Analysis of smallholder farmers’ vulnerability to climate change and variability in south Wollo, north east highlands of Ethiopia: An agro-ecological system-based approach

Background: Climate change and variability has been signicantly affecting the Ethiopian agricultural production and thereby smallholder farmers livelihoods. The level of vulnerability varied across agroecological zones (AEZs). Identication of difference in the level of vulnerability of a system is important in selecting appropriate and effective adaption options to climate change. Therefore, the aim of this study was to analyze the level of vulnerability of agricultural communities to climate change and variability at micro-scale level in ve agro-ecological zones (AEZs) of south Wollo, north east highlands of Ethiopia. Data was collected from a representative of 502 sample households from ve AEZs through multi stage random sampling methods. Focus group discussion and key informant interviews were also carried out to supplement and substantiate the quantitative data. The indicator based approach was used to empirically calculate vulnerability. Principal Component Analysis (PCA) was applied to give weight for indicators and generate index of vulnerability contributing factors. Results: The results showed that each of the vulnerability contributing factors (exposure, sensitivity and adaptive capacity) varied across the AEZs. M3, SM2 and SM3 are the most exposed AEZs but having a relatively better adaptive capacity whereas M1, M2 and M3 are the most sensitive AEZs with relatively low adaptive capacity to climate variability and change. Overall, SM2 is the most vulnerable AEZ which exhibited high sensitivity and low adaptive capacity followed by M1. Conclusions: The study explored sources and levels of vulnerability to each agro-ecology. Since the study is conducted at micro-scale level, it helps decision makers and development partners to have context-specic understanding of the impact of climate change and variability and design appropriate adaptation measures to address the specic situations.

Ethiopia is heavily dependent on rain-fed agriculture which supports more than 85% of the population, and contributes about 50% of the country's gross domestic product (GDP) (GFDRR, 2011;Gizachew and Shimelis, 2014). Ethiopia is a country that is highly vulnerable to frequent climate variability and change (World Bank 2010;Conway and Schipper 2011) and it will be negatively affected by the changing climate patterns in the coming decades (McSweeney et al., 2010a).
Regional projections of climate models do not only predict a substantial rise in average temperatures over the twenty-rst century but also an increase in rainfall variability with a rising frequency of both extreme ooding and droughts which signi cantly affect the Ethiopian agricultural production .
As a result, Ethiopia's agricultural productivity remains lower in the coming decades (Belay et al., 2014). A 10% decrease in seasonal rainfall from the long-term average brings a 4.4% reduction in the Ethiopian food production (Von Braun 2007). Unless steps to build resilience are effective, climate change will reduce Ethiopia's GDP growth up to 10% by 2045 between 0.5-2.35 per year (World Bank 2010, EPA 2011. The Climate Resilient Green Economy (CRGE, 2011) also indicated that climate change has the potential to hold back economic progress, or reverse the gains made in Ethiopia's development and could exacerbate social and economic problems.
Erratic and insu cient rainfall particularly during the short rainy season, lack of improved crop varieties, poor farming practices, land degradation and crop diseases were major challenges of crop production. Poor performance of local breeds, lack of animal fodder and prevalence of animal diseases also constrained livestock productivity. As a result of challenges related to biophysical and socio-economic issues, smallholder farmers are not able to produce enough to satisfy their annual food requirement. The national Productive Safety Net Program supported the majority of the community to complement their livelihoods (Mekonnen et al., 2013).
One of the key messages emerging out of the recent 5 th IPCC reports is that the climate change is real, happening and will continue to happen for the foreseeable future. The reports also estimate with high con dence that the negative impacts on agriculture outweigh the positives which makes adaptation an urgent and pressing challenge. However, adaptation planning requires accurate information about where, when and how the impacts are going to be felt and who will be more vulnerable.
