Climate Risk Management 20 (2018) 64–77

Contents lists available at ScienceDirect

Climate Risk Management

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

Understanding resilience of pastoralists to climate change and

T
variability in the Southern Afar Region, Ethiopia
⁎
Muluken Mekuyiea, , Andries Jordaana, Yoseph Melkab
a
University of the Free State, Disaster Management Training and Education Center for Africa, P.O. Box 339, Bloemfontein, South Africa
b
Hawassa University, Wondo Genet College of Forestry and Natural Resource, P.O. Box 128, Shashemene, Ethiopia

AR TI CLE I NF O AB S T R A CT

Keywords: Change in climate and climate extremes are acknowledged as a vital challenge to pastoral pro-
Asset duction systems. Alternative systems that are accessible to a household in order to make a living
Livelihood could determine the household’s resilience at a given point in time. This study was conducted in
Climate shock the Southern Afar region in Ethiopia to understand the resilience of pastoralists to climate change
Pastoralist
and variability. A household questionnaire survey and focus group discussions were employed to
Resilience
collect primary data at household level. A total of 250 pastoral households were sampled using
stratified random sampling. The data obtained were analysed using descriptive statistics and
principal component analysis. The resilience of households to climate shocks and stresses was
determined using a two-step modelling approach by clustering households into livelihood groups,
gender and districts. The results indicated that agro-pastoral households were more resilient than
pastoralists to climate-induced shock. Furthermore, households in the Gewane district were more
resilient than those in the Amibara district. Female-headed households were less resilient than
male-headed households. Enhancing livestock assets and productivity, social safety nets, access
to market, credit, extension services and education, improving irrigation crop farming, and
providing farm inputs significantly enhanced the resilience of pastoralists to climate change and
variability.

1. Introduction

In pastoral communities of Ethiopia, climate-induced shocks and stressors such as drought, rising temperature and irregular
rainfall reduce pasture and water availability leads to animal loss through hunger and disease (Conway, 2000). The weather-related
natural disasters frequently occur in pastoral areas of Ethiopia, which has been exacerbated by the depletion of the natural resources
and destruction of ecosystems due to anthropogenic activities (Tadege, 2007). Ethiopia is particularly susceptible to drought, making
drought the most significant disaster influencing the country over time (Seleshi & Zanke, 2004). Rainfall anomalies and the delayed
onset of the rainy season along with rising temperatures, lead to impoverished grassland, lack of feed and water, and heat stress to
livestock. This has, in turn, increased the mortality rate of herds, susceptibility of livestock to disease and emaciation as a result of the
long distances they travel in search of pasture and water (Muluneh & Demeke, 2011). Although the drought may occur all over the
globe, in general its harm is not as intense as in Africa, particularly in Ethiopia (Funk et al., 2008; Seleshi & Zanke, 2004; Williams &
Funk, 2011). Droughts, heat waves and floods have increased in Ethiopia over the past decades. Excessive floods due to the high
intensity of rainfall in the Ethiopian highlands caused loss of life and damaged properties of the people who inhabited arid and semi-

⁎
Corresponding author.
E-mail address: [email protected] (M. Mekuyie).

https://doi.org/10.1016/j.crm.2018.02.004
Received 16 June 2017; Received in revised form 9 February 2018; Accepted 14 February 2018
Available online 15 February 2018
2212-0963/ © 2018 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/BY-NC-ND/4.0/).
M. Mekuyie et al. Climate Risk Management 20 (2018) 64–77

arid areas (Tadege, 2008).

It has been observed that although change in climate happens all over the world, its influence and extent differ across multiple
levels and scales. Its impacts are not the same at district, regional, national and global level. Although changes in climate and climate
extremes will be the greatest challenge for people in Ethiopia, few studies have been undertaken on the resilience to climate change.
Most literature investigated seasonality, poverty and food insecurity (Dercon & Krishnan, 2000). Studies conducted expressly in the
context of farmers’ resilience to climate change and climate variability is limited. Deressa et al. (2008) assessed the vulnerability of
households to climate-induced shocks and stresses at national level in Ethiopia. However, insights into resilience to climate per-
turbation vary with the scale of analysis. Resilience to climate-induced shocks assessed at national level can conceal variations in
local resilience of households (Parkins & MacKendrick, 2007). Accordingly, the national-level (macro-scale) assessment by Deressa
et al. (2008) could have overlooked variations in vulnerability at the local level since the vulnerability level may vary even among
households at district level. Households at district level can vary in terms of level of food insecurity, coping and adaptation capacity,
access to credits, public services, safety nets and natural resources. In such conditions, variability at local level is usually ignored in
nationwide resilience studies. Therefore, it is difficult to precisely understand the spatial aspects of households’ resilience from
nationwide resilience assessments. This shows the significance of scale in resilience studies and ensures the necessity of resilience
studies at micro-level.
A study at district, regional, national and global level is essential to integrate worthwhile adaptation strategies in development
policies. The reason for this is that adaptation/coping capacities to climate change and variability can vary at all these levels, taking
into account households’ level of income, local exposure, and education level, to mention but a few. It is on the basis of these premises
that the present study was conducted to understand the resilience of pastoralists to climate change and climate variability in the
Southern Afar region of Ethiopia.

2. The resilience concept

Resilience is the capability of a social–ecological system to continue after a shock and to reorganise while sustaining a funda-
mentally similar function (Folke, 2006; Holling and Meffe, 1996; Walker et al. 2004). The idea of resilience was initially used by
Holling and Meffe (1996) to define ecosystem resilience, but is currently applied to other contexts. The concept is progressively used
in social sciences to explain people or household resilience (Levin et al., 1998), thus a broader concept of resilience is adopted. The
wide variety of resilience concepts makes it complicated to detect common characteristics. However, nearly all descriptions stress the
ability to adapt effectively against shocks. Norris et al. (2008) suggests an overall agreement on two significant characteristics of the
resilience definitions, namely (i) it is better conceived as a capability or a process than as a result, and (ii) it is well-conceptualised as
adaptability rather than steadiness. A first step in the direction of understanding the resilience concept in a learning environment is to
discourse the important characteristics and regulations of the system. In order to enhance a common understanding of resilience in
the situation of diverse systems, Norris et al. (2008) identified the most important principles, namely (i) a changing environment is
given, (ii) systems are too complex to know or map all interdependencies, and (iii) there is not only one stable state in reality –
alteration is the common state. Resilience therefore, is a learning process and no stable state exists in reality.
There are two contrasting resilience concepts. The first concept is described by Gunderson et al. (1995) as resilience in en-
gineering and by Cutter et al. (2008) as the ability to persist and survive a disaster with minimal influence and destruction. It includes
the ability to lessen or evade damages, encompass the impacts of hazards, and bounce back with slight disturbances (Cutter et al.,
2008). Rose (2009) also describes engineering resilience as the time taken by a system to recover to its earlier state after a dis-
turbance. Furthermore, not only the time required for bouncing back, but also the pattern of bouncing back should be considered.
According to the engineering resilience framework, opportunities to adapt or learn from a previous disturbance and shift to an
alternative state are often not taken into account. The second resilience concept, called ecological resilience, is the amount of
perturbation that a system can accommodate without redefining its structure and function (Holling and Meffe, 1996; Walker et al.,
2004). A concept regularly quoted when referring to resilience of an ecosystem is the adaptive renewal cycle, primarily developed by
Holling (2001). The adaptive renewal cycle is an informative model made from long-time measurements of ecosystem changes over
time, such as the succession of species in four phases of change forced by periodic disturbances and processes (Folke, 2006). Resi-
lience refers to persistent or robustness of a system to disturbance and about the possibility that disruption may lead to the occurrence
of new trajectories. Therefore, resilience offers the ability of the system to adapt to disturbances, which allows for sustainable
development. It does not mean that resilience has always been a positive characteristic of the system (Folke, 2006).

