Ria de Aveiro is a shallow coastal lagoon that comprises the estuary of the Vouga River (3500 km² of catchment area), having a single artificial connection to the Atlantic Ocean (1.3 km length, 350 m wide, and 20 m depth). The lagoon has semi-diurnal tides with a tidal range from 0.6 m (neap tides) to 3.2 m (spring tides); tidal circulation and wind are the main hydrological forcing functions. The tidal phase lag is in the order of 6 hours in the upper reaches of the channels; whilst the water residence time in the lagoon varies from less than 2 days near the ocean boundary, to more than 1 week in the upstream channels (Dias et al., 2000). Under the EU Water Framework Directive (WFD), Ria de Aveiro includes five transitional water bodies (WB): WB1 (Mira), WB2 (São Jacinto-Espinheiro), WB3 (Ílhavo), WB4 (Murtosa), WB5 (Ovar). Except for WB2, which is heavily modified, the others are natural water bodies. In addition, there are three coastal water bodies (CWB) (CWB-II-1B; CWB-I-2; CWB-II-2) here grouped as WB7. These, together with the ‘Bocage’ landscape that is characterised by small agriculture fields surrounded by living hedges and a network of water channels, are defined as the seven landscape units in this study (Figure 1). The management boundaries of the Ria de Aveiro Natura 2000 were set within 500 m from the margin, as specified for the river basin and estuary management plans under the EU Water Framework Directive. Both Plans aim to establish priorities, rules, and measures for the integrated management of water resources and associated ES (Lillebø et al., 2015; Lillebø et al., 2020).

The Ria de Aveiro comprises a variety of biotopes, including natural to highly modified habitats, such as large areas of intertidal sand and mud flats, seagrass meadows, salt marshes, open fields, forests, freshwater lakes, and urban areas. It also comprises the ‘Bocage’ landscape, a human-shaped habitat with water channels and treelined riparian corridors divided by living hedges for agriculture and livestock production. ‘Bocage’ landscape is located at the Baixo Vouga Lagunar (3000 ha), where the Vouga River meets the lagoon water body. The selected Ria de Aveiro Natura 2000 coastal area is classified as a Special Protection Area of ca. 515 km² from which ca. 40% are considered marine areas and is classified as a Site of Community Importance. This recognises its importance to the conservation status of the natural habitats of the Mediterranean and, to a minor extent, Atlantic biogeographical regions to which Ria de Aveiro belongs. This coastal area offers a set of ES with significant associated socio-economic activities of high economic value. Relevant examples are agriculture and livestock, industry, traditional activities (e.g., salt production and bait digging), tourism and recreation (e.g., sailing and surf), maritime port activity, aquaculture, and commercial fishing, where the Aveiro region plays an important role in the Portuguese fisheries market (Lillebø et al., 2015; O’Higgins et al., 2019).

The InVEST HRA tool, apart from assessing the cumulative risk posed to habitats by human activities, enabled the exploration of where these pressures might compromise biodiversity and ES. To obtain data on ES availability, we used the values reported by Lillebø et al. (2019). Coastal-marine services included the biotic and abiotic outputs from ecosystems following the Common International Classification of Ecosystem Services (CICES V5.1 in Haines-Young and Potschin, 2017) and were aggregated as a list of 10 ES types for stakeholders’ elicitation purposes. ES were prioritized by stakeholders’ elicitation. First, a questionnaire was prepared as an online Google form to be filled in by each participant anonymously to make pairwise comparisons of the ES to derive a ranking of criteria for the different stakeholder groups. The scaling method for ranking ES preferences was based on a Likert-type scale, using a five levels bidirectional ordinal scale, with an equivalent number of negative and positive statements: much less important (1/4); less important (1/2); equally important (1); more important (2); and much more important (4). Each participant should qualitatively rank the importance of each ES against the others (details on the stakeholders’ elicitation process in Martínez-López et al., 2019). ES values by stakeholders were published for the provisioning, regulation and maintenance, and cultural ES divisions (Supplementary Table 2).

The vulnerability score for each cell in the grid is obtained by multiplying cumulative risk by availability for each cell in the grid:

where V is the vulnerability, R is the cumulative risk score measured by the HRA model (see Equation 2), and A the expert judgement score on availability. The vulnerability score implied that the highest vulnerability was obtained when a habitat with a high potential to deliver ES was subject to higher levels of risk. Thus, the higher the vulnerability score for a given grid cell, the greater the potential loss of ES in that area (Willaert et al., 2019). We calculated cumulative vulnerability and individual (V1 – V11) for each of the 10 ES groupings.

