We conducted a systematic literature review following the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) protocol (Moher et al. 2015, Page et al. 2021). We developed a search query that was executed upon the SELINA project internal literature database to ensure SELINA internal consistency in terminology and understanding (Burkhard et al. 2023, Seguin et al. 2025). The SELINA database is derived from the Scopus and Web of Science (WoS) databases and contains 108,064 entries published from 2018 to 2022 (Seguin et al. 2025). The primary motivation for selecting this time-frame was its alignment with recent policy and research attention around ecosystem assessments, including the consolidation of MAES reporting in the EU (Maes et al. 2015, Maes et al. 2018, Maes et al. 2020c), the emergence of the IPBES methodological guidance (IPBES 2019) and increasing experimentation with EC frameworks (Czúcz et al. 2018b, Czúcz et al. 2021a, Vallecillo et al. 2022). Since the SEEA EA was only published towards the end of the review period (United Nations 2025), the review only allows for the detection of early references or methodological signals linked to its release. Furthermore, the time-frame allowed us to obtain a representative pre-SELINA baseline of the state of the art, capturing the methodological and conceptual landscape immediately preceding the project.

The relevant search query (Suppl. material 1) was structured by combining ecosystem-related terms with those referring to EC, ES and Ecosystem Accounting, as well as terms and synonyms related to indicators. Generally, an indicator is understood as being a “number or qualitative descriptor generated with a well-defined method, which reflects a phenomenon of interest (the indicandum)” (Heink and Kowarik 2010). Even though indicators are based upon variables, oftentimes, these terms are used as synonyms (Müller and Burkhard 2012, SELINA 2025). Therefore, despite recognising that, in the context of SEEA EA, EC variables and EC indicators have very distinguished definitions (United Nations 2025), we applied a broad understanding of the term indicator, encompassing “proxies”, “variables” and “metrics” (Czúcz et al. 2021a, Czúcz et al. 2021b). Moreover, in this review, terms and respective concepts, such as ecological condition, biological condition, ecosystem state, status, health, integrity, quality, functionality and capacity, are treated as conceptual equivalents to EC (see i.a. Roche and Campagne 2017, Keith et al. 2020 and references within). By using Boolean operators, establishing logical combinations of keywords and using truncations and wildcards to capture word variations in title, abstract and author keywords, we iteratively improved and refined the search logic (Suppl. material 1 and Seguin et al. 2025).

In total, 3,430 studies were initially found. The database was cleaned to remove duplicates and filtered to include only English-language scientific peer-reviewed publications (n_p = 2,720 papers). In the first screening phase, 23 reviewers checked publications by all reviewed documents title, abstract and author keywords on seven inclusion criteria (n_p = 664 papers; Fig. 1). For the second screening phase, based on the full text of the articles, we refined the inclusion criteria to only include studies addressing a link between EC and ES. This means that the study had to quantitatively or qualitatively connect the assessment of EC (or related concepts) with an assessment of ES or Nature’s Contributions to People (NCP; Díaz et al. 2018). This includes, for example, the comparison of the results from both assessments, indicators, direct input-output relations by integrating EC results into the ES assessment and the integration of EC and ES results for a third purpose.

Five articles were excluded because their full texts were not accessible (n_p = 659 papers). Ultimately, 128 publications were included in the full review (see list of included publications in Suppl. material 2) from which data were extracted, based on specific questions.

To collect, systematise and analyse the content of the reviewed literature items, we developed a structured coding template (Suppl. material 3) of several tabular sheets within Microsoft Excel 2019 (Version: 1808), that incorporated various functionalities to ensure a user-friendly, consistent and comparable review and evaluation process. Only a few entries allowed free text to ensure high inter-reviewer consistency and standardisation of entries. Throughout the fields and predefined response options, we prioritised and built upon standardised methodologies, common classifications and definitions, amongst which the SEEA EA Ecosystem Condition Typology (ECT; United Nations et al. 2021, Czúcz et al. 2021a), which was used to reclassify EC indicators (for further information see Section “Reclassification of ecosystem condition indicators”), the Common International Classification of Ecosystem Services (CICES v.5.1; Haines-Young and Potschin 2018), which was applied to reclassify recorded ES and the ES method type according to the ESMERALDA MAES Method Explorer (ESMERALDA 2018) and the SELINA Ecosystem Assessment Explorer (EASE; Brander et al. 2024a).

The review template consisted of four tabs:

The first tab included bibliographical details and a final set of inclusion criteria that had to be answered by each reviewer after a study was assigned to be included or excluded;

During the review, each contributor was asked to follow the original statements as made in the papers to minimise interpretation bias. For example, the position of the publication on whether an indicator represented EC or ES was followed, regardless of the reviewer’s interpretation. If five or more ecosystems were assessed in a study, the reviewers were asked to state “various” ecosystem types, based on the hypothesis that indicators that were applied to assess five or more ecosystems may be regarded as giving an broader integral or holistic understanding of the landscape interactions emphasising its multifunctionality.

Several quality checks were implemented to design the review process as transparently, consistently and objectively as possible. Most reviewers were involved in co-creation when developing the search objectives, inclusion criteria and review template. Every reviewer received training before starting and each screening phase included a pilot run, during which the task and material were tested and adjusted if necessary. For the choice of the ES method according to the list defined in the MAES Method Explorer (Burkhard et al. 2018a), reviewers indicated their uncertainty; during the post-review cleaning phase “highly uncertain” and “very highly uncertain” results were double-checked and corrected if necessary.

