Ecosystem accounts measure the contribution of ecosystems to human well-being and the economy (Obst et al. 2015). This demands regular and standardised stocktaking activities that record the extent of the ecosystems, their condition and the services they provide to society, which need to be supported by consistent guidelines (Obst et al. 2013, Polasky et al. 2015). This is the objective of the Experimental Ecosystem Accounting (EEA) protocol of the System of Environmental-Economic Accounting (SEEA) developed by the United Nations (UN). The SEEA EEA defines an integrated accounting framework for organising biophysical data, measuring ecosystem services, tracking changes in ecosystem assets and linking this information to economic and other human activity. This framework for ecosystem accounts was formally adopted by the UN in March 2013 (United Nations (UN) et al. 2014) and technical recommendations are available to help set up and quantify accounts in a standardised way (United Nations (UN) 2019). These SEEA EEA guidelines are presently under revision. The aim is to adopt a revised standard for ecosystem accounting in 2021. This paper is part of this revision process and contributes in particular to an updated set of recommendations for reporting ecosystem condition accounts.

SEEA EEA defines ecosystem condition as the overall quality of an ecosystem asset in terms of its characteristics (United Nations (UN) et al. 2014, paragraph 2.35; Keith et al. 2020). Condition together with extent is expected to describe the state of an ecosystem asset (see Glossary in United Nations (UN) et al. 2014 and Maes et al. 2014). Ecosystem characteristics are the system properties of the ecosystem and its major biotic and abiotic components. This term is intended to encompass all of the various perspectives taken to describe the long term ‘average behaviour’ of an ecosystem. On the other hand, variables, indicators and indices are concrete quantitative metrics with precise definitions and measurement instructions, which can be used to describe these characteristics. In the context of this paper, any quantitative metric reflecting a phenomenon of interest can be seen as a variable. Indicators are variables with a strong direct normative interpretation (i.e. distinguishing “good” from “bad”) for policy and decisions, which are preferably rescaled to a standard dimensionless scale (e.g. ranging from 0: bad to 1: good). An index is an aggregated indicator constructed from other indicators, which thus represents relatively-broad aspects of the studied system in a single number. Finally, the term metric is used to refer to any of the previous three categories (see also Keith et al. 2020).

Ecosystems have many apparent and hidden characteristics, which are influenced by each other in complex ways. Accordingly, ecosystem condition is inherently multidimensional, many metrics being needed to give a comprehensive characterisation of the condition of an ecosystem (Schoolmaster et al. 2012). This multidimensionality has to be reflected in the structure of condition accounts, which need to handle a potentially high number of metrics in a consistent way (Keith et al. 2020). This paper focuses on the typology (classification) used for grouping the various condition characteristics and metrics in the account, which ensures that the condition accounts created in different countries for different ecosystem types would be compatible and comparable. While classifying ecosystem types has a century-long history (Keith et al. 2020b) and ecosystem service classifications were also intensively studied and standardised in the last two decades (La Notte et al. 2017), the complex concept of ecosystem condition has attracted far less ‘taxonomical’ attention so far. Nevertheless, an operative SEEA EEA requires that each element in the chain comprising the path from ecosystems to final beneficiaries has its own classification (La Notte and Rhodes 2020). In this paper, we lay out the foundations of a simple classification system, the SEEA EEA Ecosystem Condition Typology (SEEA ECT), which can be applied as a standardised reporting structure for ecosystem condition characteristics, variables, indicators and indices, across multiple countries and biomes, all over the world.

A typology or classification (system) is the operation of distributing objects into classes or groups that are less numerous than the original objects. This operation is very broadly and frequently used in science, as it can create an order amongst the “chaotic and muddled multiplicities” of life and thus can reduce the complexity of the problems (Parrochia 2019). Classifications are therefore the essence of accounting systems. Classifications need to be exhaustive and mutually exclusive : classes should not overlap and their union should restore the divided concept. As each division (class) can be further subdivided, classifications can also be hierarchical.

An ecosystem condition typology is a hierarchical classification for the metrics (variables and indicators) used to describe the condition of the ecosystems. Nevertheless, as these metrics are supposed to reflect the underlying reality of the ecosystem, the condition typology can also be applied for ecosystem characteristics, thus defining the relevant “information structure” of the ecosystem itself. This way, the typology for ecosystem condition can create a meaningful order for the accounting tables, which can have multiple advantages:

• it can help to establish a common language and a shared understanding;
• it can make different studies (assessments, countries etc.) more comparable;
• it can be used as a structure for aggregation; and
• it can be used as a template for the selection of variables and indicators.