The risk assessment classified the 50 m² results for each habitat as either low, medium, or high based on the cumulative risk level (Figure 4). Habitat areas at the highest risk level (relative to their total area) were seagrasses (25.4%) and beaches and sand flats (25%), followed by mud flats (23%), ‘Bocage’ (17%) and salt marshes (7%). Four of the five habitats showed between 50-60% of their area at medium risk, whereas beaches and sand flats were only 32%. The habitats with the highest proportion of area in the lowest risk classification were salt marshes and beaches (both 43%). Regarding the seven landscape units, São Jacinto-Espinheiro WB2, Mira WB1, and the coastal WB7 experienced the highest habitat risk levels per unit area (Figures 4A–C). The exposure-consequence risk plots revealed a significant driver of risk to mud flats, seagrasses, and salt marshes (Figures 5B–D) from the environmental management pressures (S4), either existing or to be implemented by the year 2030. Interventions to maintain critical economic activities such as shipping and transport are likely to have unintended impacts to these habitats following dredging activities (increase in water velocity and in tidal prism in submerged periods that may experience further coastal squeeze, Martínez-López et al., 2019). Recreation activities were found to pose higher threats to beaches and sand flats and mudflats (Figures 5A, B) due to high intensity of use for water sports, engines of tourist boats, and fishing. The highest risk posed by the three invasive (alien) species was identified for seagrasses and mud flats, driven by high habitat suitability for the invasive (alien) lugworm Arenicola spp. With one exception, most of the pressures evaluated showed limited to no risk to the ‘Bocage’ habitat. The invasive plant species, Cortaderia selloana, showed higher level of exposure driven in part by the high overlap since it is in this habitat where this species is most present (Figure 5E and Table 4). Similarly, pressures from the services sector (i.e., related to ballast water discharge, runoff from roads and emissions) were found to be of concern in beach and mud flat habitats due to poor management of marine shipping and transportation activities (Figure 5A, B; Table 4). Agriculture pressures posed a substantial risk in salt marshes and seagrasses because of nutrient runoff in certain channels of the lagoon.

Provisioning, regulation and maintenance, and cultural services are depicted as maps in Figure 6. Results showed the highest potential for the regulation and maintenance category, while the supply of provisioning services was lower throughout the Ria. ‘Bocage’ was cited by stakeholders to supply fewer provisioning services but was highly valued for regulation and maintenance services. This high potential was also attributed to salt marsh and seagrass habitats. Beaches were also highly valued, especially for cultural services. Mudflats were the least valued habitat in terms of the full suite of potential ES in Ria de Aveiro.

The vulnerability maps for the provisioning, regulation and maintenance, and cultural ES indicated where the stocks of these potential services are at greatest risk now and into the future. ES vulnerability hotspots across the three ES divisions were identified in the northern sections of the landscape (units 2, 5, and 6, corresponding to the São Jacinto-Espinheiro WB2, Ílhavo WB3, and Mira WB1 subareas; Figure 7). The supply of cultural ES was found to be most vulnerable in beach areas located south of the lagoon inlet. This was driven by risk from recreational activities (e.g., tourism boats, water sports) that, paradoxically, both threaten and support the diverse cultural values that beaches offer. These beach-going activities relate to physical and mental health benefits and aesthetic enjoyment. The ‘Bocage’ showed intermediate vulnerability for cultural services despite being a highly valued habitat in terms of ES availability. Its vulnerability was higher for regulating and cultural services than for provisioning. Mudflats were also found to be at the intermediate vulnerability level for the provisioning services, which are important for the regional food supply through harvesting activities such as bait digging and shellfish collecting. Regulation and maintenance services showed higher vulnerability scores in seagrass habitats than salt marshes, despite high potential in the latter for this service category.

Previous applications of the InVEST HRA leveraged diverse data sources and even used the tool in combination with other methodologies, e.g., ES and coastal zone management in Belize coast (Arkema et al., 2014); marine spatial planning in the Northeast and Mid-Atlantic Ocean regions in the United States (Wyatt et al., 2017); statistical analysis of ecology and conservation policies interactions in South Korea coastal area (Chung et al., 2015); InVEST HRA and coastal vulnerability models in Brazil (Elliff and Kikuchi, 2017); Marxan habitat prioritization software in California coastal area (Studwell et al., 2021); spatially explicit criteria based on marine biodiversity habitat suitability maps (Verutes et al., 2020) or habitat connectivity metrics in Iran inland (Ghehi et al., 2020). All these approaches confirm the diversity of options when applying the InVEST HRA tool. In this study, the InVEST HRA tool was applied to assess the risk that habitats involving coastal and terrestrial are due to the pressures produced by human activities, and to identify opportunities for EBM that can be applied to improve adaptative management. Yet, there are known limitations regarding the direct comparison of results because the risk outputs are relative ranks and use a unitless scoring system. This confirms the to’l’s baseline characterization, monitoring, and evaluation utility.

