4.1. Livelihood Vulnerability Index
Communities in all three watersheds were vulnerable to the adverse effects of climate change variability because of high exposure, high sensitivity, and low adaptive capacity. In relative terms, Aba Gerima (midland) watershed was found to be the most vulnerable and Guder (highland) the least. The total score for Aba Gerima was 0.37, with 0.15, 0.14, and 0.08 contributed by adaptive capacity, sensitivity, and exposure, respectively. The corresponding scores for Dibatie were 0.35, 0.13, 0.13, and 0.09 and for Guder, they were 0.34, 0.14, 0.13, and 0.08. Interestingly, despite the very different contexts on each of the three communities, they showed relatively similar scores in the dimensions of the IPCC-LVI. Conversely, in their national scale study Ferede, Ayenew, Hanjra, and Hanjra [
43] noted that the highland part of Ethiopia is more vulnerable compared to other agroecological zones. A possible reason for the two differences could be emanated from the scale of unit of analysis, whereby we used a watershed scale which helped us to gain the necessary details about livelihood vulnerability. Moreover, we also showed that adaptive capacity had relatively higher contribution for the overall livelihood vulnerability compared to its counterparts.
Aba Gerima was found to be the most vulnerable area, mainly because of its limited adaptive capacity and higher sensitivity. The sensitivity may be a result of the area’s severe soil erosion status and the fact that households are not sufficiently participating in sustainable land management activities. In the same study area, References [
46,
59] found that the community’s lower economic adaptive capacity affected its adoption of natural resource management practices. Ecological indicators such as soil depletion have also been shown to contribute to the vulnerability of village communities in Tanzania and South Africa [
60]. In contrast, Guder was relatively less exposed to climate-related shocks. We found few drought episodes and an increasing rainfall trend in this area, so it is possible that communities in this watershed were less exposed to drought episodes and did not experience water shortages. An agroecological-based climate change vulnerability study in the Ethiopian highlands also found that highland agroecosystems are relatively less vulnerable to climate change shocks [
31].
4.2. Exposure
Although the study area has had increasing rainfall, the trend was not significant at any of the sites. The watersheds studied have a unimodal rainfall pattern, and the rainy months occur mostly during the summer season (usually from mid-June to mid-September). Temperature did show a significant increasing trend at the study sites. Berihun, Tsunekawa, Haregeweyn, Meshesha, Adgo, Tsubo, Masunaga, Fenta, Sultan, and Yibeltal [
47] also found that station data showed a significant upward trend in temperature but not rainfall in north-western Ethiopia. Moreover, Fenta et al. [
61] also did not observe a monotonic trend in rainfall time series data in Ethiopia, and Teshome [
62] found an increasing temperature trend in the Dembia District in the Upper Blue Nile basin. Furthermore, our findings are in line with other similar studies, which revealed observed trend changes in the Ethiopian highlands and elsewhere, with a mean annual temperature increase range between 0.028 °C and 1.65 °C from 1955 to 2016 [
63,
64]. In contrast, Samy et al. [
65] reported a significant decreasing trend in rainfall in the southwestern part of the Upper Blue Nile basin. SPI results indicated that all the three watersheds were affected by drought episodes in 1984/1985, 1987/1988, 1992/1993, 2000/2001, 2010/2011, and 2015/2016, which is in line with other research results of drought experiences in Ethiopia [
9,
66]. Similarly, in their work in the northern part of Ethiopia, Kasie et al. [
67] indicated that household livelihood systems were very much connected to the increment in their income, and this has been hurdled by the recurrent drought episodes (e.g., the 2015 El Niño).
Communities in Guder were found to be less exposed to climate-related shocks in five indicators (mean annual rainfall, mean annual maximum temperature, mean annual minimum temperature, number of shocks, and access to warning information). This could be partly attributed to the fact that the area has received more rainfall than the others (
Figure 3). Moreover, despite experiencing drought episodes in 1984/1985 and 1992/1993, the watershed had a positive SPI and an increasing rainfall tendency in the study period. In Ethiopia, rainfall is crucial for predominantly rain-fed agriculture; hence, having more years of normal rainfall and more rainy days means better production and productivity for small-scale farmers [
38]. In turn, having better agricultural productivity may enhance the adaptive capacity of the people. This result is in line that of with Simane, Zaitchik, and Foltz [
31] who reported that highland agroecological systems were relatively less exposed to the effects of climate variability. Despite the higher amount of rainfall, Guder has had less exposure to soil erosion problems, which could be attributable to the unprecedented expansion of
A. decurrens plantations across the watershed in the past decade [
49]. The plantations might have helped the watershed by restoring the ecosystem of degraded hillsides and intercepting rainfall [
68]. In addition, the contribution of agroforestry to rural livelihood resilience was noted in Quandt et al. [
69], who showed the contribution of agroforestry in response to the impacts of climate-related hazards like drought and flooding. For example, during drought, many tree species still produced fruit for household consumption and sale, while staple food crops, such as maize, did not survive.
As compared to Guder and Aba Gerima watersheds, Dibatie watershed has been more exposed to climate-induced shocks and experienced more anomalies in the last 35 years. Despite that this watershed is located in a regional state known for its forest resources, it has been subject to overgrazing, deforestation, and poor farming practices, which has advanced desertification in the Nile basin (UNDP, 2017) [
70]. