Introduction

Coastal ecosystems host complex physical and biological processes and are characterized by an enormous richness in species and biodiversity1. At the same time, they face significant threats from human activities2. The cumulative impacts of human activities and climate change often lead to the degradation of these ecosystems3. Over half of the world’s coastal ecosystems can be considered threatened by human pressure4.

The Mediterranean Sea is among the regions most exposed to human pressure, which increasingly threatens its biodiversity5,6. Despite its relatively small size, the Mediterranean Sea hosts several habitats, high levels of endemism, and overall marine biodiversity, earning it the designation of a “biodiversity hotspot”7,8. It is characterized by important ecological benthic biocenoses such as coralligenous bioconstructions and ecosystems such as seagrass meadows (e.g., Posidonia oceanica meadows), recognized as worthy of protection by the European Union (EU) Habitat Directive (Directive 92/43/EEC). P. oceanica meadows are highly susceptible to human disturbance (e.g., direct physical damage and water quality deterioration)9, and extensive regression of P. oceanica seagrass beds have been recorded throughout the Mediterranean Sea10,11. Coralligenous bioconstructions are negatively affected by many issues such as global warming [e.g.,12], ocean acidification [e.g.,13], and human-induced nutrient enrichments in seawater [e.g.,14].

To accurately evaluate the ecological status of underwater habitats and predict the effects of human disturbances, it’s crucial to establish sustainable and cost-effective observation systems. These systems need to incorporate affordable technologies that allow easy and extensive data collection15. Unfortunately, the availability of user-friendly, affordable marine monitoring instruments is currently a constraint16. Furthermore, effective observation systems must include dedicated measurement platforms with sufficient spatial and temporal resolution to address the varied requirements of monitoring efforts17.

Worldwide, studying coastal benthic habitats typically involves the help of scientific operators and the use of remote sensing technologies such as echosounders (single beam, multi beam, and side scan sonar) [e.g.,18,19,20,21,22] and Remote Operated Vehicles (ROVs) [e.g.,23,24,25]. Acoustic backscatter data are by far the most widely used type of remotely sensed information for habitat characterization and mapping, ranging from simple acoustic survey systems employing single-beam transducers to high-resolution multibeam echosounder (MBES) surveys26. Among the available methodologies, integrated multiplatform survey approaches (e.g., MBES surveys combined with ROV dives) provide essential high-resolution data for producing accurate habitat maps in shallow coastal waters27. Building on these advances, methodological developments focus also on integrating high-resolution data acquisition technologies with machine learning algorithms to automate analytical processes, thereby enhancing both accuracy and cost-efficiency [e.g.,28].

Recent technological innovations, such as Unmanned Surface Vehicles (USVs), offer advantages over traditional oceanographic platforms, including enhanced maneuverability, autonomy, cost-effectiveness, and endurance29,30. These autonomous technologies equipped with various sensors contribute to characterizing the marine environment, reducing survey times in large areas, and proved valuable in gathering detailed information about coastal seabed, especially in heavily anthropized coastal areas with multiple sources of disturbance31.

In this context, the development of new monitoring methodologies involving the use of innovative autonomous platforms allows rapid data collection facilitating assessments in hard-to-reach sites (e.g., Marine Protected Areas - MPAs, extremely shallow waters, areas forbidden to ordinary navigation)32,33,34. This also increases the amount of acquired data that needs to be managed.

The collection of large and diverse datasets presents a significant challenge for data integration, highlighting the importance of developing and applying ecological quality indices. Various indices have been designed to assess the condition of different benthic habitats, including seagrass meadows [e.g.,35,36,37], soft-bottom substrates [e.g.,38,39], and coralligenous bioconstructions [e.g.,23[,40–43]. For Posidonia oceanica meadows, commonly used indices include the POMI (Posidonia oceanica Multivariate Index)35, which employs multivariate analysis of physiological, morphological, structural, and community-level descriptors, and the PREI (Posidonia oceanica Rapid Easy Index)36, which is based on five key structural terms. Soft-bottom habitats are typically assessed using indices that integrate pollutant levels with benthic species composition35,36. For coralligenous bioconstructions, widely adopted indices include the ESCA (Ecological Status of Coralligenous Assemblages)42 and the MAES (Mesophotic Assemblages Ecological Status Index)23, both of which rely on non-destructive photographic surveys and incorporate parameters such as taxa abundance, species richness, and assemblage heterogeneity.