In Europe, Willaert et al. (2019) considered the EBM perspective and, like Cabral et al. (2015), by assessing the vulnerability of marine benthic habitats (EUNIS) and their potential ES to physical, chemical, and biological pressures identified by the Marine Strategy Framework Directive. Both studies translated these data into indices that characterize the vulnerability of these habitats when supplying ES in France and Portugal, respectively. Further, Caro et al. (2020) analysed the InVEST HRA and ES abundance data as a resilience descriptor in a Portuguese estuary. Our approach extended these European studies by applying the model at the local scale (the Ria de Aveiro Natura 2000 coastal zone comprises 330 km²) and producing high-resolution data (outputs at 50 m² grid cell size) to inform EBM management cycles with a spatiotemporally explicit decision-support tool.

Building on the initial ES vulnerability assessment in Ria de Aveiro, the methodology is meant to be replicated as new information becomes available. To showcase an approach to EBM in the region that can be guided by an open-source, spatially explicit risk assessment tool, we followed a stepwise approach to address the following questions: (1) Which human activities pose the greatest risk to habitats in the near term (2030)? (2) How might these pressures compromise the future delivery of ES in a Natura 2000 site? (3) What management options might minimise risks? Below we summarise the most salient findings during the three steps of our analytical approach.

Step 1. The human activities that pose significant pressures to the habitats of Ria de Aveiro by 2030 are primarily driven by environmental management objectives. These measures are in place to support economically essential activities such as the dredging of navigation channels and flood control infrastructure extension, as detailed in Lillebø et al. (2019) and Martínez-López et al. (2019). Despite these activities, they all comply with existing environmental policies and regulations, including monitoring program requirements for flora, fauna, and water quality. Agriculture activities also act as pressure due to nutrient runoff to the lagoon contributing to the nitrogen load, which is usually the limiting nutrient in brackish/marine aquatic environments This pressure is most notably in the upper reaches of the lagoon system. This finding aligns with Lopes et al. (2017), where low freshwater input into the lagoon and certain oceanic conditions can lead to exposure from non-point nitrogen sources with current land use and water management. On a unit-area basis, seagrass habitats were estimated to have some of the highest risk levels in the study area (Figure 4C). Caro et al. (2020) also identified seagrasses as habitat at high risk in the Mondego estuary anticipated impacts on ES, located ca. 50 km south of Ria de Aveiro. Both results align with the downward trend in seagrasses distribution globally (Turschwell et al., 2021). The highest cumulative habitat risks were revealed for three landscape units: the coastal one, the São Jacinto-Espinheiro, and Mira (bar charts in Figure 4C). In addition to direct human pressures, beaches and sand flats of the coastal zone are exposed to storm-induced erosion and, consequently, are at risk of collapse r’due to overtopping induced by storm waves (Duck and da Silva, 2012). The São Jacinto-Espinheiro water body is primarily exposed to ecohydrological effects of dredging the inlet that increase the tidal prism in the entire system, aggravated by the backwaters (Picado et al., 2010; Azevedo et al., 2013). In these conditions, the flood bank will inhibit the saltmarshes’ adaptation to changes in the hydrological regime, because they cannot shift landward, endangering them due to increasing submergence times (Martínez-López et al., 2019). Mira Channel is described as a subsystem: narrow and shallow as an estuary, where bait-digging activities and oyster production occur (Cunha et al., 2005). Invasive-alien species of lugworm (Arenicola marina) was first recorded in Ria de Aveiro in 2009 (Pires et al., 2015) and have driven seagrass fragmentation, degradation, and loss. The lugworm distribution (in our study, two lugworm species are grouped by the genus Arenicola) at Ria de Aveiro has been increasing since then and, therefore, restoration actions have been implemented to decrease the pressure due to bioturbation activity of Arenicola spp and restore fragmented seagrass meadows (Costa et al., 2022).

