Science of the Total Environment 799 (2021) 149505

Contents lists available at ScienceDirect

Science of the Total Environment

journal homepage: www.elsevier.com/locate/scitotenv

Drought risk for agricultural systems in South Africa: Drivers, spatial

patterns, and implications for drought risk management
Isabel Meza a,⁎, Ehsan Eyshi Rezaei b,h, Stefan Siebert b, Gohar Ghazaryan c,h, Hamideh Nouri b, Olena Dubovyk c,
Helena Gerdener e, Claudia Herbert d, Jürgen Kusche e, Eklavyya Popat d, Jakob Rhyner f, Andries Jordaan g,
Yvonne Walz a, Michael Hagenlocher a
a
United Nations University, Institute for Environment and Human Security (UNU-EHS), Bonn, Germany
b
Department of Crop Sciences, University of Göttingen, Göttingen, Germany
c
Center for Remote Sensing of Land Surfaces, University of Bonn, Bonn, Germany
d
Institute of Physical Geography, Goethe University Frankfurt, Frankfurt-am-Main, Germany
e
Institute of Geodesy and Geoinformation, University of Bonn, Bonn, Germany
f
University of Bonn, Bonn, Germany
g
Résilience Globale Pty Ltd, University of the Free State, South Africa
h
Leibniz Centre for Agricultural Landscape Research (ZALF), Germany

H I G H L I G H T S G R A P H I C A L A B S T R A C T

• South Africa is highly susceptible to

drought impacts on agriculture, given
its high water reliance.
• Drought risk varies substantially be-
tween irrigated and rainfed agricultural
systems.
• The most extreme drought for rainfed
croplands is observed in Northern
Cape, North West and Limpopo.
• Highest drought risk on time series for
irrigated crops is across Limpopo and
Eastern Cape.
• Our methodology to assess drought risk
is transferable to other regions.

a r t i c l e i n f o a b s t r a c t

Article history: The regular drought episodes in South Africa highlight the need to reduce drought risk by both policy and local
Received 30 March 2021 community actions. Environmental and socioeconomic factors in South Africa's agricultural system have been af-
Received in revised form 2 August 2021 fected by drought in the past, creating cascading pressures on the nation's agro-economic and water supply sys-
Accepted 3 August 2021
tems. Therefore, understanding the key drivers of all risk components through a comprehensive risk assessment
Available online 5 August 2021
must be undertaken in order to inform proactive drought risk management. This paper presents, for the first time,
Editor: G. Ashantha Goonetilleke a national drought risk assessment for irrigated and rainfed systems, that takes into account the complex inter-
action between different risk components. We use modeling and remote sensing approaches and involve na-
tional experts in selecting vulnerability indicators and providing information on human and natural drivers.
Keywords: Our results show that all municipalities have been affected by drought in the last 30 years. The years
Vulnerability 1981–1982, 1992, 2016 and 2018 were marked as the driest years during the study period (1981–2018) com-
Risk assessment pared to the reference period (1986–2015). In general, the irrigated systems are remarkably less often affected
Irrigated by drought than rainfed systems; however, most farmers on irrigated land are smallholders for whom drought
impacts can be significant. The drought risk of rainfed agricultural systems is exceptionally high in the north,

⁎ Corresponding author.
E-mail address: [email protected] (I. Meza).

https://doi.org/10.1016/j.scitotenv.2021.149505
0048-9697/© 2021 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
I. Meza, E. Eyshi Rezaei, S. Siebert et al. Science of the Total Environment 799 (2021) 149505

Rainfed central and west of the country, while for irrigated systems, there are more separate high-risk hotspots across the
Disaster risk reduction country. The vulnerability assessment identified potential entry points for disaster risk reduction at the local
municipality level, such as increasing environmental awareness, reducing land degradation and increasing
total dam and irrigation capacity.
© 2021 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://
creativecommons.org/licenses/by/4.0/).

1. Introduction profiles and responses compared to the commercial farming sector

(Thamaga-Chitja and Morojele, 2014). While commercial farming under-
Drought is a recurrent feature of all climates and among the most pins South Africa's food security, subsistence farming provides income
complex, damaging, and least understood of all so-called “natural haz- and food security on a household scale for much of the population.
ards” (Dai, 2013; Heim, 2002). It is generally defined as a period of ab- With the projected increase in the frequency, severity, and duration of
normally low precipitation (compared with the long-term average droughts (WMO World Meteorological Organization, 2020), subsistence
climate of a given region), which is long enough to severely impact farmers growing rainfed crops are particularly susceptible to drought as
the hydrological resources (IPCC, 2014). This complex phenomenon they highly depend on climate-sensitive resources (Schreiner et al.,
often leads to major impacts on the environment, society and economy 2018).
(Naumann et al., 2014), often with cascading effects. Moreover, with the South Africa has extensive disaster risk reduction (DRR) legislation
added pressures of climate change, the frequency, severity, and dura- (e.g. the National Disaster Management Act, 2002), which has evolved
tion of droughts will likely increase in many regions across the globe over the decades (Vogel and Van Zyl, 2016). Thus, various policy docu-
(Asadieh and Krakauer, 2017; Trenberth et al., 2014). The long-lasting ments, assessments and strategies for DRR have been compiled (e.g. the
impacts of droughts are felt in many sectors, including public water sup- 2004 National Climate Change Response Strategy, the 2010 National
ply, energy production, tourism and agriculture, the last often being the Climate Change Response Green Paper, and the 2011 National Climate
most heavily affected sector (Dilley et al., 2005; UNDRR, 2019). This is Change Response (Baudoin et al., 2017). Efforts to implement risk re-
more noticeable in countries with a large agricultural share of GDP or duction approaches are also supported through global frameworks
a large percentage of the labour force employed in agriculture, with such as the Sendai Framework for DRR (UNDRR, 2015), and various
the rural population particularly affected (Carrão et al., 2016). This dem- reporting commitments to international organizations (e.g. UNFCCC,
onstrates that the negative impacts associated with droughts are not UNCCD). The South African National Disaster Management Framework
only linked to the frequency, severity, and duration of drought events (NDMF) clearly states the need for disaster risk assessments (drought
but also the degree of exposure, susceptibility and coping capacity of a in this case) as one of the key performance areas for any DRR strategy
given socio-ecological system (SES) (Meza et al., 2020). Furthermore, (Jordaan et al., 2017a, 2017b). However, the South African government
the combined impacts of climate change, accelerated population has historically responded to drought with drought relief schemes that
growth, and several declining socioeconomic factors will intensify focus mainly on addressing the farmer's immediate needs rather than
drought hazards, exposure, and vulnerability in the long-term preemptively building resilience to possible future droughts (Ngaka,
(Ahmadalipour et al., 2019). This highlights the need to understand 2012; Jordaan, 2011).
and manage drought from a complex system perspective. It is necessary There is significant literature in South Africa regarding the assess-
to consider climate and environmental drivers along with socioeco- ment of drought impacts on agriculture, e.g. at national level
nomic factors that determine how susceptible a community, region, sys- (Masupha and Moeletsi, 2020; Muyambo et al., 2017; du Pisani et al.,
tem or sector is to drought and their capacity to cope. 1998), quaternary catchment level (Magombeyi and Taigbenu, 2008)
Global assessments focused on drought risk of impacts on agricul- and regional level (Kamali et al., 2018). However, when assessing the
ture have shown that southern Africa is at particularly high risk risk of drought impacts specifically for agricultural systems, there is
(Carrão et al., 2016; Meza et al., 2020). South Africa is recognized as a one assessment at national level (Schwarz et al., 2020), and there are
drought-prone country (Baudoin et al., 2017; Gibberd et al., 1996; only few studies at local level (Jordaan et al., 2013; Walz et al., 2018).
Jordaan et al., 2017a) that has experienced several “severe” drought Most of the drought risk assessments in South Africa still miss the con-
events (as occurred in early 1980s and 1990s, the period 2014–16 nection between holistic consideration of socio-ecological vulnerability,
(Baudoin et al., 2017), and the recent ongoing drought since 2018 exposure, and hazard from the local to the national scale. A comprehen-
(Mahlalela et al., 2020). During these years, environmental and socio- sive drought risk assessment is crucial to inform drought policies that
economic factors in the agricultural system of South Africa were im- foster proactive drought management (Sivakumar et al., 2014). So far
pacted by the drought, creating cascading pressures on the nation's a national drought risk assessment that integrates hazard, exposure
agro-economic and water supply systems. and vulnerability to risk for irrigated and rainfed agriculture separately
Agriculture is a core component of the economy and has major im- at the sub-national scale is lacking.
plications for job creation, food security, rural development and foreign Distinguishing the risk components for irrigated and rainfed agricul-
exchange (National Treasury, 2003). The agricultural sector directly ture is important because: i) rainfall deficit is the main factor impacting
contributes 3% to the national GDP (DAFF, 2018; Schreiner et al., drought hazard for rainfed systems while for irrigated systems, avail-
2018), and indirectly (through manufacturing, textiles, food processing) ability of irrigation water is more relevant, ii) spatial patterns of irri-
at least 14% (WWF (World Wide Fund for Nature), 2018). Approxi- gated and rainfed systems and growing periods of irrigated and
mately 8.5 million people (i.e. 14% of the population) are either directly rainfed crops are diverse resulting in different exposure of irrigated
or indirectly dependent on agriculture for employment and income and rainfed systems, iii) factors and weights affecting the vulnerability
(DAFF, 2018; Schreiner et al., 2018). of the systems differ for irrigated and rainfed systems as the vulnerabil-
The agricultural sector in South Africa is composed of commercial ity levels may constantly change due to changes in farming systems and
farmers as well as subsistence farmers. These sectors experience drought associated technologies, so that even in the same region vulnerability
risks differently. Historical root causes such as development support and can vary greatly (Downing and Bakker, 2000).
economic reforms have favoured and benefited commercial farmers Efforts to assess drought risk for agricultural systems at sub-national
who are largely exporters (FAO, 1997), exacerbating the difference in cop- level for specifics regions in the world have increased over the past
ing capacity and socio-environmental susceptibilities between the two years (Chen et al., 2017; Deng et al., 2018; Han et al., 2016;
groups. Therefore, subsistence farmers have fundamentally different risk Kamruzzaman et al., 2018; Ortega-Gaucin et al., 2021; Pei et al., 2018;