As also emphasised by SEEA EEA (United Nations (UN) et al. 2012), different ecosystem types have different relevant characteristics, which should be described by different indicators. Nevertheless, in order to facilitate communication, as well as comparisons and aggregation across ecosystem types, an ecosystem condition typology should be universal at least at the top levels (i.e. it is expected to be relevant for all major ecosystem types). On the other hand, the typology also needs to be able to host ecosystem type specific metrics at the lower levels.

There are already several classifications in scientific literature, which aim to provide a meaningful and comprehensive reporting structure for ecosystem characteristics, variables and indicators. The majority of structured lists of ecosystem characteristics/indicators, which can be found in scientific papers, can, in principle, be applied or adapted to the concept of ecosystem condition. This includes conceptual papers (e.g. Müller 2005, Kandziora et al. 2013, Haase et al. 2018); systematic reviews (e.g. Smith et al. 2017, Rendon et al. 2019); or policy reports (e.g. Maes et al. 2018). Existing condition accounts also often include some kind of grouping structure (e.g. White et al. 2015). Nevertheless, most of these papers do not give any description or justification for the groupings they use (e.g. Flint et al. 2017, Maes et al. 2018); they are just applied as an enumeration structure, sometimes in a more or less inconsistent way. Many of the rest are just reusing classifications from earlier publications and/or lack generality by focusing only on a specific ecosystem type or biome (e.g. Schwarz et al. 2017, Okey 2018, Flint et al. 2017).

For the purposes of this study, we have identified three “prototype” classifications which reflect a balance between conceptual clarity and practical usefulness, and which have become influential in the community of ecological/enviromental sciences:

• the essential ecosystem characteristics (EEC, published by Harwell et al. (1999) based on lessons from the US Man and Biosphere programme);
• the ecological integrity typology (EI, proposed by Andreasen et al. (2001) and further elaborated by Tierney et al. (2009) and Faber-Langendoen et al. (2012));
• the classification of essential biodiversity variables (EBV, outlined by Pereira et al. (2013) and further elaborated by Kissling et al. (2017) and others).

We used these three “prototype” classifications as starting points for designing our proposal for an operative condition classification for the SEEA EEA. In doing so, we sought the largest common denominator between these prototypes, taking into account both the conceptual background of SEEA EEA and the practicalities of the existing condition accounts (Keith et al. 2020, Maes et al. 2020).

We propose the hierarchical classification shown in to be used as the SEEA EEA Ecosystem Condition Typology (SEEA ECT). This classification contains three major groups (abiotic, biotic and landscape-level characteristics) and six classes nested in these groups. Table 1 shows the proposed six classes. Crosswalks to relevant classifications, including the three prototypes, can be found in Suppl. material 1. In Fig. 1 and Suppl. material 2, we map the condition indicators used in the already existing condition accounts to the SEEA ECT typology, based on the work of Maes et al. (2020).

Figure 1.  

The prevalence of SEEA ECT classes in existing condition accounts. The length of the bars shows the number of "case studies" (from Maes et al. 2020) which apply condition indicators that fall into the studied type (see also Suppl. material 2 for more details).

The first group of the SEEA ECT typology embraces those abiotic elements of the physico-chemical environment which are in direct interaction with the biosphere. Such abiotic elements have traditionally been considered as a component (compartment) of the ecosystems and they are also represented with three (out of seven) main classes of the EEC typology (Harwell et al. 1999). The SEEA ECT class physical state characteristics (A1) hosts the physical descriptors of the abiotic components of the ecosystem (soil, water, air…). Chemical state characteristics (A2), on the other hand, include metrics related to the chemical composition of the abiotic ecosystem compartments. Physical and chemical characteristics may play a particularly important role for freshwater and marine ecosystems. Variables should describe the state (e.g. “stocks” of pollutants), rather than the flows (emission of pollutants). This way, these SEEA ECT classes can accommodate pressures in a way that is compatible with accounting. The majority of the case studies (Maes et al. 2020, Fig. 1) include metrics of chemical state, while identifying relevant physical state characteristics has proven much more difficult (3 of the 23 case studies, see Fig. 1 and Suppl. material 2).

