Nested archetypes of vulnerability in African drylands: where lies potential for sustainable agricultural intensification?

Major dryland studies (Reynolds et al 2007, Safriel and Adeel 2008, Sietz et al 2011a, Kok et al 2016) have suggested typical socio-ecological factors affecting farming systems' vulnerability (see section SI.1 in the supplementary material available at stacks.iop.org/ERL/12/095006/mmedia), which we used to select nine socio-ecological indicators (table 1) to assess archetypes of vulnerability. Data derived from integrated assessment models such as environmental resource indicators received particular attention in order to facilitate future investigations of changes in vulnerability and scenario analysis. Our analysis focused on a sub-national resolution. Aiming at both finding a common ground for working with sufficient data quality for all the indicators considered and facilitating the use of data derived from integrated assessment models such as IMAGE (MNP 2006) and GISMO (Hilderink and Lucas 2008, Stehfest et al 2014), we worked on an intermediate resolution (0.5°). This choice is a compromise that balances differences in indicator resolution and enables future scenario analysis considering environmental and well-being dynamics (MNP 2006, Hilderink and Lucas 2008, Stehfest et al 2014).

Regarding food security (FAO 2000), we focused our analysis on food availability and access as those dimensions that are at least partly within farmers' and pastoralists' control. Dryland regions were selected according to aridity including hyper-arid to dry sub-humid areas (aridity index: 0–0.65), based on dryland characterisation used by the Convention on Biological Diversity (Sörensen 2007). The year 2000 is the reference year for which all relevant data were available. Thus, this study presents a baseline assessment suited as a starting point when analysing current or future trends in vulnerability using dynamic indicators as proposed by Lüdeke et al (2014).

Resource quality was indicated by (i) water availability modelled as surface runoff and shallow groundwater recharge at river basin level (Alcamo et al 2000) depicting water scarcity as a major constraint in dryland development (see section SI.1 in the supplementary material), (ii) agropotential including soil quality (e.g. organic carbon content, soil texture, depth and water holding capacity), precipitation and topographic conditions approximated as the ratio of actual grassland productivity to maximum productivity under ideal conditions (MNP 2006) and (iii) erosion sensitivity given by the soil's water erosion sensitivity index proposed by Hootsmans et al (2001). Resource use intensity was indicated by (i) population pressure expressed as population density (Klein Goldewijk et al 2010) to capture food, water and other demands to be satisfied and (ii) urbanisation given by the share of urban population (Klein Goldewijk et al 2010) pointing towards a decreasing importance of agriculture in people's livelihoods. Socio-political constraints were indicated by (i) remoteness from markets and decision making represented by the distance to the nearest city (Letourneau et al 2010) and (ii) governance depicted as the national average of governance indicators proposed by Kaufmann et al (2008). The remoteness indicator further differentiates the broad governance information. For example, in more easily accessible regions governments can more effectively implement policies and regulations and ensure the quality of public and civil services, and citizens can more easily participate in elections. Finally, well-being was indicated by income and undernourishment, reflecting both causes and consequences of vulnerability. Due to a lack of spatially well-resolved data, income was indicated by gross domestic product available at a national level (UNSTAT 2005, World Bank 2006). Undernourishment was indicated by the percentage of underweight children represented at a sub-national resolution (CIESIN 2005), enabling a better understanding of income distribution. If many children were underweight in areas with a high national average income, we assumed that income was very unequally distributed.

In preparing the cluster analysis, we resampled population, urbanisation and undernourishment data to a 0.5° × 0.5° resolution. Moreover, country-level governance and income data were assigned to all grid cells in a given country such that all data were available at a grid-cell level (0.5° × 0.5°). To enable comparison, all indicators used to identify the broad archetypes were normalised to values between 0−1, reflecting their minima and maxima within the drylands of sub-Saharan Africa. Complementing the broad archetypes, the nested sub-clustering presents a more focused analysis considering only those indicator values that describe the regional sub-set that belonged to the broad archetypes with higher agropotential and mainly agricultural livelihoods. That sub-set of data spans a range that is smaller or equal to the range that is spanned by the data set used in the broad analysis. The indicator values were re-scaled to values between 0 and 1, on the basis of the minimum and maximum values over that regional sub-set. This resulted in a linear re-scaling of the normalised indicator values that were used in the broad archetype analysis.