2
I. Meza, E. Eyshi Rezaei, S. Siebert et al. Science of the Total Environment 799 (2021) 149505

Zeng et al., 2019; Zhang et al., 2011); however, none of these assess- irrigated and non-irrigated crops by local municipality. The vulnerabil-
ments considered the inherent differences between irrigated and ity component was assessed through a composite-indicator based ap-
rainfed cropping systems. Frischen et al. (2020) analysed drought risk proach, where drought experts in South Africa weighted each
for irrigated and rainfed systems at the sub-national scale in indicator (Section 2.5). Then, the drought hazard, exposure and vulner-
Zimbabwe, however, the only differentiation in the methodology for ability information was compiled into a final drought risk assessment
each cropping system was considered at the exposure component (Section 2.6), which resulted in integrated risk maps for both rainfed
while the hazard and vulnerability indicators were the same for both and irrigated agricultural systems, respectively (Section 3). Lastly, the
systems. paper discusses the results (Section 4) and identifies potential ways for-
This paper aims at addressing the above gaps by conducting a sector- ward, including future research needs.
specific assessment of the drivers and spatial patterns of drought risk for
rainfed and irrigated agricultural systems in South Africa in order to 2. Data and methods
identify entry points for action. This is the first integrated drought risk
assessment for South Africa at the sub-national level, which considers 2.1. Case study region
spatio-temporal consistent hazard-specific indicators, complemented
by drought exposure and socio-ecological vulnerability factors – South Africa is located in the southern part of Africa, spreading over
weighted by local experts - at the local municipality scale, specifically 122 million ha with approximately 12% croplands (FAO, 2020a). The
for irrigated and rainfed agricultural systems. country is composed of nine provinces and has a wide range of climates
The paper presents a risk assessment based on a mixed-method ap- from arid to subtropical, temperate, and mediterranean (Fig. 1)
proach, starting from the hazard assessment (Section 2.3), which is (Waldner et al., 2017). About 91% of South African territory is arid or
based on composite drought hazard indicators calculated for irrigated semi-arid, with only 10% of the land generating half of the annual run-
and rainfed crop systems separately using drought indices based on his- off (Le Maitre et al., 2018). The country has uneven rainfall distribution
torical climate conditions (1986–2015). The exposed elements are de- with a mean annual rainfall of 550 mm and annual mean temperature of
scribed in Section 2.4 and were derived from a dataset differentiating 18 °C (FAO, 2020a). The potential annual mean evaporation for the

Fig. 1. a) Köppen-Geiger climate classification map for South Africa (1980–2006) (Beck et al., 2018). b) South African provinces. c) and d) Rainfed and irrigated areas per municipality,
respectively. e) Ratio between irrigated and total agricultural area per municipality. f) Irrigated and rainfed agriculture in South Africa at pixel level. Maps are based on data from the na-
tional land use/land cover dataset 2018 (Thompson, 2019). Black lines indicate provincial boundaries.

3
I. Meza, E. Eyshi Rezaei, S. Siebert et al. Science of the Total Environment 799 (2021) 149505

whole country is about three times greater than its annual rainfall, Agriculture (2020) over 60% of South Africans did not have their land/
1800 mm per year (WWFW (World Wide Fund for Nature), 2018). Ac- property rights recorded or registered.
cording to the general household survey performed in 2018 almost 15%
of the households were active in agricultural activities, of which more 2.2. Risk framing and workflow
than 75% are involved in order to ensure an additional source of food
(DALRRD (Department of Agriculture, 2020). Following the IPCC (2014) definition, risk results from the interac-
The agricultural economy comprises technically developed commer- tion between hazard, with exposure of human and natural systems
cial farming on the one hand and more subsistence-based production in and the systems' vulnerabilities. In this paper, exposure is defined as
the remote rural areas on the other hand (Waldner et al., 2017). The the presence of agricultural systems that could be negatively affected
dominant activities include: i) intensive crop production and mixed by hazards. Vulnerability is the predisposition or propensity to be ad-
farming in areas characterised by winter and summer rainfall, ii) cattle versely affected by drought. It encompasses a variety of concepts and el-
ranching in the bushveld and iii) sheep farming in the arid regions ements, including social-ecological sensitivity or susceptibility to harm
(Waldner et al., 2017). Considering climate and soil properties, only and lack of capacity to cope (IPCC, 2014). Also, following the IPCC
12% of the country is suitable for crop production; of which 22% is con- (2014) definition, susceptibility is understood as the likelihood of suf-
sidered as high potential land in terms of production capacity (Waldner fering harm in the event of a drought hazard process, and coping capac-
et al., 2017; WWFW (World Wide Fund for Nature), 2018). ities refer to the use of available skills, opportunities, and resources to
In general, rainfed agriculture prevails in South Africa, accounting for address, manage, and overcome adverse conditions in order to achieve
the majority of the harvested area (Fig. 1) (Hardy et al., 2011). This basic functioning in short to medium terms. The workflow for the
means that only 1.35 million ha (8.5%) of the potentially arable land is three risk components and risk aggregation is visualized in Fig. 2; the in-
irrigated (DAFF, 2019). Nevertheless, irrigated agriculture contributes dicators and data sources for hazard, exposure and vulnerability are pre-
30% to agricultural production (FAO, 2020b). Irrigation application in sented in Tables 1 and 2.
South Africa can be permanent, supplementary, or occasional. Most of
the commercial irrigation schemes are located in large river basins 2.3. Hazard assessment
(e.g. Orange, Lower Vaal, Fish) and in the Western Cape region (FAO,
2016). 2.3.1. Rainfed hazard composite index
South Africa has been frequently affected by droughts in the last four The rainfed hazard indicator was computed using the ratio between
decades. Major drought periods include 1982–1984, 1991–1992, actual evapotranspiration (AET) and potential (PET) evapotranspiration
1994–1995, 2004–2005, 2008–2009, 2015–2016, and the most recent of crops in the crop growing season for the period 1981–2018. AET re-
in 2018–2020 (Mahlalela et al., 2020; FAO, 2019; Walz et al., 2020; fers to the amount of water consumed by a crop and evaporated from
Unganai and Kogan, 1998). During those years, drought not only im- the soil under actual soil moisture calculated by performing a soil
pacted the environment, but also the social and the economic systems. water balance in daily time steps, while PET assumes no limitation in
The 1992 drought affected around 250,000 people, with an estimated crop water availability. The ratio is highly associated with crop yield
50,000 job losses in the agriculture sector, and 20,000 additional jobs and is widely used as a drought indicator for cropland (Peng et al.,
losses in related sectors (AFRA (Association for Rural Advancement), 2019). The Global Crop Water Model (GCWM) (Siebert and Döll,
1993). In 2007–2008, the South African government spent over R285 2010) was employed to simulate AET and PET for specific crops grown
million (19 million US dollars) on drought relief measures for the agri- in South Africa based on prescribed crop calendars and cropping pat-
cultural sector, primarily on the purchase and supply of subsidised fod- terns derived from the MIRCA2000 dataset (Portmann et al., 2010).
der depending on farms' sizes (Ngaka, 2012). Recent droughts such as The ERA5 global reanalysis (Hersbach et al., 2020) and ISRIC-
the one in 2015–2016 revealed the cascading impacts of the drought. WISE30sec v1.0 (Batjes, 2016) were used as the climate and soil input.
The BFAP (Bureau for Food and Agricultural Policy) (2016) reported The spatial resolution of GCWM's is five arcmin (8.3 km). Drought haz-
that the area of maize planted for the 2016–17 season was 25% lower ard in specific years was defined as deviation from the long-term mean
than the area planted in the 2015–16 season, which was reflected in condition in the reference period 1986–2015 (Meza et al., 2020). The
the year-on-year declines in seasonally adjusted sectoral GDP. In addi- annual hazard indicator for rainfed agricultural systems CH_RfAgy was
tion to the direct impact on agriculture, general economic indicators calculated as:
pointed to an aggravated situation (e.g. input providers were hard hit
due to the lack of purchasing power in the agricultural sector; given AETy=PETy
CH_RfAgy ¼ 1− ð1Þ
the suppliers' import propensity and the local currency depreciation AET=PET
(BFAP (Bureau for Food and Agricultural Policy), 2016)). Inflationary
pressures resulting, inter alia, from drastic increases in food prices where AETy and PETy are annual sums of actual and potential evapo-
drove up interest rates, which had a negative effect on farming enter- transpiration of all cultivated crops in year y (m3 yr−1). AETand PETare
prises' debt servicing costs and further restricted access to credit in the long-term annual mean of actual and potential evapotranspiration
the sector (BFAP (Bureau for Food and Agricultural Policy), 2016). (m3 yr−1) in the reference period 1986–2015. Consequently, positive
Drought policy and strategies have included efforts from as early as values of CH_RfAgy represent conditions dryer than usual, while nega-
the 1920s, concentrating on land use change, land reforms, soil manage- tive values indicate wet years. The long term hazard during the study
ment and agricultural practices (e.g. kraaling of stock) (Bruwer, 1993; period at grid level was computed as the frequency (percentile rank)
Hassan, 2013). The most recent strategy towards drought is compiled of years in which the AET/PET ratio was at least 10% lower than the
in the National Development Plan which sets a vision of eliminating mean AET/PET ratio in the reference period 1986–2015 (Meza et al.,
poverty and reducing inequality by 2030 (DALRRD (Department of 2020). A long-term hazard of 0.5 means therefore that in every second
Agriculture, 2020). However, a rethinking of drought governance is year the AET/PET ratio is lower than 90% of the long-term mean AET/
still required, which should look back in time and critically reflect on PET ratio.
past drought experiences, perceptions and needs of drought risk reduc-
tion and how local context influences drought response (Baudoin et al., 2.3.2. Irrigated hazard composite index
2017; Vogel and Olivier, 2019). The government is still challenged to The irrigated hazard index CH_IrrigAg y (-) is defined based on
change the unbalanced land-ownership patterns while sustaining eco- the annual difference between the water resource available for
nomic growth, food security and implementing effective drought man- irrigation and irrigation water requirement. The water resource
agement plans; as by 2018 according to the DALRRD (Department of available for irrigation was simulated using the WaterGAP model

4
I. Meza, E. Eyshi Rezaei, S. Siebert et al. Science of the Total Environment 799 (2021) 149505

Fig. 2. Workflow for the drought risk assessment for irrigated and rainfed agricultural systems in South Africa. The workflow is explained in detail in Sections 2.3–2.6.