Vulnerability assessments can play a vital role in designing appropriate adaptation strategies targeted towards climate change and its impacts on ecosystems, and those who depend upon these resources for their livelihoods. Assessing which particular individual, group, community, region or nation is vulnerable, and in what ways, enables to formulate clear and effective policy and strategy that promotes adaptation. More speci cally, vulnerability assessment helps set three policy measures. First, it is used to specify long-term targets for mitigation of climate change; second, to identify vulnerable places and people and to prioritize resource allocation for adaptation; and, nally, to put forward speci c adaptation recommendations for speci c places and groups (Füssel and Klein 2006). Although Ethiopia is highly vulnerable to the impacts of climate variability and change, there are no detailed studies conducted at local level, except the works of Simane et al., (2014), Dechasa et al., (2016); G/Egziabher et al., (2016); Sisay (2016) and Tesso et al., (2012). These reports showed the complex distribution of vulnerabilities across agro-ecologies since the Ethiopian highlands are highly dissected and fragmented. Therefore, prevailing climate conditions and the sensitivity of agricultural systems to climate variability can change over a distance of even a few kilometers (Simane 2014). As a result, an estimate of vulnerability at the national or regional scale is insu cient to have an overall understanding of climate vulnerabilities and adaptation interventions within the Ethiopian highlands.
This study was, therefore, aimed to assess the level of smallholder farmers' vulnerability to climate change and variability at micro-scale level in ve AEZs of south Wollo. The study area is selected for this research because no similar study was conducted in the area before. Thus, it is important to understand the level of vulnerability of smallholder farmers to climate change and variability at the local level. Since the study is conducted at micro-scale level, it helps decision makers and development partners to have context-speci c understanding of the impact of climate change and variability and take appropriate adaptation measures accordingly.
2 Methodology 2.1. Description of the study area Biophysical setting South Wollo zone is located in the north eastern part of Ethiopia with in 10°12′N, and 11°40′N and 38°30′E and 40°05′E (Fig. 1). Its zonal capital, Dessie, is found 400 kms North of Addis Ababa. It is one of the drought-prone and aid-dependent areas in Amhara Regional State (Bewket and Conway, 2007). The study area covers a total area of 17,053.45 km 2 with eighteen rural and two urban woredas with a projected total population of 2,980,912 (Male 1,475,525 & Female 1,505,387) for 2014 (CSA, 2015). The digital elevation model (DEM) reveals that larger parts of south Wollo are characterized by rugged topography that consist of very high mountains, deeply incised canyons and gorges, valleys and plateaus ( Fig.2). The elevation ranged from the dry plains at slightly less than 1000 m a.s.l. in the east to the high peaks above 4000 m a.s.l in the west. However high-land areas ranging between 1500 m and 3500 m a.s.l are the dominating feature of south Wollo (Rosell and Holmer, 2007;Rosell and Holmer, 2015). The areas covered by slopes < 2%, 2 to 15%, 15 to 30% and >30% cover 1.7%, 30.2%, 31.3% and 36.8 % of the total area, respectively (Fig. 2). Land with slopes less than 15% is most suitable for agriculture (Hurni 1988). However, this area accounts for only 32% of the total area.
Because of the diverse topography, the study area experiences different climatic zones that ranged from hot and arid lowlands in the eastern part to cold and humid highlands in the central part. It is broadly classi ed into 5 agro-ecologies (M1,M2,M3,SM2 and SM3) (MoARD 2000).  (2000) The average annual temperature varies from less than 10 o C in the western highlands to 25°C in the eastern lowlands. The annual rainfall varies between less than 1000 mm in the eastern part to more than 1200 mm in the western part (Rosell and Holmer 2015). Eastern part has a relatively dry climate and western parts have moderate to large amount of water surplus in the main rainy season, which results in high run-off, soil erosion and water-logging depending on the soil type and topography. Bimodal rainfall pattern characterizes the study area which leads to two growing periods (with single maxima during summer). The main rainy season summer (locally called kiremt) is from June-August and the small rain season spring (locally called belg) is from March to May. The spring season is noticeable for mid to high latitude areas whereas the mid lands to lowland areas mainly depend on the summer rain. Crop production is rain-fed. The small rainy season is erratic and highly variable and experienced frequent failure which hampers spring harvesting (Rosell, 2011;Dereje et al., 2012). Delayed onset and early cessation as well as poor spring performance make the region food insecure and highly reliant on food aid (Lakew et al., 2000;Bewket, 2007;Amare et al., 2011;Daniel, 2011;Kahsay, 2013). The study area has various soil types, the major ones being Leptosols, Cambisols, Vertisols, Andosols and Luvisols (FAO, 1984).