Despite significant progress in marine survey technologies and synthetic indices formulation for habitat quality assessment, integrating these procedures for a comprehensive habitat status evaluation remains a fundamental objective to achieve the marine environment management actions required by the legislation.

The European Union committed Member States to assess the environmental status of their territorial waters under the Marine Strategy Framework Directive (MSFD 2008/56/EC), as well as to implement an integrated management of the use of the seas through the Maritime spatial planning (MSP 2014/89/EU). At Mediterranean scale the Contracting Parties of the UN Barcelona Convention have agreed on the Mediterranean Action Plan (UNEP MAP) towards an “Ecosystem Approach” to assess quality status and measures on marine waters and environment.

In this study, we evaluated the effectiveness of an integrated remote sensing approach that combines cost-effective technologies with multimetric indices to assess the ecological status of benthic habitats in three coastal areas in Italy (Capo Linaro – CL; Macchiatonda – MT; Tor Paterno Marine Protected Area – TP MPA) located in the northeastern Tyrrhenian Sea. Surveys conducted using an echosounder-equipped Unmanned Surface Vehicle (USV) provided information on habitat extent and substrate characteristics. Simultaneously, video footage from a Remote Operated Vehicle (ROV) was used for ground-truthing and assessing habitat quality. The data obtained from both technologies were essential for applying multimetric indices.

Results

The ROV video acquisitions in the studied sites revealed the presence of eight taxa: plants, algae, sponges, hydrozoans, anthozoans, polychaetes, ascidians, and bryozoans.

The results of USV surveys and representative bottom types in ROV photos are presented for each study site (Figs. 1 and 2, and 3). Table 1 displays score classes and Reference Conditions (RC) for each metric used, while Table 2 summarizes the metrics’ values for Index-A and Index-B in each study area, obtained by the integration of ROV photo analysis and USV data. Tables 3 and 4 document the final computed scores and the seabed habitat quality assessment in CL, MT, and TP areas.

Fig. 1
figure 1

CL area bottom elevation and bottom type classification (panel A), visualized on a Google Maps basemap (image created using Visual Aquatic software, BioSonics Inc., https://www.biosonicsinc.com/products/software/); ROV photos of the identified bottom types (panel B).

Full size image
Fig. 2
figure 2

MT area bottom elevation and bottom type classification (panel A), visualized on a Google Maps basemap (image created using Visual Aquatic software, BioSonics Inc., https://www.biosonicsinc.com/products/software/); ROV photos of the identified bottom types (panel B).

Full size image
Fig. 3
figure 3

TP area bottom elevation and bottom type classification (panel A), and ROV photo of the identified bottom types (panel B). Elaborations were made using Surfer 8 (Golden Software Inc., USA; https://www.goldensoftware.com/products/surfer/).

Full size image
Table 1 Scores metrics used in Index-A and Index-B and reference conditions; *Maximum P. oceanica shoot density was chosen as RC.
Full size table
Table 2 Summary of the metrics’ values for each surveyed site, accounting for all the analyzed photos. 
Full size table
Table 3 Score ranges used to define habitat quality according to Index-A and Index-B. 
Full size table
Table 4 Final Index-A and Index-B scores for each study site.
Full size table

Although the Index-A metric, as outlined in the materials and methods section, included EZ and HZ metrics, the presence of megazoobenthos organisms was only detected in videos recorded in the TP area (Table 2).

