Since sustainability was introduced as a term at the global high level by WCED (1997), it has become a widely used objective guiding the actions of both governments, industries, and consumers (Portney, 2015; UN, 2015), although also being contested and problematized (Aarset et al., 2020). Most common frameworks of sustainability identify three sustainability dimensions: economic, environmental, and social (e.g., Bracco et al., 2019; Eustachio et al., 2019). Many frameworks also include a governance dimension covering institutional sustainability. Sustainability is a complex phenomenon that is difficult to observe directly (FAO, 1999; UNDESA, 2007), and many indicator-sets and abstract indicators have been constructed to capture it (Singh et al., 2012). Such indicator-sets can contain many indicators, like those suggested for fisheries (FAO, 1999; Anderson et al., 2015) or for aquaculture (Valenti et al., 2018). Certification schemes for sustainable aquaculture can be seen as a type of abstract indicator, based on several indirect sub-indicators (Osmundsen et al., 2020a). It seems that the social dimension of sustainability has been the hardest to grasp (Vifell and Soneryd, 2012; Hicks et al., 2016; Alexander et al., 2020).

Indicators in general provide information on the status or development of a phenomenon, usually one that is difficult or impractical to observe directly (Bracco et al., 2019). Some indicators give precise information on the phenomenon in question, like an indicator light telling that the temperature in a freezer is above a chosen threshold. For more complex phenomena, an indicator may only be able to indicate the status or development. Indicators can be qualitative or quantitative, and they can indicate a change, a trend, or status (Bracco et al., 2019). For an indicator to be able to say something on status, reference values for the indicator must be defined. For sustainability indicators, both the choice of reference values for indicators and the selection of indicators per se are inherently normative and political choices (Levett, 1998).

Bracco et al. (2019) distinguishes between direct, indirect, and proxy indicators, depending on how precisely an indicator relates to the underlying phenomena of interest. Direct indicators can be used when the phenomenon of interest is rather straightforward. Indirect or proxy indicators are useful when the phenomenon is abstract and cannot be measured directly or it would require complex and resource demanding efforts to measure it well. An indirect indicator does not directly represent the actual phenomenon of interest, but other phenomena related to it. If the phenomenon is escaped salmon from salmon farming, the estimated number of escaped salmon may be adequate as an indicator. If the phenomenon is the quality of aquaculture governance, indirect, or abstract indicators is required. Often a set of indirect indicators is deemed necessary to sufficiently illuminate the actual phenomenon of interest. Abstract indicators are typically constructed from a set of indirect indicators, calculating weighted or unweighted averages or similar, and mathematically normalizing to end up with values for the abstract indicator between 0 and 1 or 0 and 100, like for example in the Ocean Health Index (Halpern et al., 2012).

Choosing which sustainability themes and indicators to present is to some extent a value-laden and political choice (Levett, 1998), even though science can provide guidance, not least on the relevance and validity of specific indicators for various themes. Issues and themes that are used in actual regulation will obviously be relevant. But also themes that come up in public discourses and political discussions must be considered, as well as indicators identified from scientific work. What is considered important and relevant themes will always be constant developing, driven by changes in production technology and industry structure, culture, demography, and general economic activities in society, and also in climate, biology, and ecology. Hence, a set of indicators must also evolve to remain relevant. Another aspect is the balance of indicators between the major sustainability dimensions. Major aquaculture sustainability certification schemes have many indicators for environmental aspects, while social and economic aspects are less covered (Osmundsen et al., 2020a, Alexander et al., 2020). Many sustainability themes are however connected across the main sustainability dimensions, so individual indicators can be relevant for several dimensions.

Given the variation and complexity of choosing and creating indicators, criteria to guide the selection of indicators have been proposed, including for sustainability indicators (e.g., FAO, 1999; UNDESA, 2007; Brown, 2009). Central general criteria for indicators include scientific validity, data availability, robustness, precision, practical feasibility, cost efficiency, ability to communicate information, understandable, acceptance by stakeholders, and relevance for policy priorities.

The selection of themes and indicators can benefit from the involvement of relevant stakeholders (FAO, 1999; Consensus, 2006; Brown, 2009). Their first-hand knowledge can help in achieving scientific validity, robustness and precision as well as general acceptance of indicators. The general public, which can be seen as people without or with limited statistical knowledge, can further contribute to assessing how understandable the indicators are and how they are communicated (Eurostat, 2017, p. 22) Given that aquaculture can be a contentious issue, where available information can shape decisions and affect stakeholders, having an expert-led process can contribute to the trustworthiness of information and assessments (Servaes et al., 2012).