Socio-economic setting
Although agriculture is the major economic sector, it is still underdeveloped. The agriculture is largely based on cereal-livestock farming systems (ANRS-BFED, 2008) where they have complimentary and competitive effects to each other (Haileslassie et al., 2005). Farmlands used to supply crop residue as animal feed without putting manure back to farmlands while fallowing practices have been greatly reduced. These increased soil nutrient depletion, worsens the degradation processes and gradually contributed to crop production failure (Haileslassie et al., 2005). Due to the increasing demand for agricultural land, marginal lands have been converted to cultivation and grazing (Tekle 1999). As a consequence, land degradation has alarmingly increased (Tekle 1999;Descheemaeker et al., 2010). Land holdings are also reduced from time to time. As a result, agriculture in the zone has been unable to provide su cient food for its population.
The study area is highly degraded and deforested in terms of indigenous trees but does have substantial eucalyptus plantations. Alpine species unique to extreme highland areas are found in central parts of the study area. Local people in the area are engaged in subsistence agriculture for their livelihood. Unlike other areas of Ethiopia where highlands generally are the most food secure, the opposite is often the case in the study area. In fact, the crowded, steep-sloped highlands above 2000 meters a.s.l, including large parts of south Wollo, are among the Ethiopia's most famine-prone areas (Little et al., 2006).

Method
Sampling techniques and procedures The primary data sources for this study were smallholder farmers and multistage sampling techniques were used to select sample respondents. The rst stage involves selection of south Wollo purposively as it was vulnerable to climate variability and change. The area is highly prone to frequent drought (Bewket and Conway 2007;Spielman et al., 2011;Yimer et al., 2017). Total sample size was determined using the following formula.
Where, n = Total sample size, N= represent the total rural household heads, Z= Standard variation at a given con dence level (95%=1.96) P= Estimated characteristics or proportion of the target population that is 0.5 q = 1-p which is equal to 0.5 e = Margin of error (5%) Secondly, representative woredas from each agro-ecology zone were selected using the probability proportional to size (each agro-ecology) method. Third, the kebeles (the smallest administrative units) in each woreda were selected using simple random sampling procedure. Fourthly, the number of sample respondents in each kebele was determined using the probability proportional to size method.
where n= total number of samples to be taken from each kebele, Ni= the total number of HH in each kebele, N= total number of HHs in all selected kebeles ni= sample size determined by the researcher and; lastly household heads in each kebele were selected on simple random sampling basis. Household heads with more than 45 years of age were subjects of this study. This helps see any visible change in microclimate of a particular place with in almost 30 years of time span. Therefore they are assumed to provide more accurate data in local climate study. The scale of analysis was agro-ecological zone.

Data sources and collection methods
The study used a mixture of data collection methods (questionnaires, key informant interviews and focus group discussions) A cross-sectional survey was conducted from October to December 2015 involving 502 sample households in 13 kebeles spread across all 5 agro-ecological zones (Fig 1, Table 2). The survey questions were designed to address each indicator of exposure, sensitivity and adaptive capacity.
To substantiate the quantitative information gathered through household survey, key informant interviews with elderly who have in-depth knowledge about their environment and focus group discussions with the household heads were held. One focus group discussion and two key informant interviews were held in each sample kebele.

Constructing vulnerability index
There are two broad approaches to empirically calculating vulnerability: i) econometric and ii) indicator methods. The former expresses vulnerability as expected poverty, low expected utility and uninsured exposure to risk (Hoddinott and Quisumbing 2003). The latter tries to assess vulnerability by integrating indicators to form a composite index, which can be at a local level (Adger 1999;Hahn et al. 2009). This study adopted the indicator approach in measuring the vulnerability of smallholder farmers to climate change and variability. The basic challenge in constructing indices is the lack of standard ways of assigning weight to each indicator. The two most common weighting methods used to combine indicators are equal and unequal weighting schemes. The former method assigns equal weight to each indicator. The latter method, we adopted, assigns different weights to various indicators using (i) expert judgment (Brooks et al. 2005;Moss et al., 2001) or statistical methods such as principal component analysis (PCA) (Cutter et al. 2003;Thornton et al. 2006).