The coast-to-offshore USV route in the CL area revealed a maximum depth of 13.95 m (highlighted in blue, Fig. 1A) approximately 1000 m from the coast. The site (Fig. 1) exhibited vegetated rocky bottoms (highlighted in blue, Fig. 1A) and the presence of Posidonia oceanica (highlighted in green), forming extensive meadows in this area. Specifically, rocky bottoms accounted for 64.02%, and P. oceanica meadows for 35.98% (Fig. 1A). The echosounder data indicated that the P. oceanica canopy reached a maximum height of 1.2 m with an average percentage coverage of 40.26%. Echosounder data were consistent with observations by scientific divers in control sites (blue dots, Fig. 1A). In addition to P. oceanica, among the most encountered marine species were the red alga Jania rubens, and the brown alga Padina pavonica. These findings indicate the presence of a mosaic of P. oceanica and photophilous algae on hard bottoms biocenoses in the CL site. Figure 1B shows two representative bottom type ROV photos in the CL area. Overall, the seabed habitat quality in the CL area was classified as moderate using both Index-A and Index-B (Table 4).

Within the MT area, autonomous echosounder surveys via USV showed depths reaching 25 m (Fig. 2A) approximately 3000 m from the coast. The MT area was highly heterogeneous (Fig. 2A), characterized by rocky bottoms (31.69%), soft bottoms (7.12%), P. oceanica (32.55%), and coralligenous bioconstruction (28.64%). In the MT area, P. oceanica exhibited a patchy distribution, a canopy with a maximum height of 0.61 m, and an average percentage coverage of 39.09%, characteristics confirmed by scientific diver observations (blue dots, Fig. 4). In addition to P. oceanica, among the most encountered marine species in the MT area were the red algae of the genus Asparagopsis, and the brown alga Padina pavonica. The species presence and distribution in the MT site indicate the presence of a mosaic of P. oceanica, photophilous algae on hard bottoms biocenoses as well as coralligenous habitat in an enclave in the infralittoral, and infralittoral soft sediment. The MT area showed moderate quality using both Index-A and Index-B (Table 4).

Fig. 4
figure 4

Study sites, USV survey lines, ROV data acquisition areas, and scientific dives. The map was created using Surfer 8 (Golden Software Inc., USA; https://www.goldensoftware.com/products/surfer/) and is expressed in geographical coordinates (WGS84). The USV survey lines were represented on a satellite image using the Google Earth Pro software (7.3 version, https://www.google.com/intl/it/earth/about/versions/).

Full size image

The echosounder survey of the TP MPA focused on the shallower reserve area, reaching a minimum depth of 19.64 m (Fig. 3). This area was characterized by both rocky and soft bottoms, seagrasses, and coralligenous bioconstructions. In comparison to the CL and MT sites, TP exhibited higher biodiversity. The site was highly heterogeneous (Fig. 3A), characterized by rocks, hosting coralligenous bioconstructions (highlighted in light blue, Fig. 3A) and patches of P. oceanica (39.31%), as well as soft sediment (light yellow; 60.69%). Validation of vegetation height and coverage data in the TP area was not possible due to the absence of in-situ scuba observations (Fig. 1). In addition to P. oceanica, macro benthic species in the TP area included the gorgonian species Paramuricea clavata, Eunicella cavolini, and Eunicella singularis, as well as the red algae of the genus Lithophyllum. The species presence and distribution in the TP site indicate the presence of a mosaic of P. oceanica and coralligenous habitat in an enclave in the infralittoral and infralittoral soft sediment. Concerning the quality assessment, the habitat quality was moderate (Table 4).

Discussion

Assessing the quality of benthic habitats in coastal environments is a challenging task. It requires data on both the biotic and abiotic characteristics of the study site. Gathering this data often involves time- and resource-intensive methods (e.g., substrate sampling and analysis of the chemical and physical characterization of the water column). The challenge becomes even greater in areas with highly heterogeneous seabeds, where the morphological and hydrodynamic conditions create a mosaic of different habitats that can shift quickly or overlap within relatively small areas, as seen in the CL, MT, and TP regions. In these sites, applying habitat quality indices is also difficult. These indices are typically designed to assess the quality of specific biocoenoses or habitats (e.g., coralligenous bioconstructions)40, making them less useful for an overall assessment of highly heterogeneous environments. The integrated methodology proposed in this work and applied at the CL, MT, and TP sites aimed to address these challenges by giving a first benthic habitat quality assessment.

