There is a wide range of studies varying in scope, methods or region that conclude that green-blue areas contribute to physical and mental health of people living nearby and visitors. Access to natural environments improves overall mental health, physical fitness level, cognitive and immune functions and can lower mortality rates in general (WHO 2016, WHO 2017, NRPA 2017, Frumkin et al. 2017, Twohig-Bennett and Jones 2018, Zhang and Tan 2019). These benefits are explained by different mechanisms including stress reduction and relaxation, more physical activity, improved social interaction and community cohesion. Another set of explanations point to the impact of green-blue areas on the quality of the environment in terms of better air quality or lower noise levels. In the context of the SEEA EA, this impact is already covered in the contribution of ecosystems to, for example, air quality (and regulating services). The literature review confirmed that the benefits of exposure to green-blue areas are in addition to those regulating ecosystem services as several (although not all) studies accounted for air and noise quality as confounders.

Twohig-Bennett and Jones (2018) performed a meta-analysis of about 100 physical health indicators. They found that green space exposure was linked to reduced incidence for diseases, such as stroke, asthma, diabetes and coronary heart disease and to lower disease precursors, such as heart rate, blood pressure or cholesterol. Some studies have indicated that there is a positive correlation between higher levels of green area and lower levels of obesity (Nielsen and Hansen 2007, Bell et al. 2008)

Access to natural environments can also improve overall mental health. There is an indication that experiencing the natural environment reduces stress levels (Bowen et al. 2015). The impact of green areas on mental health also include improved general mood (Tillmann et al. 2019), reduced depressive symptoms, enhanced cognitive functioning of adults and children, short-term memory performance and enhanced creativity (Van den Berg et al. 2007, Bowen et al. 2015, WHO 2016, Tillmann et al. 2018, van Dijk-Wesselius et al. 2018). At the individual level, the lack of contact with natural areas is linked to an increase in the incidence of mental illnesses (Van den Berg and van den Berg 2015).

As there are no specific studies for Flanders, we build on results from literature to quantify the impacts. Notwithstanding the large amount of studies, most studies aim to test if there is an impact of the amount of (contact with) green areas on health, whereas there are few studies that report dose-effect relationships. In addition, the wide range of different indicators for green areas and health in these studies hampers comparison of results and meta-analysis. We found that the amount of green areas is relevant within at least 1 km of the residence. We have found sufficient evidence for dose-effect relationships for effects on mortality and for a selection of two mental and five physical diseases (Buekers et al. 2020). As elaborated in more detail below in the section on Materials and Methods, the selected dose-effect relationships for mortality are based on a meta-analysis (Gascon et al. 2016). For morbidity, we selected the dose-effect relationships of a study in the Netherlands (Maas et al. 2009) if they were confirmed by other studies. Selection of dose-response functions is subject to uncertainty due to the wide range of methods and indicators in the studies and the lack of meta-analysis for the effect on mental and physical morbidity (Buekers et al. 2020).

We recognise that negative effects of contact with green-blue areas can also occur, such as tick bites and Lyme disease. However, these specific health outcomes are not covered in the studies, nor do we have dose-effect relationships to quantify these impacts.

For the monetary valuation of these effects, we build on the literature economic valuation of health impacts and distinguish, for each disease, medical costs (e.g. hospitalisation), absenteeism and productivity loss and welfare loss due to suffering and reduced life expectancy (NCAVES and MAIA 2022aOECD 2011). As further detailed in Table 3, it is possible to value for each health impact the medical costs and productivity losses, based on statistics and specific studies for Belgium and/or Europe. The welfare losses due to suffering can only be estimated for mortality impacts, based on a European study. In the next section, the relationship with NCA is discussed.

Approaches to measure the stocks of natural resources that yield benefits as natural capital have gained considerable attraction in recent decades. By providing regular, objective data that are consistent with wider statistical data, natural capital accounting can provide the fundamental evidence base required for providing information for economic and environmental decision-making that delivers on these ambitions for natural capital (EEA 2018).

The impacts of green-blue space on public health is measured in terms of health outcomes and their monetary valuation is challenging in the context of natural accounting. The SNA aims to value outputs in terms of exchange values and market prices. As this is not possible for health outcomes, the contribution of the health sectors (e.g. doctors and hospitals) to GDP is valued in terms of their inputs (costs), similar to approaches used, for example, for education or defence (NCAVES and MAIA 2022b). The contribution of green-blue space to public health is assessed in terms of health outcomes for mortality and morbidity and this can be valued, based on methods used, for example, for cost benefit analysis of health or environmental policies. As elaborated above, these methods distinguish between avoided health expenditure (medical costs), avoided productivity losses and avoided suffering. The contribution of green-blue space to health outcomes can, thus, be estimated, based on the method of avoided damage costs. This is an alternative, second best method that can be used in case no market prices are available (NCAVES and MAIA 2022b). It is, for example, common to value some regulatory ecosystems services, such as flood control, climate regulation or air filtration from trees, based on avoided damage. Valuation of avoided medical costs builds on the same market data (e.g. hospitalisation costs, medicine) used in SNA to value health care outputs. Avoided absenteeism is valued, based on market prices for labour. The valuation of welfare losses from suffering and reduced life expectancy builds on willingness-to-pay studies that may not be consistent with SNA if they include a consumer surplus component. To ensure a better consistency with SNA that requires market prices or exchange values, we follow the guidelines in NCAVES and MAIA (2022b) and use willingness-to-pay studies that estimate simulated exchange values (Hein et al. 2016). In addition, literature on the expenditure per DALY of ongoing health care programmes are an indicator of the willingness-to-pay of governments and society for health outcomes and is also used as a benchmark indicator for cost-effectiveness analysis in health policy (Ryen and Svensson 2014). These estimates can be used for a simplified, less data intensive approach to monetary valuation.

