2002; Frantzis et al., 2003), and for the unique stranding record in the 2012 in the Marmara Sea (Dede et al., 2013) appear to confirm the minor prediction in these areas.

Findings of this study confirmed some of the existing knowledge on the long-finned pilot whale (Gm). The species is known to be found almost exclusively in the WMED (Verborgh et al., 2016; ACCOBAMS, 2021) with a strong preference for deep pelagic waters. Relative higher densities were reported in the Strait of Gibraltar and Alboran Sea (Cañadas et al., 2005; De Stephanis et al., 2008) and lower in Balearic and Corso-Ligurian-Provençal Seas (Raga and Pantoja, 2004; Gómez de Segura et al., 2006; Azzellino et al., 2008; Praca and Gannier, 2008). The ACCOBAMS survey of 2018 (ACCOBAMS, 2021) also observed larger groups of Gm in the Alborán Sea, along the coast of Morocco and in the Gulf of Lion, and relatively smaller pods in the Ligurian Sea. The species was never recorded in the central Tyrrhenian Sea (Arcangeli et al., 2013; Arcangeli et al., 2017), but a stable pod has been recurrently sighted in the Pontine Archipelago since 1995 (Mussi et al., 2000). In accordance with the literature, the ODR in this study for Gm was exclusive of the WMED, but with a tendency to shift towards offshore waters during recent years, especially in the Sardinian-Balearic basin. Suitable habitats were also mostly predicted in the Alboran Sea and along the continental shelf of the Balearic Archipelago, Gulf of Lion and the Corso-Ligurian-Provençal basin with a similar shifting trend towards offshore as the Observed Range. Smaller areas were predicted in the Pontine Archipelago, supporting the stable presence reported by Mussi et al. (2000), and around Sardinia Island. In the Tyrrhenian Sea instead, a relevant potentially suitable area was highlighted during both periods, although no sightings have been reported either from this study or by literature (e.g. Arcangeli et al., 2017). Further investigation could be directed to determine whether anthropogenic activities or other pressures are operating there as limiting factors for the species. During the second period, a reliable enlargement of suitable habitat was predicted in the WMED Subregion, especially over the Alboran Sea and the Strait of Gibraltar, most likely as a result of the new added monitored transects representative of the westernmost part of the basin and intercepting the Strait of Gibraltar sub-population (Verborgh and Gauffier, 2021). A large Ecological Potential area stretching from Gibraltar towards the northern African coast was indeed predicted by this study in the second period, consistent with the ACCOBAMS (2021) sightings of large pods and by some reported strandings in Morocco (Bayed, 1996; Masski and De Stephanis, 2018), Algeria (Boutiba, 1994; Bouslah, 2012) and Northern Tunisia (Attia El Hili et al., 2010; Karaa et al., 2012). The species was never detected either in the Central MED and in the Adriatic Subregions, and no EPR was predicted here, while the large prediction stretching from the Aegean to Libya seems unreliable given the current knowledge on the species distribution.

Known habitats of Cuvier’s beaked whale (Zc) were highlighted by the study in the WMED Subregion, while the south Adriatic and Hellenic Trench of the eastern Ionian Sea were only predicted during the second period likely due to the effort performed in those areas that allowed including some environmental features not considered by the environmental variability of the WMED effort area only. Zc is considered to inhabit both the western and eastern basins of the Mediterranean Sea (Podestà et al., 2016), and this species is mostly found in canyon areas in the Ionian Sea, the Hellenic Trench, the deep southern Adriatic Sea (Frantzis et al., 2003; Carlucci et al., 2020b), the central Tyrrhenian Sea (Gannier, 2015; Arcangeli et al., 2016), the Balearic and the Alborán Seas (Cañadas and Vázquez, 2014; Cañadas et al., 2018), and the Ligurian Sea (Moulins et al., 2007; Azzellino et al., 2008; Tepsich et al., 2014). The ACCOBAMS survey of 2018 confirmed the existing knowledge on the basin wide presence of the species and at the same time showed how Zc occur in relatively small patches at low densities (ACCOBAMS, 2021). In accordance with literature, this study highlighted the importance in particular in the WMED of the Alboran Sea, the central Tyrrhenian Sea and Ligurian Sea and also a permanent area of suitable habitat in correspondence with the Spanish-French continental slope coast and stretching offshore. However, despite being recognised by some studies (Raga and Pantoja, 2004; Gannier and Epinat, 2008; Praca and Gannier, 2008; Podestà et al., 2016; Arcangeli et al., 2017) and the records of the Accobams survey (ACCOBAMS, 2021), this latter area was not considered among the important areas for the species. This discrepancy could indicate either an underrepresentation of scientific literature or a minor occupancy of Ecological Potential habitat for the species.

In general, the persistence over time of presence and suitable habitat of Gg, Gm and Zc in the WMED confirmed the importance of this Subregion for the species. However, the changes in the extent (whichever a contraction or expansion) and the shift highlighted on both the observed distribution and the suitable areas indicate changes in spatial distribution of the species across time periods ( Table 6 ). This could be the result of exploitation of new potential suitable areas or an adaptation forced by existing pressures or changes in the distribution of habitat over time. In particular Gg enlarged the proportion of occupied area over the ecological potential by almost 50% distributing also outside the predicted suitable areas (i.e., in the Corso-Ligurian-Provençal basin). In addition, the new areas that emerged in the centre of the Sardinian Balearic basin or eastern Corsica coast, together with the contraction of the areas in the south Tyrrhenian Sea and around the Sicilian coasts, revealed changes that need further investigation. Moreover, results highlight a concurrent enlargement of the area of distribution of Gm and Zc, even if for a minor extent, that is not yet reported by other studies. If confirmed, this would be a signal of a general tendency towards a more dispersed distribution that surely deserves attention.