Cluster analysis was used to categorise the drylands in sub-Saharan Africa in groups or clusters with similar socio-ecological characteristics, here referred to as 'archetypes of vulnerability'. We used an established clustering approach based on the sequence of hierarchical (hclust) and partitioning (k-means) algorithms (Sietz et al 2011a, Janssen et al 2012, Kok et al 2016). Cluster analysis was conducted in the nine-dimensional data space spanned by the indicators detailed in section 2.1. Firstly, clustering was performed for all drylands in sub-Saharan Africa, which revealed broad archetypes. Secondly, only those broad archetypes encompassing regions with higher agropotential in which people relied mainly on agriculture were clustered to identify nested archetypes.

The optimal number of clusters was determined by investigating the stability of cluster partitions using different starting points (see section SI.2 in the supplementary material). The overall importance of indicators for a given cluster partition was estimated by assessing a partition's sensitivity to the omission of indicators, expressed as a Fraiman Index (Fraiman et al 2008, see figures SI.2 and SI.4 in the supplementary material). Low values of the Fraiman Index depict important indicators while high values mean less important indicators. While the Fraiman Index delivered an overview of the most important indicators at the level of cluster partitions, the average indicator values for the clusters provided the basis for discussing cluster-specific mechanisms. The spatial distribution of clusters offered insights into regional combinations of vulnerability drivers and outcomes.

Cluster analysis revealed eight broad archetypes of vulnerability as the consistency measure showed a maximum for partitions with eight clusters for all drylands in sub-Saharan Africa (see figure SI.1 in the supplementary material). Four groups of broad archetypes can be identified including (i) extremely dry and resource-constrained areas, (ii) dry areas with better agropotential, (iii) better governance, higher income and better nourished people and (iv) urbanised areas. According to the Fraiman Index (see figure SI.2 in the supplementary material), the most important indicators distinguishing the broad archetypes were erosion sensitivity, governance and urbanisation (in order of decreasing importance). In contrast, water availability and population pressure hardly differentiated the clusters. We named the archetypes according to the differentiating vulnerability dimensions that often, but not always, coincided with the most important indicators revealed by the Fraiman Index (e.g. erosion sensitivity and governance). The broad archetypes of vulnerability are discussed in more detail in the following sub-sections.

3.1.1. Extremely dry and resource-constrained areas

The broad archetypes encompassing extremely dry and resource-constrained areas described the most severe resource constraints in terms of water availability and agropotential, combined with very low income and medium to most severe undernourishment (red and orange clusters, figure 1). Contrasting levels of governance and remoteness differentiated vulnerability in these clusters. Representing the most isolated and sparsely populated regions, this group of clusters encompassed only a minority of people (4%) living in the sub-Saharan drylands (table 2).

Zoom In Zoom Out Reset image size

Table 2. Broad archetypes of vulnerability: dryland population and area share.

Groups of broad archetypes	Broad archetypes	Populationa (in thousands)	Population share (%)	Area (1000 km2)	Area share (%)
Extremely dry and resource-constrained areas	Poor governance and severest undernourishment	5624	3	1873	15
	Better governance and highest remoteness	1846	1	2376	19
Dry areas with better agropotential	Highest erosion sensitivity	41 205	20	2135	17
	Low erosion sensitivity	32 030	16	1789	14
	Poorest governance and high erosion sensitivity	15 033	7	1342	10
Better governance, higher income and better nourished people	Rural areas and high income	4042	2	1751	14
	Higher urbanisation, highest income and least undernourishment	30 317	15	685	5
Urbanised areas	Urbanised areas	75 048	37	879	7
Total		205 145	100	12 649	100

^aData source: Klein Goldewijk et al (2010).

