Based on the full extent of accounting required for the SEEA EA, the KwaZulu-Natal monetary ecosystem accounts aimed to include ecosystem services from all types of cultivated areas, tree plantations, urban areas and man-made waterbodies, as well as natural (or semi-natural) terrestrial, freshwater and estuarine ecosystems within the accounting area, which excluded marine ecosystems.

While the concept of ecosystem services is well established (Ehrlich and Mooney 1983), the development of a standardised approach to classify ecosystem services remains a serious challenge (United Nations et al. 2014a, Potschin et al. 2016, La Notte et al. 2017). Inconsistency across concepts and terminology has resulted in ambiguity and the SEEA Technical Recommendations (United Nations 2017) flagged the “definition and classification of ecosystem services” as a key area for further research. In this study, a list of services was generated, based on a range of existing classification systems, such as CICES (Haines-Young and Potschin-Young 2017) and knowledge of the study area (Table 1). This is similar to the "reference list" subsequently published in the SEEA EA (United Nations 2021). The study only considered biotic services. Water was not included as a provisioning service, since it is not produced by ecosystems and is accounted for separately as a resource account. Rather, we regard ecosystem services pertaining to water supply as being those that regulate the timing and location of water flows and those that affect water quality, both of which affect the costs of collecting and producing potable water for use. Unlike CICES, within crop and animal production (eco)systems, we considered the ecosystem service to be the in situ environmental input to production, rather than the value of crop and animal production. This also means that we could account for pollination and pest control services as an input from surrounding ecosystems. While broad coverage of inland ecosystem services was attempted and key services were prioritised, not all services could be fully covered. The study also excluded estuary inputs to marine ecosystems (see Turpie et al. 2017b).

While many authors argue that one should only value final services, ignoring intermediate services (the provision of a service from one ecosystem type to another, such as crop pollination) would lead to a spatial bias in the valuation of ecosystems. Our approach was to estimate the contribution of the intermediate service and attribute that value to the habitat providing the service. Thus, the pollination contribution to agricultural production was subtracted from the estimated value of ecosystem inputs in agricultural land to avoid double counting.

For each ecosystem service, we selected valuation methods that are conceptually valid and that produce values that are consistent with the SNA. We attempted to value actual use (rather than capacity to supply). We expressed the value of ecosystems in terms of exchange values (consistent with the SNA) rather than welfare values, but point out that these go a large part of the way to providing information for welfare values. The benefits derived from ecosystem services were expressed in terms of annual flows. These were then summed across all benefit flows to estimate a total annual flow of value from each spatial unit. This total value flow was then used to estimate the asset value of that spatial unit in terms of its net present value (NPV). We used a social discount rate of 3.66% (Addicott et al. 2020) and a time period of 25 years, noting the high level of uncertainty in projecting values into the future (see review by Badura et al. 2017) and that discounting mitigates this concern to some extent. In most cases, it was assumed that ES flows were constant, but for provisioning services, future ES flows were modelled, based on estimated levels of offtake relative to their stocks and productivity. All values are presented in 2010 Rands, which was the most recent national accounting base at the time of the study.

We used the national 100 x 100 m (1 ha) Basic Spatial Unit (BSU) grid, constructed by Statistics South Africa for the purpose of ecosystem accounting (Stats SA, see Anderson 2019). Base raster layers (e.g. land use, biomes, census areas) were first projected and then snapped to this grid, ensuring consistency across all ecosystem services and no overlaps for any given area per land-cover class or ecosystem type. We used the 2005 and 2011 KwaZulu-Natal Land Cover datasets, which have 47 classes (including a measure of condition for major natural land-cover classes) and a nominal resolution of 20 m.

The supply and use tables only account for ecosystem services used, such that the sum of supply of a particular service must equal the sum of use. For wild biomass, the amount used would also include illegal use and amounts exceeding sustainable yield. In the case of some regulating services, accounting only for the service used is easier to achieve in monetary than physical terms because of the spatio-dynamic complexity of the service and, thus, for certain services, the physical accounts have reported on the service capacity, irrespective of whether it is demanded. For certain cultural services, only the monetary accounts are provided, since physical measures were not available (see Table 1). The ecosystem monetary asset account records the monetary value of opening and closing stocks of all ecosystem assets within an ecosystem accounting area and any additions or reductions in these stocks.