Each indicator of exposure, sensitivity, adaptive capacity doesn't have equal impact on computation of vulnerability indices of different agro-ecological zone. As a result, different weights were assigned for these indicators. Therefore, Principal component Analysis (PCA) was selected to generate the weights for each indicator component in this study (G/Egziabher et al., 2016;Thornton et al., 2006;Cutter et al., 2003;Filmer and Pritchett 2001). PCA was run for the selected indicators of exposure, sensitivity, and adaptive capacity separately using STATA Software.
Once one or more indicators are selected, the selected indicators were integrated into sub-indices and composite index was produced. The data sets measured using different scales or measurement units were made comparable by transforming them into a common scale by normalization (Nelson, et al., 2010b;Gbetibouo & Ringler, 2009;Hahn 2009;Vincent, 2004).
The loadings from the rst component of PCA are used as the weights for the indicators. The magnitude of the weights describes the contribution of each indicator to the value of the index. A stepwise (nested) PCA was run separately for the indicators of exposure, sensitivity and adaptive capacity to get the index of these three vulnerability determining factors.
For example, the adaptive capacity index has been generated by using four nested PCAs. The inner most PCA was performed to produce an index for fertilizer consumption, improved seed, soil and water conservation and access to weather forecast. These four categories have been used in the second PCA to generate index for technology. At the third PCA, we have generated index for nancial capital, physical capital, human capital and social capital. The results of the third PCA were used to generate the fourth index for adaptive capacity (PCA 4). The index of the adaptive capacity is computed by using loadings of nancial capital, physical capital, human capital and social capital. The indices of sensitivity and exposure are also calculated in similar way. Finally the outer most PCA was performed to generate index for overall vulnerability.
Vulnerability index was calculated as: where, LVI-IPCC is the vulnerability index of the region, E r is the calculated exposure score for AEZ r AC r is the calculated adaptive capacity score for AEZ r S r is the calculated sensitivity score for AEZ r.
Determining the scale of analysis is very important in measuring vulnerability. Analysis of vulnerability to climate change ranges from household to global level (Adger 1999;Brooks et al., 2005). The decision of scale selection is determined by many factors such as the objective of the study, the methodologies set, and the available data for the analysis. Although the most appropriate scale of analysis for vulnerability study is at the lowest administrative units, the scale of analysis, for this study was AEZs.
This index does not give the absolute measurement of vulnerability; rather it gives a comparative ranking of study agro-ecologies. Test of a one way analysis of variance (ANOVA or F-test) was conducted to compare the averages of major indicators of vulnerability among the ve AEZs. ANOVA tells whether there was an overall difference between the groups, but it does not tell which speci c group was differed. In order to con rm where the differences occurred between groups, post hoc tests were used. It should only be run when an overall statistically signi cant difference in group averages occurred (i.e., a statistically signi cant one-way ANOVA result).

Results And Discussion
In this chapter, the contribution of each major component to vulnerability was analyzed. In this study, eight major components of vulnerability were used to generate the vulnerability index.

Adaptive Capacity Pro les
Adaptive capacity pro le is composed of many socio-economic pro les which include human capital, wealth, technology, infrastructure, institutions/social capital. The detail explanation of each pro le together with their indicators was presented in Table 5. The relationship between most of the pro les and vulnerability is believed to have negative relationship.
Human/demographic characteristics: The demographic and socioeconomic characteristics of the study population constituted the human capital which was indexed under adaptive capacity. From a total of 502 sample smallholder farmers, 77.9% were male and 22.1% were female; 76.7% were married; 10.6%divorced and 11% widowed. In terms of age category, majority of respondents (83.5%) were within the active working age group (45-64), while 16.5% respondents were above 64 years of age. The education level of the respondents was low where 30.9% of them were illiterate and the remaining 70.1% were literate. Household heads with higher levels of education believed to have a better level of planning, access and effectiveness of early warning information from different sources, better reactions and rehabilitation skills during and after natural shocks, alter agricultural operation and adopt extension strategies (Legese et al., 2016;Muttarak and Lutz 2014). Education tends to improve the ability of smallholder farmers to better comprehend issues affecting them and therefore look for possible solutions (Etwire et al., 2013). The One-way ANOVA analysis reveals that there was a signi cant difference of educational status among the sample households in the study AEZs (Table 4). The level of education was the least in SM3 (Table 5).