3. Methodology

3.1. Description of study areas

The Afar region is situated in the north-eastern part of Ethiopia and comprises an area of about 72,053 km2 (CSA, 2008) between
39° 34′ and 42° 28′ East Longitude, and 8° 49′ and 14° 30′ North Latitude. The Afar Region has a population of approximately 1.4
million people, of which approximately 87% are living in rural areas (CSA, 2008). The study was conducted in the Southern Afar
region, in Amibara and Gewane districts (Fig. 1). Agro-ecologically, the Amibara is semi-arid with a temperature ranging from 25 °C
to 35 °C and an average annual rainfall of 530 mm. The altitude of Amibara ranges from 720 m asl to 1100 m asl. Gewane is arid and
semi-arid with a temperature ranging from 28 °C to 42 °C and an average annual rainfall of 450 mm. The altitude of Gewane ranges
from 550 m asl to 650 m asl (CSA, 2008).

65
M. Mekuyie et al. Climate Risk Management 20 (2018) 64–77

Fig. 1. Map of the study areas.

3.2. Research design

A stratified random sampling approach was adopted to select villages and households. Stratification was based on livelihood
activities practiced by households in Amibara and Gewane districts. Firstly, villages were identified and stratified into pastoralists
(those households solely dependent on livestock production) and agro-pastoralists (those households dependent on both livestock and
crop production). Accordingly, ten pastoralist villages and five agro-pastoralist villages from the Amibara district was included from
which two pastoralist villages (Andido and Bure) and one agro-pastoralist village (Melka-Worer) were randomly selected. Similarly,
from the Gewane district, six agro-pastoralist and two pastoralist villages were included from which two agro-pastoralist villages
(Yigle and Urafita) and one pastoralist village (Arogew Gewane) were randomly selected. Overall, six villages were sampled from the
two districts. A total of 250 households were selected for interviews using semi-structured questionnaire to elicit responses between
December 2015 and April 2016. A random start was used in selecting the first household to be interviewed. For the selected
households whose heads were absent, next household was chosen and interviewed.

3.3. Data collection and analysis

A semi-structured questionnaire with open-ended, multiple-response and dichotomous questions was administered during data
collection. Information on various aspects was collected through interviewing of the selected household head. The survey addressed
information about household characteristics, access to basic services, annual income, sources of income, climate change information,
adaptation and coping strategies, farm labour, social networks and remittances. To avoid misunderstanding, the household interviews
were undertaken in the local language by local field assistants. To complement the household questionnaire data, 29 individuals from
the sampled villages and various organisations were interviewed as key informants.
Four focus group discussions were conducted separately with gender equality (six men and six women) from each district. Clan
leaders and elders who had long experience about the region’s climate, livelihood sources, and politics were included in the focus
group discussions. The discussions aimed at capturing the local knowledge on climate change and variability and its impacts on local
communities’ adaptation and coping strategies. The collected data were coded and thereafter analysed using SPSS (version 20) and
STATA software (version 13.0). The Principal Component Analysis [PCA] was also employed to achieve the objectives of this study.

66
M. Mekuyie et al. Climate Risk Management 20 (2018) 64–77

3.4. Quantitative assessment of pastoralists’ resilience to climate-induced shocks

In recent years, the quest to quantify individuals, households and society's resilience to disasters or extreme events beyond their
control remains a challenge from a quantitative point of view. Given the fact that the resilience of individuals, households and
societies are not observed by the analyst, quantifying resilience demands a detailed modelling approach. Various authors have
employed different modelling approaches (Alinovi et al., 2008, 2010; Carter et al., 2006; Demeke & Tefera, 2010). Alinovi et al.
(2008, 2010) as well as Demeke and Tefera (2010) employed a latent variable approach to quantify resilience among households in
Kenya and Palestine. Carter et al. (2006) adopted a proxy variable approach by employing an observable variable to represent
resilience. These authors assessed resilience and expressed it in six dimensions, which included social safety nets, public service
accessibility, assets, income and food access, stability and adaptive capacity. They expressed these dimensions as latent variables.
Following this, resilience was estimated in a two-stage procedure. The first estimation procedure involves the use of factor analysis to
identify the set of latent variables explaining resilience. The identified latent variables in the first stage are then used to calculate an
index for resilience for the sampled households.
Additionally, the latent variable approach and PCA were employed by Demeke and Tefera (2010) to examine resilience to food
insecurity in Ethiopia. These authors used panel data at a household level. They first identified four components that explained
resilience to food security using PCA. These components included household access to food, liquid assets, level of education, and
social network. Secondly, the authors employed panel fixed effect and dynamic panel modelling approaches to find factors that
influence household resilience.
In Ethiopia, different coping strategies have been adopted by resilient households. Carter et al. (2006) revealed that resilient
households in Ethiopia relied on their assets and also depleted their stock in periods of drought, especially during the 1998–1999
drought periods. Similar strategies were employed in the Honduras in 1998 to minimise the impact of the Mitch hurricane.
Households, who are not resilient, tend to cope by cutting down on their consumption level as a way of keeping their assets or
properties.
Based on the above discussion, it is clear that using a proxy variable approach is the straightforward method for calculating
resilience. One limitation of this approach is the difficulty in identifying variables, which will act as proxies for resilience. This is
because resilience is a multifaceted concept. Additionally, there is difficulty in defining the levels or categorising resilience and
finding its determinants if the proxy variable approach is employed. For these reasons, it was decided to adopt the multi-stage
modelling approach developed by Alinovi et al. (2010), building on its flexibility to adapt to very different real cases. The use of this
approach depended on the premise that the alternatives accessible to a household to make a living, would determine the household’s
resilience at a given point in time.
Empirically, the resilience index (RI) for a household (i) is described as:

RIi = f (Ai ,ACi,SSNi,APSi ,Si,IFAi ) (1)

where:

RI = resilience index
A = assets
AC = adaptive capacity.
SSN = social safety nets
APS = access to public services
S = stability
IFA = income and food access

During the research, household resilience was not observed by the researcher, and hence it is regarded as a latent variable. In
measuring resilience, it was essential to calculate IFA, S, SSN, APS, A and AC separately. These indicators were also latent, because
they were not directly observable during the survey, but could be computed using multivariate techniques. The household resilience
model diagram is presented in Fig. 2, where boxes denote the observed variables and circles refer to latent variables.
According to Alinovi et al. (2010), two options are adopted to estimate resilience of households. The first approach is structural
equation models, which are the most suitable models to determine resilience of households based on latent variables. This approach is
based on regression and factor analysis. According to this model, observed variables are employed to estimate the latent variables
using a factor analysis model, and concurrently a regression analysis is employed to determine relations amongst the latent variables
(Bollen, 1989). A multi-stage strategy is the second alternative proposed by Alinovi et al. (2010) to estimate the latent variables
separately using observed variables.
In this study, a two-step approach was employed to estimate the RI of households. The first procedure involved the identification
and measurement of observed variables or indicators for the estimation of dimensions of the resilience. Secondly, the RI for each
household was determined based on the estimated values of the latent variables (dimensions of resilience). A PCA was used to
examine the components of resilience and the percentage variance explained by each of the components, as well as their com-
monalities. According to the approach proposed by Alinovi et al. (2010), the factor variance obtained for each factor from the PCA
was multiplied by the generated factor to develop the RI of each household (see equation iii). The formula is described as follows:
RI = V1 ∗F1 + V2 ∗F2 + Vn ∗Fn (2)

67
M. Mekuyie et al. Climate Risk Management 20 (2018) 64–77

Fig. 2. The household’s resilience index model diagram.

where:

RI = resilience index
V1 = variance of proportion explained by factor 1 (F1)
V2 = variance of proportion explained by factor 2 (F2 )
Vn = variance of proportion explained by factor n (Fn ) with eigenvalues > 1

4. Results and discussions

This study determined the current status of the resilience of Afar pastoralists to climate-induced shocks based on cross-sectional
data. However, due to a lack of panel data sets, the present study did not analyse the dynamics of pastoral household’s resilience to
climate variability and change. As discussed in Section 3.4, household resilience to climate-induced shock was determined using a
two-step approach developed by Alinovi et al. (2010). The first procedure involved the identification and measurement of observed
variables or indicators for the estimation of RI. Secondly, the RI for each household was determined based on the estimated values of
the latent variables using Thompson’s regression method as proposed by Alinovi et al. (2010). The RI is discussed by comparing
resilience of households based on livelihood strategies (pastoralists versus agro-pastoralists), gender (male-headed household [MHH]
resilience versus female-headed household [FHH] resilience) and districts (Amibara district versus Gewane district). In the next
sections, demographic characteristics and public services of the study area, the six dimensions used to estimate the RI and resilience
of households with respect to livelihood groups, gender and districts are discussed.

4.1. Descriptive statistics

The results indicated that the average size of the family in the study area was 7.9. This was relatively higher than the national
average rural household size. According to Central Statistical Agency & the World Bank (2013), the average household size in rural
areas of Ethiopia was 5.1 persons. Such large family size in the region might be associated with the polygamous culture that is
commonly practiced in the Afar region. The average age of household heads was 52.7 years and most were males (58.8%). We
acknowledge that our findings have lower male headed households percentage compared to other studies. This might be as a result of
the fact that data was collected during a drought period and most male of households out-migrated to other areas in search of pasture
for their livestock. Furthermore, female head of households very often graze around their homestead. A considerable number of the
respondents (72.2%) were illiterate, which meant that only 28.8% of the respondents could read and write with formal educations
ranging from 1 to 10 years. Furthermore, the results revealed that only 28.8% of households had access to extension services,
indicating that access to extension services in the Southern Afar region was very poor.
Access to basic services in the Southern Afar region was generally poor. For example, only 22.2% of the households had access to
markets near their villages. The rest of the households (77.8%) usually travelled long distances to sell their cattle and camels.
According to the respondents, on average they travelled more than 12 h across other adjacent districts or regions (Amhara and
Oromia regions) to get to the market for selling large animals (cattle and camels), which was costly in terms of time and labour.
During such long-distance travelling, pastoralists lost their livestock due to stresses associated with prolonged travel time, feed and
water scarcity and car accidents. Finally, the remaining animals that reached the market lost weight with an adverse effect on prices.

68
M. Mekuyie et al. Climate Risk Management 20 (2018) 64–77

Sheep and goats were sold daily in nearby small village markets, but because of a lack of bargaining power and not enough pur-
chasers they could not sell their livestock at a reasonable price. Less than a third (30%) of the households had access to veterinary
clinics and services and 32% households had access to the health centre and health services.
The results further indicated that households’ access to credit (14.8%) was very poor in the study area. Furthermore, 90% of
households received market information about the price prior to selling through an informal local information exchange system
(locally called Dagu). Only 2% of households received formal market information through district extension officers and the radio.
Therefore, there is a need to provide affordable credit access and market information in order to enhance the resilience of pastoralists
towards climate related shocks and stresses.

4.2. Observed variables and their factor loadings used to compute components of resilience

This section discusses the findings on how the observed variables contributed to measuring the value of the latent variables
representing the resilience components. The PCA was used to estimate the components of resilience (the latent variables). The
approach depended on the premise that the alternatives accessible to a household to make a living determined the household’s
resilience at a given point in time. The RI was determined based on six dimensions (latent variables) including assets (A), adaptive
capacity (AC), social safety nets (SSN), access to public service (APS), stability (S), and income and food access (IFA).