Hundreds of wild plant and animal species are harvested in the wild by large numbers of households who rely on harvesting natural resources on a subsistence or small-scale basis (Dovie et al. 2002, Shackleton et al. 2002a, Shackleton et al. 2002b, Shackleton et al. 2007, Twine et al. 2003, Kaschula and Shackleton 2009, Turpie et al. 2010, Turpie et al. 2014). These were grouped into ten broad types of resources (Table 2). Their stocks were estimated, based on land cover, vegetation type (South African National Biodiversity Institute 2018) and corresponding information from literature and availability was based on land tenure. Demand was estimated at the census sub-place (~ village) level, based on household survey data (mainly from Turpie et al. (2010), Dovie et al. (2002), Shackleton et al. (2002a), Shackleton et al. (2002b), Twine et al. (2003), Cocks and Wiersum (2003)Turpie et al. (2010)). Household use was estimated using a purpose-built spatial model, under the assumption of a 5-10 km range of collection, limited by the availability of stocks (see Turpie et al. in prep). Resource was mapped in physical terms (see example for fuelwood (Fig. 4). Values were based on prices reported in the surveys and collection costs were assumed to be negligible. The asset value was calculated as the net present value over 25 years taking sustainability into account where harvesting was compared to the corresponding sustainable yield at the level of the BSU. Where harvesting exceeded the estimated sustainable use, stocks were eroded at the corresponding rate, affecting future use and values. In some cases, this effectively resulted in the asset life being less than 25 years.

Figure 4.  

Estimated spatial variation in the informal harvesting of fuelwood across KwaZulu-Natal.

KwaZulu-Natal accounts for about 20% of cattle in the country (Meissner et al. 2013). These are found in both high input production systems on private lands and ‘low input-low output’ systems on communal lands. The ecosystem service is the land’s contribution to livestock or wildlife production (for meat, live sales or hunting), which includes natural fodder provision. As a proxy for the physical measure, we quantified the amount of livestock supported in standardised large stock units (LSUs, Mokolobate et al. 2017). District level stocking rates, production and value on private and communal lands were derived from the Census of Commercial Agriculture 2007 (Statistics South Africa 2011), quarterly provincial statistics, the agricultural household survey (Statistics South Africa 2012) and other studies (Kunene and Fossey 2006, Turpie et al. 2010, Turpie et al. 2014, Mahlobo 2016, Nkosi 2017). Wildlife stocking and offtake rates were not available, but production value was calculated using information from Ezemvelo KwaZulu-Natal Wildlife and literature (Taylor et al. 2015). Reared animal production was valued using the resource rent method (gross income less intermediate expenditure, labour costs and user costs of fixed capital), based on the agricultural census.

We used production in tonnes of produce or m³ of timber as a proxy for the physical measure of land inputs to cultivated production. The service was valued in terms of resource rent, less the contribution of pollination services from adjacent natural ecosystems. Commercial crop production, prices and input costs were taken from the Census of Commercial Agriculture (Statistics South Africa 2006, Statistics South Africa 2011), using groupings that could be aligned to land-cover classes. Silviculture production and value were only available at provincial level (Forestry South Africa 2017). Communal crop production, prices and input costs were taken from household studies in northern KwaZulu-Natal (Turpie et al. 2014) and used as a single average value per hectare.

While cultural values are generally understood to comprise both use and non-use values, the SEEA EA only accounts for use values, which we term "experiential value". We developed a framework for the consideration of experiential benefits, as accruing in three ways: to people who live close to the site used (herein characterised as “locals”), to people who come from nearby (“visitors”) and to people who travel relatively far to visit the site (“tourists”), with methods of valuation typically being different for each (Fig. 5Fig. 6). This study only covered value to locals as manifest in property investment and value to tourists. There were no data or valuation studies (e.g. travel cost studies) on which to estimate the value to nearby visitors.

Figure 5.  

A typology of values associated with experiential uses of ecosystems (source: authors).

Figure 6.  

Nature-based tourism value for the year 2011 across KwaZulu-Natal, based on the distribution of geo-referenced photographs uploaded to Flickr.

