South African context
South Africa has a diverse environment ranging from semi-desert to sub-tropical forest and exceptional biodiversity (Driver et al. 2012) making it one of 17 mega-diverse countries in the world (UNEP 2014). It is the 30th driest country in the world (DWA 2013) and only 12% of land is capable of supporting rain-fed crop production (Collett 2013). Climate change projections for South Africa show significant warming, as high as 5–8 °C over the interior by 2100, and a risk of drier conditions in the west and south, and wetter conditions in the east (DEA 2013a). The country has rich mineral deposits, including gold, platinum, iron ore, diamonds and coal. The mining sector has played a key role in the economy for 140 years, making South Africa the most industrialised country in Africa (Chamber of Mines of South Africa 2013). South Africa is also the biggest greenhouse gases (GHG) emitter, and is responsible for 38% of Africa’s carbon emissions (Boden et al. 2011).
Despite being the largest economy in Africa, roughly half of the population of 55 million live below the national upper-bound poverty line (DPME 2015a), and more than 10% of people live on less than $1.25 per day (DPME 2013). Over 38% of the labour force (including discouraged jobseekers) is unemployed (StatsSA 2016a) and South Africa’s labour force participation rate (58%) is among the lowest in Africa (World Bank 2016). South Africa has one of the world’s highest levels of income inequality (Palma 2011) with a Gini coefficient of 0.65 in 2010 (DPME 2015a). It has spatial inequality across multiple aspects of social deprivation (Wright and Noble 2009), a legacy of the racial segregation of Apartheid.
South Africa has a unitary but decentralised state with cooperative governance between three spheres of government—national, provincial and local (Republic of South Africa 2012). The nine provinces—Eastern Cape, Free State, Gauteng, KwaZulu-Natal, Limpopo, Mpumalanga, Northern Cape, North West and Western Cape—were created as part of the transformation to democratic rule in 1994. They were based on a set of ‘development regions’ aimed at planning across previous racially based administrative boundaries and were given considerable functions in the Constitution (Wittenberg 2006). In 2015/16 the provinces received 43% of the national budget with significant autonomy to allocate resources to respond to provincial priorities and meet national objectives (National Treasury 2015). The provinces, therefore, have the mandate and in theory the ability to address many of the environmental and social challenges highlighted in our national barometer. The provinces are shown in Fig. 1 and summarised in Table 1.
In creating our national barometer, we developed a decision flowchart to assess the environmental and social dimensions, indicators and boundaries that make up the ‘safe and just operating space’ and adapt them to the national level. The aim was to ensure repeatability and consistency so that it could be used in other countries or at other scales (Cole et al. 2014).
The criteria used for selecting dimensions were ‘Is this relevant at the national scale?’ and ‘Does the set of dimensions include the main environmental and social concerns in South Africa?’. The criteria for indicator selection were (a) ‘Is the indicator the best available direct measure of that dimension?’, ‘(b) Are there sufficient reliable data that are measured on a regular basis?’ and (c) ‘Can a national boundary be determined?’ If the existing global dimension or indicator did not meet the criteria then it was removed or replaced with a more appropriate national-scale choice. These criteria are similar to the proposed criteria for SDG indicators, which should be relevant, methodologically sound, measurable, easy to communicate and access, limited in number and outcome-focused (UNSD 2015a). The data were taken from relevant national databases and reports, international databases and academic papers. We also sought expert judgment on indicators and boundaries through semi-structured interviews with 43 South African experts from national, provincial and metropolitan government, national research institutes, universities and international NGOs.
To create the provincial barometers, we did not use the decision flowchart to select new indicators, as we wanted to explore sub-national heterogeneity in those indicators we had already chosen. Instead we used three methods of disaggregation of national data for the dimensions in our national barometer: (a) share the national total amongst the provinces, (b) aggregate local data to the provincial scale, and (c) fit data reported by ecological units into administrative borders. These methods are described further below.
We updated the data sources where new data were available, or where sub-national data sources could be found. The data were used to produce nine provincial barometers for both environmental stress and social deprivation. We also plotted the average annual change since 1994 (or since data collection for each specific indicator began) for all the dimensions in two graphs. We did not plot the yearly status due to space constraints, as it would require 20 graphs.