The survey also indicated that household size ranged from two to fourteen household members, with an average of 5.8 with slight difference across agro-ecologies. From the total 502 interviewed farmers, 46 had small family size (1-3), 300 had medium family size (4-6) and 156 had large family size (>6). The average family size of the study area is higher than the regional average (4.6) (CSA 2014). Household size is related with poverty and consequently, with climate change vulnerability. A study conducted by Cutter et al., (2009)  Wealth: Number of livestock, food availability and dependency ratio were considered as major indicators of wealth in the study area. Livestock production was one of the most important livelihood systems in the study area. However, climate change brought a negative impact on livestock production. More than 70% of the farmers in the study area reported that the number of their livestock was decreasing through time.
Shortage of animal feed attributed to the reduction of their number. They also added that availability of water for livestock is diminished in recent times. Respondents (88%) in M1 stated that their livestock number is signi cantly decreased (Table 5). The analysis indicates that M1 and SM3 had signi cant difference in the number of livestock than others AEZs. SM3 is the most vulnerable in terms of livestock number with average value of 0.000 (Table 5). Different reasons were provided for this. First, it was due to the extremely cold nature of the area and unavailability of pasture and water; second, since the area is accessible to all weather roads, farmers could sell their products to towns and gave attention to quality than quantity. The other reason was that their nearby grazing lands were delineated for area closure under watershed management program. However, the livestock market was showing a remarkable improvement in its price in recent times compared to the past years. This helps them minimize their vulnerability to climate change.
It is worth to cite the statement of the following key informants' response from the M3 agro-ecological zone (Keregimba) as follows: "We used to practice livestock rearing and crop production equally. Before two or three decades, most of us had large number of livestock because of unlimited access to abundant grazing lands and excess crop residues available on our farms. But, today, neither the grazing lands nor crop residues are available. Grazing lands were either changed into area closure or converted to crop lands and we do not leave plant residues during harvesting. This has created an increased shortage of feed for livestock and this became a challenge for us and we supplement the gap by purchasing livestock feed from other remote areas. Hence, the number and variety of livestock population gradually decline as the result of lack of grazing lands and water resources".
Although varied in its magnitude, food insecurity was common in the study zone. 62% (313) of the respondents con rmed that their household did not have adequate food in some months of the year. They were assisted by food aid. The nding showed that households remained with food shortage in a year for long periods ranging from 1-6 months. Households signi cantly differ (P<0.001) in their food security status across AEZs (Table 4). M3 and SM3 were vulnerable in terms of food availability ( Table   5).
The average dependency ratio in the study area was 0.33 (33%). However, it was very high in M3 and M1 compared to the remaining regions. It has a strong implication in vulnerability reduction efforts. A higher dependency ratio could lead to labor shortages. A higher dependency ratio is likely to reduce productivity.
As the number of dependent family members (children and elders) becomes large, the active population would not support the family instead the degree of vulnerability worsen. They spent their time to cover livelihood costs of their unproductive family member. Siagian et al., (2014), reported that high number of dependency ratio increases sensitivity and may limit their ability to bounce back or recover from the impacts of natural hazards, like climate extremes.
Technology: The households widely used fertilizers to improve the productivity of their small plots of land. Considerable number 82% of smallholder farmers in the study area indicated that their fertilizer consumption is being increased for the last 20 and 30 years (Table 5). However, 82% smallholder farmers in SM3 reported a signi cant reduction in using fertilizers compared to the others. They did not afford the price of fertilizers. SM3 is found in high altitude and it is susceptible for degradation due to water erosion.
Smallholder farmers were aware of new technologies through agricultural extension workers. As food insecurity prevails in the study area, smallholder farmers were showing strong willingness to accept what extension workers told to them. They tried to increase productivity by using improved seeds. About 90% of interviewed farmers revealed that the use of improved seeds was increasing in recent years when compared to the past 20 and 30 years.