The SEEA ECT group of biotic ecosystem characteristics comprises characteristics that are typically associated with ecosystems and biodiversity. Biotic characteristics are central in all previous condition typologies and the EBV typology (Pereira et al. 2013) entirely focuses on characteristics that belong here. To subdivide this large group of characteristics, we follow Andreasen et al. (2001) and a strong ecological tradition (e.g. Noss 1990) and distinguish composition, structure and function.

The SEEA ECT class compositional state characteristics (B1) comprises a broad range of ‘typical’ biodiversity variables, describing the composition of ecological communities from a biodiversity perspective. Characteristics in this class are typically derived from species data, like the presence/abundance of a species or species group or the diversity of species groups at a given location and time. From a location-based perspective, the distribution of a species is based on species composition (local presence). Compositional metrics can characterise the local “biodiversity quality” of sampling sites (Feest 2006) through the presence/absence or abundance of individual species, taxonomic groups (birds, butterflies) or non-taxonomic guilds (e.g. soil invertebrates, macrozoobenthos). In fact, the most frequent SEEA ECT class in existing condition accounts are compositional state characteristics (Fig. 1).

Nevertheless, not all relevant characteristics of ecosystems are derived from species data. We distinguish two further important classes: structural state characteristics (B2) which are aggregate properties (e.g. mass, density) of the whole ecosystem or its main biotic compartments, while functional state characteristics (B3) include summary statistics (e.g. frequency, intensity) of the biological, chemical and physical interactions between the ecosystem compartments (cf. Boyd and Banzhaf 2007). This interpretation of functions might seem restrictive, given that many authors use the term “ecosystem function” in an utilitarian way, emphasising the “functioning” directly leading to ecosystem services (e.g. Mace et al. 2012, cf. Heink and Jax 2019). In SEEA EEA, however, most of this service-orientated “functioning” should probably be considered under the ecosystem service accounts, which leaves the “internal interactions between the ecosystem components” as the definition of function in the context of the condition accounts. Structural and functional characteristics are relatively rarely quantified in the case studies (Fig. 1).

The last SEEA ECT group, landscape-level characteristics (C1) has a single class covering the characteristics of entire landscapes (or waterscapes, seascapes) consisting of multiple ecosystem types. This involves landscape metrics (e.g. diversity, connectivity or fragmentation), which can describe the integrity of landscapes at 'local' landscape scales (~ 10-1000 km², Pagella and Sinclair 2014). It is important to note that, while such metrics characterise landscapes, the structure of condition accounts expects metrics to be linked to concrete ecosystem types. This apparent conflict can be resolved by a relatively simple and straightforward accounting decision: the landscape-level metrics (calculated, for example, with a moving window) should be assigned to the focal ecosystem type. In other words, the ‘landscape diversity’ of a forest should be interpreted as the diversity of the landscape in which the forest is situated. The importance of landscape-level metrics for condition accounting practitioners can be seen from the relatively-high prevalence of this class in the case studies (Fig. 1).

To ensure the mutual exclusivity of the classification system, it is important that all variables can be linked unequivocally to a single SEEA ECT class. This needs well-defined classes, supported by definitions that carefully eliminate overlaps and borderline cases in a consistent way. Nevertheless, the short definitions of the classes, as outlined above, allow for several potential overlaps (Table 2). Several key terms (e.g. structure or function) are actually ‘boundary objects’ (Star and Griesemer 1989, Steger et al. 2018) that are used and interpreted in different ways by different researcher communities. For example, in the case of freshwater and marine ecosystems, community structure and composition are sometimes handled as synonyms (e.g. Dudgeon 2010).

In principle, most of the borderline cases could be resolved in any direction without doing harm to the integrity of the condition accounts. Ideally, these decisions should be consistent and if, for example, soil organic carbon is considered to be chemical (A2) in a Chinese forest, then it should be classified similarly in US farmlands, too. Ensuring this level of consistency (e.g. through detailed guidelines in a SEEA EEA annex) might seem a daunting task. Nevertheless, the range of potential condition variables is highly restricted by data availability and conceptual considerations (selection criteria, see Keith et al. (2020) and Czúcz et al. (in press)), an overview of the relevant cases that occur in practice might already be feasible (Table 2).