The poor governance and severest undernourishment archetype was primarily found in the hyper- to semi-arid zones in eastern Africa and the eastern and central Sahel zone (red cluster, figure 1). This archetype represented the most critical development trajectory outlined in the Dryland Livelihood Paradigm in which inherently low primary productivity significantly limits well-being (Safriel and Adeel 2008), shown by the worst undernourishment among all broad archetypes. For example, recurrent droughts, food shortages and violent conflicts are major causes of vulnerability and food insecurity in the northern part of Afar Regional State in Ethiopia (Tesfay and Tafere 2004). The predominant pastoralists in this region had frequently been excluded from development programmes compared with other pastoral groups in Ethiopia, and government support for rangeland management remained largely ineffective, partly as a result of centralised, top-down approaches in project planning and implementation (Tesfay and Tafere 2004).

The better governance and highest remoteness archetype was mainly found in western Africa where, despite better governance, the largest distances to markets and institutions created serious barriers to trading and effective policy implementation (orange cluster, figure 1). Extreme water scarcity further challenged farmers and pastoralists in achieving food security, reflected in the severe undernourishment. For example, the Mauritanian government prioritised food security improvements and designed national policies and development programmes directly aimed at preventing, monitoring and reducing child undernourishment (Wuehler and Ould Dehah 2011). However, their implementation was lacking in less accessible areas, and greater efforts are needed particularly in regions with extremely low productivity due to a lack of rainfall and surface water such as in central and northeastern Mauritania (Ould Soule 2006).

3.1.2. Dry areas with better agropotential

3.1.3. Better governance, higher income and better nourished people

3.1.4. Urbanised areas

In analysing nested archetypes, we were particularly interested in those agricultural regions where sustainable intensification can contribute towards reducing vulnerability. Three dimensions captured in this vulnerability analysis (see section 2.1) are especially suited to exploring possibilities for sustainable intensification. Firstly, agropotential and water availability jointly indicate primary productivity, depicting an environmental potential for intensification. Lack of environmental potential precluded regions from being considered for sustainable intensification efforts. Secondly, erosion sensitivity points towards the risk of land overuse. It directly links environmental conditions and land management, allowing a more differentiated discussion of intensification opportunities. Thirdly, we considered effective planning and implementation processes, specified by governance and remoteness, to be institutional prerequisites for success in sustainable intensification. Among the broad archetypes, the clusters depicting dry areas with better agropotential (see group of blue clusters, figure 1) demonstrated higher primary productivity together with a varying risk of land overuse and planning as well as implementation effectiveness. In these regions we identified nested archetypes to refine vulnerability insights and investigate sustainable intensification opportunities. In contrast, primary productivity was strongly limited in the archetypes representing extremely dry and resource-constrained areas, and agricultural livelihoods did not prevail in the archetypes indicating better governance, higher income and better nourished people as well as urbanised areas.

We identified six nested archetypes (figure 2) embedded in the dry areas with better agropotential according to the consistency measure (see figure SI.3 in the supplementary material). The consistency measure showed the global maximum for partitions with three clusters. This reconfirmed the existence of the three broad archetypes identified among all drylands in sub-Saharan Africa. In each of the broad archetypes two finer archetypes were nested, mainly differentiating farming systems' vulnerability according to erosion sensitivity, governance and undernourishment (see figure SI.4 in the supplementary material).

Zoom In Zoom Out Reset image size

The two archetypes nested in the regions with highest erosion sensitivity differed in undernourishment and agropotential (dark and light brown clusters, figure 2) but otherwise largely maintained the broad archetype's characteristics including highest erosion sensitivity and better governance (dark blue cluster, figure 1). The nested archetype depicting severe undernourishment and lower agropotential was mainly found in the western and central Sahel region (dark brown cluster, figure 2), representing about one third of the population in nested archetypes (table 3). For example, agro-constraints including low soil fertility and precipitation scarcity restricted crop and pasture productivity in northern Burkina Faso (Ingram et al 2002). As a result, agro-pastoralists living in this dry region, reliant on cattle production as their main source of cash income, faced difficulties in producing sufficient food and generating income (Ingram et al 2002). The nested archetype showing low to medium undernourishment and better agropotential depicted another vulnerability mechanism primarily in southeastern African drylands (light brown cluster, figure 2) encompassing 17% of the nested archetypes' population (table 3). For example, farmers in southern Mozambique who diversified their livelihoods to include farming, livestock keeping and natural resource collection for market sale could generate some cash income (Hahn et al 2009). Moreover, food storage and seed savings helped these farmers to reduce risks of food shortage (Hahn et al 2009), potentially leading to the lower undernourishment depicted in the light brown cluster (figure 2).