In our national barometer, we used the Environmental Sustainability Indicators (ESI) technical report (DEA 2013b) published annually by the Department of Environmental Affairs (DEA) as a starting point for our analysis. The ESI was developed based on a comprehensive review of potential national indicators, Yale’s Environmental Performance Index (Hsu et al. 2016) and the DPSIR framework (e.g. Hammond et al. 1995; Gabrielsen and Bosch 2003). We then reviewed relevant national policies, reports and assessments, and academic literature to identify the most suitable dimensions, indicators and boundaries and tested these with experts. While we adapted three of Rockström et al’s (2009a, b) dimensions we adjusted all of the indicators and boundaries to suit national scale and circumstances. For the provincial barometers, we reviewed the most recent provincial State of Environment and State of Biodiversity reports.
Table 2 shows the environmental dimensions, indicators, data sources, level of confidence, and the method of disaggregation used in the provincial barometers. Table 2 also shows the type of safe environmental boundary for each dimension, as defined in our national barometer. Type A is an internationally agreed target based on a global biophysical threshold, which varies by country based on differences in national capability and responsibility. Type B is a national biophysical limit for the sustainable use of land or freshwater resources, which can include or exclude human intervention such as infrastructure and technology, and uses local biophysical thresholds to define the boundary. Type C is a local biophysical threshold based on established research and expert judgment in the country being studied, and is unaffected by scale (i.e. national and provincial boundaries are the same). Each dimension is briefly explained below with further details given in the SI.
Rockström et al. (2009a) based their climate change indicator and boundary on global atmospheric carbon dioxide (CO2) concentrations. As this cannot be disaggregated to the national level, we used CO2 emissions for our national indicator. Our safe boundary is based on the emissions trajectory of the ‘Required by Science’ scenario in the Long Term Mitigation Scenarios, LTMS (Scenario Building Team 2007), which South Africa uses for its national commitments to the United Nations Framework Convention on Climate Change. In 2011 South Africa emitted 477.7 MtCO2 (UNSD 2015b) and the safe boundary is calculated as 453.7 MtCO2. South Africa’s national inventory (DEAT 2009) reports sub-national data by sector, not by region, with only four provinces having their own emissions inventories (but only for different years) (Gauteng 2007, Eastern Cape 2008, Western Cape 2009, Free State 2012).
For the provincial status we, therefore, had to share national CO2 emissions for the status and boundary between provinces. As a province’s share of the population can be quite different to its energy use, it would not be equitable to use population as the basis for disaggregation. Instead we used provincial electricity consumption (StatsSA 2012a) to allocate provincial emissions (see Table S3 in the SI) as it has the largest share (46%) of national CO2 emissions. We used consumption (and not production) as it is reported at provincial level. Although this overestimates CO2 emissions in provinces with low carbon energy sources such as wind and nuclear, we did not have the necessary data to adjust the figures. Our calculated figures correlate reasonably well with the four provincial inventories that are available (see SI). We shared the national boundary using the provincial contribution to GDP (StatsSA 2012b) (see SI Table S3) to measure the energy intensity and thus mitigation responsibility of each provincial economy. To analyse the trends we used the year 2002 as this was the furthest back we could obtain electricity use by province (StatsSA 2002). As the LTMS baseline year is 2003 there is no ‘required by science’ target for the year 2002, hence we shared the actual national emissions of 347.7 MtCO2 between the provinces (see SI Table S4).
Rockström et al. (2009b) based their ozone depletion indicator and boundary on the global ozone concentration. As this cannot be disaggregated to the national level, we used consumption of hydro-chloro-fluoro-carbons (HCFCs) for our national barometer. In line with the Montreal Protocol, South Africa has phased out the production and consumption of all ozone-depleting (ODP) substances except HCFCs (DEA 2014a) and is a consumer rather than a producer of HCFCs. For the provincial status, we aggregated individual company HCFC-22 and HCFC-141b consumption data for 2010 (NEDLAC 2012). We then projected it to 2015 based on the latest national HCFC consumption figure of 238.6 ODPt reported by the UNEP Ozone Secretariat (UNEP 2016) (see SI Table S5).
This showed that distributors in Gauteng, Western Cape and KwaZulu-Natal consume all the HCFCs. The national boundary is based on the government commitment to reduce HCFC consumption to 332.7 ODP tonnes by 2015 and eliminate it by 2040 (NEDLAC 2012). We shared this between these three provinces based on their share of HCFC consumption. As historical sub-national data do not exist, for the trend analysis we used the 2010 provincial ratios of HCFC consumption to share the 103.3 ODPt of HCFCs consumed in 1990 (UNEP 2016). We used the government target to freeze consumption at 370 ODPt in 2013 as no limits are defined before 2013.