Moreover, after completion, each phase of the review underwent manual inspection by a dedicated team of experts to validate that the entries meet the requirements. Semi-automated processes in R (4.3.2 (2023-10-31 ucrt); R Core Team (2023)) streamlined the validation process of the data entries. Furthermore, a Fleiss-Kappa consistency test was applied to evaluate the literature review's quality and assess the reviewers' inter-rater reliability (Fleiss 1971, Fleiss et al. 2003, McHugh 2012, Kassambara 2019). It was calculated, based upon the reviewer’s answers on inclusion criteria for two randomly selected publications that were screened by all reviewers (κ: 0.706, p < 0.0005*). A second Fleiss-Kappa coefficient was calculated, based on one publication that had garnered a consensus for inclusion in the review amongst a large majority of 21 out of 23 reviewers (κ: 0.856, p < 0.0005*). By examining these two groups separately, the aim was to gain insights into the consistency of reviewer assessments, with a particular emphasis on understanding the reliability and agreement in the screening and inclusion decisions. The Kappa values indicate a substantial to strong level of agreement reinforcing the robustness and validity of the screening process.

The ecosystem characteristic represented by EC indicators, following the SEEA EA Ecosystem Condition Typology (ECT) described in Czúcz et al. (2021a), was recorded during the review when specified in the publication. Additionally, all EC indicators were reclassified to match the SEEA EA ECT by trained review team members following data extraction to aid in aggregating the data. Indicators that did not represent any of the main ecosystem characteristics described in the ECT were assigned to additional classes (ECT+), largely based on the different types of ancillary data sources detailed in Czúcz et al. (2021a) (Suppl. material 4). In case an indicator could not be assigned to any ECT or ECT+ class, it was classified as ‘Other’. Two review team members reclassified each indicator independently, with conflicting classifications discussed and resolved within pairs.

The data processing, as well as analysis, was fully conducted in the R environment using various packages like “tidyverse” (Wickham et al. 2019), “revtools” (Westgate 2019a, Westgate 2019b), “readxl” (Wickham and Bryan 2023), "writexl” (Ooms 2023), “ggsankey” (Sjoberg 2024), “extrafont” (Chang and Bertrand 2023) and “cowplot” (Wilke 2024).

The collected data were compiled into one database. All EC indicators that did not have a link to a specific ES – either directly or implicitly as part of a composite indicator – were removed from the database (nᵢ = 82). Subsequently, the database was analysed using descriptive statistics, based on the number of papers (n_p) or the number of indicators (n_i) and their relative frequencies. Additional evaluation for specific parts of the dataset, such as a spatial assessment and a Sankey diagram, were added to visualise and better understand the connections between EC indicators and ES classes. The spatial analysis was conducted to detect asymmetries in the global distribution of studies. We firstly examined their distribution at the country level using ArcGIS Pro (version 3.5). For this, publication counts were assigned per country: if a publication included multiple case studies within the same country, it was counted once, whereas publications covering multi-national assessments or multiple case studies in different countries contributed to the count of each country assessed. Publications that presented an assessment conducted only at the continental level, without precise details of the countries included in the study, were excluded from this spatial analysis (n_p = 5).

In the following, we provide insights into the general publication characteristics and focus, amongst others, on the spatio-temporal distribution of the researched studies. Out of the 128 reviewed publications, 119 dealt explicitly with the ES concept, while in nine publications, the conceptual frame was not mentioned. In the majority of cases, the ES potential was assessed (n_p = 74), followed by the ES flow (n_p = 52); ES demand was addressed far less frequently (n_p = 7). Four publications applied the Ecosystem Accounting approach, thus compiling accounts for ES (n_p = 3) or for EC and ES (n_p = 1). In terms of temporal distribution, we observed a continuous increase in the number of publications over the years, with a dip in 2020 and a more prominent increase in 2022. The spatial scale applied in the publications shows a clear focus on smaller scales with regional to local assessments (Fig. 2). Five publications focused on a continental scale (including Europe n_p = 2; Africa n_p = 1; Africa, Asia and America n_p = 2). In total, 64 countries were recorded, while marine ecosystems were addressed in two publications (Fig. 2). No publications addressed Antarctica and only a few included assessments located in Africa, Central America and Oceania. In contrast, the majority of publications included assessments in North America, Europe and Asia. More precisely, most publications included assessments located in China (n_p = 30), followed by the USA (n_p = 15). Most Chinese assessments were conducted on the regional scale, while US assessments were predominantly local. In Europe, Spain was the most studied country.

The focus of the EC assessments were diverse and often overlapping. Most frequently, they related to the mapping and/or assessment of ES (n_p = 89), followed by mapping or assessment of EC (n_p = 50), while nearly 20 publications were associated with the planning or evaluation of restoration actions. Most publications included the same number or more EC indicators than ES, typically ranging from two to five. An exception are publications addressing marine ecosystem types, where the total number of ES was on average more than double the number of EC indicators.

In this section, details are given on the EC and ES indicators identified across the reviewed studies. In total, 1636 indicators were identified. Hereof, 929 have been identified as EC indicators with a direct (nᵢ = 744) or implicit (nᵢ = 185) link to one or more ES and 707 have been identified as ES indicators. In the following, we present the specifics of the EC indicators, notably focusing on the distributions according to the extended ECT classification (including both ECT and ECT+ classes) and the ecosystem types. Thereafter, the specifics of the recorded ES indicators are shown.

Even though the majority of EC indicators (nᵢ = 700; 75.3%) were not classified according to the SEEA EA ECT by the respective publication authors, most recorded indicators could, nonetheless, be matched to the extended ECT classification. In total, 895 indicators (96.3%) were successfully assigned, with 74.1% classified as ECT (nᵢ = 688) and 22.3% as ECT+ (nᵢ = 207); the remaining 3.7% (nᵢ = 34) were classified as “other”, which included, for example, Terrain Location, Snow Gliding Distance, Average Mutual Information and System Entropy (Fig. 3). The predominance of ECT over ECT+ classes was consistent across all ecosystem types, with the exception of the “various” category, where the distribution between ECT and ECT+ indicators was more balanced.