The contribution of ecosystems to tourism value was estimated using tourism statistics and spatial data on activity (Fig. 6). First, the tourism expenditure attributed to visiting attractions, as opposed to other activities, such as visiting family and friends, was estimated for different categories of domestic and foreign tourists, based on information collated from the South African Tourism (SAT) annual performance reports (South African Tourism 2005, South African Tourism 2006, South African Tourism 2011) and from data collected in regional tourist offices. Resource rent was estimated using ratios from the SAT Satellite Accounts (Statistics South Africa 2010, Statistics South Africa 2015). This value was spatially allocated to the BSU grid in proportion to photo user density using the InVEST Recreation Model 3.5.0 which obtains data on geotagged photographs uploaded to flickr.com. Geotagged photographs are a reliable proxy for visitation rates (Wood et al. 2013, Turpie et al. 2017a, Turpie et al. 2017b, Lee and Tsou 2018, Barros et al. 2019).

The amenity value of ecosystems to locals was estimated in terms of the contribution of urban green open space areas to property value. The property value of urban green open space areas in KwaZulu-Natal was estimated, based on the hedonic pricing study of eThekwini Municipality which was derived from very detailed property and open space data (see Turpie et al. 2017b). Such data were not available for other urban areas in the Province. Thus, a simple benefits transfer model was derived for the latter areas, based on the relationship between suburb-level census data on household income and premiums paid for open space in eThekwini. This was applied to the ten urban centres identified using the Functional Town Typology by van Huyssteen et al. (2018). The aggregated premium value was annualised for the supply and use tables.

In South Africa, total ecosystem carbon was mapped at national scale in 2015 (Department of Environmental Affairs 2015). This was used to estimate the 2005 and 2011 carbon stocks, based on average carbon values per land cover type. We sought to value the service in terms of the avoided climate change damage costs, both to South Africa and the rest of the world. This was considered preferable to using market values, since there is very little trade in ecosystem carbon in South Africa (Remme et al. 2015). However, estimates of the global social cost of carbon (SCC)- the net present value of avoided costs from the release of 1 t of CO₂ - are numerous and vary greatly (for example, see Tol 2008). Estimates have also been increasing over time, as the more recent studies have tended to be more comprehensive. Estimates now range from US$10 to US$1000/tCO₂ (Ricke et al. 2018). In their critical review of literature, van den Bergh and Botzen (2014) suggested a lower bound value of US$125/tCO₂. Disaggregated estimates have also started emerging. For example, Nordhaus (2017) provided an updated median estimate of global SCC as US$31/tCO₂ (in 2010 US$) and estimated that 3% of this would be borne in Africa. Ricke et al. (2018) produced a far higher median estimate of global SCC (US$417/tCO₂ in 2018 US$) and disaggregated this to country-level, yielding a unit cost of US$3.31 for South Africa. In this study, we used the more conservative estimate from Nordhaus (2017) with a global SCC of US$31/tCO₂ and an estimate of US$3.31 for South Africa’s SCC, which is 0.8% of the global SCC estimate produced by Ricke et al. (2018).

The value of SCC is expected to increase over time as populations and per capita incomes grow and should correspond to the year of the account, as ecosystem carbon retained will increase in real value over time. Therefore, the SCC estimates from the literature were escalated at a rate of 3% per year (Nordhaus 2017) to derive different estimates for 2005 and 2011. The annualised social cost of carbon was estimated using a social rate of discount of 3.24%.

Crop pollination by insects increases both yield and quality of crops (Melin et al. 2014). In KwaZulu-Natal, most commercial crops are wind pollinated and it is assumed that vegetable and fruit farmers tend to actively pollinate their crops. We, therefore, focused on the likely benefits to communal land farmers. The location and extent of vegetable and fruit gardens was estimated using census data (Statistics South Africa 2012), community survey data (Statistics South Africa 2016) and information from literature (Shisanya and Hendriks 2011, Ogundiran et al. 2014). We used land-cover data to calculate the amount and type of natural vegetation surrounding each of the settlement areas and we used a value function transfer approach, based on the panel model of Tibesigwa et al. (2019) to predict the value added to income from pollinator-dependent crops by natural pollinators.