In developing accounts, it is also common practice to reuse data originally collected for other policies and reporting schemes. Such data are often available in a highly aggregated format, combining data points from several ecosystem accounting (spatial aggregation) areas or SEEA ECT classes (thematic aggregation; see “pre-aggregated indices” in Fig. 1). In such cases, the original data points would be much more useful than the aggregated index (Broszeit et al. 2017, Moriarty et al. 2018). While the efforts to reuse what is already available are understandable, the ideal practice would be to add all relevant characteristics individually and perform the appropriate aggregations within the condition account itself. One of the main functions of the SEEA ECT typology is to provide a standardised aggregation scheme that can be meaningfully used across ecosystem types, countries and continents. Such ‘overarching’ pre-aggregated indices might violate this function.

Not all environmental variables are appropriate for measuring ecosystem condition (Keith et al. 2020). Consequently, the SEEA ECT typology does not aim to cover all policy-relevant environmental metrics, it just aims to be exhaustive for the variables that are valid to use in the context of a SEEA EEA condition account. Nevertheless, the scarcity of available data sometimes necessitates compromises or suboptimal choices. Even the case studies (see Maes et al. 2020; Suppl. material 2) contain a few variables that cannot be hosted in the SEEA ECT classes as presented above. Such variables violate aspects of SEEA EEA condition accounts (Keith et al. 2020, Czúcz et al. (in press)) in some way, often because the creators of such accounts used them as ‘proxies’ for something that they could not measure. In the next few paragraphs, we give a short overview of the main types of such ancillary data found in the existing ecosystem condition accounts tested by Maes et al. (2020) (see also Fig. 1 and Suppl. material 2).

• Accessibility (six case studies, see Suppl. material 2): Distance from roads or human population centres appears in a high number of condition accounts (typically from the UK, for example, White et al. (2015)). Nevertheless, accessibility is not a characteristic of the ecosystem per se and it can be better conceptualised as a factor behind ecosystem service demand. Furthermore, if an ecosystem becomes more accessible (e.g. there is a new highway), would we like to see this as a condition improvement or degradation?

• Protected areas (five case studies): Administrative land designations (including the status and degree of nature protection) do not reflect the state of an area, but rather a human response to degradation or perceived land value. Using protection status as an ecosystem condition indicator will also compromise the ability of the condition accounts to evaluate the efficiency of protection measures.

• Pressures (four case studies): Pressures (e.g. pollutant fluxes) do not reflect the state of the system, they are rather external forces influencing future state. They are still popular in some ecosystem condition accounts, as they are typically easier to measure than the underlying state variables that are affected by the pressure. Even the SEEA EEA Technical Recommendations suggest that pressures can be considered a useful surrogate (United Nations (UN) 2019: 4.9), as long as the relationship between the two is well understood and justified (Bland et al. 2018). Using pressures as an ecosystem condition indicator will, nevertheless, definitely compromise the ability of the condition accounts to evaluate the impact of those pressures (Broszeit et al. 2017). Furthermore, pressures are already covered in a different SEEA account (in the SEEA Central Framework, United Nations (UN) et al. 2012), so including them in the condition accounts would be a duplication.

• Natural resource management (two case studies): Similar to pressures, human management (e.g. grazing, felling, fishing, agriculture...) is also sometimes considered in the context of ecosystem condition (Maes et al. 2018). Human management, however, should not generally be seen as an internal part of the studied ecosystems, even if it influences a broad range of services (Santos-Martín et al. 2019). Including management in an ecosystem condition account will also compromise its ability to evaluate the impact of management.

• Certificates, audits (two case studies): Certificates (e.g. the ‘blue flag’ certificate for EU beaches or the ‘green flag’ certificate for UK urban parks; Office for National Statistics 2016, Office for National Statistics 2018) and audits are market products, which are only available for locations with solvent demand. Hence, the absence of a certificate does not mean that the location in question would not meet the necessary qualifications. In principle, certificates are based on measured data. If such data are available, the ideal practice would be to encode all relevant characteristics individually in the condition account and create appropriate aggregated indices within the account (similarly to the recommended treatment of pre-aggregated indices).

• Stable environmental characteristics: Climatic and other environmental variables are occasionally also proposed for inclusion in condition accounts (e.g. de Jong et al. 2016). In principle, it is not useful to include very slowly changing variables into accounting tables (e.g. climate, slope, aspect or geology), even if climate has started to change recently. Furthermore, these variables are largely external to the ecosystems. Climate is also already well-covered by a number of international conventions and data repositories, so it might be reasonable to keep the ecosystem condition accounts for ‘more ecological’ variables.