A number of countries take the lead in natural capital accounting (e.g. UK ONS (2019), The Netherlands (Horlings (2020)). To the best of our knowledge, an overall analytical framework for health effects of contact with green-blue areas has not yet been applied. Most natural capital accounts focus on estimating and valuing the contribution of green-blue areas to recreation and/or tourism and real estate value.

It is recognised that there are few methods and tools available to assess these health benefits in the context of natural capital accounting. Quantification is challenging due to the wide variation in methods (indicators to define green areas and exposure), scope (green areas, health outcomes) and contexts. In addition, assessment of health benefits requires region specific health data for a wide range of health outcomes. Economic valuation requires accounting for a wide range of region-specific data (costs of illness, labour productivity etc.) and integration of different methods (market data, valuation studies for longer life expectancy etc.).

The scoping study to develop urban natural capital accounts for the UK focused on health benefits of (extra) exercise in green areas, which have been estimated, based on local data for recreation in green areas and dose-effect relationships regarding the impact of extra exercise on health outcomes (EFTEC 2017, ONS 2019). The impacts on health were further calculated, based on literature and key figures regarding health benefits per visit (DALYs/visit). This overall approach is interesting, but is limited to one of the mechanisms that explain health benefits (more physical activity) and does not account for other mechanisms that provide benefits.

The specific impact of green-blue areas is also not available in regular SNA as, in health accounts, the largest focus is on curing practices not prevention (Horlings 2020).

The impact of health was included in the assessment of ecosystem services of NATURA 2000 areas in Flanders, applying in a simplified way the dose-effect relationships for morbidity from a health impact study in the Netherlands (Maas et al. 2009) and aggregate and value health impacts in DALYs and valued the outcome, based on an average value for a DALY from literature, mainly based on willingness-to-pay studies (Broekx et al. 2013). This approach is further elaborated in the online tool to assess ecosystem services in Flanders and Belgium (Liekens et al. 2021). The overall approach is interesting, but the selection of dose-effect relationships was based on a quick review of literature (up to 2012) and does not include mortality, did not use local data on prevalence for diseases and the monetary valuation is simplified and not in line with the requirement for natural capital accounting (exchange values) (United Nations 2021).

The literature review concluded that it is possible to develop physical and monetary accounts for health impacts of exposure to green-blue space for Flanders. The selected dose-response relationships also impose limitations, especially related to the level of detail in indicators to measure contact with or exposure to green space. It also limits the number of health outcomes taken into account.

The indicators for green space refer to all green-blue land-use types, including parks, agriculture, private and public gardens, small informal green areas and all kinds of surface waters. This is in line with the selected dose-effect relationships and the literature review that indicated that also informal green areas, agriculture or private gardens contribute to health benefits. Although there is less evidence for the impact of blue space (surface water) in the health impact studies, we included it as part of green space. This is also in line with other studies that indicate the added value of water for quality and attraction of landscapes and amenity value.

Exposure to or contact with green-blue space is also defined in a broad sense, including recreation and sport activities and gardening, but also more a passive form, such as the view on green space from home or during local transport. This is in line with literature and the wide range of possible mechanisms that explain these benefits. However, based on the available, data exposure is only measured at the vicinity of the residence and not at other locations such as school, work or trips or holidays away from home.

Health outcomes are defined as both avoided physical and mental illness and longer life expectations. The health outcomes generated by exposure to green-blue areas are in addition to other health benefits generated indirectly by the delivery of other ecosystem services, such as air pollution removal, noise and heat stress control. The latter are not taken into account here. Within the ecosystem services classification, these are part of the regulating services with their own accounts. However, in gathering data on dose-effects relationships, it is important to correct for these health benefits of nearby green spaces to avoid double-counting.

The monetary accounts will look into the three major components of total health costs, as defined by WHO, i.e. (avoided) costs of illness (health care costs, for example, hospitalisation costs), productivity gains (less absenteeism) and welfare gains from life years gained (mortality). The first two components will be valued using market prices and the life years lost will be valued using simulated exchange values, in line with the exchange value methodology of national accounts (United Nations 2021).