Ecosystems can reduce seasonal variation in downstream river flows through infiltration and temporary storage in catchment areas (relative to the variation in rainfall), reducing the built storage capacity needed to achieve a given yield through the year (Vogel et al. 1999, Vogel et al. 2007, Guswa et al. 2017). Therefore, water supply infrastructure and reservoir capacity, in particular, can be treated as a substitute for the service provided by ecosystems.

For this study, a hydrological model was set up for all of the catchments of KwaZulu-Natal using the Soil and Water Assessment Tool (SWAT) model. The service was mapped in physical terms as the difference in infiltration relative to a bare landscape, in m³ per ha. The benefits generated from the service were considered in terms of the avoided additional storage capacity required to meet the yield of the existing supply systems, which was estimated, based on the theoretical relationship between storage, yield and reliability for a standardised reservoir (Kuria and Vogel 2015) and costed, based on data from a national inventory of reservoirs. We also estimated the avoided costs of obtaining water for people who depend on instream flows for their domestic water supplies, based on monthly water demands by these households within each sub-catchment (derived from Statistics South Africa 2012).

Erosion and sedimentation within watersheds can become a major issue as it causes structural damage to reservoirs, causes flooding, affects the quality of drinking water and increases water treatment costs (Pimentel et al. 1995, Basson 2009). Natural vegetation and crops can reduce erosivity by stabilising soils and intercepting rainfall (de Groot et al. 2002). Vegetated areas also capture eroded sediments transported in surface flows, preventing them from entering rivers (Blumenfeld et al. 2009, Conte et al. 2011). While some level of sedimentation of dams is expected and planned for under natural conditions, elevated catchment erosion either incurs dredging costs or shortens the lifespan of dams and related infrastructure.

The extent to which ecosystems retained and/or captured sediments relative to a bare landscape was estimated and mapped using the InVEST Sediment Delivery Ratio model. Due to the potentially large and costly damages of sedimentation, we assumed that the service would be fully demanded. We used the replacement cost of lost storage capacity to estimate its value.

The impacts of natural vegetation and cultivated land on water quality were estimated using the SWAT hydrological model. The model was set up to estimate changes in phosphorus loads at raw water treatment extraction points relative to a bare landscape in which the retention/absorption capacity of the vegetated areas was reduced. The value of the service was then estimated in terms of the avoided costs to water treatment works using a water treatment cost model developed by Turpie et al. (2017b).

This study found that ecosystem services made a significant contribution when compared to GDP. Only a fraction of this value is currently recorded in the national accounts, primarily associated with formal agriculture and forestry production and ecosystem contributions to property value and tourism. The value of the additional service flows valued in the ecosystem accounts is significant and includes informal use of provisioning services, all regulating services and the recreational value not estimated in this study.

The study also showed, however, that the volumes of all provisioning and regulating services from natural landscapes decreased over the accounting period. These losses in services were likely due to a combination of the overharvesting of resources, overgrazing leading to denudation in some areas and bush encroachment in other areas, the spread of invasive alien plants and the loss of habitat due to expanding cultivation and human settlements. While these concerns have been well noted (Driver et al. 2015), this study has shown that their aggregate economic impact can be substantial. Furthermore, the losses were not fully portrayed in the asset valuation. While the effects of overexploitation on provisioning services were taken into account, it was implicitly assumed that the future ecosystem capacity to supply regulating services and reared animal production would remain unchanged. This is unrealistic given current rates of degradation. Future studies would also need to consider the sustainability of crop production. Given the significant losses in value of ecosystem services from natural ecosystem types over only six years, it is clear that further research is required to validate these findings and to seek urgent solutions. Indeed, the degradation of natural ecosystems could have significant impacts for the most vulnerable communities in the Province, who are reliant on them for their livelihoods, water and food security.

The interpretation of value changes is important. In spite of decreases in most of their services, the values of most natural systems increased as a result of the increased demand for tourism and carbon. In the case of nature-based tourism, value increased in spite of general evidence of a decline in the extent and condition of natural areas. Over time, it would be useful to consider the counterfactual or what tourism value might have been in the absence of the negative ecological trends. Similarly, for carbon, the increase in service value would have been greater if carbon retention had not decreased.