EC indicators were identified for each ECT class. Within the ECT classes, chemical state characteristics (nᵢ = 168), such as carbon, nutrients and heavy metal concentrations, as well as pH measurements, were most frequently used, whereas functional state characteristics, for example, indicators related to leaf traits, were least represented (nᵢ = 69, Fig. 3). The distribution of indicator types varied according to the ecosystem type. For woodland and forest ecosystems, structural state indicators, for example, indicators related to vegetation cover, were most common (nᵢ = 51), while grasslands were predominantly assessed using compositional state indicators (nᵢ = 28), i.e. indicators related to specific species composition, diversity, richness and abundance. Cropland ecosystems were primarily assessed using chemical state indicators (nᵢ = 104), followed by physical state (nᵢ = 66) indicators, for example, those related to soil texture, composition and structure, whereas indicators recorded for the “various” ecosystem type most frequently referred to landscape characteristics (nᵢ = 57), such as landscape connectivity or complexity, as well as patch density and size. Within the ECT+ classes, pre-aggregated indicators were most common (nᵢ = 57, Fig. 3), mainly related to “various” ecosystem types. Additionally, 34 indicators represented pressures acting on ecosystems, for example, indicators related to pollution, contamination and land use and another 34 referred to stable environmental characteristics, such as topography and climatic conditions (Fig. 3).

Generally, EC indicators were identified for all ecosystem types, but they were unevenly distributed (Fig. 4, Table 1). The majority of EC indicators were used for one specific ecosystem type (nᵢ = 504), while 20.1% (nᵢ = 187) were applied for 'various' ecosystem types. The highest number was used to quantify the condition of cropland ecosystems (nᵢ = 331), followed by woodland and forests (nᵢ = 212) and “various” ecosystem types (nᵢ = 187, Fig. 4). Few EC indicators were applied to describe heathland and shrub, sparsely vegetated land, wetlands and marine and coastal ecosystems.

In most cases, input data were used to directly quantify the respective EC indicator (nᵢ = 433; 46.6%, Table 1). For one third of the EC indicators (nᵢ = 293; 31.5%), the input data were used for an index calculation. Further data processing or modelling was required for 17.5% (nᵢ = 163). However, this general pattern did not apply to marine, river and lake ecosystems and various ecosystem types, where input data were most commonly used for index calculation. The assessment of EC indicators relied predominantly on field data (nᵢ = 517, Table 1). Additional data sources included remote sensing (nᵢ = 150), other processed spatial data (nᵢ = 137) and statistical data (nᵢ = 135). Across ECT classes, landscape state characteristics were the only exception here, as their EC indicators were primarily derived from remote sensing and other processed spatial sources. This specific pattern was also observed for indicators assigned to “various” ecosystem types. Almost one third of the EC indicators were not spatially explicit (Table 1). The remaining two thirds were distributed fairly evenly across fully spatially explicit indicators and those aggregated either at the ecological or administrative scale. Only in very few publications (n_p = 20; nᵢ = 90), reference conditions or levels were applied to the EC indicators. In cases where they were used, this was most commonly done following a fixed-year approach (n_i = 25), using reference levels defined by a policy target (n_i = 21) or a natural reference condition (n_i = 12). Reference levels and conditions were most frequently used in the calculation of indicators in freshwater ecosystems (nᵢ = 33). The majority of collected indicators were individual metrics (n_i = 647; 69.6%), while 24.4% (n_i = 227) of the collected metrics were part of a composite indicator (n_i = 40; 4.3%). Relatively, EC indicators applied to “various” ecosystem types (n_i = 187) had the highest share of composite indicators (n_i = 28) and part of composite indicators (n_i = 127).

In the following, we focus on the assessment of those EC indicators that have been directly linked to one or more ES (n_i = 744; 80%). Overall, their distribution largely mirrors the general pattern described above with a few differences. There are relatively few landscape characteristic indicators, particularly those derived from remote sensing or other processed spatial data. Furthermore, relatively few EC indicators applied to various ecosystem types were directly linked to ES and they exhibited an even stronger predominance of field data as input.

The EC indicators were most frequently linked to ES from the Regulation & Maintenance section (n_i = 291) and multiple ES classes (n_i = 222), followed by Provisioning (n_i = 158) and then Cultural ES (n_i = 72). EC indicators applied in croplands were predominantly linked to Provisioning ES, whereas those used in wetlands, rivers and lakes were more frequently associated with Cultural ES. For composite indicators, 68.9% (n_i = 40) were explicitly linked to ES, while, for the parts of the composite indicator, only 18.4% (n_i = 42) were recorded showing this direct connection. The majority of these indicators (n_i = 341) were quantitatively compared to ES, for example, through statistical analysis, while 73 were qualitatively compared (Fig. 6a). In other cases, EC indicators were directly integrated into the ES assessment (n_i = 256). Here, we found the largest share of ECT+ indicators (Fig. 6a). Finally, in 90 cases, the EC and ES results were assessed individually and afterwards combined for a third purpose, such as conservation or restoration measures, for example. However, when differentiated by ecosystem type, the pattern deviated from this overall trend. In cropland, grassland as well as woodland and forest ecosystems, EC indicators were most frequently integrated directly into the ES assessments, whereas in heathland and shrub, sparsely vegetated lands, wetlands and “various” ecosystem types, EC indicators were predominantly combined with ES assessments for a third purpose. In 14 cases, EC indicators could only be linked to general ES rather than to specific CICES v.5.1 classes or sections. In most of these instances (n_i = 11), the EC indicator and ES were combined for a third purpose, for example, a Life Cycle Impact Assessment (Chen et al. 2021).