The indicators here tested helped to describe the main consistencies or changes in short-term range trends between periods. Results highlighted the advantages and weaknesses of each indicator and of the approach tested.

The Observed Distributional Range (ODR) indicator has the advantage of preventing difference biases by data processing, analysis settings or approximations and is closely related to the real observed distribution of the species. On the other hand, results are only representative of the area where the effort is performed, introducing the need for specific planning of the sampling design of the data collection if used as representation of species distributional range. Spatially extensive surveys covering the whole range of species would deliver an adequate baseline for detecting ODR, but they are cost-expensive and may lack the temporal resolution needed to detect the natural species variability avoiding output linked to occasional or seasonal fluctuations. Continuous local scale surveys could provide long-term series but lose the spatial representativeness. Local and large scale surveys could be merged to increase the spatial representation of outputs providing that appropriate metric is used to match data collected with different methodologies. Time extensive large-scale monitoring data collected in sampled areas spatially representative of regional ecological conditions could represent a suitable balance and can be used as an index of the real species range. A prior assessment of the ecological variability representativeness of monitored transects is needed to avoid bias in underrepresented regions.

With regard to the methods to represent the distributional range, if compared to the species occurrence mapped in a 10 x10 km² grid as suggested by HD and MSFD, the Kernel density smoother proved to be a feasible tool to spatially generalize the distribution of species and define the area where the species is found. It is adaptable to the spatial scale (grain) and resolution of data through the adjustment of search radius and cell size resolution while still remaining relatively simple to apply. Moreover, when using high quality spatial data as those of this study, the use of KDE could be considered as more accurate than other coarser methods such as grid of occurrence or the Minimum Convex Polygon used by some EU Member States. Other approaches such as the Kriging could also apply to the same purpose and are worth exploring.

Finally, care must be taken when calculating the trend in the extent of ODR in cases when the monitored area changed between time periods. In this study, the trend was calculated as percentage of change of ODR vs Extent of Effort (i.e., it was normalised by the Effort), and the percentage of change didn’t vary if considering the entire effort areas for each period or the common area only. However, the second approach was chosen as more conservative. Indeed, a change in the investigated area could produce a bias if, for example, an area completely outside, or, vice versa, in the core of the species range, is surveyed during one period only. Given the long-term monitoring required by the legislative framework at the large-range spatial scale needed for cetacean species, changes in the monitored areas over time could occur for example in the case of new organisations or countries joining an international effort. This aspect should be carefully considered, and the trend detected should be investigated with a conservative approach within the common effort area only.

The Ecological Predicted Range (EPR) based on sites of known occurrences and extrapolated through habitat maps models proved to be able to generalize the spatial distribution of the species also outside the area of effort providing meaningful outputs especially in the WMED Subregion where sampling was spatially representative of regional ecological conditions. Results of this study further confirm that sampling effort must be designed in order to assure representativeness of the regional ecological variability, and the SDM outputs in not surveyed regions (e.g., as in the case of the Aegean-Levantine basin in this study) should be taken with caution. In addition, predictions and extrapolations should be validated whenever possible by independent datasets as soon as new data become available. Results of this study indicate a general correspondence of trends detected in the Observed and Predicted Range both in terms of shifts (e.g., towards offshore areas in the Western-Mediterranean Subregion for all the species) and extent of areas (e.g., enlargement recorded for Gm in both ODR and EPR). These results confirm the potential for using the EPR to indirectly determine the AOO as suggested by the IUCN Guidelines (IUCN, 2001). However, some differences were also detected such as the new areas detected by the ODR in the Sardinia channel for Gg that were not predicted by the EPR in the corresponding period. Thus, careful consideration is needed to correctly discriminate the meaning of the range predicted on the basis of SDM to investigate the species conservation status, as the Potential Range does not always correspond to the actual distributional range of the species. Output must be carefully validated and adjusted using the estimated proportion of ODR/EPR as suggested by IUCN (2001).

On the other hand, Suitable Habitat Maps can be directly used to define the extent, trend and pattern of the suitable habitats to answer the parameter/criteria ‘Habitat’ for the species (e.g., for HD and MSFD). By including information on the main ecological factors that drive their distribution, these models can also be used to investigate the “Habitat conditions” requirement if the pressures are added to the models.

Provided SDMs accurately reflect potential ranges, EPR can also be used to compare the Observed versus the Potential Range (IUCN, 2001; IUCN, 2022) as they indicate the area of occupied habitat and describe unoccupied habitats of suitable quality allowing the long-term survival of the species (DG ENV, 2017). If appropriate data are available, the comparison between the Observed and the Potential Range can also help to identify potential suitable areas that are not used by the species due to the influence of anthropogenic pressures or other limiting factors. Alternatively, EPR can also be used to determine if the species is pushed outside of the preferred suitable habitat as a consequence of a pressure, change in the distribution of habitat or the exploitation of new resources. Trend in the ratio between Observed vs Potential range could then be used to correlate the detected changes with other environmental or anthropogenic parameters and/or assess the effectiveness of mitigation measures.