This study has estimated the value of a range of ecosystem services, covering most broad types. We did not include all ecosystem services, some are only partially valued and the geographic coverage excludes the marine environment. Future iterations should expand these accounts to include other services such as local recreation, local climate regulation and air quality amelioration, as well as expanding the area to include coastal and marine ecosystems and their services. In addition, some of the methods used in this study are innovative and require further refinement and validation. For some services, compiling the biophysical aspects are the most limiting factor, while for others, economic data are limiting. Nevertheless, the study provides a solid beginning from which to progress.

In many cases, data were not ideal from a temporal or spatial perspective, especially for provisioning services. For example, the low spatial resolution and commercial focus of agricultural data, as well as lack of regularly updated data, created difficulties. There were very little data on wildlife ranching and forestry and none on bioprospecting. These types of data issues will be relatively straightforward to resolve once countries start planning for ecosystem accounting. Addressing other data shortcomings may require considerable effort, such as ecosystem condition, ecosystem capacity for resource provision, illegal offtake of endangered species and the sustainability of various types of ecosystem use. Our study showed that, taking sustainability of resource use into account, had a significant effect on asset value. It is easy to see that this could produce a more appropriate policy response than an account that assumes that values can be sustained in perpetuity.

Our valuation of cultural services focused on the use value aspect, which we termed experiential value. Both the aggregate tourism estimates and the estimated contribution of urban green space to property value were considered reliable and relatively complete estimates. However, we missed an entire component pertaining to the relatively local use of ecosystems, such as recreation, religious or cultural ceremonies. Such values can be estimated in future, based on travel cost surveys (e.g. Ezebilo 2016) or mobile phone or other big data (e.g. Jaung and Carrasco 2020) and would add value to ecosystems close to any rural or urban settlements, especially where inhabitants are wealthier.

The valuation of hydrological services involved modelling at a far greater scale than is typically the case. Clarification of these services is critical since they are widely interpreted in literature. We value the role that ecosystems play in saving the costs of water supply to people, including their influence on climate, but not as a source of water per se (see further comment below). Since the role that tropical forests play in influencing local or regional rainfall (Duku and Hein 2021) does not apply in the study area, we focused on the timing and quality of flows. We, therefore, developed a relatively rapid method for quantifying the flow regulation service which obviated the need for detailed water resources modelling in each water supply area (Turpie et al. in prep). Nevertheless, scaling the modelling up to national scale will likely be onerous and may require development of simpler hydrological models. Not all benefits were quantified in this study, including flow regulation for irrigation supply and flood attenuation outside eThekwini Municipality. Furthermore, the SEEA EA (United Nations 2021) now recommends that water supply is included as a provisioning service. Although not all authors see water itself as being supplied by ecosystems, its inclusion will greatly increase the overall asset value of ecosystems.

Data limitations meant that pollination had to be valued using a benefits transfer method, using a relationship developed elsewhere on the continent, which was far from ideal. It is almost certain that many of these estimates will eventually be replaced by better estimates, based on better data and better models. As this happens, it will be important to update the earlier estimates as best as possible in order to ensure continuity of the accounting dataset. While this may require an adjustment of policy responses as better information comes to light, it is still preferable to begin with some estimate than with none. Where it comes to natural ecosystems for which a change in land cover or use could lead to permanent damage, preliminary estimates should probably have been less conservative.

The valuation of carbon stored in ecosystems is a critical concern in monetary ecosystem accounting. There has been a long-standing debate within the SEEA with regards to the framing and treatment of the carbon service and a consistent approach to its valuation (Remme et al. 2015, Edens et al. 2019, Keith et al. 2021). Accounting for the climate regulation service in monetary terms commonly involves using one of two approaches: the carbon trading price or the social cost of carbon, as recommended in the updated SEEA EA (United Nations 2021). Both of these represent a measure of avoided damage, a cost-based method which assumes there is a willingness to pay to avoid the associated damage (United Nations 2021). While using the carbon price seems preferable over the SCC because it comes from actual market transactions (Horlings et al. 2020,United Nations 2021), many countries (including South Africa) do not yet have mature carbon trading systems and the SCC is the only suitable alternative.