For EC indicators compared to ES (Fig. 6b), either quantitatively or qualitatively, the majority of relationships were positive (n_i = 124), followed by mixed (n_i = 119), referring to cases that showed positive and negative relations and unclear/not specified relationships (n_i = 93). In contrast, no relation (n_i = 38) and negative relationships (n_i = 40) were comparatively rare. This overall pattern holds when distinguishing between quantitative and qualitative comparisons; however, in the qualitative comparisons, only a small number of mixed relationships were addressed (Fig. 6b).

When the assessed directionality of links is differentiated by ECT class, several patterns become apparent (Fig. 6c). For EC indicators related to physical and chemical state characteristics, most relationships were reported as unclear/not specified or mixed. EC indicators, for which solely negative or negative and no relation was found, were mostly reported as chemical state characteristics. Indicators linked to compositional state were most frequently associated with mixed and positive relationships. In contrast, indicators related to structural and functional state characteristics predominantly showed positive relationships. For landscape-related indicators, no dominant trend emerged; instead, positive, negative and no relationships were more evenly distributed. Negative relationships to ES were primarily found for landscape indicators related to (mean) patch size, length, cohesion or aggregation. By contrast, landscape indicators showing positive relationships to ES were more diverse and related to aspects, such as connectivity, complexity, density and continuity.

When differentiating the directionality of EC–ES links by ecosystem type, further patterns can be observed. Woodland and forest ecosystems, wetlands, coastal and marine ecosystems were clearly dominated by positive relationships. In croplands, positive relationships predominated, with negative and no relations occurring at lower, but comparable frequencies. In urban ecosystems, links were most frequently reported as unclear/not specified, followed by positive relations. Grassland, as well as heath and shrubland ecosystems, showed a more even distribution across positive, mixed and no-relation outcomes. The pattern for rivers and lakes was more balanced than in other ecosystem types.

To gain deeper insights into the role of EC indicators in ES assessments, we examine different aspects of these indicators and their links to specific ES (i.e. CICES v.5.1 classes). We differentiate their distributions according to the CICES v.5.1 sections that they are linked to – namely, Provisioning, Regulation & Maintenance and Cultural ES (Fig. 7). We focus on trends and distributions that depart from the general patterns described above and draw attention to particularly noteworthy findings. It is important to note that a single EC indicator may be associated with multiple ES and, thus, with multiple CICES v.5.1 classes across different sections. Therefore, in the following paragraph, the assessment differentiates between EC indicators linked to Provisioning, Regulation & Maintenance and Cultural ES, as well as those linked to ES from multiple CICES v.5.1 sections. The latter includes those EC indicators that were linked to “Ecosystem Services” in general (Fig. 7). Generally, links were identified from EC indicators from each ECT class to all CICES v.5.1 sections.

EC indicators that were only linked to Provisioning ES (nᵢ = 158) mainly related to chemical and physical state characteristics (nᵢ = 64 and 32, respectively; Fig. 7). Within this group of EC indicators, the most frequently linked ES classes were “cultivated terrestrial plants (including fungi, algae) grown for nutritional purposes” (CICES v.5.1 code: 1.1.1.1; nᵢ = 79), “wild animals (terrestrial and aquatic) used for nutritional purposes” (CICES v.5.1 code: 1.1.6.1; nᵢ = 28) and “wild plants (terrestrial and aquatic, including fungi, algae) used for nutrition” (CICES v.5.1 code: 1.1.5.1; nᵢ = 20). Additionally, the spatial scale of these EC indicators was mostly unclear/not specified (nᵢ = 82), followed by “not spatially explicit” (nᵢ = 42). In many cases, the direction of the relationship (in both quantitative and qualitative comparisons) was reported as unclear or not specified (nᵢ = 72), while positive relations were the next most frequently noted pattern (nᵢ = 21).

EC indicators that were only linked to Regulation & Maintenance ES (n_i = 291) mainly related to structural state characteristics (n_i = 77), followed by chemical state and compositional state characteristics (Fig. 7). Comparatively high was also the relation to landscape characteristics (n_i = 23). Within this group of EC indicators, the most frequently linked ES classes were “maintaining nursery populations and habitats (including gene pool protection)” (CICES v.5.1 code: 2.2.2.3; n_i = 116), “decomposition and fixing processes and their effect on soil quality” (CICES v.5.1 code: 2.2.4.2; n_i = 80) and “regulation of chemical composition of atmosphere and oceans” (CICES v.5.1 code: 2.2.6.1; n_i = 77). The EC indicators were predominantly not spatially explicit (n_i = 111) or aggregated at the ecological scale (n_i = 98). Naturally, most of the linked EC indicators were not (part of) a composite indicator (n_i = 274). However, the share of EC indicators reported as “composite indicator” (n_i = 13) outnumbered the share of “part of composite indicators” (n_i = 2). Across both quantitative and qualitative comparisons, positive relationships were detected most frequently (n_i = 71).

EC indicators exclusively linked to Cultural ES (nᵢ = 72) commonly related to chemical and physical state characteristics (nᵢ = 15 and 13, respectively), followed by compositional state characteristics (Fig. 7). Within this group of EC indicators, the most frequently linked ES classes were “characteristics of living systems that enable activities promoting health, recuperation or enjoyment through active or immersive interactions” (CICES v.5.1 code: 3.1.1.1; nᵢ = 66) and those enabling such activities through passive or observational interactions (CICES v.5.1 code: 3.1.1.2; nᵢ = 21), as well as “characteristics of living systems that enable aesthetic experiences” (CICES v.5.1 code: 3.1.2.4; nᵢ = 5). For the EC indicators linked to Cultural ES, diverse spatial scales were reported. They were all not (part of) a composite indicator. Regarding the comparison between EC indicators and ES, mixed relationships were observed most frequently (nᵢ = 46).