(Müller Schmied et al., 2020) as annual sum of discharge Q at a spatial was computed as the frequency of the years with an irrigated hazard in-
resolution of 30 arcmin for the period 1981–2018. The irrigation water re- dex CH_IrrigAgy of bigger than 0.5 meaning that the deficit in the annual
quirement IWR was simulated using GCWM as the volume of water difference between discharge and irrigation requirement exceeded half
needed to increase the AET of irrigated crops to their PET (Siebert and of the long-term median of annual discharge. A long term hazard for ir-
Döll, 2010). Drought hazard for irrigated crops CH_IrrigAgy was computed rigated conditions of 0.2 means then that such a deficit occurs every 5
for each year as: years.


ðQ−IWRÞmed − Q y −IWRy
CH_IrrigAg y ¼ ð2Þ 2.4. Exposure assessment
Q med
Based on the drought risk assessment workflow (Fig. 2), agricultural
where (Q − IWR)med is the median of the difference between discharge land (irrigated and rainfed) was used to analyse drought exposure. The
and irrigation water requirement (m3 yr−1) in the reference period estimation of exposed agricultural land was based on the South African
1986–2015, Qy and IWRy are discharge and irrigation water require- National Land Cover dataset 2018 (Thompson, 2019), from which irri-
ments in year y (m3 yr−1), and Qmed is the median of the annual dis- gated and rainfed land were extracted as separate classes. The SANLC
charge in the reference period 1986-2015 (m3 yr−1). Positive values of 2018 map has 20 m spatial resolution and was generated using multi-
CH_IrrigAgy indicate drought, while negative values indicate that the dif- seasonal Sentinel 2 satellite time series data acquired during the period
ference between water resources and water demand for irrigation is 01 January 2018 to 31 December 2018, 20 m spatial resolution and
larger than usual (wetness). Both models (GCWM and WaterGAP) used 90.14% accuracy (Thompson, 2019). Rainfed systems are mostly located
the same soil and climate input data and the same simulation period in the North Eastern provinces, as well as in Northern and Western Cape
(1981–2018). The outputs of GCWM (for crops grown in South Africa) (DAFF, 2018). The hazard indicators - CH_RfAgy and CH_IrrigAgy - were
were aggregated to 30 arcmin to match the spatial resolution used by aggregated from pixel to municipality level as average of the pixel
WaterGAP. The long-term hazard for irrigated conditions at grid level values, using the rainfed or irrigated area within each pixel derived

Table 1
Hazard and exposure indicators used for the irrigated and rainfed assessment and the origin of the input data.

Risk component Agricultural system Indicator Data source Processed data

Drought hazard Irrigated Water availability WaterGAP (Müller Schmied et al., 2020) Annual time series of the difference between discharge Q
Water requirement GCWM (Siebert and Döll, 2010) and irrigation requirement IR compared to the long-term
(1986-2015) mean of that difference (Eq. (2))
Calculated for period 1981-2018
Rainfed Crop drought stress GCWM (Siebert and Döll, 2010) Annual time series of the deviation of the ratio AET / PET
from the long-term (1986-2015) mean of that ratio (Eq. (1))
calculated for period 1981-2018
Exposed elements Rainfed or irrigated Area rainfed or irrigated Thompson, 2019 National land use/land cover dataset 2018 (DEA, 2019)
in the local municipality differentiating between rainfed and irrigated agriculture

5
I. Meza, E. Eyshi Rezaei, S. Siebert et al. Science of the Total Environment 799 (2021) 149505

Table 2
Final list of indicators used to perform the vulnerability assessment with expert weighting for irrigated and rainfed systems. The weights with a value close to 1 are highly relevant, whereas
indicators with a value close to 0 indicate lower relevance. Only indicators with selected values were used for the respective vulnerability assessment (Irrigated, Rainfed).

Indicator Direction Data source Expert weight Expert weight

irrigated rainfed

Social susceptibility
Unemployment rate (%) + StatSA, 2011 1.00 0.91
Population with assistive devices and + StatsSA census 2011 (Boundaries 2016) - Disability 0.95 0.76
medication-Chronic medication
Population with inadequate sanitation/sewerage/toilet + StatsSA (Community survey 2016) 0.91 0.75
services
Population with environmental awareness by district – StatsSA (Labour Force Survey) 0.89 0.88
Dependency ratio (population at the age of 0-14 and > + StatSA (Agricultural Household survey 2016) 0.88 0.79
65)
Accessibility to high-density urban centers by travel + Weiss et al., 2018 0.85 0.79
time
HH with alternative on farm income + StatsSA (Agricultural Household survey 2016) 0.84 1.00
People skipping a meal for five or more days in the past + StatsSA (Community survey 2016) 0.83 0.88
30 days
Population that have experienced violence and crime + StatsSA (Community survey 2016) 0.81 0.78
Debtors by municipality (%) + National Treasury (Balance Sheet) Municipal Finance Data Tables 0.73 0.95
Hydropower installed capacity [MW] + World Bank (Global Dams Database, 2020) and the Global Reservoir 0.71 No selected
and Dam Database (GRanD)
Gender inequality (gender parity) + SatsSA 2016 Gender Series Empowerment 0.69 0.70
Population per municipality that rate the overall quality + StatsSA (Community survey 2016) 0.63 0.68
of the water services poor
Population that has experience of crime - Theft of + StatsSA (Community survey 2016) 0.63 0.61
livestock; poultry and other animals
Population with ill-health (mental) (%) + StatsSA (Community Survey 2007) - Disability No selected 0.73

Environmental susceptibility
Farm land ratio + StatsSA (Agricultural household survey 2016) 0.89 0.85
Land Degradation Index (LADA) + Department of Agriculture, Forestry and Fisheries, 2016 DAFF) 0.87 0.86
Clay content (0-2 micro meter) in (g/100 g) (w%) at – Hengl et al., 2015 0.80 0.81
depth 0-5 cm
Maximum fertilizer application rate kg/h – Mueller et al., 2012 and West et al., 2014 for mineral fertilizer data and 0.79 0.74
manure and atmospheric deposition.

Coping capacity
Total dam storage capacity in million cubic meters – Lehner et al., 2011 for GRanD 0.87 0.70
Borrowed money from total municipality liability + National Treasury (Balance Sheet) Municipal Finance Data Tables 0.80 0.84
People that receive social grants – StatsSA (Welfare - Community Survey 2007) 0.75 0.85
Road density m/km2 – GloBio (Global Roads Inventory Project (GRIP) dataset) 0.72 0.75
Area equipped for irrigation expressed as percentage of + FAO, 2020a, 2020b, 2020c No selected 1.00
total area

from the SANLC 2018 dataset for weighting. From this point, the com- array of methods for assessing vulnerability to natural hazards exists,
bined components of hazard and exposure are referred to as ‘hazard/ indicator-based approaches are among the most common to represent
exposure’. the multi-dimensional nature of vulnerability (Hagenlocher et al.,
The simulated hazard/exposure for rainfed conditions was validated 2019; de Sherbinin et al., 2019). For this assessment, composite indica-
using the remotely sensed AET/PET ratio in the period 2001–2018. AET tors were developed according to the impacted sector: i) irrigated agri-
and PET values were extracted from the Moderate Resolution Imaging culture and ii) rainfed agriculture, considering a wide array of
Spectroradiometer (MODIS) product (MOD16A2.006) which provides environmental, social, and economic indicators.
data at 500 m spatial resolution (Running et al., 2017). The dataset is de- Relevant indicators were identified through a combination of litera-
rived from meteorological reanalysis data coupled with remotely ture review and expert consultation. The review was conducted based
sensed products of land cover and vegetation properties (Huang et al., on pre-defined search terms (Table S1) in Web of Science and Scopus.
2017). The dataset was preprocessed based on the quality control The selected articles (n = 17) were coded with MAXQDA software
layer, and pixels with low quality were excluded. The original data set (VERBI Software, 2019) to extract suitable indicators. Later, these indi-
provided the AET and PET in 8 days intervals, which were summed up cators were compared and complemented with the ones identified by
to yearly values. The CH_RfAgy was recalculated for model results and Hagenlocher et al. (2019) in their review of existing drought risk assess-
remote sensing observations considering the reference period 2001- ments, and within South Africa at a local municipality level by Walz
2018 to account for the limited availability of remote sensing observa- et al. (2018) and a quaternary catchment level by Jordaan et al.
tions. Both datasets were aggregated to the municipality level consider- (2017a, 2017b). In total, 44 suitable indicators for rainfed and irrigated
ing the extent of the rainfed growing area in each pixel. The Pearson systems in South Africa were identified (Fig. S2).
correlation coefficient was calculated between model and remote sens- To assess which of those 44 indicators are the most relevant for
ing driven CH_RfAgy at the municipality level. representing vulnerability of these two systems towards drought, an
online expert survey was conducted as a joint effort with the National
2.5. Vulnerability assessment Disaster Management Centre (NDMC) of South Africa. A total of 33 ex-
perts representing all provinces of South Africa participated in this sur-
Drought impacts are often associated with drought hazard severity, vey. They selected 36 relevant indicators for irrigated systems and 40 for
but the degree of the impact is mediated by the vulnerability of the ex- rainfed (Fig. S2). These experts were from multiple sectors including ac-
posed agricultural system, i.e. its susceptibility and the (lack of) capacity ademia (n = 4), private sector (n = 5), NGO (n = 1), government (n =
to cope with drought events (IPCC, 2014; World Bank, 2019). While an 20), international organizations (n = 1) and others (n = 2). The final