Soil and Water Conservation: In order to reduce soil erosion, a number of soil and water conservation structures have been constructed in the study area including physical measures (terracing, check dams, eye brow basin, and area closure). Farmers' land use/ownership right helps farmers invest more on their land. The study shows that 74.7% and 16.3% of the respondents possessed land as owner and both as owner and share cropping respectively. Large number of the respondents (92%) exercised the physical soil and water conservation (SWC) measures. However, the practice is signi cantly noticeable in relatively lower altitude AEZs (M1, M2 and SM2) than high altitude AEZs (M3, SM3) due to the fact that the problem is serious in lower altitude areas. There exists a signi cant difference in the use of fertilizer, improved seed, soil and water conservation across AEZs (Table 4).
Infrastructure: Infrastructure is very essential for providing any social services to the rural community. When roads are easily accessible, the provision of any services will be facilitated. There were considerable efforts in different AEZs to make various infrastructures accessible. Almost 78% of interviewed farmers explained that they had road access. Smallholder farmers have signi cant difference (P<0.001) in accessing all weather roads. Relatively, SM3 had better road access than others and SM2 had less road access than others (Table 4) Access to credit was showing progressive trend in recent years. In terms of access to credit there was a signi cant difference (P<0.001) between agro ecological zones. However farmers in M1, M2, and SM2 had a better access to credit than SM3 and M3. It has strong implication to household vulnerability.
Farmers who received credit will have high probability to reduce vulnerability as compared to those who do not have access within that category.
Accessibility of primary schools was more or less similar in all AEZs (Table 5). Majority (75%) of the respondents indicated that children have to travel on average less than 30 minutes to reach nearby primary school. However, the post hoc test analysis con rmed that children in SM3 were signi cantly taking more time than others. More than 90% of farmers in SM3 reported that their children took long time to reach primary schools ranging from half an hour to one hour to reach nearby primary school. to reach nearby health posts between AEZs (Table 4) where farmers in M3 travel the longest distance (Table 5).

Sensitivity pro les
Landholding size: Almost 34.46% and 48.2% of the respondents own less than 0.5 hectare and 0.5-1 hectare respectively. It is in line with the ndings of CSA (2013) and Little D. et al., (2006) which reported landholding size of south Wollo to be 0.51-0.75 ha and 1.72 ha respectively. This shows that the land holding size of the study region is less than the national average (1.2 ha). There was no signi cant change in landholding size across all AEZs. Households who owned a better size of land were less vulnerable/ less sensitive and thereby increase their adaptive capacity. The larger the size of land, the higher the diversity of crops that are grown, the better their access to improved technology, and the better their access to nancial services and extension services (Dechasa et al., 2016;Gutu et al., 2012).
Irrigable land: Access to irrigation is negligible in the study area. Only 12.5% of the total respondents had an irrigable land. There was a signi cant difference across AEZs in having irrigable land (Table 4) where SM3 had the least access to irrigable land (Table 5). It is believed that the availability of rivers for irrigation was not common in areas where located in very high altitude. SM3 is found on the highest altitude of the zone in which Amba Ferit (4246 m.a.s.l) is found. Households having more irrigable land will be less sensitive to climate change impacts by maximizing the cropping intensity. Farmers reported that they harvested twice in a year from their irrigable land and be less vulnerable.
Land productivity: There was low agricultural productivity in the study area. More than half (55 %) of the respondents explained the reduction of their land productivity. The crop yield was dwindling from time to time. They tried to avoid the decline in crop yield by applying fertilizers and improved seeds. However, there was a signi cant (P<0.001) variation in land productivity between AEZs (Table 4). Poverty and fastgrowing population, coupled with land degradation poses a serious threat to household food security (Simane et al., 2013). In the study area, 63.7% of the respondents also perceived progressive land degradation. They claimed that loss of soil fertility, soil erosion, and steepness of the slope attributes to less productivity. More than 75% of the respondents perceived that the status of soil fertility was deteriorating from time to time. This problem was worse in central and western highland woredas of south Wollo where it is dominated by worst terrain with very stony, steep and bare lands (Little et al., 2006;Castro et al., 1999). This fully characterizes M3 and SM2 agro-ecological zones.