Rockström et al. (2009a) measured the consumption of freshwater by humans, the global aggregate of local use. In our national barometer we used South Africa’s freshwater consumption reported in the National Water Resource Strategies (DWAF 2004; DWA 2013). Our safe boundary was the available water supply, which takes ecological requirements into account. For the provincial barometers, we could use the demand and supply of the 19 Water Management Areas and 87 sub-areas (see SI Table S6), however, these figures are only available for the year 2000. We considered using the government’s current Water Allocation Registration Management System (WARMS) database, but this would only provide water allocation not demand and supply.
We decided to use demand and supply figures found in the Department of Water and Sanitation’s (DWS) 840 reconciliation strategies for all towns in 2008 and Water Supply Systems that supply the metropolitan areas. The All Town Studies provide the first comprehensive water use information at the local level across South Africa and are aimed at informing water resource investment and management decisions (DWA 2013). As the reconciliation strategies do not account for ecological requirements, we reduced the supply using the ecological requirements for the year 2000 to provide a more accurate picture of the stress on freshwater supply (see SI Table S6). While this may overestimate the reserve as total supply includes groundwater, the reserve figures in 2000 did not include estuaries, which usually have higher ecological requirements (DWAF 2004). As the reconciliation strategies focus on domestic water demand, agriculture and heavy industry are not included in our results. This is not ideal but it is the best available dataset. In addition, annual progress reports are published for the Water Supply Systems and the town strategies are being updated, so more recent data will become available which will allow the calculation of long-term trends.
Arable land use
Rockström et al. (2009a) focused on land use change and its detrimental effects on biodiversity and climate change. However, South Africa’s land cover has remained relatively stable since 1961 (Niedertscheider et al. 2012; Schoeman et al. 2013). The only national land degradation study was done by Hoffman et al. in 1999 and is qualitative not quantitative (DEAT 2006). South Africa is largely a semi-arid country with very limited land capable of supporting sustainable crop production (Collett 2013). We therefore focused on land capability, i.e. the ‘total suitability for use, in an ecologically sustainable way, for crops, for grazing, for woodland and for wildlife… exclusive of social and economic variables’ (Schoeman et al. 2002). The national land capability classification defines eight classes based on a combination of climate, soil and terrain. Arable land (i.e. land that can be used for crop production) is termed ‘arable land of acceptable quality for crop production’ (Classes I-III) or ‘marginal arable land’ (Class IV).
Our indicator for land use is total arable land (Classes I–IV) converted to cropland and our safe boundary is acceptable arable land (Classes I–III). We excluded marginal arable land from the boundary as it is more prone to crop failures in low rainfall years (Biggs and Scholes 2002) and requires irrigation to be sustainable in the long-term. Data at the provincial level is available in the draft Preservation and Development of Agricultural Land Framework Bill (DAFF 2015) which improves on previous datasets as it measures cultivated land for each land capability class. We aggregated cultivated land for the status (Class I–IV) and boundary (Class I-III) (see SI Table S7). Cropland in the non-arable classes (Classes V–VIII) is termed ‘unique farmland’, e.g. Cape Winelands in Class IV and VI which can be sustainably farmed despite shallow natural soil depth (Collett 2013). As the specific figures for unique farmland are not provided we excluded it from the analysis, although this does mean that the Western Cape exceeds its boundary. We could not calculate the trend over time as the cultivated land per land capability class has not been reported before.
Rockström et al. (2009a) argued that the additional phosphoros (P) and nitrogen (N) activated by humans is disturbing the global cycles. Eutrophication of freshwater resources is a global concern (Steffen et al. 2015) and is widespread in South Africa (van Ginkel 2011). South Africa’s National Eutrophication Monitoring Programme measures levels of chlorophyll and phosphorus at over 1,200 monitoring points in 16 drainage basins. In our national barometer we used mean annual total phosphorus (P) concentrations in freshwater as the indicator. We used South Africa’s critical threshold, and effluent discharge limit for wastewater treatment plants of 0.10 mg/l (Oberholster and Ashton 2008) for the safe boundary.
For the provincial barometers, we aggregated total P concentrations reported by drainage basin and calculated weighted averages using gross drainage basin volumes (DWA 2014). We then matched basins to provinces so that each province was an average of weighted total P values (see SI Table S8). Where basins were shared by provinces, we included them in all the relevant provinces. We used the national boundary for all provinces as it is a local threshold. We calculated the trend from 2000 to 2012 using the same dataset and boundary.