4.2.1. Asset (A)

Assets are one of the most important capitals households have as an adaptation/coping mechanism to hazards. Hence, assets
should be taken as a key factor in measuring resilience. The indicators for this latent variable are discussed as follows:

i. Livestock owned in tropical livestock unit (TLU): This is a materialised asset owned by a household. By having more assets a
household’s resilience to climate-induced shocks increases. Those households with more livestock will have more opportunities to
recover after climate-related hazards.
ii. Farmland owned: Households practicing crop farming using irrigation or opportunistic rainfall agriculture and livestock pro-
duction may have better resilience to climate shocks than households solely dependent on livestock production.
iii. Adult labour: Those households with a better working force can have better livelihood outcomes and more resilience.
iv. Access to irrigation water: This is also a crucial asset, especially for pastoralists who are vulnerable to frequent and prolonged
drought and where there is increasing climate variability. Therefore, households having access to irrigation water can have the
opportunity to diversify their livelihood strategies and easily adapt to climate-induced shocks. It is a dummy variable equal to 1 if
the household has access to irrigation water in the previous year; 0 otherwise.
v. Access to irrigation equipment and improved seed varieties: This is critical for agro-pastoralist households. Water pumps for irrigation
and improved seed varieties, which can resist droughts and diseases as well as early maturation for harvest, increases efficiency of
households to adapt to prolonged and recurrent droughts. The value of this indicator is 1 if the household had access to irrigation
equipment and improved seed varieties: 0 otherwise.
vi. Access to credit: The value of this indicator is 1 if the household had affordable credit access; 0 otherwise.

As shown in Table 1, all the values of the variables under agro-pastoralist and districts are characterised by a fair degree of
similarity. The results indicated that all variables had positive impacts on the assets of households. However, the impacts of irrigation
equipment and improved seed supply, access to irrigation water, and farmland-owned had a negligible on the assets of pastoralists.
This was explained by pastoralists exhibiting little or no participation in crop farming. On the other hand, labour availability and
livestock ownership were more relevant and important influences on pastoralists’ assets. Similarly, labour availability and livestock
ownership were also significant for the assets of agro-pastoralists as well as households in the Amibara and Gewane districts. Access
to credit also influenced assets of both livelihood groups and districts to a similar degree. Farmland ownership, access to irrigation
water and inputs such as access to irrigation equipment, and improved seed supply were more important for agro-pastoralists fol-
lowed by households in the Gewane district.

Table 1
Contribution of observed variables to estimate asset (A) by livelihood groups and districts.

Variables Livelihood groups Districts

Agro-pastoralist Pastoralist Amibara Gewane

Irrigation equipment and improved seed supply 0.68 0.236 0.504 0.743
Adult labour 0.71 0.775 0.746 0.783
Livestock owned (TLU) 0.754 0.945 0.793 0.610
Access to irrigation water 0.879 0.121 0.515 0.792
Access to credit 0.718 0.717 0.64 0.672
Farmland owned (ha) 0.907 0.350 0.607 0.803

Source: Author’s field survey (2016).

69
M. Mekuyie et al. Climate Risk Management 20 (2018) 64–77

Table 2
Results of the factorial analysis to estimate adaptive capacity (AC) of households.

Variables Livelihood groups Districts

Agro-pastoralist Pastoralist Amibara Gewane

Herds composition change 0.456 0.955 0.855 0.604

Mobility and herd splitting 0.687 0.927 0.916 0.801
Combining livestock production with crop farming 0.912 −0.17 0.72 0.832
Access to extension & credits 0.665 0.597 0.508 0.609
Level of education 0.632 0.515 0.607 0.678
Number of income sources 0.752 0.543 0.602 0.678

Source: Author’s field survey (2016).

4.2.2. Adaptive capacity (AC)

Adaptive capacity shows the ability of a household to adapt and cope with a hazard, in this case, climate-induced shocks and
stresses such as drought, floods and climate variability. It enables households to continue performing their basic functions. The
following observed variables were included to estimate adaptive capacity for this specific study area.

i. Diversity of income sources: Refers to the number of income sources for the household. The more income sources, the more
resilience the pastoralists have to a given disaster risk.
ii. Changing herd composition: The value of this indicator was 1 if the household had changed his/her herd composition from keeping
higher numbers of cattle and sheep, but lower numbers of camels and goats to higher numbers of camels and goats production but
lower numbers of cattle and sheep in the previous year; 0 otherwise.
iii. Herd mobility: The value of this indicator was 1 if the household used mobility as coping strategy to climate shock; 0 otherwise.
iv. Irrigation farming: The value of this indicator was 1 if the household used irrigation farming as an adaptation strategy to climate
shock; 0 otherwise.
v. Access to extension service: The value of this indicator was 1 if the household had access to extension services in the previous year;
0 otherwise.
vi. Education level: In this case, households were grouped based on their education level as illiterate (the value of this indicator was
0), only read and write (the value of this indicator was 1), those who completed primary school (the value of this indicator was 2),
secondary school (the value of this indicator was 3), and above (the value of this indicator was 5).

Table 2 indicates the results of the factorial analysis to estimate the adaptive capacity of households by livelihood groups and
districts. The results revealed that changes in herd composition, mobility and herd splitting were more important for pastoralists and
households in Amibara district followed by households in Gewane district. However, herd composition change was considered as low
adaptation strategies by agro-pastoralists. The results further indicated that combining livestock production with crop farming and
income diversification were more important for the resilience of agro-pastoral households followed by households in the Gewane and
Amibara district. On the other hand, access to education, extension and credits influenced the latent variable, adaptive capacity in a
similar degree across the livelihood groups and districts.

4.2.3. Social safety nets (SSN)

This is an important dimension of household resilience, particularly the poor, because social safety nets assist in reducing crises.
Recently, because of mounting poverty, dependence of households on support from charities, non-governmental organizations
(NGOs) and international agencies have increased in the Afar region, while the mutual support from relatives and friends was still
significant. Therefore, it is important to consider social safety nets to represent the system’s ability to reduce climate shocks, and
indicators for social safety nets should be involved in the measurement of household’s resilience. The indicators or observed variables
used to estimate the social safety nets were:

i. Free distribution of food/grain from government or NGOs: The value of this indicator was 1 if the household received food/grain
support; 0 otherwise.
ii. Remittance: The value of this indicator was 1 if the household received cash and in-kind assistance from friends, communities, and
relatives; 0 otherwise.
iii. Cash-for-work: Included transfers received by households from international agencies, charities and NGOs. The value of this
indicator was 1 if the household received cash from international agencies, charities and NGOs; 0 otherwise.