6
I. Meza, E. Eyshi Rezaei, S. Siebert et al. Science of the Total Environment 799 (2021) 149505

selection of relevant indicators for each agricultural system based on the using equal intervals from the maximum, and then those classes
survey results followed a two-step approach as proposed by the were combined to obtain the final risk for each agricultural system
(European Commission, JRC, 2019): i) Indicators were kept if more (Supplementary material Fig. S8).
than half of the experts considered them a medium-high or highly rele-
vant and ii) Z-scores with a 95% confidence interval were used to ensure 3. Results
that there was high level of agreement among the experts. The data was
then standardized to give each indicator a value between 0 and 1 in each 3.1. Drought hazard and exposure of agricultural systems
category (i.e. not relevant, low relevance, low-medium relevance, me-
dium high relevance and highly relevant). The average was then calcu- Our results demonstrate a large variability in drought hazard and ex-
lated by dividing the total number of replies given for each indicator by posure among provinces and local municipalities. The most extreme
the total number of answers given for each indication. Indications with a drought hazard/exposure for rainfed conditions is observed in the
value near 1 are extremely relevant, while indicators with a value near 0 North Cape, North West and Limpopo provinces during the study period
are less relevant (Fig. S2). (Fig. 3). On the other hand, the lowest hazard and exposure in the pe-
Open-source data for the selected indicators was retrieved (Table 2, riod 1981-2018 is computed for Kwazulu Natal province (Fig. 3).
e.g. statistics from StatSA (2011, 2016); National Treasury (2019), Western and central parts of Eastern Cape and Mpumalanga provinces
World Bank (2019, 2020)) in order to ensure that the final results can also have a low level of rainfed drought hazard/exposure (Fig. 3). The
be validated and reproduced in a different context - as recommended time series analysis of drought hazard and exposure showed that
by Naumann et al. (2014). Following the methodological suggestions 1992 and 2016 were the driest years during the study period under
by Hagenlocher et al. (2018), Meza et al. (2020), Naumann et al. rainfed conditions (Figs. 4 and S4). The year 2000 and 2006 are classi-
(2014), and OECD (2008), statistical operations were performed to pre- fied as wettest years across South Africa (supplementary material
pare an indicator dataset to perform the vulnerability assessment (S1 & (Figs. 4 and S4). The frequency of dry years for rainfed systems remark-
Fig. S1): i.e., i) imputation of missing data, ii) normality test, iii) outlier ably increased after year 2010.
detection and treatment, iv) multicollinearity assessment, In general, the irrigated systems are less often affected by drought
v) normalization and vi) expert weighted aggregation. than rainfed systems, with larger areas exposed to drought in Limpopo
The selected vulnerability indicators were normalized to make them and Eastern Cape provinces of South Africa (Fig. 3). These areas have
comparable. A linear min-max normalization was applied to create a semi-arid to arid climates and are characterised with less annual precip-
range between 0 (lowest vulnerability) to 1 (highest vulnerability) itation than the rainfed growing areas of the country. For irrigated crop-
(Beccari, 2016; Carrão et al., 2016). lands, larger areas were affected by drought hazard/exposure since
The final step to build the composite vulnerability index (CVI) for 2012, even in areas that have low share of irrigated croplands, such as
each agricultural system (irrigated and rainfed) was the weighted arith- north western municipalities in the Northern Cape (Figs. 3 and 5). De-
metic aggregation for each vulnerability component (SOC-ENV_SUS and spite smaller areas of hazard/exposed irrigated land compared to
lack of COP) based on the normalized indicators (Zi) and the weights ob- rainfed areas, the impacts can be significant due to the number of af-
tained from the expert survey (Wi). fected people. Roughly about 230,000 irrigation farmers were affected,
mostly smallholders often with very small plots for self-consumption
n n
CVIIrrigated ¼ ∑i¼1 ðZi ∗ WiÞ CVIrainfed∑i¼1 ðZi ∗ WiÞ ð3Þ (FAO, 2016). The highest hazard/exposure was found in years 2015-
2016 and the lowest in year 2001 (Figs. 5 and S5).
The accuracy of simulated hazard/exposure for rainfed agricultural sys-
2.5.1. Reliability analysis tems was tested by comparing modeling outputs with remotely sensed
In order to increase the transparency on the data quality used to exposure data in the period 2001-2018 (Fig. 6). There was a strong corre-
perform the vulnerability assessment, a metric to calculate the reliabil- lation (0.5 to 0.9) between remotely sensed and simulated drought expo-
ity of the data for each local municipality was developed. Following sure for rainfed conditions for most of the municipalities across South
suggestions of the European Commission, JRC (2017) in their Index for Africa. The lowest correlation (0 to 0.2) was obtained in a limited number
Risk Management (INFORM) and Hagenlocher et al. (2018), the reliabil- of municipalities mainly in KwaZulu-Natal and Eastern Cape provinces,
ity metric included two dimensions i) average year of the data sources which are largely covered by natural grasslands. The annual drought signal
(recency) and ii) percentage of missing data across all indicators. Each obtained by remote sensing may therefore deviate considerably from the
dimension score was then normalized to a scale from 0 to 1, aggregated conditions in the cropping period considered in the model.
and averaged in order to have the final reliability scores. Where the ten- Moreover, we assessed the relationships between annual drought
dency to 1 indicates that the vulnerability score for that particular local exposure simulated for rainfed systems and yield/production reported
municipality is based on more reliable data, while the tendency to 0 in- at the country scale (FAO, 2021). The correlation coefficient among sim-
dicates less reliable data (Supplementary material Fig. S3). ulated drought exposure and reported yield and production anomalies
The reliability metric was computed separately for each of the were −0.32 and −0.41, respectively (Fig. S6) which means that drought
two agricultural systems considered in this article (irrigated and resulted in lower yields and production. The model reproduced the
rainfed). drought for the years (1992–2015–2016) which showed the largest
yield/production reduction. As a second analysis, we performed the as-
2.6. Risk assessment sessment for maize production anomaly in South Africa in the period
1986 to 2018 and its relationship with the annual rainfed hazard/
Drought risk, in any particular area, is composed of hazard, expo- drought simulated for rainfed systems across five most important
sure, and vulnerability (IPCC, 2014). For this paper, hazard/exposure maize production provinces in South Africa (Fig. S7). The results
and vulnerability were combined through a matrix approach showed a remarkable overlap between negative production anomalies
(Fig. S8). Two different drought risk assessments were performed - and simulated drought hazard for all provinces, e.g. in years 1992-93,
one for irrigated agricultural systems and one for rainfed systems - 2007, 2013 and 2016. In contrast, positive production anomalies were
at the municipality level. Following methodological suggestions of recorded in all provinces in years with low drought hazard such as
the International Standard on Risk Norm ISO/IEC 31010 (IEC 1991, 2006 or 2014 (Fig. S7). It is important to note that the FAO and re-
(International Electrotechnical Commission), 2019), Frigerio and gional yield/production data did not distinguish between rainfed and ir-
De Amicis (2016) and Tung et al. (2019) the CVI and hazard/expo- rigated systems. Therefore, we expected even higher correlations when
sure for each agricultural system was classified into seven classes separate data would become available.

7
I. Meza, E. Eyshi Rezaei, S. Siebert et al. Science of the Total Environment 799 (2021) 149505

Fig. 3. Long-term drought hazard and combined hazard/exposure for rainfed (top row) and irrigated (bottom row) cropping systems across South Africa at grid and local municipality
levels in the period 1981–2018.

3.2. Vulnerability and risk of rainfed and irrigated systems relevant are the lack of area equipped for irrigation, which affects the
coping capacity of the system, followed by a low fertilizer application
The vulnerability assessment shows heterogeneity across the coun- rate.
try (Fig. 7) for both systems. Our assessment highlights that crops According to the experts (Table 2 and Fig. S2), the most relevant vul-
under rainfed systems are more vulnerable to drought than irrigated nerability indicator for irrigated systems is unemployment rate (%). This
systems. Several indicators contribute to the difference, but the most is also recognized as a relevant indicator by the scientific community in

Fig. 4. Annual drought hazard/exposure for rainfed cropping systems across local municipalities of South Africa in the period 1981–2018.

8
I. Meza, E. Eyshi Rezaei, S. Siebert et al. Science of the Total Environment 799 (2021) 149505

Fig. 5. Drought hazard/exposure for the irrigated cropping system across local municipalities of South Africa for the period 1981–2018.

the South African context as the country suffers from deep structural on their farm income (~97%); this could result from a drought period
unemployment having a direct impact on poverty levels (Chibba and that requires compromising their entire livelihoods. Having an alterna-
Luiz, 2011). Agriculture proved to be the best way to reduce rural pov- tive income may increase their coping capacities as they do not depend
erty according to the rural development literature, besides, in most de- solely on the agricultural income derived from crop sales.
veloping countries, agriculture and agriculture-related activities The experts assigned to the two indicators “population with assistive
provide most of the rural employment (Machethe, 2004). Irrigation devices and medication (disability)” and “total dam storage capacity”
schemes have had great impact in South Africa, not only in food produc- high weights for irrigated systems but much lower weights for rainfed
tion but also alleviating poverty. One notable example is the one caused systems. In contrast, the indicators “households with alternative farm
by the Great Depression by resettling of returning soldiers that reduced income” and “debtors” received high weights for rainfed systems and
the unemployment rate in the country (FAO, 2016). Irrigated agricul- much lower importance for irrigated systems.
ture employs between 10% and 15% of the total agricultural workforce The vulnerability maps display high values particularly on irrigated
(DWA, 2002). systems for the Western Cape municipalities and for rainfed agricultural
The most relevant indicator for rainfed systems according to the ex- systems in KwaZulu-Natal. Our findings underline that determining fac-
perts (Table 2) is the percentage of households with an alternative to tors of vulnerability vary depending on the sector which is susceptible
farm income. Low harvests threaten the households that only depend to the negative impacts of drought. For instance, the main indicators
which shape the vulnerability for irrigated systems and are potential
entry points for the drought risk reduction is the lack of environmental
awareness, poor water quality, and low total dam storage capacity. In
the South African context this is due to the limited access to extension
services (e.g. geographically remote farmers tend to have little network
coverage), and very limited financial resources to invest in technologies
or utilities. Resulting in a lack of accessible, relevant, and practical infor-
mation to share, as well as few or no opportunities to expand the irriga-
tion farmers' capacities (FAO, 2020b).
For rainfed agricultural systems, the key indicators shaping the
socio-environmental susceptibility and the coping capacities of the
local municipalities are the small fertilizer application rate, the lack of
area equipped for irrigation, and land degradation. This last indicator
is relevant for both systems; land degradation is linked to different fac-
tors in the context of agricultural systems in South Africa, one of them is
the lack of environmental awareness that led to unsustainable farming
practices (Rother et al., 2008; Schulze, 2016).
Fig. 6. Correlation coefficient between drought exposure of rainfed systems obtained by The drought risk assessment highlights its context-specificity and
modeling and remote sensing. how different communities of a country experience different levels of

9
I. Meza, E. Eyshi Rezaei, S. Siebert et al. Science of the Total Environment 799 (2021) 149505

Fig. 7. Drought vulnerability and risk in South Africa at local municipality level for rainfed (top row) and irrigated agriculture (bottom row). Tendency to dark blue shows lower levels of
vulnerability and risk, the tendency to red shows higher vulnerability or risk values.