The low erosion sensitivity archetype (light blue cluster, figure 1) was further differentiated by one nested cluster capturing severest undernourishment and medium governance (light olive cluster, figure 2). This nested archetype was present in eastern and southern Ethiopia, some parts of southern Chad and southern Niger, assembling 14% of the nested archetypes' population (table 3). People living in these regions were the most severely undernourished among all nested archetypes and had very low incomes. Marginalised agro-pastoralists with limited access to extension services and markets in southern Ethiopia illustrated this nested archetype. Since these farmers and pastoralists mainly subsisted on rain-fed land, the precipitation decrease (1980–2010) that concurred with a rapid population increase in this region largely undermined their food security (Lòpez-Carr et al 2014). The other nested archetype included regions with medium undernourishment, better governance and lowest erosion sensitivity (dark olive cluster, figure 2) was found in the western Sahel region and some parts of eastern and southeastern Africa. Poor soil fertility illustrates an agro-constraint that restricted smallholder farmers' production of staple crops such as maize, sorghum and millet in southwestern Zimbabwe (Ncube et al 2009). Better resource-endowed farmers who owned livestock used manure to partially improve soil fertility and crop productivity while resource-poor farmers relied on remittances and governmental support, including food for work schemes and food assistance, to cope with rain shortages.

In the regions with poorest governance and high erosion sensitivity (turquoise cluster, figure 1), one nested archetype showed absence of governance, lack of income and medium undernourishment in Somalia (yellow cluster, figure 2), representing 5% of the nested archetypes' population (table 3). For example, on-going violence, debt repayment and recurrent droughts have largely exhausted the asset base of politically marginalised agro-pastoralists in Somalia (Le Sage and Majid 2002). In particular, reduced herd sizes had decreased the pastoralists' capacity to sell livestock as a key strategy for coping with dry periods, resulting in a reduced number of daily meals (Le Sage and Majid 2002), thus contributing to the undernourishment depicted by this nested archetype. In contrast to the more agriculture-based economies addressed so far, there is one nested archetype representing most remote areas, weak governance, higher income and severe undernourishment (beige cluster, figure 2) that clearly distinguished the economies of Sudan and Angola. Weak governance and remoteness limited people's capacity to produce sufficient food and prepare for climate, market and other risks. The combination of higher income and undernourishment again suggested strong inequalities in wealth distribution, which is corroborated by a higher GINI Index, above all in Angola (World Bank 2015). Industries constituted the main share of gross domestic product in Angola (72%) while agriculture (42%) and services (37%) were important in Sudan (FAO 2014).

Agricultural productivity is a major driver of well-being (Irz et al 2001) such that sustainable agricultural intensification provides essential opportunities for reducing farming systems' vulnerability (Pretty et al 2011). The archetypes identified in this study help to identify those regions where sustainable intensification is a viable strategy for reducing vulnerability. They provide entry points for discussing sustainable intensification strategies at regional scales and identifying potential out-scaling domains, that is to say regions with similar socio-ecological potentials and constraints where interventions can be tested (Coe et al 2014). Within such scaling domains, intensification approaches can be suggested, but the implementation of particular practices will require in-depth knowledge of the local socio-ecological circumstances in which farmers operate.