The factor loadings of the observed variables used to estimate social safety nets are indicated in Table 3. The latent variable was
not influenced by the observed variables to a similar degree across the livelihood groups and districts. The observed variables were
relatively more important to influence the social safety nets of pastoralists than agro-pastoralists. This implied that pastoralists were
more dependent on social safety nets than agro-pastoralists. Similarly, social safety nets were more important for households in the
Amibara than the Gewane districts. More dependency on social safety nets indicated a low level of livelihood outcome, like low food

70
M. Mekuyie et al. Climate Risk Management 20 (2018) 64–77

Table 3
Factor loadings of the observed variables to measure social safety nets (SSN).

Variables Livelihood groups Districts

Agro-pastoralist Pastoralist Amibara Gewane

Food aid from government or NGOs 0.512 0.79 0.69 0.607

Remittance 0.578 0.738 0.637 0.525
Cash-for-work 0.502 0.884 0.784 0.569

Source: Author’s field survey (2016).

security.

4.2.4. Access to public services (APS)

Access to public services provides pastoral households with numerous essential factors for improving their resilience, for instance,
by enhancing the efficiency of pastoral households’ access to public services. Consequently, access to public services impacts a
pastoral household’s ability to manipulate risks and react to climate shocks. The following indicators were considered to estimate
access to public services:

i. Time taken to nearest health centre: In this case, the value of the indicator is the time that it takes to walk to the nearest health
centre. Those households near to the health centre can have better access to health, which is important in building resilience.
Household members can accomplish their livelihood activities successfully if they are healthy.
ii. Time taken to nearest market centre: The value of this indicator is the time that it takes to walk to the nearest market centre.
Distance to the market also influences resilience of households negatively or positively. Households who are near to the market
centre can easily buy and sell their agricultural inputs and outputs. This, in turn, increases their adaptation to shocks and stresses.
iii. Time taken to nearest veterinary clinic: The value of this indicator is the time that it takes to walk to the nearest veterinary clinic.
Households with better access to a veterinary clinic can have a better livestock asset as they have a more opportunities to reduce
livestock death associated with disease outbreaks, than households with low access to a veterinary clinic.
iv. Time taken to the nearest school: The value of this indicator is the time that it takes to walk to the nearest school. Access to schools
increases the number of children obtaining an education and this, in turn, increases the human capital of households. The higher
the human capital of the household, the higher its flexibility and its ability to diversify its livelihood strategies such as non-
pastoral/non-farm activities and, hence, the higher the resilience of households.
v. Access to early warning information (EWI): This is an important element of disaster prevention and management, which provides
information to pastoralists before the occurrence of a climate-related hazard. For example, those households who are better
informed about a coming drought may destock their livestock at a good price before the occurrence of drought and can have a
better recovery capacity after drought. The value of early warning information was 1 if the household received such a warning
from the Disaster Prevention and Management Centre of the region; 0 otherwise.

Accesses to public services are beyond the household’s control, but are important in building household resilience to climate-
induced shocks by enhancing economic connectivity through market access. The economic connectivity of a household is its ability to
be connected to several markets and income-generating opportunities. Access to public services also increases the household’s access
to assets by increasing human capital of households through increasing access to health and education. Table 4 reveals that all
observed variables in both livelihood groups and districts have similar patterns and influenced access to public services positively.
However, the observed variables were more relevant and relatively important for enhancing resilience of agro-pastoralists than
pastoralists. The observed variables were also more important for households in the Gewane district than the Amibara district. This
implies that there were inequalities in accessing public services among livelihood groups and districts. For example, the factor
loading for the time taken to the nearest school was 0.49 for pastoralists implying access to school for pastoralists were poor as
compared to agro-pastoralists. However, access to early warning systems was almost equally important in enhancing the resilience of

Table 4
Access to public services (APS) and their factor loadings across livelihood zones.

Variables Livelihood groups Districts

Agro-pastoralist Pastoralist Amibara Gewane

Time taken to nearest health centre 0.603 0.525 0.581 0.700

Time taken to the nearest school 0.551 0.49 0.53 0.574
Time taken to nearest market centre 0.654 0.589 0.686 0.756
Time taken to nearest veterinary clinic 0.705 0.635 0.607 0.670
Access to early warning systems (EWI) 0.691 0.622 0.618 0.638

Source: Author’s field survey (2016).

71
M. Mekuyie et al. Climate Risk Management 20 (2018) 64–77

Table 6
Observed variables and their factor loadings to estimate income and food access (IFA).

Variables Livelihood groups Districts

Agro-pastoralist Pastoralist Amibara Gewane

Total income 0.532 0.46 0.501 0.568

Per capita expenditure 0.623 0.538 0.602 0.657

Source: Author’s field survey (2016).

both livelihood groups and households in both districts.

4.2.5. Stability (S)

Stability is an important component for resilience of pastoral households, which shows how the livelihood options of the
household differ over time. The asset losses experienced by a household were taken into account to measure the value of this variable.

i. Livestock loss: Refers to the number of livestock that a household lost as a result of a climate-induced hazard such as drought.
ii. Crop loss: Refers to crop loss due to climate-induced hazards such as drought and floods.
iii. Other shocks: Refer to market shocks, illness and death of a household family member associated with climate-induced shocks.

The observed variables used to measure stability were instability indicators. Subsequently, each indicator was multiplied by
negative 1 in order to make them consistent with the meaning of the latent variable stability. Table 6 reveals that stability across
clusters was highly influenced by livestock losses, crop losses except in primary pastoralists, and other shocks such as market in-
stability. This implies that climate-induced livestock and crop loss as well as market failure were common and stable in the study
area.

4.2.6. Income and food access (IFA)

Income and food access is directly linked to the resilience of a household. Those households with better resilience to shocks have
better access to food (Alinovi et al., 2010). Usually, food access is estimated by income; however, to better measure the general
feature of access to food, per capita expenditure was included in estimating income and food access:

i. Total income: Refers to the total annual income of a household from all income sources.
ii. Expenditure: Refers to the annual expenses of a household for consumption.

The communalities of the observed variables to measure income and food access in Table 5 show that the share of per capita
income to food access was relatively low for pastoralist groups. This indicated that the role of per capita income was too low to
estimate the income and food access indicator in the case of pastoralists. Similarly, the values of per capita expenditure were char-
acterised by a fair degree of homogeneity across clusters, except for pastoralists, which meant that all values of per capita expenditure
contributed in a similar degree to estimate the income and food access dimension.

4.3. Measuring resilience

Following the methodology used by Alinovi et al. (2010), the factors of the resilience components were used to compute the
overall RI of households. The latent variables measured in the above sub-topics came to be the explanatory variables to measure RI.
Using the iterated principal factor method, factor analysis was run to re-estimate communalities iteratively.
Table 7 indicates the first factors that seem to represent the resilience of households fairly well, although social safety nets were
not positively correlated to the first factors indicating its negative correlation with other variables. As households become poorer, the
availability of social safety nets from members of the society, government or NGOs increases. The social safety nets had a positive
contribution to the household’s resilience to climate-induced shocks and hence became positive in the second factor. Adaptation

Table 5
Factor of stability (S) and their relative loadings.