The different types of indicators can have different purposes. Fundamentally, indicators allow for comparisons, either across geography, time, units (companies, activities, etc.), or with reference values. Direct indicators can lead to immediate action, especially if clear reference values have been defined. Indirect indicators will inform on different aspects of a phenomenon of interest and can thus help guide the selection and priority of actions. Abstract indicators can be used to qualify or rank actors and may thus guide the selection among a set of actors. This could be the authorities deciding who should get an aquaculture license or consumers deciding who to buy from. It can also motivate those that score poorly to do something about the situation. But unlike the direct and indirect indicators, the abstract indicators themselves cannot guide what to do if the indicator score is too low. For that, one needs to analyse the underlying data and indirect indicators.

When choosing reference values for indicators there are several principal options (UNAIDS, 2010). One option is to choose a historical situation, a baseline value, as the reference point. This will show historical development. Historical trends could also be used as reference, or some measure of stakeholders' expectations. Comparison with similar activities elsewhere and how they develop, or using expert opinions and research findings are other options. The challenge with assessing sustainability is that what is ultimately to be considered is not a historical development, but how the future will be (Stiglitz et al., 2010, p. 61). This is obviously difficult.

Many types of sustainability assessments are described in the literature (e.g., Singh et al., 2012). They can be said to be of four main types: (1) Dashboards of indicators, (2) Composite indices, (3) Footprints of resource use, and (4) “Sustainability-adjusted” measures of welfare and wealth (Stiglitz et al., 2010; GGKP, 2016; Bracco et al., 2019). Dashboards present sets of indicators that directly or indirectly relate to sustainability, without ranking or weighting them. Creating a broad set of indicators is a necessary first step in any analysis of sustainability, since sustainability is complex by nature and a list of potentially relevant variables must be established (Stiglitz et al., 2010, p. 63). The dashboard framework makes a broad assessment across all dimensions of sustainability possible. However, the link between the value of an indicators and sustainability may not always be clear (Stiglitz et al., 2010, p. 63), and it can be difficult to compare situations when individual indicators vary in amplitude and direction of change (Bracco et al., 2019).

Composite indices can help such comparisons, but weighting, aggregation, and normalization, which is required to go from several sustainability indicators to one, requires implicit value-judgements, and cannot always be scientific (Böhringer and Jochem, 2007; Bracco et al., 2019). If a clear outcome can be defined and measured, composite indicators can be calculated by scientific methods, but for sustainability this is difficult (Nardo et al., 2005). While the authors behind composite indices of sustainability often are very explicit on how the weighting is done, the normative foundations or implications are rarely justified or made explicit (Stiglitz et al., 2010, p. 65). Also, a difference in score on a composite index between two entities does not give information on why this has come about—rather, it is like an invitation to study the underlying components closely (Stiglitz et al., 2010, p. 65), thus returning to a dashboard of indicators. Both dashboards of indicators and composite indices on sustainability are criticized by Stiglitz et al. (2010) for lacking a well-defined notion of what sustainability means.

The “footprints”-approach to measure sustainability is about estimating the over-use, under-investment in, or pressure on resources (Stiglitz et al., 2010, p. 67). Such indicators tend to consider the use or flow of one or a few resources that affect stocks of resources that future generations' welfare will depend on, like the stock of greenhouse gases in the atmosphere. For broader footprint-estimates, one needs to find an appropriate metric and ways to aggregate, which gives the same sort of challenges as in making other composite sustainability indices (Stiglitz et al., 2010, p. 67). The conceptual link between the estimated footprint and sustainability is however usually clearer than with the two types of assessments explained above. The same goes for the fourth type of sustainability assessment; measures of GDP (gross domestic product) and wealth that try to systematically correct for elements that matter for sustainability and which are not included in standard GDP calculations (Stiglitz et al., 2010, p. 65). While footprints can be calculated for activities by individuals or industries, the adjusted GDP estimates typically estimate status for countries or even larger entities, and are thus less relevant for indicating the sustainability of aquaculture as an activity.