EC indicators linked to ES from multiple CICES v.5.1 sections (nᵢ = 222), show a less pronounced trend when it comes to related ECT classes (Fig. 7). A slight peak in compositional state characteristics can be identified (nᵢ = 38), followed by landscape characteristics (nᵢ = 28) and the ECT+ class “Pre-aggregated indicators” (nᵢ = 24). Furthermore, the distribution of the nature of the relation between the EC indicators and the ES showed an emphasis on the integration of EC indicators into ES assessments (nᵢ = 92).

Of the 744 EC indicators linked to ES, 256 were integrated into ES assessments. Within this subset, several distinct patterns emerged compared to the general pool of EC indicators with any link to ES presented above. The predominance of EC indicators linked to Regulation & Maintenance ES, as well as to multiple CICES v.5.1 sections, was even more pronounced. In terms of ecosystem types, EC indicators were most frequently applied in croplands (nᵢ = 167), followed by grasslands (nᵢ = 125) and woodland and forests (nᵢ = 90). Notably, the relative share of indicators applied in rivers and lakes was considerably lower (nᵢ = 16). The share of EC indicators with unclear or unspecified spatial resolution dropped substantially compared to all linked EC indicators. Fully spatially-explicit indicators were slightly less common (nᵢ = 25), while indicators aggregated at the administrative scale became more prevalent (nᵢ = 59). EC indicators integrated into assessments of Cultural or Provisioning ES were largely not spatially explicit. Regarding ECT classes, chemical state characteristics remained dominant (nᵢ = 45), but physical state characteristics were also well represented in this subset (nᵢ = 36). A relatively higher proportion of indicators were classified as ECT+ (nᵢ = 87), particularly those related to stable environmental characteristics, ecosystem services and ecosystem extent (whereas pre-aggregated indicators were less prominent). Overall, fewer types of input data were used, mainly for index calculations. When reference levels were applied, simple data-driven approaches became less common, while policy-orientated and safety/health-related approaches relatively increased – especially amongst EC indicators integrated into Cultural ES assessments. No “part of composite” indicators were integrated into ES assessments; instead, either stand-alone metrics or a few composite indicators were used. Examples of those composite indicators were “Forest structure” aggregating “Stand density” and “Mean tree size” (González-Díaz et al. 2019) and “Ecological environmental quality index” aggregating information on the “Normalised difference vegetation index”, “Leaf area index” and “Net primary productivity” (Zhou et al. 2021). These composite indicators were mainly applied in assessments involving multiple CICES v.5.1 sections and, to a lesser extent, in assessments of Regulation & Maintenance ES. Finally, seven EC indicators were not only integrated, but also directly compared to a specific ES. Of these, six showed a positive relationship; amongst those six, there were multiple EC indicators related to soil nutrient concentrations and microbial biomass carbon with an identified positive relation to the ES “Regulation of soil quality” (CICES v.5.1 codes 2.2.4.1 and 2.2.4.2).

With regard to the temporal distribution, a noticeable increase in publications was observed in 2022. However, given the short review period, it is not possible to determine whether this rise reflects a coincidence, a delayed publication effect from the COVID-19 pandemic or a response to major policy developments, such as the release of the SEEA EA framework (United Nations 2025), the EU Biodiversity Strategy for 2030 or advances under MAES/KIP INCA initiatives (La Notte et al. 2022, Inácio et al. 2025b). Since the SEEA EA ECT was only published towards the end of the review period, it was not surprising that the majority of papers did not apply this typology.

A spatial asymmetry was evident in the literature, with most studies being conducted in the United States, China and Europe – a pattern also found in previous reviews (Runting et al. 2017, Pinto et al. 2022). Additionally, differences could be detected with regard to the applied methodological approaches between regions. For example, several studies from China employed total ES value approaches and composite Ecosystem Health Index models, based on the Vigour–Organisation–Resilience framework (e.g. Ma et al. 2022, Xu et al. 2022).

Only a small number of publications identified in this review dealt explicitly with ecosystem accounting. Most operational ecosystem accounting applications – particularly those aligned with the SEEA EA framework (United Nations 2025) – are implemented by national or regional statistical offices and other public authorities, who typically disseminate results through official reports rather than peer-reviewed scientific publications (Comte et al. 2022, Lange et al. 2022). Second, the temporal window covered (2018–2022) largely precedes or overlaps only partially with the formal release of SEEA EA (United Nations 2025), meaning that many accounting initiatives may still be in early development or not yet published.

The EC and ES indicators, discussed in this review, represent a specific subset where EC is explicitly linked to ES. Thus, we only focused on indicators that directly or implicitly connect the two concepts, so the patterns we observed are influenced by this research scope and might not mirror the broader methodological landscape assessed in other studies (e.g. Nicholson Thomas et al. 2025).