Variables Livelihood groups Districts

Agro-pastoralist Pastoralist Amibara Gewane

Livestock loss 0.762 0.931 0.753 0.611

Crop loss 0.678 0.238 0.732 0.521
Other shocks 0.602 0.684 0.677 0.621

Source: Author’s field survey (2016).

72
M. Mekuyie et al. Climate Risk Management 20 (2018) 64–77

Table 7
Factor loadings of the components of resilience.

Variable Factor 1 Factor 2 Factor 3

Asset 0.7460 −0.0065 0.1011

Adaptive capacity 0.8690 0.0340 0.0045
Access to public services 0.5740 0.1400 0.0245
Social safety net −0.4680 0.6670 0.1340
Stability 0.6930 0.3450 0.0040
Income and food access 0.5601 0.0790 0.0450

Source: Author’s field survey (2016).

strategies followed by the households’ assets were the most important components of the resilience of pastoral households.
Table 8 reveals that factor 1 alone explains 56.3% of the variance, while factors 2, 3, 4 and 5 explains 18.2%, 12.1%, 10.1% and
3.26% of the variance, respectively. For resilience measurement, the first four factors that explained 96.7% of the variance were
included.
The Thompson’s regression method developed by Alinovi et al. (2010) was employed to generate the four factors in order to
estimate resilience of pastoral households. Once the four factors had been generated, each factor was multiplied by its own proportion
of variance explained as:
Resilience = 0.563 * Factor 1 + 0.182 * Factor 2 + 0.121 * Factor 3 + 0.101 * Factor 4
Households were clustered with livelihood groups, gender of the household head and districts, in order to discuss the estimated
RI. Fig. 3 shows Epanechnikov’s kernel density estimates of the resilience distribution in the Gewane and Amibara districts of the
Southern Afar region.
Fig. 3 reveals that more households with the highest RI, approximately 1.6 < RI ≤ 2, were found in the Gewane than in the
Amibara district. However, more households with moderate levels of RI, approximately 1.3 ≤ RI ≤ 1.6, were found in the Amibara
than in the Gewane district. On the other hand, almost similar proportions of households were found in both districts having lower
levels of RI, 0 < RI < 1.3. The results further indicated that household resilience in the Amibara district revealed the most irregular
distribution or high variability, while relatively uniform resilience distribution of households was observed in the Gewane district.
The t-statistics test indicated that, on average, households with a higher RI were observed in the Gewane district than in the
Amibara district (Table 9). The higher resilience of households in the Gewane district can be explained by the relatively short
duration of drought occurrence, better per capita income level and assets and better access to agricultural inputs. The findings
indicated that there was a prolonged drought in the Amibara district, which started in 2014 and lasted until the beginning of 2016
(more than two years), while the Gewane district was hit by the drought from the middle of 2015 until the beginning of 2016 (less
than a year). The results further showed that the asset endowment of the Gewane district, such as the number of livestock owned per
household, farmland owned per household, access to irrigation equipment, improved seed varieties and per capita income level were
relatively better than that of the Amibara district. For example, the average per capita income of households in the Gewane district
was 4 492.70 Ethiopian Birr [ETB], while the average per capita income of households in the Amibara district was 3 897.5 ETB.
Hence, households with a better income level can easily diversify their livelihood options and become more resilient.
Analysing resilience by livelihood groups indicated that agro-pastoral households were more resilient than pastoral households
(Fig. 4). The higher resilience of agro-pastoral households can be explained by their small-scale irrigation crop farming activities in
addition to livestock production. According to respondents, since livestock numbers and productivity declined over time due to
recurrent droughts in the area, some households started cropping using small-scale irrigation along the banks of the Awash River.
Pastoral households practiced small-scale crop cultivation using traditional irrigation methods such as furrows and channels using
ground gravity. However, recently, due to the number of dry spells and recurrent droughts in the study areas, the Awash River flow
was decreasing, making traditional irrigation difficult. Thus, subsidising of water pumps for irrigation, which was not affordable to
pastoral households, was in demand. The results indicated that out of 250 households, 40% practiced crop cultivation along the banks
of the Awash River. On average, the agro-pastoral households owned 1.94 ha plots of farmland. The respondents noted that crop

Table 8
Eigenvalues and variances explained by the components of the resilience.

Eigenvalues % variance

Factor 1 2.816 56.321

Factor 2 1.910 18.202
Factor 3 0.906 12.111
Factor 4 0.705 10.105
Factor 5 0.1975 3.261

Factor 1 includes adaptive capacity (AC) and assets (A), while Factor 2, 3, 4 and 5 include social
safety nets (SSN), access to public services (APS), stability (S) and income and food access (IFA),
respectively.
Source: Author’s field survey (2016).

73
M. Mekuyie et al. Climate Risk Management 20 (2018) 64–77

Fig. 3. Distribution of resilience of households in the Amibara and Gewane districts.

Table 9
Means variation for resilience between the Gewane and Amibara districts.

District Observed Mean Standard error Standard deviation df T Sig

Amibara 150 1.227074 0.0192557 0.2358331 248 −2.6540 0.004

Gewane 100 1.317569 0.0301738 0.3017376
Combined 250 1.263272 0.016906 0.2673073
Difference −0.090495 0.0340979 −0.1576534

Source: Author’s field survey (2016).