Investigations via USV equipped with echosounder in the CL, MT, and TP areas provided important information regarding the type of coastal benthic habitats. Given the sampling plan’s design to offer a “snapshot” of the habitats, it’s important to note that the absence of a particular species along the USV route does not necessarily imply its total absence in the area. The USV and ROV surveys enabled the detection of characteristic species and the attribution of observations to specific benthic biocoenoses. However, further investigation could be useful to better characterize the benthic communities, due to the patchy distribution of the different biocoenoses within the study sites. The number of taxa as diversity proxy used in the calculation of multimetric indices23 is derived from ROV videos, but the taxa in the areas may encompass a greater number of species. Consequently, the results obtained, in terms of species richness, may be subject to underestimation and dependence on the surveyed area’s extent and absence of replicates. It should be considered that the index results may vary due to seasonal fluctuations of certain components (e.g., SHP, SCEP, and SEN). Therefore, surveys conducted in different seasons would likely yield variations in the calculated values. These limitations could be addressed through extensive, regular, and continuous monitoring, which would allow a more accurate assessment of the evolution of benthic biocoenoses.

The application of the multimetric indices revealed that all three areas exhibited a moderate habitat quality using both Index-A and Index-B. However, looking at the Index-A (Table 4) value, which considers more parameters than its reduced form Index-B, the TP area tended towards the ‘good’ class compared to the CL and MT sites. The validity and sensitivity of the original MAES and qMAES indices were proved by23. The updated versions of Index-A and Index-B have been developed for broader applications in heterogeneous seabeds, including coastal marine areas characterized by photic environments and various environmental impacts. To date, no additional studies applying multimetric indices for a comprehensive assessment of habitat quality in the study sites are available in international literature. Given that the outcomes of Index-A and Index-B are considered representative of the quality of individual habitat types, we compared the results obtained at the CL, MT, and TP sites with those derived from habitat-specific indices applied to Posidonia oceanica meadows and coralligenous bioconstructions. For Posidonia oceanica meadows, the moderate quality classification is consistent with the findings of the MSFD Summary Report44 for the Latium region. Regarding coralligenous bioconstructions, Ranaldi45 reports high habitat quality for the TP seabeds based on the ESCA42 and ISLA43 indices, and good quality according to the COARSE index41. Moreover, the qualitative results from Index-A and Index-B align with those reported by3,6, who also classified the habitats in the same areas as being in moderate condition, based on cumulative assessments of anthropogenic pressure.

Our attempt to apply multimetric indexes in highly heterogeneous sites suggests the importance of developing habitat quality assessment indexes that are both practical and cost-efficient, while still providing precise results. These indexes should consider the inherent characteristics of each site (e.g., hydrodynamic conditions) and the impacts of human activities [e.g.,43].

Overall, our results highlighted a substantial difference between TP and the other two study sites located further north. The TP site showed a higher group biodiversity (T value of 8; Table 2). The high level of biodiversity observed may be attributed to several factors. Firstly, this area was investigated in greater spatial detail than the CL and MT sites, with a higher number of USV survey lines, which may have influenced the results. Secondly, the TP site may be subject to different hydrodynamic conditions (e.g., river inputs, wave action) compared to CL and MT, as it is also located farther offshore. Another important factor is the variation in environmental protection regimes among the three sites: TP is a Marine Protected Area (MPA), established in 2000 (Ministerial Decree of 29 November 2000), whereas CL and MT were designated as Special Areas of Conservation (SACs) in 2017 and 2016, respectively (Ministerial Decrees of 2 August 2017 and 6 December 2016). The potential role of MPAs in maintaining the ecological integrity of coastal areas, therefore, cannot be ruled out46,47. However, these hypotheses require further investigation to be confirmed.

In conclusion, coastal areas host habitats of inestimable value which are heavily impacted by human activities and climate change. To tackle these issues and improve our knowledge of the resistance and resilience of impacted benthic biocenoses, appropriate monitoring programs, conservation, and restoration efforts are needed48.

In this study, we assessed the effectiveness of an integrated monitoring approach that combines cost-effective technologies with multimetric indices, evaluating the ecological status of benthic habitats in shallow waters.