We observed a clear imbalance in ecosystem type representation, with cropland and “various” ecosystems being most frequently assessed, followed by woodland and forest, grassland and river and lake ecosystems. In contrast, exclusively marine ecosystems were substantially under-represented, indicating a strong bias towards terrestrial assessments, in particular for EC indicators. This trend is in line with Maes et al. (2016), Vári et al. (2024), Zhou et al. (2024) and also shows similar patterns to the EC consideration in ES valuation studies from the ESVD (Hernández‐Blanco et al. 2022). Assessments covering “various” ecosystem types differed from ecosystem type-specific assessments. Whereas the latter often aim to understand ecological functioning or rather specific management needs, multi-ecosystem assessments tend to pursue broader comparative, prioritisation or policy-support objectives, such as spatial conservation planning or land-use optimisation (e.g. Maes et al. 2020b). Following the EU-wide MAES assessment (Maes et al. 2020b), these indicators are referred to as cross-cutting and, for EC, they include “landscape fragmentation, water quality related indicators and the share of the ecosystem that is managed as a protected area”. Indicators related to “various” ecosystem types often rely more heavily on generalised or transferable indicators, which can be consistently applied across heterogeneous landscapes. In this context, Nicholson Thomas et al. (2025) highlight that, while standardisation supports alignment across studies, excessive uniformity can neglect ecosystem-specific realities and undermine stakeholder legitimacy. Within the ECT classes, EC indicators applied in “various” ecosystem types were mostly related to landscape characteristics and, within ECT+, pre-aggregated indicators were most common. The dominance of remote sensing and other processed spatial data for this share of EC indicators reflects the need for wall-to-wall, spatially standardised inputs (Turner et al. 2015, Nicholson Thomas et al. 2025), while the frequent use of value transfer methods in ES assessments aligns with literature that highlight the importance of respective scalable approaches (Johnston et al. 2021, Grammatikopoulou et al. 2023, Brander et al. 2024). In contrast, the more ecosystem type-specific assessments, such as those focusing on croplands or forests, are more likely to use field-based or functional data and methods, reflecting domain expertise (Binder 2017, Burkhard et al. 2018a, Bank et al. 2025, Nicholson Thomas et al. 2025).

We identified a general predominance of chemical state indicators, while process-orientated indicators, for example, those describing functional state characteristics, remain less operationalised. This pattern is consistent with the broader challenge of translating ecosystem functioning into standardised indicators with relevance, for example,, for decision-making and policy (Smit et al. 2021, Nicholson Thomas et al. 2025). The comparatively high use of reference conditions in freshwater ecosystems likely reflects the strong regulatory and monitoring frameworks that exist for rivers and lakes already since the turn of the millennium. Instruments, such as the EU Water Framework Directive (WFD; European Commission 2000) and comparable national policies mandate the definition of ecological status classes and require assessments against established reference conditions or policy-defined targets (Solimini et al. 2009, Maes et al. 2012, Reyjol et al. 2014, Lyche Solheim et al. 2025). This has led to the widespread development of standardised water quality indices and long-term monitoring datasets, facilitating diverse reference level approaches. For example, Apostolaki et al. (2019) provide a conceptual approach how to combine the EU WFD requirements with the ES approach and EC assessment, although not explicitly considering reference conditions. In contrast, structured reference frameworks are less consistently available for other ecosystem types, which may explain the more limited use of reference levels in those systems.

More generally, this review confirms that EC is not used in a consistent or conceptually disciplined way across scientific literature, particularly when implemented through indicators. As also reported by Bank et al. (2025) and Nicholson Thomas et al. (2025), EC is applied across a broad spectrum of interpretations, ranging from ecosystem properties or environmental characteristics (slope, relief, climatic variables) to direct biophysical state metrics, composite indices, pressure proxies and even outcome-based measures that quantify ES supply rather than EC. To bring structure to this heterogeneity, we classified all EC indicators recorded in this review according to the SEEA EA ECT and assigned remaining indicators to an extended classification (ECT+), following Czúcz et al. (2021b). Around one fourth of the reviewed EC indicators were not linkable to the original ECT classes recognised by SEEA EA. In contrast to this diverse and sometimes ambiguous usage found in the reviewed literature and also identified by recent studies (Bank et al. 2025, Nicholson Thomas et al. 2025), both SEEA EA (United Nations 2025) and the EU-wide methodology for EC assessment (Vallecillo et al. 2022) define EC more narrowly. Under these frameworks, EC refers to the quality of an ecosystem measured in terms of its abiotic and biotic characteristics and ES are the contributions of ecosystems to the benefits that are used in economic and other human activity. Therefore, we recommend that indicators labelled as EC in literature, but in fact describing pressures or ES (thus having been reclassified as respective ECT+ class) should not be treated as condition metrics in a strict sense (United Nations 2025). Their inclusion in EC assessments may be practical in applied contexts, but, for analytical clarity, condition, pressure and service roles, should be clearly distinguished.

As outlined in the Introduction, the relationship between EC and the supply of ES remains conceptually established, but empirically unresolved. While numerous studies report positive associations between EC or related concepts (such as ecological integrity) and ES, agreement is still lacking on the specifics or contexts in which these links hold (Strong et al. 2015, Roche and Campagne 2017, Townsend et al. 2018, Smit et al. 2021). Against this backdrop, we here provide a synthesis of how EC–ES connections are currently implemented through indicators across ecosystem types, ECT classes and CICES v.5.1 sections.

Implementation across ecosystem types

The observed pattern of EC indicators being most frequently linked to Regulation & Maintenance ES likely reflects both ecological dependencies and long-lasting research efforts (Müller 2005, Quijas et al. 2010, Rendón et al. 2022). Regulation & Maintenance ES build upon crucial processes for the ecosystems’ long-term functioning and maintenance and are often closely tied to measurable biophysical functions, which makes them more amenable to indicator-based linkage from EC (Müller and Burkhard 2007, Harrison et al. 2014, Martínez Pastur et al. 2017, Sutherland et al. 2018). Furthermore, ecosystem type-specific patterns emerged. EC indicators in croplands were predominantly linked to Provisioning ES, reflecting the production-orientated framing of agricultural landscapes (Power 2010, Wiggering et al. 2016, Inácio et al. 2025a). Conversely, in wetlands, rivers and lakes, EC indicators were more commonly associated with Cultural ES, consistent with the recreational, aesthetic and identity-based values, emphasised in the Ramsar Convention (Carpenter et al. 2006, Lankia et al. 2023). These findings suggest that EC–ES linkages, reported in literature, are not only neutral reflections of ecological interdependencies, but are also shaped by disciplinary norms, policy priorities and the availability of operational indicators across ecosystem types, as well as CICES v.5.1 classes and sections.