risk. Drought risk varies substantially for rainfed and irrigated systems The Northern-Cape province has the lowest population dependent
(Fig. 7). There is a high-risk pattern towards the North provinces for on crops. However, it is one of the provinces with more local municipal-
rainfed agricultural systems. Meanwhile, high-risk hotspots for irrigated ities on high rainfed risk, as this province has arid climate which exposes
agricultural systems can be found in some local municipalities of rainfed crops to high drought hazard. In contrast, the Limpopo province
Limpopo (e.g. Modimolle, Polokwane local municipalities), North has a higher amount of population dependent on crops, but more local
West (e.g. Merafong, Rustenburg) and Gauteng (e.g. Merafong city, municipalities are at high risk for irrigated systems.
Rand West city) provinces.
When analyzing the risk for rainfed systems, among the local munic- 4. Discussion
ipalities in the Northern Cape, Emthanjeni has the lowest risk score than
other provinces despite its high hazard and exposure levels; it is ex- The dependency of agriculture on water resources (approx. 60% of
plained by a lower social susceptibility (e.g. overall quality of water ser- the total water demand (Schreiner et al., 2018) is making water avail-
vices, less population have experienced crime and theft of livestock), ability one of the key factors for the agricultural system, furthermore,
and higher coping capacities (e.g. access to credits). In contrast, the the predominance of rain-fed agriculture in South Africa makes the
local municipality of Khai-Ma in the same province has lower vulnera- country extremely susceptible to drought. Despite this, drought risk
bility than other local municipalities, but its high hazard and exposure management remains ambiguous and mainly reactive (Hornby et al.,
scores result in a high risk. 2016; Baudoin et al., 2017; Vogel and Olivier, 2019). Drought is a recur-
In order to identify priority areas for disaster risk management, the rent phenomenon in South Africa's climate and is one of the most rele-
risk assessment of each agricultural system was plotted against the vant hazards (Gibberd et al., 1996; Jordaan et al., 2017a). In fact, all local
crop dependent population in each local municipality (Fig. 8). The com- municipalities were affected by drought during the last 30 years (Figs. 4
parison shows that the local municipalities with higher irrigated and and 5). The dependency of South Africa's economy on agricultural prod-
rainfed systems are not among the highest in terms of crop dependent ucts emphasises the importance of drought risk assessments and the
population. The city of Johannesburg presents a higher crop depen- identification of potential entry points for reducing its vulnerability.
dency, but also has high risk for both systems. Its drought hazard and An integrated hazard, exposure and vulnerability assessment of the ag-
exposure are high, and the vulnerability analysis reveals that their lack ricultural sector (irrigated, rainfed) specifically was lacking so far for
of environmental awareness, fertilization rate and land degradation South Africa at national level, and it is presented here for the first
are key factors contributing to their overall very high risk; highlighting time. Furthermore, the methodology can be transferable in other re-
the relevance to take actions in this municipality. Johannesburg, the gions, the hazard and exposure assessment can be reproduced in any
largest city in South Africa, is facing enormous challenges which reflect country, however the vulnerability assessment is context specific and
on the drought vulnerability level. Challenges like urbanisation's impact some indicators that might be relevant for South Africa will not be for
on the soil and water quality and availability, and facing non- another country, therefore, we suggest to identify key indicators follow-
sustainable growth paths (SACN, 2016) have significant impacts on ing the methodology applied on this paper.
the magnitude of Johannesburg's vulnerability towards drought.
In contrast, the city of Tshwane has a high number of crop depen- 4.1. Limitations
dent population, but it presents a medium rainfed risk and very low ir-
rigated risk. Its medium risk is explained by its medium-low Our innovative methodology to simulate hazard indicators captured
vulnerability as a result of better performance in nutrition level, good the spatiotemporal pattern of the drought for a long-term period (back
water quality and road density, among others. to 1980s); the time that remote sensing was not available (generally

10
I. Meza, E. Eyshi Rezaei, S. Siebert et al. Science of the Total Environment 799 (2021) 149505

Fig. 8. Local municipalities contrasted with drought risk for rainfed (x axis) and irrigated (y axis) systems. The size of the bubbles represent the amount of crop dependent population by
local municipality (Data from Statsa, 2016).

available from early 2000s). Our results show that exposure to drought Cape's vulnerability pattern follows an east-west descending gradient
in croplands varies for rainfed and irrigated systems, spatially and tem- reported by Walz et al. (2018). Jordaan et al. (2013) showed that the
porally. A time series of exposure for irrigated and rainfed agriculture coping capacities such as the land ratio and management, access to
shows different patterns; this proves the necessity for separate analysis credit and markets are key in determining the level of risk in the North-
for these two cropping systems. The hazard indicators for rainfed and ir- ern Cape. Similar to the results of this study, Schreiner et al. (2018) sug-
rigated systems were computed in different ways; for rainfed systems, gests that expanding the storage capacity of existing dams and water
we assume a strong impact of meteorological drought on the system conservation practices would reduce drought risk, especially for irri-
while for irrigated systems, we assume a strong impact of hydrological gated agricultural systems. Furthermore, the low drought risk values
drought on the system. Therefore, hazard indicators and, subsequently, identified for the Northern Cape for irrigated systems also agree with
risk indicators for irrigated and rainfed systems should not be directly previous drought risk assessments performed by Jordaan (2011) and
compared. Jordaan et al. (2013).
To better manage and mitigate drought risk, it is necessary to im- It is necessary to analyse and interpret the drought risk through sys-
prove the response to drought impacts, the preventive actions and ac- tems perspective (Vogel and Olivier, 2019), as extreme droughts and its
tively address the root causes of vulnerability as well as build impacts are not a result of a linear equation, rather they reflect the dy-
capacities as in the local communities and the government. The vulner- namic and complex realities of the socio-ecological system. To address
ability assessment helps to identify potential entry points to reduce the the complex realities in this assessment, we considered the nature of
level of drought risk for both irrigated and rainfed systems; which in- farming in South Africa in terms of climate and social factors (e.g. de-
clude better water quality, reduction of land degradation, and increas- pendency ratio, unemployment rate). An enhancement for future as-
ing the dam storage capacity. Specifically for rainfed systems with sessments could be the integration of temporal dynamic exposure and
high risk could become irrigated if they are located in regions where ir- vulnerability with the hazard data. As Schreiner et al. (2018) stated,
rigated risk is low as areas equipped for irrigation can help in supporting the South African government knows that drought is a recurrent hazard,
the livelihood of rural communities and food production. However, it is and particularly with climate change, it is critical to implement the nec-
relevant to consider the water availability, the access to the water essary structures to support the diverse makeup of the agricultural sec-
source, the soil and topography conditions, among others before tor. Further, it is necessary to plan actions according to specific needs of
installing any irrigation system. the system, irrigated or rainfed. We also need to understand better how
The contribution of relevant experts on selecting and weighting vul- severe, prolonged and repetitive drought events might shift policies,
nerability indicators is an added value of this assessment. However, the local and rural economies, and actions (Schreiner et al., 2018).
expert survey consultation could be enhanced by expert interviews, Despite the wealth of climate change and drought policies and re-
where more details and the rationality behind the ranking of the differ- sponses in South Africa, recent droughts are a stark reminder of the real-
ent indicators could be further explained. Another point of improve- ities of climate variability and the difficulty of effectively responding.
ment is the number of experts who responded to the survey. With Notwithstanding the examples and legislation mentioned, recent re-
more time and resources, more experts could participate. sponses to drought reveal a lack of awareness and a need for a broadly in-
As this study is the first to separately assess the drought risk for formed assessment of drought in a rapidly changing socio-environmental
rainfed and irrigated agricultural systems, there is no comparison of context (Vogel and Olivier, 2019). So far, while the changes on policy over
our findings with other national assessments. However, the drought time have had the goal to improve drought risk management, the focus is
risk analysis results and its components agree with other studies con- still largely on relief and emergency support instead of implementing pro-
ducted at the local level for agricultural systems. For instance, Eastern active policies (Vogel and Van Zyl, 2016; Bruwer, 1989; Vogel et al., 2010;

11
I. Meza, E. Eyshi Rezaei, S. Siebert et al. Science of the Total Environment 799 (2021) 149505

South African Weather Service, 2017 in Baudoin et al., 2017). Interdisci- CRediT authorship contribution statement
plinary drought risk assessments like the one presented here can be
used in decision-making processes. These assessments help to identify IM conducted the risk assessment with the support of EER and SS
potential pathways and actions towards proactive drought risk reduction under the supervision of MH. SS, JK, HG, CH, EP, EER and HN developed
policies such as the increasing access to finance, increasing extension ser- the hazard indicators and EER conducted the hazard/exposure analysis.
vices and programs in order to improve the environmental awareness, re- IM conducted the expert survey. IM and MH were responsible for the
ducing land degradation and increasing farmers' capacities towards a vulnerability assessment. All authors have contributed to drafting the
sustainable agroecosystems. manuscript, the interpretation of the results, and the process of approv-
Limitations in data availability impact the accuracy of our research ing the final manuscript.
like many others. For instance, the hazard and exposure analysis is
based on the land cover data from one timestep (static input data), Declaration of competing interest
which can impact the results (i.e., as cropping patterns are dynamic
and often can change over time). The authors declare that they have no known competing financial
Furthermore, future analysis can be improved by accounting for risk interests or personal relationships that could have appeared to influ-
differences of individual crop types, and exposed farmers. ence the work reported in this paper.