If better agropotential, low erosion sensitivity and better governance are regarded as prerequisites for success, sustainable intensification can particularly support vulnerability reduction in the medium undernourishment, better governance and lowest erosion sensitivity archetype (dark olive cluster, figure 2). These regions constituted 21% of the nested archetypes' area (table 3) and comprised 9% of the broad archetypes (table 2). Favourable socio-political conditions are especially valuable for improving information access, knowledge co-creation and trust-building among all stakeholders as a prerequisite for successful intensification (Pretty et al 2011). For example, effective agricultural extension services and well-enforced environmental legislation were cornerstones for sustainably intensifying smallholder agriculture in southwestern Niger (Reij and Smaling 2008), where this nested archetype was found. Extension services had provided information about half-moon structures and contour stone bunds for capturing rainwater and reducing nutrient loss in order to overcome water scarcity and soil infertility.

If erosion can be controlled, the nested archetypes depicting low to medium undernourishment and better agropotential and severe undernourishment and lower agropotential (light and dark brown clusters, figure 2) would also be suited for sustainable intensification. These areas represented 41% of nested archetypes (table 3) and 17% of broad archetypes (table 2). The slightly better agropotential in the light brown cluster may be more attractive but would require careful consideration to avoid potential soil erosion. Moreover, the relatively good governance and somewhat lower remoteness in this archetype indicated supportive socio-political conditions. Finally, the severest undernourishment and medium governance archetype (light olive cluster, figure 2) may also provide suitable conditions for sustainable intensification given relatively good agropotential and low erosion sensitivity, though requiring governance improvement. In total, these four nested archetypes represented 31% of the drylands in sub-Saharan Africa encompassing 36% of the dryland population (tables 2 and 3). This constitutes a substantial dryland area well-suited to complementing sustainable intensification efforts in the humid parts of Africa. A next step would be to estimate these regions' potential for sustainably increasing agricultural production taking into account (i) the socio-ecological causes of yield gaps and ways of closing them (Tittonell and Giller 2013, Jägermeyr et al 2016), (ii) spatial interactions between land use and agri-food systems (Verburg et al 2013, Zimmerer et al 2015) and (iii) the potential of particular intervention options such as irrigation to support the transition towards sustainably intensified agriculture (Amjath-Babu et al 2016).

Moreover, the archetypes provide useful insights for transferring intensification approaches, assuming that suitable interventions are comparable among similar vulnerability situations (Sietz et al 2011a, Kok et al 2016, Václavík et al 2016). Ingram et al (2002) studied agricultural practices and the potential of seasonal weather forecasts to enhance productivity in three types of farming systems in Burkina Faso. For example, subsistence-oriented farmers frequently used soil conservation measures such as stone bunds to improve the productivity of heavily eroded upland soils in parts of the Central Plateau categorised in the erosion-sensitive light brown cluster (figure 2). Stone terraces also contributed to overcoming erosion as a major productivity constraint in drylands in southern Kenya (Ifejika Speranza et al 2008), classified in the same archetype (light brown cluster). In this region soil erosion was one of the major concerns smallholder farmers raised regarding food security. At the same time, precipitation variability undermined food production in both central Burkina Faso and southern Kenya, rendering seasonal weather forecasts an important risk-reduction strategy (Ingram et al 2002, Ifejika Speranza et al 2008). In particular, forecasts about the onset and length of the rainy season could assist farmers in both regions in timing their cropping activities and using faster-maturing varieties (Ingram et al 2002, Ifejika Speranza et al 2008).

These empirical insights highlight the value of the nested pattern approach presented in this study. They demonstrate that the membership in a nested archetype can help to better understand sustainable intensification opportunities in order to reduce vulnerability in a given location. The broad archetypes reflected the empirical findings only partly. With regard to implementation, attention needs to be given to socio-ecological synergies and trade-offs (Sendzimir et al 2011, Lahmar et al 2012, Descheemaeker et al 2016), as well as key dynamics pertaining to the uptake, modification, abandonment and replacement of intensification strategies (Sietz and Van Dijk 2015). These finer-scale interactions remain beyond the scope of a continental assessment due to the lack of appropriate data, yet they are important for mainstreaming sustainable intensification and climate adaptation in policy planning (Sietz et al 2011b, Wright et al 2014).