Implementation across CICES v.5.1 sections

Across CICES v.5.1 sections, the EC indicators linked to ES revealed differences in how they are implemented depending on the ES context. When EC indicators were linked to Provisioning ES, they were predominantly chemical and physical state characteristics, based on field measurements. This mirrors long-established practices in agroecosystem monitoring (OECD 2001, Bank et al. 2025, Nicholson Thomas et al. 2025). Most EC–ES relationships were reported as unspecified or positive, echoing a general dependence of Provisioning ES on EC (Harrison et al. 2014). When EC indicators were linked to Regulation & Maintenance ES, a broader spectrum of EC types was involved — structural, chemical, compositional and landscape characteristics. This reflects the inherently multi-dimensional nature of Regulating ES, which depend on system organisation rather than single-state variables (Müller and Burkhard 2007, Kandziora et al. 2013). Most EC-ES relationships were positive, aligning with the theory that higher ecological integrity underpins stronger supply of Regulating ES (Kandziora et al. 2013, Harrison et al. 2014). EC indicators linked to Cultural ES again drew mostly on chemical and physical state characteristics, but these were more frequently transformed into composite indices and benchmarked against policy or safety standards. This may reflect the need for methodological plurality, as well as the need to aggregate biophysical metrics into socially interpretable values for culture, recreation or aesthetics (Fish et al. 2016). Reported EC-ES relations were most frequently mixed, consistent with evidence that cultural appreciation can peak at intermediate landscape openness or heterogeneity (e.g. Kandziora et al. 2013, Dronova 2017). Kandziora et al. (2013) also found that Cultural ES are condition-dependent, but socially mediated, while Graves et al. (2017) highlighted that the use of species richness as a single proxy for landscape aesthetics and the assumption that more species confer greater Cultural ES value are insufficient. Finally, EC indicators that were linked to ES across multiple CICES v.5.1 sections did not cluster around a specific EC type and were more frequently integrated directly into ES assessments (e.g. Laporta et al. 2021, Marull et al. 2021). This is consistent with observations from broader literature, where aggregated indicators are often used when assessments aim to capture multifunctionality rather than a single service pathway (Andreasen et al. 2001, Rowland et al. 2020).

Implementation of directionality by ECT class and ecosystem types

Healthy ecosystems are recognised as fundamental providers of ES, contributing to human well-being (Kandziora et al. 2013, Burkhard et al. 2018b, Haines-Young and Potschin 2018, Grizzetti et al. 2019, Hatziiordanou et al. 2019, Martini et al. 2024, United Nations 2025). In this perspective, the enhancement or degradation of EC influences ES supply. Ideal EC variables offer a univocal, clear and broadly accepted “directional interpretation” (Heink and Kowarik 2010, Czúcz et al. 2021a, La Notte et al. 2022). Our research revealed that EC–ES relations are complex and can have different directions. This is in line with findings from a review on various linkages between Biodiversity and ES (Harrison et al. 2014).

While structural and functional EC indicators more consistently exhibited positive relations to ES, physical and chemical indicators showed more context-dependent or even contrasting relationships. This finding mirrors earlier observations that functional and structural metrics capture ecological capacity more directly, while some chemical or physical indicators tend to reflect stress gradients whose relationship to service delivery may be non-linear or even inverted (Keith et al. 2020). Additionally, some chemical state characteristics relate to pollutant concentration in ecosystems, for example, high sulphate or heavy metal concentrations in rivers and lakes (López Sardi and Larroudé 2020, Lomnicky et al. 2021). For landscape characteristics diverse relations were identified. This variability can be partly attributed to the interdependence of landscape metrics themselves. Many commonly used indicators describe opposite aspects of spatial structure – for example, fragmentation versus cohesion or patch size versus density. As a result, positive relations of certain metrics with ES may mirror negative relations of their counterparts.

EC indicators should be interpreted in light of their ecological pathways rather than assumed to correlate uniformly with ES. Additionally, these findings support the need for a balanced EC indicator set covering all ECT classes (Czúcz et al. 2021a, Bank et al. 2025, United Nations 2025). For croplands, woodland and forest ecosystems, wetlands, coastal and marine ecosystems, we found dominantly positive relationships between EC indicators and ES, which is in line with earlier findings. For croplands and grasslands, high biodiversity or soil condition have been linked to increased pollination, erosion control and carbon storage (Egoh et al. 2009, Bai et al. 2011, Maes et al. 2012, Rendón et al. 2020, Rendón et al. 2022). For woodland and forest, EC indicators, such as species richness and tree cover density, have been positively linked to ES, such as nursery population and habitat maintenance services, recreation, wood provisioning and global climate regulation (Albrich et al. 2018, Martini et al. 2024). Intact wetlands, having a better EC condition than those intensively used, are found to provide more ES (De Groot et al. 2010, Bonn et al. 2016). For freshwater, coastal and marine ecosystems, a positive relationship between EC and the supply of regulating and cultural ES, such as coastal protection and nursery population and habitat maintenance services, is described (Liquete et al. 2016, Grizzetti et al. 2019). Urban ecosystems represent a special case within EC and ES assessments. Within our review, EC indicators applied in urban systems mainly related to structural state and landscape characteristics. Unlike predominantly natural ecosystem types, they function as socio–ecological–technological systems (McPhearson et al. 2016, Pickett et al. 2016, Babí Almenar et al. 2023, Babí Almenar et al. In press). As shown in the European pilot on urban ecosystem accounting (Babí Almenar et al. 2023), applying condition indicators in cities often requires adjusted metrics and reference logics. This might explain why EC–ES relationships in urban systems are reported more frequently as unclear/not specified in our study.