Variety of crops: Smallholders were diversifying crop varieties they grow to avoid the effect of climate change and variability. There was signi cant difference (P<0.001) in diversifying crop varieties between AEZs where M3 is the lowest in crop diversity. This is due to the fact that this AEZ experiences very cold climate which is conducive to grow only barley. The focus group discussion held with farmers living in SM2 agro-ecology zone (Haroge kebele) revealed that they were able to diversify the type of crops grown. In the abnormal rainfall seasons, all of the crops may not be affected equally. Some may be e cient in absorbing the shock and the others not. This increases the adaptive capacity of farmers and lessens their vulnerability to climate change impacts. Crop diversity has an implication to climate change adaptation. Different crops have different capacity to respond to the changing climate. They need water for their growth in a varying degree. A few may grow with less water and other need excess water for their optimum growth. Some may provide yield under harsh environment than others.
Pests and diseases: These problems became common threats to all AEZs because of the changing climate. The ndings from the focus group discussions held with farmers living in M1, M2 and SM2 agro ecological zones revealed that leguminous crops like Beans (Vicia faba) and Peas (Pisum sativum), which were widely grown in the study area are no longer cultivated today. The discussants told that they have noticed this challenge in recent years and related it with climate change. These leguminous crop starts to be damaged by pests and diseases at its owering stage. As a result, the production of such crop has been worsened from time to time to the extent of disappearing and their market value is growing alarmingly when compared to other crops. Most farmers mentioned that the loss of such crops have brought a negative effect on their livelihoods.
The increment of CO 2 and temperature concentrations associated with climate change has a substantial impact on plant-disease interaction (Yanez- Lopez et al., 2012;Rosenzweig and Tubiello, 2007). The rising levels of CO 2 and temperatures are having direct effect on pests and diseases in crops. Exposure to high CO 2 generally increases the carbon-to-nitrogen ratio (C: N) of plant tissues (Rogers et al., 2004;Hamilton et al., 2005), reducing the nutritional quality for protein-limited insects. Some insects and pests may increase consumption rates to compensate reduced leaf nitrogen content. Concentration of CO 2 can increase levels of simple sugars in leaves and lower their nitrogen content (Dermody et al., 2008). These can increase the damage caused by many insects, who will consume more leaves to meet their metabolic requirements of nitrogen. Thus, any attack will be more severe. Higher temperatures from global warming, mainly due to elevated CO 2 , will mean that more numbers of pests will survive. Besides, warmer temperature lowers the effectiveness of some pesticides.

Exposure Pro le
This was mainly composed of rainfall and temperature variability and change. This includes six indicators (Table 5). All indicators were signi cantly (P<0.001) showing variation between AEZs. Farmers living in different AEZs had different biophysical and socio-economic characteristics. It is believed that farmers perceive climate change differently due to disparities in local climate impacts, as well as difference in socio-economic perspectives on these impacts (Simane et al., 2013). Sampled households were asked to point out the manifestations of temperature and rainfall changes for the last 30 years. 73 % of the respondents pointed out that temperature was showing increasing trend for the last 30 years. This perception is consistent with scienti c temperature analysis done for the study area, which showed a warming trend of average monthly maximum temperature between 0.07-0.49 o C/decade (Yimer et al., 2018). With regard to rainfall change for the last 30 years, 78.39 % of respondents perceived declining of annual rainfall, however, differed signi cantly across AEZs. This was consistent with the analysis of meteorological data which showed a declining tendency of spring rainfall, ranging from 8.55mm/decade to 50.52 mm/decade in all study stations of south Wollo (Yimer et al., 2018). In addition to the reduction of its amount, the late onset and early cessation of rainfall affected their livelihood.
The analysis of the pro les revealed that M3 and SM3 AEZs were signi cantly vulnerable in terms of reduction of rainfall amount, increment of temperature, late onset of rainy season and highest frequency of drought than other AEZs. These AEZs were also more vulnerable than others in the frequency of drought. These two AEZs are typically very cold highlands which were known for using spring rain (become erratic and variable in recent years) for their main growing season.