Nitrogen is essential for food production. However, nitrogen fertiliser use can have a range of local negative effects (Rockström et al. 2009b; de Vries et al. 2013). Sustainable fertiliser use for crop production can be measured using the nitrogen balance or the nitrogen use efficiency (Brentrup and Palliere 2010). Both indicators are calculated using nitrogen (N) applied to the soil through fertilisers and nitrogen removed from the soil by crop production. In our national barometer we used the nitrogen use efficiency (N removed divided by N applied) in maize production, which uses 62% of all nitrogen in fertiliser in the country (FertASA 2013). Sub-national data on fertiliser consumption for maize or any other crop is not available. Sharing the national total between the provinces by crop area or yield would not take variations in soil and climate into account. We, therefore, could not populate this indicator for the provinces.
Rockström et al. (2009a) measured the extinction rate of species, which saw a massive acceleration in the twentieth century. In 2004 the South African National Biodiversity Institute (SANBI) started to assess biodiversity by ecosystem, rather than species, threat status. The methodology was improved in 2011 and we used the percentage of critically endangered (CR) and endangered (EN) ecosystems for our national biodiversity loss indicator. Our safe boundary was that no ecosystems should be endangered or critically endangered.
For the provincial status, each ecosystem type required a slightly different approach. Estuarine ecosystems were reported at district level (van Niekerk and Turple 2012) and had to be aggregated. Inshore marine and coastal ecosystems were reported by habitat type and geographic region (Sink et al. 2012) and had to be matched to the four coastal provinces. Terrestrial ecosystems were reported by province (DEA 2011). Interviews with experts at SANBI suggested we convert each total area to a percentage and average the three ecosystem types by area to obtain a single value for percentage CR and EN ecosystems per province (see SI Table S10). We kept the safe provincial boundary the same as the national boundary. We did not determine the threat status for freshwater ecosystems (rivers and wetlands) as they are reported by the old 19 Water Management Areas (Nel and Driver 2012), which do not match well to the provinces. We could not calculate trends as the methodology changed from 2004 to 2011.
In our national barometer we replaced Rockström et al.’s (2009a) ocean acidification with marine harvesting due to the lack of understanding of the process in South Africa’s marine environment (CSIR 2012). Our national indicator was depleted marine fisheries (below the biomass level at which maximum sustainable yield is obtained) and our safe boundary was zero depleted marine fisheries. Recently a new Ocean Acidification Indicator (ACID-I), defined as the aragonite saturation state, has been defined for the west coast of South Africa (DEA 2015a) but is not comprehensive enough to be used here. As marine harvesting is only relevant for the four coastal provinces (Eastern Cape, Western Cape, Northern Cape, KwaZulu-Natal) we considered changing the dimension to ‘aquatic harvesting’ to include inland fisheries. However, there are almost no data on inland harvesting rates or stock status (McCafferty et al. 2012). For marine harvesting at provincial level, we estimated the depleted status (percentage of total number of species with known status) per province based on the geographic location of the fisheries (DAFF 2014) (see SI Table S12). Our safe boundary is zero. We calculated the trend from 2009, when reporting started, to 2013, which is the most recent data.
In our national barometer, we replaced Rockström et al’s (2009a) atmospheric aerosol loading with the more relevant dimension air pollution. Particulate matter less than 10 microns (PM10) is the ‘greatest national cause for concern in terms of air quality’ and is used for the National Air Quality Indicator, NAQI (DEA 2013c, 2015b). Annual PM10 concentrations for monitoring stations in mining or industry hubs, coal-fired power stations and very large urban centres are reported in ‘State of the Air’ reports. We used this data and indicator in our national barometer. We used the national PM10 limit of 50 μg/m3 (DEA 2009) as our safe boundary.
For the provincial barometers, we aggregated the monitoring station data in the six provinces (Gauteng, KwaZulu-Natal, Limpopo, Mpumalanga, North West and Western Cape) used in the NAQI to determine provincial averages (see SI Table S13). The national PM10 limit decreased to 40 μg/m3 in 2015 (DEA 2014b) and we used this for our provincial boundaries. Although monitoring began in 1994, it was not comprehensive and we calculated the trend from 2003 to 2014 to ensure all relevant provinces were covered.
Similarly to Rockström et al. (2009a) and Steffen et al. (2015), we did not identify a national indicator for chemical pollution due to the lack of detailed and accurate data. Although South Africa’s National Waste Information Baseline Report (DEA 2012a) provides an estimated baseline, reporting is voluntary and measurement is incomplete.