Scientific consensus increasingly holds that EC must form the basis for ES assessment and accounting efforts (Burkhard et al. 2018b, Hein et al. 2020, Rendón et al. 2022, Kokkoris et al. 2024, Martini et al. 2024, Hinsch et al. 2024). In recent years, great progress has been made in the development and use of EC indicators, for example, in relation to the SEEA EA framework (Czúcz et al. 2021a, La Notte et al. 2022, La Notte et al. 2022, Nicholson Thomas et al. 2025, United Nations 2025), helping to move the field from theoretical proposals towards operational, evidence-based applications. In particular,Nicholson Thomas et al. (2025) provide a comprehensive review of over 300 studies to derive a harmonised typology of condition indicators per ecosystem types. Meanwhile, Bank et al. (2025) apply SEEA EA selection criteria, followingCzúcz et al. (2021a), to select the most meaningful and applicable EC indicators for agroecosystem in a case study area in Lower Saxony, Germany. Their methodology underscores the trade-offs between ecological significance and data practicality in indicator selection. Together, these studies illustrate how the indicator selection process is maturing — moving beyond ad hoc choices towards evidence-based, transparent criteria applied across ecosystem types. These advancements can be taken up in future integrated ecosystem assessments (Lange In press), that build upon EC-enabled ES assessment approaches.

Our results indicate that the implementation of EC-enabled ES assessment approaches up to 2022 is most advanced in cropland, grassland and woodland/forest ecosystems. This likely reflects the management- and production-orientation of these ecosystems (Graves et al. 2017, Maes et al. 2018, Bethwell et al. 2021, Bruzón et al. 2023), while our findings suggest that other ecosystem types remain comparatively under-represented in EC-enabled ES assessments. This imbalance suggests that methodological development should prioritise expanding EC-enabled ES assessments beyond the traditionally well-instrumented boundaries.

The types of EC indicators that are directly integrated into ES assessments reveal a strong dominance of chemical and physical state characteristics, suggesting that practitioners in these cases gravitate towards those components of EC that are relatively easy to quantify and to monitor through established environmental reporting frameworks (Grizzetti et al. 2019, Maes et al. 2020b, Vallecillo et al. 2022). We also observed a comparatively high representation of EC indicators integrated into ES assessments, that related to ecosystem extent, ES outputs themselves or stable environmental characteristics. Following the SEEA EA guidance (United Nations et al. 2021) and the conceptual clarifications by Czúcz et al. (2021a) and La Notte et al. (2022), such indicators should be treated with caution when used as proxies for EC. Extent and service-based metrics do not capture the underlying ecological functions or resilience of ecosystems and stable environmental characteristics are not sensitive to change (Czúcz et al. 2021a, Czúcz et al. 2021b). This suggests that a substantial fraction of current EC-enabled ES assessments may still rely on indicators of limited diagnostic value for the actual assessment of EC. A striking pattern is the prevalence of positive relationships between EC indicators (that are related to ECT classes) and ES whenever these relations were explicitly tested. This generally supports the prevailing notion that ecosystems in a better condition provide more ES (Haines-Young and Potschin 2010, Kandziora et al. 2013, Burkhard et al. 2018b, Hatziiordanou et al. 2019, Martini et al. 2024). Despite the progress in EC assessments described above, many EC indicators identified here remain spatially non-explicit, which constrains their utility for fine-scale assessment and limits their integration into spatial planning. These findings indicate that the field continues to face critical gaps related to, for example, ecosystem type coverage, EC indicator diversity and spatial explicitness. Addressing these limitations will be essential for advancing robust, integrated assessments capable of supporting ecosystem-based management and decision-making processes, as well as for a meaningful and fit-for-purpose national and European implementations of SEEA EA.

This section provides insights into the strengths and limitations of this review and suggests areas for further study. This review specifically focuses on the applied links between EC and ES as reported in the assessed literature published in 2018-2022. Our focus has been on documenting how researchers have implemented this linkage in practice — that is, which EC indicators have been connected to which ES and what the nature of that link was. Importantly, we did not evaluate whether these linkages represent the most meaningful, scientifically robust or societally relevant relationships between nature and human well-being. Rather, the links reported here reflect the choices made by researchers, potentially shaped by disciplinary traditions, data availability, methodological feasibility and policy or funding priorities. Consequently, the absence or under-representation of certain EC–ES relationships in literature should not be interpreted as a lack of relevance in ecological or socio-economic terms. Instead, it may point to limitations in available indicators, gaps in understanding, challenges in quantification or mismatches between observation scales. Emerging developments in the direction of integrated ecosystem assessments (Lange In press) – such as guiding frameworks, improved functional indicators, long-term monitoring data or advances in social-ecological modelling – may enable additional EC–ES links to be assessed in future research.

An additional limitation is given by the selected time frame, 2018-2022. Consequently, some more recent developments in the integration of EC indicators into ES assessments have not been captured in our analysis. Recognising this gap, an updated review is currently underway to incorporate the latest data and insights, ensuring that our findings remain relevant and reflect the most recent advancements in the field.

The literature review faced common challenges, such as potential omissions due to limiting the search to English-language scientific publications. The use of diverse terminology and synonyms, particularly for ecosystem concepts, may have impacted the review's comprehensiveness and may have introduced selection bias. The review's reliance on scientific literature excluded potentially valuable grey literature (Yoshida et al. 2024). While precise inclusion criteria enhanced the review's reliability and rigour, bias persisted due to subjective judgement and interpretation. Fleiss’ Kappa results indicated robustness in the screening process, though sample size limitations and team-based reviewing may have affected the analysis. The varied quality of publications, with incomplete reporting and selective data, affected synthesis and conclusions. The review’s indicator-based approach risked over-representing methodological procedures in studies assessing multiple indicators.