A 48 years old man residing in SM3 stated the following about the changing climate.
"When we were teenagers, the highland area that you see in front of us (by showing his hand to the mountain peak) was covered by snow during winter. But it was decreasing from time to time and currently we do not observe any snow cover in the area. The rain used to come regularly and consistently, and we were able to plant and harvest on time. But currently, it deviates from the normal. It is unpredictable, kiremit rain starts late (mid of July) and stops early around buhe (Christian festival celebrated on August 19). We were also highly dependent on spring rainfall for our agricultural practices. Availability of water and pasture becomes scarce after the long dry season (which extends from December-February). It is the spring rainfall (March-May) which saves our livestock. Besides, we plant crops like barley (Hordeum vulgare) which need longer gestation period for harvesting. But today, we lost all these chances due to failure or/and inadequate spring rain.

LVI -IPCC contributing factors
The overall vulnerability index based on the function of exposure, adaptive capacity and sensitivity was presented in Table 6   Overall, SM2 (mid highland) and M1 (lowland) AEZs are more vulnerable to climate change and variability than other AEZs in the study area ( gure 5). As Table 6 indicates SM2 was the most vulnerable AEZ due to its greater exposure to climate change and variability as well as least adaptive capacity whereas M1 is more vulnerable due to its high exposure and sensitivity scores. SM3 and M3 were less vulnerable AEZs due to a relatively better adaptive capacity and lower sensitivity scores than others.

Conclusions And Policy Implications
This study adopted the indicator approach in measuring the vulnerability. It employed the livelihood Vulnerability Index (LVI) for measuring vulnerability of farmers in different AEZs to climate variability and change. It provides a detailed description of factors driving households to vulnerability. Vulnerability was calculated as the net effect of sensitivity and exposure on adaptive capacity. The analysis provides useful insights and captures speci c variations across agro-ecologies. The results indicated that the lowland and mid highland AEZs were found to be most vulnerable to climate change due to high level of exposure and sensitivity and relatively low level of adaptive capacity compared to highland areas. Although highland AEZs have less vulnerability index due to their better adaptive capacity, their exposure to climate variability and change was very high. The frequency of climatic variability/change has been increasing both in terms of frequency of extreme climate events and gradual changes, and consequently worsening the vulnerability of agriculture to these events. Smallholder farmers perceived declining trends of rainfall (especially belg rain) and increasing trends of temperature in the study area over the last 30 years. These perceptions agreed with the analysis made using the metrological data in the previous chapters. The results indicated that level of exposure to the long term changes and the degree of adaptive capacity was the most important components to determine the overall vulnerability of the AEZS. However, biophysical elements in the ecosystem have a multiplier effect to the deviation of the two.
The nding of the study will have important policy relevance that could enable vulnerable smallholder farmers in lowland and mid highland AEZs to better adapt to the impacts of climate change and variability. Improving the adaptive capacity of farmers has implications on improving the sensitivity of the community. For example, improving the irrigation facilities (infrastructures) decrease the sensitivity of crops to droughts which in turn increases crop diversity and farmers' resilience to climate change vulnerability. Improving credit access would also strengthen the capacity of farmers to be engaged in alternative economic activities. Creating more access to technologies (fertilizers, improved seed and soil fertility management) with affordable price would lead to improve adaptive capacity and lessen sensitivity of biophysical factors to climate change. Promoting insurance schemes as a strategy would also help farmers avoid risks during drought or ood incidences. Replication of this study in the same location over time might provide information about how the exposure, adaptive capacity, and sensitivity of the study AEZs change as adaptation practices are initiated and promoted.

Abbreviations
AEZs-Agro-Ecological Zones ANRS-Amhara National regional State Availability of data and materials: Datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.
Competing interests: The authors declare that they have no competing interests Funding: The data collection of the study was supported by Hawassa University.
Authors' contributions: YM was responsible for all activities of the research process such as the design, data compilation and entry, data analysis, and interpretation of results as well as writing up of the manuscript. The other three authors were involved in improving the quality of manuscript by providing constructive guidance, critical comments and suggestions on the study design, data analysis and interpretation. All authors read and approved the nal manuscript.