Fig. 4. Resilience index by livelihood groups.

farming was undertaken to complement pastoralism rather than to substitute livestock production. Recently, the customary in-
stitutions were marginalised by the government and replaced by formal institutions. According to local respondents, the formal
institutions made no effort towards managing and caring for rangeland resources since the government considered pastoralism as a
backward and non-viable livelihood strategy. Consequently, rangelands were deteriorated due to overgrazing and grazing during
inappropriate times. Consequently, lack of government support for traditional rangeland management and training to strengthen the
indigenous knowledge of the local people to manage the rangeland resources made pastoral households less resilient to shocks and
stresses induced by climate change and variability. In the study area, crop farming as well as livestock production has become
difficult due to expansion of Prosopis juliflora on the rangelands and farmlands. According to respondents, palatable grasses and
browse species disappeared partly due to an aggressive spreading of P. juliflora. The informants further noted that P. juliflora had not
only a direct impact on the availability of grazing pastures and farmlands, but it also had a direct negative impact on livestock health.
When livestock fed on P. juliflora pods for an extended period of time due to prolonged droughts and lack of pasture in the invaded
areas, it resulted in health problems such as constipation, disfiguration of jaws, breakdown of teeth, and loss of body condition,
reduced overall productivity and death of the animal. Thickets of P. juliflora also restricted livestock mobility in search of feed and
water.
In this study, the annual income of households was calculated and the results revealed that the minimum annual income of
pastoral households as ETB was 1 000 and the maximum was 14300, while the minimum annual income for agro-pastoral households
was 3960 and the maximum was 8 640. The average annual income of the pastoral households was 4 281, while that of the agro-

74
M. Mekuyie et al. Climate Risk Management 20 (2018) 64–77

Fig. 5. Resilience index by gender of the household head (HH).

pastoral households was 5 932.80. Significant differences were observed between the per capita income of agro-pastoral households
and pastoral households, indicating that the per capita income level of agro-pastoral households was significantly higher than that of
pastoral households. Obviously, those households with a better income level could be more resilient to climate shocks and stress than
households with a low level of income.
This study also compared resilience of MHH and FHH. The findings indicated that MHH were more resilient than FHH (Fig. 5).
Male headed households had ownership, access to and control over assets such as land; access to information, and education and
decision-making power. The results indicated that 80% of the farmlands were owned by MHH. The literacy levels were higher among
MHH compared to FHH. Thus those households with better education levels were in a better position to access and utilize information
and hence, better prepared to adapt to climate change. It was also a common practiced that MHH in Southern Afar region migrated
long distances in search of pasture and water for their livestock while FHH depended on grazing areas near their home for their
livestock. In the study area, though females were more accountable for most of the household’s well-being activities, FHH had low
access to resources, social services and credits as compared to MHH. This study is in agreement with the findings of Hassan and
Nhemachena (2008) who reported that MHH were more resilient than FHH to climate-induced shocks. Similarly, the present result is
supported by Haque et al. (2012) and Kartiki (2011) who indicated that FHH were less resilient than MHH to climate shocks and
stresses.
Furthermore, results from analysis of the resilience components do not express the entire history. In general, the findings in-
dicated that although variations among resilience strategies across clusters were observed, limited access to food and income and
poor access to public services (access to market, health, early warning information) was observed across all clusters (districts,
livelihood groups and gender of household head) (Figs. 6–8).
The results indicated that the Gewane district showed the high level of resilience, which was better in terms of assets (A) such as
livestock, farmland and labour, and adaptive capacity (AC) such as combining livestock production with crop farming, mobility, herd
splitting and herd composition change (Fig. 6). In contrast, the Amibara district showed a low level of resilience, which had better
access to social safety nets (SSN) such as food aid, cash-for-work and remittance.
Both districts showed low stability indicating a high loss of livestock and crop as well as market instability due to the recurrent
droughts and unreliable rainfalls in the area as reported by respondents. The results also indicated that households in the Gewane
district had better income and food access than the Amibara district (Fig. 6). Pastoralists showed very low levels of resilience, which
can be explained by low levels of all resilience components, except social safety nets indicating the worst situation of pastoralists
(Fig. 7). Similar findings were reported by Alinovi et al. (2010) who revealed that pastoralists in Kenya were less resilient than agro-
pastoralists, though the resilience difference was not significant. Although agro-pastoral households showed relatively high levels of
resilience, which were highly dependent on assets and adaptive capacity, they were also characterised by low stability and poor
access to public services and limited access to food and income (Fig. 7).
Fig. 8 revealed that FHH were less resilient than MHH. There were large differences between FHH and MHH in terms of access to
assets (mainly access to farmlands, labour, access to agricultural inputs and credits), ability to deal with climate-induced shock
(mostly due to large differences in terms of access to extension services, education and low capacity in irrigation crop farming), poor

Fig. 6. Resilience components by district.

75
M. Mekuyie et al. Climate Risk Management 20 (2018) 64–77

4.5
4.0
3.5
3.0
2.5
2.0 Agro-pastoralist
1.5
Pastoralist
1.0
0.5
0.0
Income and Access to Social safety Stability Adaptive Asset
Food Access Public Nets Capacity
Services

Fig. 7. Resilience components by livelihood strategy groups.

Fig. 8. Resilience components by gender of the household head.

access to public services such as access to early warning information and health, low stability and limited access to food and income.
The only better accessible capital for FHH was social safety nets, which they used to cope with climate-induced shocks such as
droughts.

5. Conclusion and recommendation

Local respondents listed the natural resource base, traditional resource management and utilisation, indigenous early warning
systems, and informal safety nets in the community as the most important components of their resilience. This suggested the need to
strengthen the natural resource base of the community, train and enhance traditional resource managers and use local people to
empower informal institutions. The results indicated that the Gewane district was more resilient than the Amibara district. This was
explained by the duration and intensity of the droughts that occurred in 2015 and 2016. The Amibara district experienced severe and
prolonged droughts while the Gewane district received some rainfall showers despite having drought for a year. Hence, those
households who lived in drought prone areas (Amibara) were less endowed in income and food access (IFA), assets (A), adaptive
capacity (AC), access to public services (APS) and stability (S) and, hence, less resilient. The findings also showed that pastoralists
were less resilient than agro-pastoralists. This was largely due to low level of adaptive capacity and assets owned by pastoralists. The
results also revealed that FHH were less resilient than MHH due to their low level of adaptive capacity and poor access to assets,
public services, food and income access.
Overall, for long-term resilience of pastoral households in the Afar region, it is better to focus on livestock intervention strategies
such as enhancing the ability of pastoral households to de-stock their livestock with a fair price before and during drought season, by
improving market access and infrastructure, increasing the productivity of livestock through better access to pasture and water, and
improving veterinary services. The study also recommends the government to develop strategies to control expansion of P. juliflora on
rangelands as well as farmlands. Water-related interventions are also paramount to enhance the resilience of pastoral households,
such as creating and rehabilitating water wells and boreholes and subsidised provision of water pump and fuel for irrigation ac-
tivities. This would assist the local people to diversify their livelihood activities into crop farming, develop forages and improve water
access for livestock and human consumption. If enhancing the resilience of pastoralists is the ultimate goal, the government and other
partner organisations should focus on long-term strategic livelihood interventions rather than emergency relief interventions by
equipping the local people with the capacity to manage and respond to climate-induced shocks in the early stage of the crisis.
Furthermore, policies with emphasis on women empowerment such as improving their access to and control over resources through a
better institutional set-up, improving irrigation facilities and skills, creating opportunities for non-farm income, giving access to
affordable credit and early warning information, supporting livestock mobility, as well as increasing access to markets, health and
veterinary services, are likely to improve resilience of pastoral households.