Table 1 gives an overview of the selected dose-effect relationships that estimate the impact of a 10% increase in the availability of green-blue areas in the total land use on the prevalence for mortality or morbidity, based on the review in Buekers et al. (2020). As explained below in more detail, the effects are only statistically significant for the distance bands indicated in the 4^th column. It must be noted that the available dose response functions are linear (Maas 2008, Gascon et al. 2016). This may feel counter-intuitive, as one would expect a greater impact in areas with less availability of green-blue space. On the other hand, it requires more or better data to find non-linear relationships and, as discussed below, the data in these studies are insufficiently detailed in relation to to accessibility, quality and use of green-blue space.

For mortality, we selected the dose-effect relationship from the meta-analysis of Gascon et al. (2016). It applies to cardiovascular mortality and for a percentage of green-blue area within 0.5 km of the place of residence. The selected relationship is confirmed in a more recent meta-analysis on general mortality and green areas (Rojas Rueda 2019).

For morbidity, we selected dose-effect relationships from the study in the Netherlands (Maas 2008, Maas et al. 2009) and checked for additional confirmation in more recent studies. In addition, the application in ecosystem accounting requires further data on DALY weights for the specific disease and data for monetary valuation. We opted for the dose-effect relationships from the original study (Maas et al. 2009) for the share of green-blue areas within 1 km range. We only selected the six endpoints that are confirmed in further literature (WHO 2016, Frumkin et al. 2017, Buekers et al. 2020). For mental illnesses, recent research confirms these positive effects (Kondo et al. 2018, Kondo et al. 2019).

We calculated the percentage of green-blue spaces within 500 m and 1 and 3 km of the residence using the Extent account also created in this project (De Nocker et al. 2020). The methodology used for this Extent account is largely based on the detailed land-use maps of Flanders (Poelmans et al. 2019).

Table 2 gives an overview of the data used to indicate the relative importance of the selected diseases for Flanders. These are the numbers of prevalence of diseases, expressed in number of cases per 1000 inhabitants. For cardiovascular mortality, yearly data are available within the Flemish Agency for Public Health (AZG 2017a, AZG 2017b). The indicators are expressed in numbers of cases and in numbers of life years lost (SEYLL, Standard expected years of life lost).

For the physical illnesses, prevalence data are available for Flanders in the Inego databank with 2015 as most recent year (Inego-databank 2021). For diabetes, the IMA databank was used with 2017 as the most recent year (IMA 2015). For mental illnesses, prevalence is based on a 6-yearly survey (Sciensano 2018), with 2018 as most recent year.

These data were translated to DALY, based on the expected healthy life years. For mortality, it is based on data for Flanders (Statistiek in Vlaanderen 2018), whereas for morbidity, data from the Netherlands are used (RIVM 2015; RIVM 2019a). DALY indicators will be updated if local specific data for Flanders become available (ongoing research).

Table 3 gives an overview of the data used to value, for each disease, the three components of total health costs, i.e. medical costs, productivity loss and welfare gains. Medical cost and productivity loss are valued, based on market data for Flanders (or European studies/countries). For mortality, life years lost are valued using data from literature, based on simulated exchange values (Hein et al. 2016), in line with the exchange value methodology of accounting (United Nations 2021, NCAVES and MAIA 2022b).

The medical costs are estimated, based on information from the health sector and include costs for hospitalisation, care facilities and patient medication. Costs for cardiovascular mortality are based on Buekers et al. (2014). Only for the Chronic obstructive pulmonary disease (COPD) this is based on Belgian statistics (costs of hospital admissions). For other diseases, this is based on one-off studies for Flanders or Belgium or on data from literature (statistics for the Netherlands, European studies). It should be noted that data for mental health are more uncertain because less studies are available. The costs of illness are borne by the government (healthcare), health insurance companies and patients. These costs are based on market prices and, therefore, consistent with the SEEA guidelines.

Productivity losses included the costs for employers and employees for absenteeism, lower employment rate (chronic illnesses), lower productivity on the job (data only for heart failure) and loss of doing domestic work. There are no systematic data on productivity loss per disease. This would require specific studies to estimate the number of working days lost. For each disease, we estimated the number of working days lost accounting for days in hospital, while accounting for the employment rate and age. Lost days were multiplied with the average gross wage cost for Belgium. The costs are borne by patients (loss of income), employers (loss of production) and the government (less taxes and higher unemployment fees). These costs are based on market prices and consistent with SEEA guidelines.

For the valuation of welfare losses for the patient and his family due to suffering and reduced life expectancy, no market prices or statistics are available and it requires specific valuation studies (OECD 2012a). For the valuation of cardiovascular mortality, we value the number of life years lost (YOLL). We use the simulated exchange value per YOLL, estimated in Hein et al. (2016), based on the results of WTP surveys that are commonly used for valuation of mortality in European environment, health and transportation studies (Desaigues 2011, OECD 2012a). There are insufficient willingness-to-pay studies available to value (avoided) suffering for the morbidity endpoints (OECD 2012a).