4.2. Recommendations and next steps Acknowledgements

There are various ways to measure drought hazard, and composite The research is part of the project GlobeDrought (grant no.
indices could make additional use of surface and ground water deficit, 02WGR1457A-F) funded by the German Federal Ministry of Education
provided that time series for these variables can be reliably derived and Research (BMBF) through its Global Resource Water (GRoW)
from hydrological modeling. In recent years the observation of surface funding initiative. The authors would like to thank the National Disaster
water volume changes from remote sensing, and of groundwater vari- Management Centre (NDMC) in South Africa as well as the experts who
ability from the GRACE and GRACE-FO satellite missions combined participated in the online expert survey for their support. Special thanks
with data assimilation, has made tremendous progress and we expect to Alexandra Dudley for her inputs and proofreading the revised paper.
that adding such indices would lend more robustness to our risk assess- Great thanks and gratitude to Dr. Siviwe Shwababa for the discussions
ment framework. and inputs to improve this paper.
Future assessments may benefit from new approaches to assess vul-
nerability beyond administrative boundaries (e.g., pixel-level vulnera- Appendix A. Supplementary data
bility data), since much of the information and effort in analyzing
hazard and exposure at the smallest possible resolution is lost when ag- Supplementary data to this article can be found online at https://doi.
gregated at administrative boundary levels reducing the capacity to ac- org/10.1016/j.scitotenv.2021.149505.
curately reflect reality. In addition to examining the environmental,
social and political processes shaping drought risk, an enhancement References
for this assessment could be developing a reliable and standardized da-
AFRA (Association for Rural Advancement), 1993. Drought Relief and Rural Communities.
tabase of losses and damages regarding agricultural systems in South Special Rep. No. 9. Pietermaritzburg, South Africa.
Africa. Such database can help better examine the medium- and long- Ahmadalipour, A., Moradkhani, H., Castelletti, A., Magliocca, N., 2019. Future drought risk
term impacts of drought and allow the comparison of impacts of similar in Africa: integrating vulnerability, climate change, and population growth. Sci. Total
Environ. 662, 672–686. https://doi.org/10.1016/j.scitotenv.2019.01.278.
hazard events in different parts of the country (e.g. drought of 2015- Asadieh, B., Krakauer, N.Y., 2017. Global change in streamflow extremes under climate
2016) (JRC (Joint Research Centre. European Commission), 2014). It change over the 21st century. Hydrol. Earth Syst. Sci. 21, 5863–5874. https://doi.
could also help identify indirect and cascading effects even after org/10.5194/hess-21-5863-2017.
Batjes, N.H., 2016. Harmonized soil property values for broad-scale modelling
the drought hazard event is finished. Moreover, loss data collec- (WISE30sec) with estimates of global soil carbon stocks. Geoderma 269, 61–68.
tions can be useful to identify trends and patterns in data over https://doi.org/10.1016/j.geoderma.2016.01.034.
time (JRC (Joint Research Centre. European Commission), 2014), Baudoin, M.-A., Vogel, C., Nortje, K., Naik, M., 2017. Living with drought in South Africa:
lessons learnt from the recent El Niño drought period. Int. J. Disaster Risk Reduct.
and to achieve consistent and coordinated implementation of risk
23, 128–137. https://doi.org/10.1016/j.ijdrr.2017.05.005.
reduction strategies. Beccari, B., 2016. A comparative analysis of disaster risk, vulnerability and resilience com-
posite indicators. PLoS Curr. 8.
Beck, H.E., Zimmermann, N.E., McVicar, T.R., Vergopolan, N., Berg, A., Wood, E.F., 2018.
5. Conclusions
Present and future Köppen-Geiger climate classification maps at 1-km resolution.
Sci. Data 5, 180214. https://doi.org/10.1038/sdata.2018.214.
Drought impacts on South Africa's agricultural sector are recurrent; BFAP (Bureau for Food and Agricultural Policy), 2016. Policy brief on 2015/2016 drought.
these drought events provide opportunities to learn and to improve Available at. http://www.nstf.org.za/wp-content/uploads/2016/06/Agri-SA-Drought-
Report_CS4.pdf. (Accessed 23 November 2020).
drought risk reduction efforts. We present, for the first time, an inte- Bruwer, J.J., 1989. Drought policy in the Republic of South Africa. Proceedings of the
grated drought risk assessment that considers hazard, exposure and SARCCUS Workshop on Drought, Alpha Training Center, South Africa P.23.
vulnerability to evaluate the impact of drought on irrigated and rainfed Bruwer, J.J., 1993. Drought policy in the Republic of South Africa. In: Wilhite, D.A. (Ed.)
Drought Assessment, Management, and Planning: Theory and Case Studies. Natural
systems (separately) at national level. In addition, we pioneer an expert Resource Management and Policy vol 2. Springer, Boston, MA. https://doi.org/10.
survey to weigh relevant indicators at national level. Our spatially ex- 1007/978-1-4615-3224-8_11.
plicit results assist to identify priority regions to take actions. Our find- Carrão, H., Naumann, G., Barbosa, P., 2016. Mapping global patterns of drought risk: an
empirical framework based on sub-national estimates of hazard, exposure and vul-
ings highlight the relevance of assessing and discussing drought risk nerability. Glob. Environ. Chang. 39, 108–124. https://doi.org/10.1016/j.gloenvcha.
in relation to specific impacts and diagnosing entry points to reduce 2016.04.012.
drought risk in a context-specific manner (i.e. irrigated and rainfed sys- Chen, J., Deng, M., Xia, L., Wang, H., 2017. Risk assessment of drought, based on IDM-VFS
in the Nanpan River Basin, Yunnan Province, China. Sustainability 9 (7), 1124. https://
tems). This ensures that relevant proactive policies and planning can be
doi.org/10.3390/su9071124.
effective even within the same sector (i.e. agricultural sector) before the Chibba, M., Luiz, J.M., 2011. P. Econ. Pap. 30, 307–315. https://doi.org/10.1111/j.1759-
worst impacts occur. While this assessment provides valuable informa- 3441.2011.00129.x.
tion at local municipality level, the assessment can be enhanced with a DAFF, 2018. Abstract of agricultural statistics. South African Department of Agriculture,
Forestry and Fisheries. http://www.daff.gov.za/Daffweb3/Portals/0/Statistics%20and
temporal dynamic exposure and more spatially explicit vulnerability %20Economic%20Analysis/Statistical%20Information/Abstract%202018.pdf. (Accessed
information. 7 October 2020).