Acknowledgment

The African Forest Forum (AFF) and the National Research Foundation of South Africa (NRF) is gratefully acknowledged for their
funding of the research.

76
M. Mekuyie et al. Climate Risk Management 20 (2018) 64–77

Competing interests

The Authors declare that they have no competing interests.

References

Alinovi, L., D’Errico, M., Mane, E., Romano, D., 2010. Livelihoods strategies and household resilience to food insecurity: An empirical analysis to Kenya. Paper
prepared for the Conference on Promoting resilience through social protection in sub-Saharan Africa, Dakar, Senegal, 28-30 June 2010. Available from http://erd.
eui.eu/media/BackgroundPapers/Alinovi-Romano-D%27Errico-Mane.pdf.
Alinovi, L., Mane, E., Romano, D., 2008. Towards the measurement of household resilience to food insecurity: applying a model to Palestinian household data.
Deriving food security information from national household budget surveys. Experiences, Achievement, Challenges. FAO. Rome, pp. 137–152.
Bollen, K.A., 1989. Structural Equations with Latent Variables. John Wiley & Sons, New York.
Carter, M.R., Little, P.D., Mogues, T., Negatu, W., 2006. Shocks, sensitivity and resilience: Tracking the economic impacts of environmental disaster on assets in
Ethiopia and Honduras. BASIS Research Program on Poverty, Inequality and Development. US Agency for International Development, Washington, DC.
Central Statistical Agency & the World Bank, 2013. Ethiopia Rural Socioeconomic Survey (ERSS). Available from. http://siteresources.worldbank.org/INTLSMS/
Resources/3358986-1233781970982/5800988-1367841456879/9170025-1367841502220/ERSS_Survey_Report.pdf.
Conway, D., 2000. Some aspects of climate variability in the north east Ethiopian Highlands-Wollo and Tigray. Sinet: Ethiopian J. Sci. 23 (2), 139–161.
CSA (Central Statistical Authority), 2008. Summary and statistical report of the 2007 Population and Housing Census, Addis Ababa, Ethiopia.
Cutter, S.L., Barnes, L., Berry, M., Burton, C., Evans, E., Tate, E., Webb, J., 2008. A place-based model for understanding community resilience to natural disasters.
Glob. Environ. Change 18 (4), 598–606.
Demeke, M., Tefera, N., 2010. Household resilience to food insecurity in Ethiopia: Panel data evidence.
Dercon, S., Krishnan, P., 2000. Vulnerability, seasonality and poverty in Ethiopia. J. Develop. Stud. 36 (6), 25–53.
Deressa, T., Hassan, R.M., Ringler, C., 2008. Measuring Ethiopian farmers' vulnerability to climate change across regional states. IFPRI Discussion Paper 00806.
Ethiopia: International Food Policy Research Institute.
Folke, C., 2006. Resilience: the emergence of a perspective for social–ecological systems analyzes. Glob. Environ. Change 16 (3), 253–267.
Funk, C., Dettinger, M.D., Michaelsen, J.C., Verdin, J.P., Brown, M.E., Barlow, M., Hoell, A., 2008. Warming of the Indian Ocean threatens eastern and southern
African food security but could be mitigated by agricultural development. Proc. Natl. Acad. Sci. 105 (32), 11081–11086.
Gunderson, L.H., Holling, C.S., Light, S.S. (Eds.), 1995. Barriers and Bridges to the Renewal of Ecosystems and Institutions. Columbia University Press, New York.
Haque, M.A., Yamamoto, S.S., Malik, A.A., Sauerborn, R., 2012. Households' perception of climate change and human health risks: a community perspective. Environ.
Health 11 (1), 1.
Hassan, R., Nhemachena, C., 2008. Determinants of African farmers’ strategies for adapting to climate change: multinomial choice analysis. Afr. J. Agricult. Resour.
Econ. 2 (1), 83–104.
Holling, C.S., Meffe, G.K., 1996. Command and control and the pathology of natural resource management. Conserv. Biol. 10 (2), 328–337.
Holling, C.S., 2001. Understanding the complexity of economic, ecological, and social systems. Ecosystems 4 (5), 390–405.
Kartiki, K., 2011. Climate change and migration: A case study from rural Bangladesh. Gender Develop. 19 (1), 23–38.
Levin, I.P., Schneider, S.L., Gaeth, G.J., 1998. All frames are not created equal: a typology and critical analysis of framing effects. Organ. Behav. Human Decision
Process. 76 (2), 149–188.
Muluneh, W., Demeke, H., 2011. Climate change and variability, its impact on rural livelihoods, local coping and adaptation strategies in Woreilu Woreda, North
Eastern Ethiopia. Ethiopian J. Develop. Res. 32 (2).
Norris, F.H., Stevens, S.P., Pfefferbaum, B., Wyche, K.F., Pfefferbaum, R.L., 2008. Community resilience as a metaphor, theory, set of capacities, and strategy for
disaster readiness. Am. J. Commun. Psychol. 41 (1–2), 127–150.
Parkins, J.R., MacKendrick, N.A., 2007. Assessing community vulnerability: a study of the mountain pine beetle outbreak in British Columbia, Canada. Global Environ.
Change 17 (3), 460–471.
Rose, A. 2009. Economic resilience to disasters. CARRI Research Report 8. Community & Regional Resilience Institute. Available from http://www.resilientus.org/wp-
content/uploads/2013/03/Research_Report_8_Rose_1258138606.pdf.
Seleshi, Y., Zanke, U., 2004. Recent changes in rainfall and rainy days in Ethiopia. Int. J. Climatol. 24 (8), 973–983.
Tadege, A. (Ed.) 2007. Climate change national adaptation program of action (NAPA) of Ethiopia. Federal Democratic Republic of Ethiopia, Ministry of Water
Resources, National Meteorological Agency. Addis Ababa, Ethiopia.
Tadege, A., 2008. Climate change and development adaptation measures. Economic Focus 11 (1), 15–24 Addis Ababa, Ethiopia: Ethiopian Economics Association.
Walker, B., Holling, C.S., Carpenter, S.R., Kinzig, A., 2004. Resilience, adaptability and ztransformability in social–ecological systems. Ecol. Soc. 9 (2), 5.
Williams, A.P., Funk, C., 2011. A westward extension of the warm pool leads to a westward extension of the Walker circulation, drying eastern Africa. Clim. Dyn. 37
(11–12), 2417–2435.

77