The SCC, used in this study, is an estimate of the current marginal damage cost of a tonne of carbon emitted. Applying this at scale to the gigatonnes of carbon in the landscape makes the assumption that this cost would still apply to the last tonne in the landscape. In aggregate, this value can rival the GDP of a country and is, hence, unrealistic in terms of proxy for exchange value. In reality, the technological replacement cost for this service, which is currently too high to be considered in valuation studies, may well decrease to well below this level by the time the last tonne of carbon is lost to the atmosphere. Thus, the carbon retention value in this study could be a gross overestimate and is one of the methods that needs a thorough review.

The SEEA-EA methods align with the SNA, which produce measures such as Gross Domestic Product (GDP). These are measures of production value, but are often misused as measures of welfare (Jorgenson 2018). Economic analysis used to guide project, programme or policy decisions often involve the estimation of changes in welfare for the gainers and losers involved. These welfare measures, which comprise the sum of producer and consumer surplus, tend to be larger than contributions to GDP, which are more closely aligned to incomes generated in production. Ecosystem accounting requires that values are expressed in terms of exchange values as per the SNA, so called since they are the incomes generated in the exchange of goods and services. This raises the concern that ecosystem accounting could be used incorrectly and lead to distorted decision-making with regard to the environment. However, it is worth noting firstly that in many cases, exchange values of ecosystem services will align closely with welfare values and, for the rest, that ecosystem accounting at a national scale will go a long way towards producing welfare estimates.

The valuation methods used in accounting are largely based on methods developed for cost-benefit analysis and, in reality, many estimates of the value of ecosystem services intended for cost-benefit analysis have, in fact, been exchange values, since they are often easier to compute. Based on the current state of the art, there would be little distinction between exchange value and welfare value estimates in the values of provisioning and regulating services. Provisioning services, as for this study, are typically estimated by subtracting costs from the value of production, focusing on either resource rents or producer surplus, but not consumer surplus. Regulating services, as for this study, tend to be valued in much the same way in both the accounts and for cost-benefit analysis, using an avoided costs approach. This has the same impact on production values and welfare values.

Cultural services are where value estimates are likely to differ most, since valuation methods for economic analysis have focused on deriving consumer surplus, especially for cases where access to (or viewing) nature is free or merely priced to cover costs. This study estimated the direct value added from domestic and international tourism expenditure (as reflected in the national accounts), as well as that from property value premiums attributed to green open space. An economic analysis would also be interested in the consumer surplus of domestic users. Estimates of international tourism value, therefore, need little or no adjustment. For the rest, revealed preference valuation studies (travel cost and hedonic pricing) can be adapted to provide welfare estimates through second stage analysis (for example, see Wolf and Klaiber 2021). More importantly, an economic analysis would also be expected to include non-use value, which is estimated using stated-preference surveys. Thus, in general, the work undertaken in compiling ecosystem service accounts is likely to be very useful in feeding into economic analysis, but will require augmentation, particularly for cultural services. Ultimately, we suggest that both should be derived during the accounting process, with the welfare values provided in supplementary tables. Meanwhile, it is important that the difference is clearly communicated to policy-makers. The monetary accounts should not be interpreted as being “the value of nature” (Hein et al. 2020).

Related to the above, there are also concerns that monetary accounting of ecosystems may send the wrong policy message since anthropogenic landscapes are often more valuable, on average, than natural landscapes, as was found in this study. Accounting tables report the aggregate and average values of ecosystems, but do not report on marginal values. Although agricultural value may be higher than that of neighbouring ecosystems, subsequent conversions are likely to return decreasing values as they expand into less suitable areas. Conversely, the marginal value of natural ecosystems will increase as they contract into critical areas. For all else being equal, the optimal allocation of land use is the point where marginal values of each become equal. The analysis of potential policy impacts, therefore, involves projection of these non-linear area-value relationships in conjunction with the conversion of accounting values to welfare values. While accounting alone does not provide the full information required for such analyses, it goes a long way to providing information for them and will become increasingly valuable as long term datasets start to emerge in the process. It will also be important for future iterations of this work to draw on additional data inputs to capture changes in the capacity of anthropogenic, as well as natural landscapes, in the estimation of future flows and asset values.