How should sustainability assessments aiming to inform a general public be different from when they are aimed at industry actors, authorities, researchers, or other experts? The Bellagio Sustainability Assessment and Measurement Principles (Bellagio STAMP) (Pintér et al., 2012) give some guidance for effective communication: One should use clear and plain language, present information in a fair and objective way that helps to build trust, use innovative visual tools and graphics to aid interpretation and tell a story, and make data available in as much detail as is reliable and practicable. That the presentation is considered to be without bias is seen as important for building trust, as is engagement early on with users of an assessment (Pintér et al., 2012).

Aquaculture is a very diverse activity globally (Garlock et al., 2020), and its development will have impact on food nutrition, human well-being, and global environmental health (Gephart et al., 2020). Monitoring its development along all dimensions of sustainability will be crucial to understand and govern that development (Krause et al., 2015). Sustainability comparisons across the diversity of aquaculture technologies and species, geographical, political, and socio-economic contexts can be relevant and useful. There are even sustainability comparisons for animal protein providers across meat and fish (Coller FAIRR, 2020). Yet, it is obvious that sustainability assessments also need to be tailored to more specific situations and contexts to give information relevant for national and regional challenges.

There have been many attempts at making comprehensive indicator sets for the sustainability of aquaculture. A workshop in 2006 proposed 78 indicators on sustainability for aquaculture in Europe, across nine different themes (Consensus, 2006). Many indicators are included in various certification schemes for aquaculture. Osmundsen et al. (2020a) mapped the indicators of eight widely used certification schemes and found that they contained altogether 1,916 indicators, ranging from 52 to 468 indicators per certification scheme. They also found that environmental indicators dominated, and that other dimensions were poorly covered.

It is quite common that companies in a value chain demand that those they buy products or services from are certified according to specific certification schemes (Gutierrez and Thornton, 2014). For the consumers, the large number of certification schemes can be confusing (Gutierrez and Thornton, 2014). Despite this, certification schemes are quite commonly used. However, seafood consumers are not the only intended target group for the aquaculture industry's use of certification schemes. Persons living in the vicinity of aquaculture production may have concern about the environmental, economic, and social impacts of aquaculture, and some certification schemes target such aspects (Aas et al., 2019; Osmundsen et al., 2020a). This may be one way for the industry to strengthen their social license to operate (Kelly et al., 2017; Mather and Fanning, 2019; Sinner et al., 2020). Even though some social and socio-economic sustainability indicators have been proposed and exist, this is the sustainability dimension that seems to have the poorest coverage generally (FAO, 2009; Alexander et al., 2020; Krause et al., 2020). For doing trade-offs between different sustainability aspects, it could help if data on the different effects of aquaculture were comparable (Zheng et al., 2009), like economic data (Knowler, 2008), but such data seems to be largely missing (Mikkelsen et al., 2020). Some indicators are used directly in the authorities' management of aquaculture, for example in in Norway (Osmundsen et al., 2020b, NFD, 2015), but there is also clear criticism of authorities in some countries being too slow to incorporate indicators in management (Milewski and Smith, 2019). The relative fuzziness of the sustainability term has also led to what some authors call a power struggle between authorities and industry actors over how it should be interpreted and have operational consequences (Aarset et al., 2020).

As this paper mainly is about making sustainability indicators for Norwegian aquaculture, which again is dominated by salmon farming, a brief introduction of that is warranted. Since the start of the Norwegian salmon farming industry in the late 1960s, it has developed into a significant industry, where Norway is now the world's biggest exporter of farmed salmon, and companies originating from Norway are also major players in the other salmon-producing countries, including Chile, Scotland, and Canada (Hersoug et al., 2019). In Norway, there are around 1,000 localities for salmon farming along nearly the full length of the long Norwegian coast, of which around 60% have production at any given time. The industry provides jobs and income (Johansen et al., 2019), especially in rural areas (Johnsen et al., 2020) where decreasing employment and population numbers in general have been observed (Iversen et al., 2020a). The impacts of salmon farming have varied over the years, as have which impacts are in focus in the public debate, including environmental (Taranger et al., 2015; Olaussen, 2018), economic, and social impacts (Hersoug et al., 2021).

The dominating open net pen concept has proven economically very successful, but production costs have increased sharply the latter years, due to increased feed prices and costs for prevention, treatment, and mortality associated with pathogens including salmon lice (Iversen et al., 2020b). Consolidation of the industry and its ownership, as well as changes in production technology and subsequent changes in labor use have led to an increasingly skewed distribution of benefits from the industry between different municipalities with fish farms (Hersoug et al., 2021). This has led to challenges with the industry's social legitimacy (Hersoug et al., 2021).