12
I. Meza, E. Eyshi Rezaei, S. Siebert et al. Science of the Total Environment 799 (2021) 149505

DAFF, 2019. Abstract of agricultural statistics. South African Department of Agriculture, Hengl, T., Heuvelink, G.B.M., Kempen, B., Leenaars, J.G.B., Walsh, M.G., Shepherd, K.D., Sila,
Forestry and Fisheries. https://www.dalrrd.gov.za/Portals/0/Statistics%20and%20Eco- A., MacMillan, R.A., Mendes de Jesus, J., Tamene, L., Tondoh, J.E., 2015. Mapping soil
nomic%20Analysis/Statistical%20Information/Abstract%202019.pdf. (Accessed 1 May properties of Africa at 250 m resolution: random forests significantly improve current
2021). predictions. PLoS ONE 10, e0125814. https://doi.org/10.1371/journal.pone.0125814.
Dai, A., 2013. Increasing drought under global warming in observations and models. Nat. Hersbach, H., Bell, B., Berrisford, P., Hirahara, S., Horányi, A., Muñoz-Sabater, J., Nicolas, J.,
Clim. Chang. 3 (1), 52. Peubey, C., Radu, R., Schepers, D., Simmons, A., Soci, C., Abdalla, S., Abellan, X.,
DALRRD (Department of Agriculture, Land Reform and Rural Development, 2020. Agricul- Balsamo, G., Bechtold, P., Biavati, G., Bidlot, J., Bonavita, M., Chiara, G., Dahlgren, P.,
ture, land reform and rural development | South African government. https://www. Dee, D., Diamantakis, M., Dragani, R., Flemming, J., Forbes, R., Fuentes, M., Geer, A.,
gov.za/about-sa/agriculture. (Accessed 20 May 2021). Haimberger, L., Healy, S., Hogan, R.J., Hólm, E., Janisková, M., Keeley, S., Laloyaux, P.,
DEA, 2019. Department of Environmental Affairs. https://www.environment.gov.za/ Lopez, P., Lupu, C., Radnoti, G., Rosnay, P., Rozum, I., Vamborg, F., Villaume, S.,
projectsprogrammes/egis_landcover_datasets. (Accessed 15 February 2020). Thépaut, J., 2020. The ERA5 global reanalysis. Q. J. R. Meteorol. Soc. 146, 1999–2049.
de Sherbinin, A., Bukvic, A., Rohat, G., Gall, M., McCusker, B., Preston, B., Apotsos, A., Fish, https://doi.org/10.1002/qj.3803.
C., Kienberger, S., Muhonda, P., Wilhelmi, O., Macharia, D., Shubert, W., Sliuzas, R., Hornby, D., Vanderhaeghen, Y., Versveld, D., Ngubane, M., 2016. A harvest of dysfunction
Tomaszewski, B., Zhang, S., 2019. Climate vulnerability mapping: a systematic review – rethinking the approach to drought, its causes and impacts in South Africa. Oxfam
and future prospects. WIREs Clim Change https://doi.org/10.1002/wcc.600. Report, Johannesburg. https://www.oxfam.org.za/wp-content/uploads/2017/03/
Deng, M., Chen, J., Huang, J., Niu, W., 2018. Agricultural drought risk evaluation based on Final-Final-OZA-Oxfam-Drought-Report.pdf. (Accessed 15 June 2020).
an optimized comprehensive index system. Sustainability 10 (10), 3465. https://doi. Huang, L., He, B., Han, L., Liu, J., Wang, H., Chen, Z., 2017. A global examination of the re-
org/10.3390/su10103465. sponse of ecosystem water-use efficiency to drought based on MODIS data. Sci. Total
Department of Agriculture, Forestry and Fisheries (DAFF), 2016. LADA Degradation Index. Environ. 601–602, 1097–1107. https://doi.org/10.1016/j.scitotenv.2017.05.084.
https://doi.org/10.15493/DEA.CARBON.10000018. IEC (International Electrotechnical Commission), 2019. Technical Committee 56, Interna-
Dilley, M., Chen, R.S., Deichmann, U., Lerner-Lam, A.L., Arnold, M., Agwe, J., Buys, P., tional Organization for Septuagint and Cognate Studies, Congress. Risk management:
Kjekstad, O., Lyon, B., Yetman, G., 2005. Natural Disaster Hotspots: A Global Risk Anal- risk assessment techniques = Management du risque -Techniques d’appreciation du
ysis (Disaster Risk Management Series; No. 5). The World Bank Group. risque.
Downing, T.E., Bakker, K., 2000. Drought Discourse and Vulnerability. Chapter 45. In: IPCC, 2014. In: Barros, V.R., Field, C.B., Dokken, D.J., Mastrandrea, M.D., Mach, K.J., Bilir, T.E.,
Wilhite, D.A. (Ed.), Drought: A Global Assessment, Natural Hazards and Disasters Se- White, L.L. (Eds.), Climate Change 2014: Impacts, Adaptation, and Vulnerability. Part
ries. Routledge Publishers, Abingdon. B: Regional Aspects. Contribution of Working Group II to the Fifth Assessment Report
du Pisani, L.G., Fouché, H.J., Venter, J.C., 1998. Assessing rangeland drought in South Africa. of the Intergovernmental Panel on Climate Change. Cambridge University Press, New
Agric. Syst. 57, 367–380. https://doi.org/10.1016/S0308-521X(98)00024-9. York - Cambridge.
DWA, 2002. Governing Board Induction Manual. Department of Water Affairs. Jordaan, A.J., 2011. Drought Risk Reduction in the Northern Cape. University of the Free
European Commission, Joint Research Centre, 2017. Index for risk management - IN- State, Bloemfontein.
FORM: concept and methodology, version 2017. Publications Office, LU. https:// Jordaan, A.J., Sakulski, D., Jordaan, A., 2013. Interdisciplinary drought risk assessment for
drmkc.jrc.ec.europa.eu/inform-index. (Accessed 18 February 2020). agriculture: the case of communal farmers in the northern Cape Province, South
European Commission, Joint Research Centre, 2019. Drought vulnerability indicators for Africa. S. Afr. J. Agric. Ext. 41, 01–16.
global-scale drought risk assessments: global expert survey results report. Publica- Jordaan, A.J., South Africa, Water Research Commission, South Africa, Department of Ag-
tions Office, LU. https://publications.jrc.ec.europa.eu/repository/handle/JRC117546. riculture, F., and Fisheries, Sakulski, D.M., Muyambo, F., 2017a. Vulnerability, adapta-
(Accessed 15 March 2020). tion to and coping with drought: the case of commercial and subsistence rain fed
FAO, 1997. The political economy of food, agriculture and irrigation development in East farming in the Eastern Cape. Volume I: report to the Water Research Commission
and Southern Africa [WWW Document]. http://www.fao.org/3/w7314e/w7314e09. and Department of Agriculture, Forestry and Fisheries Volume I: report to the
htm. (Accessed 26 January 2021). Water Research Commission and Department of Agriculture, Forestry and Fisheries.
FAO, 2016. AQUASTAT Country Profile –South Africa. Rome, Italy, Food and Agriculture http://www.wrc.org.za/Knowledge%20Hub%20Documents/Research%20Reports/TT%
Organization of the United Nations (FAO). 20716-17.pdf. (Accessed 9 May 2020).
FAO, 2019. GIEWS Update. Southern AfricaDry weather conditions reduce agricultural Jordaan, A.J., Water Research Commission, Sakulski, D.M., Muyambo, F., Africa, South,
production prospects in 2019. Available at. https://reliefweb.int/sites/reliefweb.int/ 2017b. Vulnerability, adaptation to and coping with drought: the case of commercial
files/resources/ca3071en_0.pdf. (Accessed 10 July 2020). and subsistence rain fed farming in the Eastern Cape. Vol. II. WRC. http://www.wrc.
FAO, 2020a. FAOSTAT crops dataset. Available at. http://www.fao.org/faostat/en/#data/ org.za/Knowledge%20Hub%20Documents/Research%20Reports/TT%20716-2-17.pdf.
QC. (Accessed 2 May 2021). (Accessed 29 October 2020).
FAO, 2020b. Country profile, South Africa. http://www.fao.org/3/x9751e/x9751e07.htm. JRC (Joint Research Centre. European Commission), 2014. Current status and best prac-
(Accessed 8 January 2020). tices for disaster loss data recording in EU Member States. Institute for the Protection
FAO, 2020c. Digital Agriculture Profile South Africa. http://www.fao.org/3/cb2506en/ and the Security of the Citizen. Publications Office, LU.
CB2506EN.pdf. (Accessed 2 January 2021). Kamali, B., Abbaspour, K.C., Wehrli, B., Yang, H., 2018. Drought vulnerability assessment of
FAO, 2021. FAOSTAT crops dataset. Available at. http://www.fao.org/faostat/en/#data/QC. maize in Sub-Saharan Africa: insights from physical and social perspectives. Glob.
(Accessed 2 May 2021). Planet. Chang. 162, 266–274. https://doi.org/10.1016/j.gloplacha.2018.01.011.
Frigerio, I., De Amicis, M., 2016. Mapping social vulnerability to natural hazards in Italy: a Kamruzzaman, Md., Kabir, Md.E., Rahman, A.T.M.S., Jahan, C.S., Mazumder, Q.H., Rahman,
suitable tool for risk mitigation strategies. Environ. Sci. Pol. 63, 187–196. https://doi. M.S., 2018. Modeling of agricultural drought risk pattern using Markov chain and GIS
org/10.1016/j.envsci.2016.06.001. in the western part of Bangladesh. Environ. Dev. Sustain. 20 (2), 569–588. https://doi.
Frischen, J., Meza, I., Rupp, D., Wietler, K., Hagenlocher, M., 2020. Drought risk to agricul- org/10.1007/s10668-016-9898-0.
tural Systems in Zimbabwe: a spatial analysis of Hazard, exposure, and vulnerability. Le Maitre, D.C., Seyler, H., Holland, M., Smith-Adao, L., Nel, J.L., Maherry, A., Witthüser, K.,
Sustainability 12 (3), 752. https://doi.org/10.3390/su12030752. 2018. Identification, Delineation and Importance of the Strategic Water Source Areas
Gibberd, V., Rook, J., Sear, C.B., Williams, J.B., 1996. Drought Risk Management in Southern of South Africa, Lesotho and Swaziland for Surface Water and Groundwater, Report
Africa. The Potential of Long Lead Climate Forecasts for Improved Drought No. TT 743/1/18. Water Research Commission, Pretoria.
Management. Lehner, B., Liermann, C.R., Revenga, C., Vörösmarty, C., Fekete, B., Crouzet, P., Döll, P.,
Global Dams Database, 2020. GraND Data Catalog [WWW Document]. https:// Endejan, M., Frenken, K., Magome, J., Nilsson, C., Robertson, J.C., Rödel, R., Sindorf,
datacatalog.worldbank.org/dataset/global-dams-database. (Accessed 12 March N., Wisser, D., 2011. High-resolution mapping of the world’s reservoirs and dams
2020). for sustainable river-flow management. Front. Ecol. Environ. 9, 494–502. https://
Hagenlocher, M., Renaud, F.G., Haas, S., Sebesvari, Z., 2018. Vulnerability and risk of deltaic doi.org/10.1890/100125.
social-ecological systems exposed to multiple hazards. Sci. Total Environ. 631–632, Machethe, C.L., 2004. Agriculture and poverty in South Africa: can agriculture reduce pov-
71–80. https://doi.org/10.1016/j.scitotenv.2018.03.013. erty. Overcoming Underdevelopment Conference Held in Pretoria. Citeseer, p. 29.
Hagenlocher, M., Meza, I., Anderson, C.C., Min, A., Renaud, F.G., Walz, Y., Siebert, S., Magombeyi, M.S., Taigbenu, A.E., 2008. Crop yield risk analysis and mitigation of small-
Sebesvari, Z., 2019. Drought vulnerability and risk assessments: state of the art, per- holder farmers at quaternary catchment level: case study of B72A in Olifants river
sistent gaps, and research agenda. Environ. Res. Lett. 14, 083002. https://doi.org/10. basin, South Africa. Phys. Chem. Earth 33, 744–756. https://doi.org/10.1016/j.pce.
1088/1748-9326/ab225d. 2008.06.050.
Han, L., Zhang, Q., Ma, P., Jia, J., Wang, J., 2016. The spatial distribution characteristics of a Mahlalela, P.T., Blamey, R.C., Hart, N.C.G., Reason, C.J.C., 2020. Drought in the eastern cape
comprehensive drought risk index in southwestern China and underlying causes. region of South Africa and trends in rainfall characteristics. Clim. Dyn. 55, 2743–2759.
Theor. Appl. Climatol. 124 (3–4), 517–528. https://doi.org/10.1007/s00704-015- https://doi.org/10.1007/s00382-020-05413-0.
1432-z. Masupha, T.E., Moeletsi, M.E., 2020. The use of water requirement satisfaction index for
Hardy, M., Dziba, L., Kilian, W., Tolmay, J., 2011. Rainfed farming systems in South Africa. assessing agricultural drought on rain-fed maize, in the Luvuvhu River catchment,
In: Tow, P., Cooper, I., Partridge, I., Birch, C. (Eds.), Rainfed Farming Systems. Springer, South Africa. Agric. Water Manag. 237, 106142. https://doi.org/10.1016/j.agwat.
Netherlands, Dordrecht, pp. 395–432 https://doi.org/10.1007/978-1-4020-9132-2_ 2020.106142.
16. Meza, I., Siebert, S., Döll, P., Kusche, J., Herbert, C., Eyshi Rezaei, E., Nouri, H., Gerdener, H.,
Hassan, R., 2013. Drought management strategies in South Africa and the potential for Popat, E., Frischen, J., Naumann, G., Vogt, J.V., Walz, Y., Sebesvari, Z., Hagenlocher, M.,
economic policy instruments. In: Schwabe, K., Albiac, J., Connor, J., Hassan, R., Meza 2020. Global-scale drought risk assessment for agricultural systems. Nat. Hazards
González, L. (Eds.), Drought in Arid and Semi-Arid Regions. Springer, Dordrecht Earth Syst. Sci. 20, 695–712. https://doi.org/10.5194/nhess-20-695-2020.
https://doi.org/10.1007/978-94-007-6636-5_21. Mueller, N.D., Gerber, J.S., Johnston, M., Ray, D.K., Ramankutty, N., Foley, J.A., 2012. Closing
Heim Jr., R.R., 2002. A review of twentieth century drought indices used in the United yield gaps through nutrient and water management. Nature 490, 254–257. https://
States. Bull. Am. Meteorol. Soc. 83 (8), 1149–1165. doi.org/10.1038/nature11420.