The results highlighted the reliability of the proposed methodology and its adaptability for widespread application in coastal regions worldwide, especially those with highly heterogeneous seabeds. Additionally, the method shows strong potential for broader use in coastal monitoring, supporting conservation efforts such as the MSFD (2008/56/EC), MSP (2014/89/EU), and other action plans like UNEP MAP. With further refinement, this approach could become a valuable tool for the global assessment and management of coastal and marine ecosystems. Future research will apply this methodology to more coastal areas to test its robustness.

Comprehensive habitat assessments should integrate terrestrial and marine factors, considering anthropogenic pressures both locally and in surrounding environments49. Despite data harmonization challenges, adding indicators like land use, development, sewage discharge, sedimentation, and water quality to multimetric indices will improve assessment accuracy.

The techniques employed in this study could enhance the spatial resolution of data concerning both biotic and abiotic characteristics of submerged habitats. Consequently, the methodology can also contribute to a deeper understanding of current biodiversity status and its responses to environmental changes—critical information for improving biodiversity management50.

Lastly, the potential of this approach to support and enhance the Digital Twin Ocean (DTO) will be evaluated. The deployment of autonomous and remotely operated technologies at broader spatial and temporal scales offers significant advantages, including the generation of high-resolution, near real-time data on habitat quality. This would improve both spatial and temporal coverage, enabling cost-effective and time-efficient monitoring of benthic habitats, even in areas just a few meters from the shoreline, which are typically difficult to investigate using standard monitoring platforms. The data generated could also be incorporated into numerical models, aiding in the evaluation of seagrass meadows as effective nature-based solutions.

Methods

Study sites, USV and ROV data acquisition, and analysis

The assessment of habitat extent and quality status was conducted in three coastal marine areas of the Tyrrhenian Sea, each with varying levels of conservation (Fig. 4). Proceeding from northwest to southeast, the areas investigated in this study are Capo Linaro (CL), Macchiatonda (MT), and Secche di Tor Paterno Marine Protected Area (TP) (shaded rectangles in Fig. 1). In detail:

  • CL is located approximately 79 km northwest of Rome and is part of the Special Area of Conservation (SAC) “Seabed between Punta del Pecoraro and Capo Linaro” (Habitats Directive 92/43/EEC, code IT6000006). It overlooks a rocky shoreline.

  • MT is part of the “Secche di Macchiatonda” SAC (Habitats Directive 92/43/EEC, code IT6000008) and is situated around 60 km northwest of Rome, within the Capo Linaro - Capo Anzio physiographic unit. It overlooks a sandy beach.

  • The TP area, in addition to being in SAC, is also a Marine Protected Area (MPA). TP is located approximately 40 km south of Rome and designated under the Habitats Directive 92/43/EEC with the code IT6000010 (https://www.ampsecchetorpaterno.it/). TP is located about 5 nautical miles from the sandy beach.

The three study areas host the 1120* (Posidonia oceanica meadows; Habitats Directive 92/43/EEC) and 1170 (Reefs; Habitats Directive 92/43/EEC) habitats of community importance51.

The USV surveys were conducted along coast-to-offshore lines (the sampling plans of the three sites are reported on the right side of Fig. 1) between 2022 and 2023 (CL – September 2022, MT – April 2023: TP – January 2023). The USV line at the CL site was 923.2 m long, while at the MT site it measured 2850 m. At the TP site, the lengths of the USV lines were 1050 m (A), 1088 m (B), 1071 m (C), 1025 m (D), and 1019 m (E) (Fig. 4). A USV (EchoBoat, Seafloor Systems Inc., Shingle Springs, CA, USA) equipped with a MX Aquatic Habitat Echosounder (single beam, 204.8 kHz, BioSonics Inc., Seattle, WA, USA) was used. The USV operated in autonomous mode, navigating at a speed of 0.75 m/s, collecting data on bottom depth, the presence of hard and soft bottom, submerged aquatic vegetation height, and coverage52,53,54. BioSonics Visual Acquisition and Visual Aquatic software were used for echosounder data acquisition and processing, respectively (https://www.biosonicsinc.com/products/software/). Surfer 8 (Golden Software Inc., USA; https://www.goldensoftware.com/products/surfer/) was used for data interpolation and visualization of bottom depth (m), bottom type, submerged vegetation coverage (%), and height (m). For the data analysis at the TP site, the kriging interpolation method was applied.