The public management of aquaculture in Norway has emphasized different aspects of sustainability over the years, but even from the very first (temporary) Aquaculture Act of 1973 both environmental, economic, and social sustainability concerns were included. With the Aquaculture Act of 1991 the term sustainable development (“bærekraftig utvikling”) were explicitly put in its objective (§1), and has remained there through later revisions of the legislation (Mikkelsen et al., 2018). Regular monitoring and reporting are required regarding i.a. the parasitic salmon lice, diseases, biomass, and pollution situation below pens, and the authorities can force reduced or closed production or other restrictive measures due to environmental or social concern (Mikkelsen et al., 2018). New salmon farming licenses have been issued in rounds that have emphasized as diverse priorities as rural development, industrial development, ownership, fish health, the environment, and industry's willingness to pay for new capacity (Hersoug et al., 2019). A government white paper in 2015 launched the ambition to devise a system of “predictable and environmentally sustainable growth in Norwegian salmon and trout farming” (NFD, 2015). This ended in the so-called “traffic light system,” established in 2017; an assessment of the extra induced mortality on wild salmon and trout stocks due to salmon lice originating from the salmon farms dictate in which regions salmon production capacity can increase (green light), must be reduced (red light), or stays the same (orange light), and companies' willingness to pay for more production capacity decide which companies actually get any increased capacity (Hersoug et al., 2021). The problems with social legitimacy have also led the authorities to introduce two different taxation/redistribution schemes that shall ensure economic benefits for all municipalities and counties that have salmon farms (Hersoug et al., 2021).

The process to establish the web-portal has to a large degree involved answering the same questions that are posed in this article, and thus that process largely constitutes the method to answer the questions. One important function has been to organize the input and involvement of different actors, experts, stakeholders, and the general public, with their competence and perspectives, into the considerations necessary to choose and present indicators so that relevance and trustworthiness are achieved. The overall process took several years, and went through a pre-project, phase one of the main project, and is now (March, 2021) in phase two of the main project. Table 1 gives an overview of the individual projects and major elements in them. The projects have been financed and owned by the Norwegian Seafood Research Fund (FHF). This section also specifically presents a survey to the general public and the method for theme and indicator selection.

The aims of the pre-project were to (i) identify criteria for holistic sustainability indicators to be included in the portal covering environmental, economic and social sustainability, (ii) consider data availability and data sources and efficient methods for collection and processing of data, and (iii) suggest set-ups for presentation (Andreassen et al., 2016). The process to fulfill these aims included a review of scientific and government literature, certification schemes, and sustainability reports from aquaculture companies, supplemented by a workshop with relevant stakeholders and representatives from research institutions. The result was an overview of possible criteria, indicators and data sources, ending up with 26 potential indicators across 10 “focus areas” (Andreassen et al., 2016). It was recommended to have a balance between indicators covering environmental, economic, and social sustainability in a web portal. This also became an important point when the Norwegian Seafood Research Fund decided to fund the main project.

In the first phase of the main project the aim was to get the portal established and openly available. This phase should also define the target group. In the second phase it was to run and further develop the portal. The core project group was set up with members from research institutions Nofima (project lead) and SINTEF Ocean, to ensure competence and also independence from the aquaculture industry. BarentsWatch was also member of the core project group, as partner responsible for web publication. BarentsWatch has an open information system on oceans and marine use at www.barentswatch.no, and has 10 ministries and 29 directorates and research institutions as partners. By establishing the aquaculture sustainability web-portal on the Barentswatch platform, one hoped for efficient technical production and maintenance of the portal, and that it would find users among those that already used the Barentswatch platform.

Two additional groups were formed to support the development of the portal in the first phase. The first was a quality assurance group with members from Norwegian universities and research institutes, an environmental NGO (Bellona), and from a consultancy [Teigen Consulting, Institute for Policy Analysis and Development (INPAD)]. The second was a so-called steering group appointed by the project owner and funder, with four members from aquaculture companies. The steering group should, as stated in their mandate from FHF, contribute to the project reaching its objectives, to maximize the benefits for the industry, and that the results of the project is implemented in the industry. The steering group explicitly did not have authority to “influence the project in a way that could weaken the scientific management” of it¹. Both the members of the quality assurance group and the steering group gave valuable inputs as experts and stakeholders, but Nofima and SINTEF Ocean have been responsible for the final choice of themes, datasets, indicators, and the presentation in the portal.