To determine the 12 dimensions and indicators in our national barometer, we used the South African Index of Multiple Deprivation (SAIMD) (Noble et al. 2009; Wright and Noble 2009) and the annual Development Indicators report (DPME 2013), published by the South African Presidency. Both have been informed by international good practice and adapted to South African conditions and are used by the government on a regular basis. We made a number of changes to the original 11 Raworth (2012) dimensions. We separated water and sanitation into individual dimensions, we added housing, household goods and safety, and we removed resilience, social equity and gender equality. Expert interviews suggested that resilience is a cumulative effect that is dependent on the other dimensions, and therefore, an indirect measure. Experts also felt that both social equity and gender inequality should be incorporated into the other dimensions, as they are cross-cutting. Although social equality and gender equality have dedicated SDG goals (Goal 5 and Goal 10) they are mainstreamed throughout and will be covered by data disaggregation.
The social indicators in our barometer reflect national priorities and official indicators. The social floor (boundary) for each dimension is determined by the indicator selected and the goal that nobody (0% of the population) lives in deprivation. There is usually a set of indicators to choose from that reflects a range in social deprivation. The choice of indicator, therefore, partly determines the definition of the social floor.
There are three types of indicator sets that we identified. Type 1 indicators are typically reported as a range of levels of access, as are commonly found in household surveys. For example, choosing ‘access to piped water within 200 m of the dwelling’ rather than ‘access to piped water in the dwelling’ sets a lower social floor. Type 2 indicators have a range of definitions of the same broad indicator. For example, unemployment can be defined as narrow or broad, where the latter includes discouraged jobseekers. Type 3 indicators offer diverse representations of different aspects of a dimension. For example, material deprivation can be measured by ownership of a refrigerator, washing machine, radio and/or television.
We did not define an indicator for voice in the national barometer. This is because there is a lack of a generally accepted definition of voice, a lack of consensus among experts on a single indicator, as well as a large range in values for different indicators. Without other countries to compare it to, it would not have added much value. However, for the provincial barometers, we felt that it would be worthwhile to select an indicator for voice as the comparison between provinces can circumnavigate the problem of the variation in values for different indicators. Development Indicators 2012 lists four indicators under the heading ‘Social cohesion: Voice and Accountability’ that could measure voice: membership of voluntary organisations, voter turnout, female representation in parliament and the corruption perceptions index. None of these were used, however, based on expert judgment or because the indicator is not a deprivation measure or is gender-specific. The most appropriate indicators were found in the Afrobarometer, a comparative series of independent public attitude surveys on democracy and governance run since 1990 in 35 African countries (Citizen Surveys 2013). We identified 14 possible indicators, shown in SI Table S14. There is quite good correlation between the different indicators in terms of comparing the provinces. We chose the indicator ‘people who feel they are not free to say what they think’ as it is easy to understand and shows meaningful variation between provinces.
As all social data could be found at the provincial level in existing reports or databases, no special disaggregation methods was performed. Table 3 shows the 12 social dimensions, indicators and data sources in our provincial barometers, grouped into four domains—basic services, public goods, livelihoods and living standards. We largely used the 2015 General Household Survey, GHS (StatsSA 2016b), a key data source for Development Indicators, as it had the most recent data. For the indicators not covered by the GHS, we used the Development Indicators 2014 report database (DPME 2015b), the 2014/15 Victims of Crime Survey (StatsSA 2015a), the Quarterly Labour Force Survey Fourth Quarter 2015 (StatsSA 2016c), and the South African Afrobarometer Round 5 (Citizen Surveys 2013).
To plot the trends in the social dimensions, we used the same data sources so that the figures are comparable, as sometimes other data sources used different calculation methodologies. We looked for data from 1994 or similar, as we had done in the national barometer. However, we found that the Development Indicators 2014 generally reported provincial data from 2001 onwards. In the case of water and sanitation, we used the 2001 Census data in StatsSA’s SuperWeb database (StatsSA 2014) as it was not available in Development Indicators. We also used Census 2001 for household goods as it did not appear in the recent GHS’s. StatsSA’s historical revision of Labour Force Surveys (which preceded Quarterly Labour Force Surveys) (StatsSA 2009) was used for unemployment in 2001, as Development Indicators reported the narrow rather than the broad definition by province. For safety, we used the National Victims of Crime Survey 2003 (Burton et al. 2004). We could not find provincial data for three dimensions—housing, voice and income—as the indicators we used in the barometer were not reported.