13
I. Meza, E. Eyshi Rezaei, S. Siebert et al. Science of the Total Environment 799 (2021) 149505

Müller Schmied, H., Cáceres, D., Eisner, S., Flörke, M., Herbert, C., Niemann, C., Peiris, T.A., Thompson, 2019. South African National Land-Cover (SANLC) 2018. Department of Envi-
Popat, E., Portmann, F.T., Reinecke, R., Schumacher, M., Shadkam, S., Telteu, C.-E., ronmental Affairs of the Republic of South Africa, 2019. https://www.environment.
Trautmann, T., Döll, P., 2020. The global water resources and use model WaterGAP gov.za/projectsprogrammes/egis_landcover_datasets. (Accessed 8 July 2020).
v2.2d: Model description and evaluation. Geosci. Model Dev. Discuss. https://doi. Trenberth, K.E., Dai, A., Van der Schrier, G., Jones, P.D., Barichivich, J., Briffa, K.R., Sheffield,
org/10.5194/gmd-2020-225. J., 2014. Global warming and changes in drought. Nat. Clim. Chang. 4, 17–22. https://
Muyambo, F., Jordaan, A.J., Bahta, Y.T., 2017. Assessing social vulnerability to drought in doi.org/10.1038/nclimate2067.
South Africa: policy implication for drought risk reduction. Jàmbá 9. https://doi.org/ Tung, C.-P., Tsao, J.-H., Tien, Y.-C., Lin, C.-Y., Jhong, B.-C., 2019. Development of a novel cli-
10.4102/jamba.v9i1.326. mate adaptation algorithm for climate risk assessment. Water 11, 497. https://doi.
National Treasury, 2003. Intergovernmental Fiscal Review, Republic of South Africa. org/10.3390/w11030497.
National Treasury, 2019. Balance Sheet Municipal Finance Data Tables Municipal Money UNDRR, 2015. Sendai Framework for Disaster Risk Reduction 2015-2030. Geneva,
Data [WWW Document]. https://municipaldata.treasury.gov.za/table/aged_credi- Switzerland. http://www.unisdr.org/files/43291_sendaiframeworkfordrren.pdf.
tor/?year=2019&amountType=AUDA. (Accessed 13 March 2021). (Accessed 18 October 2020).
Naumann, G., Barbosa, P., Garrote, L., Iglesias, A., Vogt, J., 2014. Exploring drought vulner- UNDRR, 2019. Global Assessment Report on Disaster Risk Reduction 2019, Chapter 6:
ability in Africa: an indicator based analysis to be used in early warning systems. Special Section on Drought. UN, New York.
Hydrol. Earth Syst. Sci. 18 (5), 1591–1604. https://doi.org/10.5194/hess-18-1591- Unganai, L.S., Kogan, F.N., 1998. Drought monitoring and corn yield estimation in south-
2014. ern Africa from AVHRR data. Remote Sens. Environ. 63, 219–232. https://doi.org/10.
Ngaka, M., 2012. Drought preparedness, impact and response: a case of the eastern cape 1016/S0034-4257(97)00132-6.
and Free State provinces of South Africa. Jàmbá J. Disaster Risk Stud. 4 (1), 1–10. VERBI Software, 2019. MAXQDA 2020, Somputer Program. VERBI Software, Berlin.
OECD, 2008. Handbook on Constructing Composite Indicators: Methodology and User Vogel, C., Olivier, D., 2019. Re-imagining the potential of effective drought responses in
Guide. https://doi.org/10.1787/9789264043466-en. South Africa. Reg. Environ. Chang. 19, 1561–1570. https://doi.org/10.1007/s10113-
Ortega-Gaucin, D., Ceballos-Tavares, J.A., Ordoñez Sánchez, A., Castellano-Bahena, H.V., 018-1389-4.
2021. Agricultural drought risk assessment: a spatial analysis of Hazard, exposure, Vogel, C., Van Zyl, K., 2016. Drought - in search of sustainable solutions to a persistent
and vulnerability in Zacatecas, Mexico. Water 13 (10), 1431. https://doi.org/10. 'wicked' problem in South Africa. In: Salzmann, N., Huggel, C.,
3390/w13101431.
Nussbaumer, S.U., Ziervogel, G. (Eds.), Climate Change Adaptation Strategies: an
Pei, W., Fu, Q., Liu, D., Li, T., Cheng, K., Cui, S., 2018. patiotemporal analysis of the agricul- Upstream-Downstream Perspective. Springer, Cham, pp. 197–211 https://doi.org/
tural drought risk in Heilongjiang Province, China. Theor. Appl. Climatol. 133 (1–2), 10.1007/978-3-319-40773-9_11.
151–164. https://doi.org/10.1007/s00704-017-2182-x.
Vogel, C., Koch, I., Van Zyl, K., 2010. “A Persistent Truth”—Reflections on Drought Risk
Peng, L., Zeng, Z., Wei, Z., Chen, A., Wood, E.F., Sheffield, J., 2019. Determinants of the ratio
Management in Southern Africa. 2. Weather, Climate, and Society, pp. 9–22.
of actual to potential evapotranspiration. Glob. Chang. Biol. 25, 1326–1343. https://
https://doi.org/10.1175/2009WCAS1017.1.
doi.org/10.1111/gcb.14577.
Waldner, F., Hansen, M.C., Potapov, P.V., Löw, F., Newby, T., Ferreira, S., Defourny, P., 2017.
Portmann, F.T., Siebert, S., Döll, P., 2010. MIRCA2000-global monthly irrigated and rainfed
National-scale cropland mapping based on spectral-temporal features and outdated
crop areas around the year 2000: a new high-resolution data set for agricultural and
land cover information. PLoS ONE 12, e0181911. https://doi.org/10.1371/journal.
hydrological modeling: monthly irrigated and rainfed crop areas. Glob. Biogeochem.
pone.0181911.
Cycles 24. https://doi.org/10.1029/2008GB003435 n/a-n/a.
Walz, Y., Dall, K., Graw, V., León, J., Haas, S., Kussul, N., Jordaan, A.J., 2018. Understanding
Rother, H., Hall, R., London, L., 2008. Pesticide use among emerging farmers in South
and reducing agricultural drought risk. Examples From South Africa and Ukraine.
Africa: contributing factors and stakeholder perspectives. Dev. South. Afr. 25 (4),
399–424. Walz, Y., Min, A., Dall, K., Duguru, M., Villagran de Leon, J.-C., Graw, V., Dubovyk, O.,
Sebesvari, Z., Jordaan, A.J., Post, J., 2020. Monitoring progress of the Sendai framework
Running, Steve, Mu, Qiaozhen, Zhao, Maosheng, 2017. MOD16A2 MODIS/Terra Net
Evapotranspiration 8-Day L4 Global 500m SIN Grid V006. https://doi.org/10.5067/ using a geospatial model: the example of people affected by agricultural droughts in
MODIS/MOD16A2.006. Eastern Cape, South Africa. Progress in Disaster Science. 5, p. 100062. https://doi.org/
SACN, 2016. State of South African Cities Report 2016. Johannesburg: SACN. http://www. 10.1016/j.pdisas.2019.100062.
socr.co.za/wp-content/uploads/2016/06/SoCR16-Main-Report-online.pdf. (Accessed Weiss, D.J., Nelson, A., Gibson, H.S., Temperley, W., Peedell, S., Lieber, A., Hancher, M.,
2 January 2021). Poyart, E., Belchior, S., Fullman, N., Mappin, B., Dalrymple, U., Rozier, J., Lucas, T.C.D.,
Schreiner, B.G., Mungatana, E.D., Baleta, H., 2018. Impacts of Drought Induced Water Howes, R.E., Tusting, L.S., Kang, S.Y., Cameron, E., Bisanzio, D., Battle, K.E., Bhatt, S.,
Shortages in South Africa: Sector Policy Briefs. Gething, P.W., 2018. A global map of travel time to cities to assess inequalities in ac-
Schulze, R.E., 2016. On observations, climate challenges, the South African agriculture sec- cessibility in 2015. Nature 553, 333–336. https://doi.org/10.1038/nature25181.
tor and considerations for an adaptation handbook. In: Schulze, R.E. (Ed.), Handbook West, P.C., Gerber, J.S., Engstrom, P.M., Mueller, N.D., Brauman, K.A., Carlson, K.M., Cassidy,
for Farmers, Officials and Other Stakeholders on Adaptation to Climate Change in the E.S., Johnston, M., MacDonald, G.K., Ray, D.K., Siebert, S., 2014. Leverage points for im-
Agriculture Sector within South Africa. Section A: Agriculture and Climate Change in proving global food security and the environment. Science 345, 325–328. https://doi.
South Africa: Setting the Scene, Chapter A1. org/10.1126/science.1246067.
Schwarz, M., Landmann, T., Cornish, N., Wetzel, K.-F., Siebert, S., Franke, J., 2020. A spa- WMO World Meteorological Organization, 2020. World Meteorological Organization
tially transferable drought hazard and drought risk modeling approach based on re- Commission for Climatology. State of the Climate in Africa 2019. World Meteorolog-
mote sensing data. Remote Sens. 12 (2), 237. https://doi.org/10.3390/rs12020237. ical Organisation, Geneva.
Siebert, S., Döll, P., 2010. Quantifying blue and green virtual water contents in global crop World Bank, 2019. Assessing Drought Hazard and Risk: Principles and Implementation
production as well as potential production losses without irrigation. J. Hydrol. 384, Guidance. World Bank, Washington, DC.
198–217. https://doi.org/10.1016/j.jhydrol.2009.07.031. World bank, 2020. Global - Dams Database | Data Catalog [WWW Document]. https://
Sivakumar, M.V.K., Stefanski, R., Bazza, M., Zelaya, S., Wilhite, D., Magalhaes, A.R., 2014. datacatalog.worldbank.org/dataset/global-dams-database. (Accessed 1 December
High level meeting on National Drought Policy: summary and major outcomes. 2021).
Weather Climate Extremes 3, 126–132. https://doi.org/10.1016/j.wace.2014.03.007. WWF (World Wide Fund for Nature), 2018. Agriculture: Facts and Trends, South Africa
StatSA, 2011. Statistics South Africa Data. http://superweb.statssa.gov.za/webapi/jsf/ [online]. World Wide Fund. http://awsassets.wwf.org.za/downloads/facts_brochure_
dataCatalogueExplorer.xhtml. (Accessed 8 May 2019). mockup_04_b.pdf. (Accessed 6 July 2020).
StatSA, 2016. Statistics South Africa Data. http://superweb.statssa.gov.za/webapi/jsf/ Zeng, Z., Wu, W., Li, Z., Zhou, Y., Guo, Y., Huang, H., 2019. Agricultural drought risk assess-
dataCatalogueExplorer.xhtml. (Accessed 8 June 2019). ment in Southwest China. Water 11 (5), 1064. https://doi.org/10.3390/w11051064.
Thamaga-Chitja, J.M., Morojele, P., 2014. The context of smallholder farming in South Zhang, D., Wang, G., Zhou, H., 2011. Assessment on agricultural drought risk based on var-
Africa: towards a livelihood asset building framework. J. Hum. Ecol. 45, 147–155. iable fuzzy sets model. Chin. Geogr. Sci. 21 (2), 167–175. https://doi.org/10.1007/
https://doi.org/10.1080/09709274.2014.11906688. s11769-011-0456-2.

14