One possible division of target groups are between “specialists” and “citizens”/“the general public” (Eurostat, 2017, p. 22), which differ in their needs for and abilities to understand detailed statistical information. It can also be specifically about their previous knowledge about the aquaculture industry. Within the general public, Eurostat (2017, p. 23) lists the following as possible subgroups: policy-makers, youngsters, University students, pensioners, families, representatives of the civil society, generalist journalists. The choice of target group(s) can affect all aspects of the web portal. The more precisely the target group can be defined, the easier it will be to approach it to get useful input regarding choice of themes and indicators, the geographic and timewise resolution for them, the format for presentation, and more. It will likely also be less need for making compromises regarding these choices.

The preproject report (Andreassen et al., 2016) emphasized the need to clarify the target group(s) for the web-portal. While leaving this to the main project, it did however mention both decision makers and “ordinary people,” and that they had received inputs especially about media, consumers of farmed salmon, and local and regional decision makers relevant for aquaculture. In the main project additional and more specific target groups were considered in a workshop with the project researchers and persons from BarentsWatch, and in discussions with members of the steering group and quality assurance group. Additional groups mentioned included politicians, NGOs, and individuals concerned with regional development and sustainability, persons and organizations from other industries, and especially also those with little prior knowledge of the aquaculture industry.

To select themes, identify and collect data, select or construct indicators and make a web-presentation of them has been an iterative process. Figure 1 sketches the workflow in the main project. The selection of themes and identification of data sources (1 and 2 in the Figure 1) used the list from the pre-project as a starting point. The project aimed to select indicators based on existing data, and thus not create or collect primary data itself. The motivation was to get a first version of the portal operative without having to wait for new data to be collected.

As data was identified and collected, and draft versions of indicators (3 in Figure 1) were constructed, certain challenges appeared. In some cases, the available data was found to be too limited for the breadth or complexity of the theme, and indicators based on them would likely give an incorrect picture of its status or development. In other cases, the problem was the high number of relevant datasets making it difficult to select a reasonable sub-set suitable for presentation on a webpage. An example of the latter was for the planned theme Safe and Healthy Food. A very large number of both nutrients and contaminants in farmed salmon are monitored regularly in Norway (IMR, 2020). Of the ca. 80 contaminants monitored, 17 have an official maximum threshold, and both were considered too many to be presented on a webpage in the portal. Another data-related challenge was data only being available from some years back. Was it likely or unlikely that they were so outdated as to give a wrong impression of today's situation? For greenhouse gas emissions, available data were from 2007, and it was easy to conclude they could not be used. Data on feed ingredients and their conversion to energy and proteins in the fish were from 2012/2013, and they were included.

To construct indicators from datasets it was necessary to consider the most relevant aspects for aquaculture sustainability. This included geographic level (national/regional/local), time periods (year, month, week) and if indicators should have values relative to how the aquaculture sector developed over time (size or activity level), or to the environmental, economic or social context the industry operate in.

When suitable indicators had been constructed, the next step was to design a webpage for the theme (4 in the Figure 1). After designing a template to be used across themes, the focus moved to the concrete content for each theme's webpage. Quality assurance (5) was conducted through several mechanisms. The quality assurance group gave input in meetings or email to selection of themes and indicators and their presentation. When the portal was approaching version 2, a user-test was carried out where the users were asked to find answers to certain issues/questions by using the portal. How they used the webpages was observed and they also commented on their experience. The user-test gave input that was useful for the specific design of the webpages, layout of diagrams, and on wording. In addition to the feedback from these groups, a nationwide survey gave important input to quality-assurance.

The nation-wide survey was conducted to get input on how important different topics on aquaculture were for the respondents, and important factors for a web-portal on aquaculture to be credible. The survey was conducted 6 March to 12 April 2018 as a web survey and was performed by the survey company Norstat based on their panel of respondents. The survey was representative with regard to age-groups and gender for the general population in six regions covering all of Norway, and had 630 respondents. This gives, with the total number of inhabitants of Norway 18 years or older as the survey population, a 95% confidence interval of maximum ± 3.9 %.

The 33 themes covered by the survey were selected based on a literature review (Andreassen et al., 2016), workshops, and meetings with representatives from research, aquaculture industry, and environmental NGOs with knowledge on aquaculture and sustainability, and a mapping of the availability of relevant data for themes.