Step 2. The second step estimates where and to what extent the future delivery of ES in a Natura 2000 site may be compromised. The coastal-marine habitats that deliver ES in Ria de Aveiro should be in a good environmental status to continue supporting the flows of life- and livelihood-giving services. The EU recognizes the natural capital stocks of this region through a designation of Sites of Community Importance, which highlight the need to maintain and enhance ES in the area. According to our results, regulation and maintenance services were the most vulnerable service values, while provisioning services were the least vulnerable. This approach is based on quantifying ES to represent vulnerability, and recently, Peng et al. (2023) proposed an indicator-based framework to face the imbalance in the representation of social and ecological systems. Spatial planning tools, such as the HRA tool employed in Ria de Aveiro, can guide habitat protection and restoration efforts to areas that will safeguard the most vulnerable services. The cumulative risk maps show spatially the area of the habitats with the highest risk, which are informative to guide where protection is urgent. This form of EBM aligns with EU and UN policy targets that aim to prevent, halt, and reverse the degradation of ecosystems worldwide.

Step 3. The third step identifies management options that are likely to reduce risk. Initially, we searched for overlap between areas of high ES availability and the greatest threat to habitats that supply these services (vulnerability assessment, step 2). Then, we confirmed which activities drive risk to those habitats and locations (HRA, step 1). Based on our findings, efforts could be prioritized to mitigate the unintended impacts of ongoing environmental management. Unfortunately, many of these impacts result from indirect effects, meaning they are not localized to the activity footprint. For example, the flood bank extension by 2030 is expected to increase coastal squeeze. This flood control measure does not violate Portuguese or EU environmental regulations despite the unintended consequences.

Alternatively, other environmental management activities, such as the direct effects of dredging, present substantial risk to proximate seagrass beds (near maximum exposure score in some areas; Figure 5C, Table 4). Management of dredging has improved significantly through improved ability to predict the spatial extent, however, the susceptibility to environmental changes caused by dredging to seagrasses is not well understood (Erftemeijer and Lewis, 2006). Given the seagrass biotype’s importance to ES (regulating and cultural values; Supplementary Table 2), management interventions should aim to reduce seagrass risks from dredging and other human pressures. Ultimately, the unintended impacts from environmental management can be further monitored and analysed to confirm their status and update the findings of this vulnerability assessment over the near term.

All models have advantages and disadvantages. When applying the InVEST HRA tool, the following limitations should be considered. Caro et al. (2020) highlighted that the HRA misses detecting the potential increase in vulnerability that comes from impacts to ES that are highly demanded by beneficiaries and decreases in susceptibility of those habitats that safeguard more abundant supplies of natural capital. This is likely due to how the HRA assumes a priori additive effects of human pressures (Arkema et al., 2014). Therefore, a future option to consider synergistic or antagonistic interactions based on multiple pressures could improve ecosystem risk assessment (Studwell et al., 2021). Samhouri and Levin (2012) applied a classic risk-assessment framework to regional populations of indicator marine species. They found it helpful to validate the risk scores with alternative sources, such as extinction thresholds or measures of irreversible harm. Another potential HRA model limitation relates to data availability, quality, and uncertainty. Wyatt et al. (2017) raised concerns that the HRA outputs are conditioned on the choice of habitats, pressures, and the area of interest and that the approach is built upon a subjective system (i.e., derived from expert elicitation) for scoring the exposure and consequence criteria.

Our study overcame many of these limitations by leveraging the best available information and detailing all sources of information (see Supplementary Table 1). Similar to the challenges experience by Cabral et al. (2015), it was difficult to harmonize certain pressure data sets (e.g., sightings of invasive species) due to disparate information sources collected at different spatial and temporal scales. Chung et al. (2015) recommends the use of in-situ field data to improve the accuracy and reliability of spatial assessments and establish accurate statistical models. When time and resource constraints make this impossible, the HRA tool enables scoring of data quality (d_i ) based on the source of information and the confidence in the data accuracy. Knowledge gaps can also be addressed through variable weights (w_i ) that impact the importance placed on a criterion through a weighted average. Our approach in Portugal leveraged diverse research methods to assemble spatially explicit information at the local scale and an interdisciplinary team with intimate socio-ecological knowledge of the region to characterize and account for data uncertainty.

While we did not analyse alternative management options in this study, habitat risk and ES vulnerability outputs provide a future roadmap for monitoring and evaluation as additional data are collected about the future 2030 scenario and beyond. With a solid foundation of expert knowledge and InVEST HRA inputs, this approach in Ria de Aveiro can be revisited over time to respond to changes in habitat-pressure interactions that reflect new knowledge and insights (Wyatt et al., 2017), as is typical with EBM management cycles. This framework offers a transparent and generalizable approach that can be applied in most places where human activities threaten biodiversity and ES.