If, despite all the issues highlighted above, such ancillary data are considered as proxies instead of measured variables in an ecosystem condition account, then this should be done in a clear and transparent way (i.e. it should be clearly documented and justified that metric X is considered to be a proxy for characteristic Y that we could not measure). Such proxies should be assigned to the category where the original variable would normally belong. Nevertheless, to ensure the consistency of the condition accounts, the inclusion of ancillary data should preferably be avoided. This does not mean that these data types would be irrelevant or worthless — on the contrary, most of these data are indispensable for a proper ecosystem assessment to be done. For example, climate, geology, management or accessibility can be key input data for ecosystem service models (see, for example, Zulian et al. 2013, Vallecillo et al. 2019), so they might need to be collected and handled anyway in the context of SEEA EEA. They should just preferably be kept out of the condition accounts. Beyond the diverse types of ancillary data listed above, there are two further important groups of data that should not be considered as condition variables: ecosystem extent and services. Extent and services have their dedicated accounts in the context of SEEA EEA and the different accounts should not repeat each other, but should convey different pieces of information.

In line with the principles discussed by Keith et al. (2020), the selection of ecosystem condition metrics should be an iterative process, reflecting a good ecological understanding, as well as practical considerations of data availability. The proposed SEEA ECT classification can play a key role in this process by highlighting major data gaps (e.g. if there are no appropriate metrics available for a specific class). This also makes it possible to implement the selection process, by proposing simple rules (like “1-3 indicators from each SEEA ECT class”). Nevertheless, as the current examples demonstrate (Fig. 1), for some ECT classes, it might be more difficult to identify/construct appropriate metrics than for other classes. The most universally-accessible SEEA ECT types are compositional state (B1), chemical condition (A2) and landscape (C1) characteristics, representing each of the three main SEEA ECT groups (abiotic, biotic and landscape-level characteristics). Accordingly, an alternative option would be to formulate the ‘minimum requirements’ at the level of ECT groups which would give more flexibility to practitioners of ecosystem condition accounts.

The SEEA ECT classes only provide a rough thematic structure for the condition accounts. To make the SEEA ECT classification more responsive to user needs, future studies should identify more concrete (families of) indicators, taking into consideration all relevant criteria (Keith et al. 2020) including global data availability. These indicator families (or “ECT subclasses”) would also create an opportunity to improve the level of standardisation for the accounts (Polasky et al. 2015). In principle, they can be specified concretely (what characteristic to measure, how to measure it), so that the variables and indicators implemented for subclasses would be comparable across countries, continents and -- for crosscutting characteristics -- also ecosystem types. Nevertheless, this would also mean that, at the level of subclasses, the SEEA ECT classification will not be exhaustive any more: the list of SEEA ECT subclasses should rather be seen as the SEEA EEA recommendation on the concrete metrics to measure, than as a comprehensive classification for all possible metrics. The broad SEEA ECT classes can be considered universal across all biomes and ecosystem types, but the lower hierarchical levels of the SEEA ECT classification will necessarily have to be ecosystem type specific. No single set of metrics can suffice for all ecosystem types (Andreasen et al. 2001).

The standardisation of ecosystem accounts (Polasky et al. 2015, Steger et al. 2018) is not possible without a comprehensive and consistent (exhaustive and mutually exclusive) classification system of ecosystem characteristics and metrics. We think that the SEEA ECT typology offers a good compromise between conceptual clarity and practical usefulness. Its construction is closely linked to ecological theory (Harwell et al. 1999, Andreasen et al. 2001, Pereira et al. 2013), while the extendible hierarchical system makes it flexible and adaptable for all biomes and ecosystem types. This hierarchical structure also creates the foundations for aggregation, which is consistent across countries and biomes, thus making them comparable, which is necessary to maximise the utility of the SEEA EEA efforts for the UN community.

The System of Environmental Economic-Accounting – Experimental Ecosystem Accounting (SEEA EEA) is going through a revision process between 2018 and 2021. The revised SEEA EEA is expected to be adopted by the United Nations Statistical Commission in March 2021. This article is based on a discussion paper that contributed to the revision process. The views and opinions expressed in this article are those of the authors and do not necessarily reflect the official position of the SEEA EEA. Neither do these views reflect an official position of the European Commission.