Validation of echosounder data (bottom type and submerged vegetation height) was achieved through ground-truth observations using ROV (Bluerobotics BlueROV2; Bluerobotics Inc., Torrance, CA, USA) and scientific divers. ROV dives were conducted in the red areas indicated in Fig. 1. To ensure high-quality video footage, the ROV moved along linear routes (100–150 m), recording continuously at a slow speed (< 0.3 m/s) and a constant height from the bottom (< 1 m) with a visual field of about 2 m. The ROV was equipped with an underwater acoustic tracking position system, depth and temperature sensors, and a compass.

To discriminate P. oceanica distribution from the rest of the submerged vegetation, a plant detection threshold of 0.2 m was applied during the analysis of the echosounder data53, considering the minimum height of P. oceanica leaves in the area during the winter season, based on ROV videos and images, and scientific divers’ reports.

Application of ecological quality indices

An in-depth analysis of ROV videos was conducted in the CL, MT, and TP areas to assess the quality of seabed habitats. To ensure an objective evaluation, given the significant heterogeneity among the three areas, we considered synthetic indices described by23, specifically the Mesophotic Assemblages Ecological Status Index (MAES) and its abbreviated version, quick MAES (q-MAES). Originally developed to assess the quality of hard bottom megabenthic mesophotic aggregates, these indices considered the parameters outlined in Eqs. 1 and 2, respectively:

$$MAES: S_{T} + S_{CB} + S_{E} + S_{H} + S_{EN} + S_{L}$$
(1)
$$q-MAES: S_{T} + S_{CB} + S_{CE} + S_{L}$$
(2)

Here ST denoted the score of the number of megabenthic taxa, SCB signified the score of the percentage of biotic cover in the basal layer, SE represented the score of the density of erect megabenthic species, SH was the score of the average height of the most abundant erect species, SEN was the percentage of colonies with epibiosis/necrosis, SL denoted the marine litter density, and SCE was the percentage cover score of erect species.

Each score was determined based on the analysis of ROV videos, defining a Reference Conditions (RC) value representing the best value detected. RC values were chosen considering the “best” values of each metric noted23,35,55. RC values were then used to calculate three Metric Value (MV) ranges and assign a score (score 1, 2, or 3): MV < 50% of RC; 50% < MV < 80% of RC; MV > 80% of RC, as previously described in detail by23. Finally, results were expressed in three quality classes (bad, moderate, and good)23,41.

Multimetric indices were applied by considering ten random photographs from each video acquisition (three videos in each ROV site highlighted with the red dots, Fig. 1), extracted from the entire video transect using the VLC program (https://www.videolan.org/vlc/).

The MAES and q-MAES indices were applied with modifications to suit the investigated habitats and make them applicable in the photic environment. Data related to vegetation height and coverage obtained from USV surveys were also used for the application of the indices. The revised MAES (Index-A) and q-MAES (Index-B) for our study sites are outlined in Eqs. 3 and 4, respectively:

$$Index-A: S_{T} + S_{CB} + S_{EZ} + S_{HZ} + S_{EP} + S_{HP} + S_{EN} + S_{L}$$
(3)
$$Index-B: S_{T} + S_{CB} + S_{CEZ} + S_{CEP} + S_{L}$$
(4)

Here ST parameter represented the score of the maximum number of flora and fauna megabenthic taxa (from ROV video analysis), SEZ and SHZ represented the scores of the density (number of organisms in each analyzed ROV photo) and average height (from ROV video analysis) of erect megazoobenthic species, SEP and SHP were the scores of the density (from scientific dives and ROV video analysis) and average height (from echosounder-equipped USV surveys) of erect megaphytobenthic species, SCEZ and SCEP were the percentage cover score of megazoobenthic (from ROV video analysis) and megaphytobenthic erect species (from echosounder-equipped USV surveys). The SEN parameter has been updated to indicate the percentage of epibiosis/necrosis on both flora and fauna. Notably, SHZ values and SEP values for TP were not available, requiring consideration of underestimated results in these cases.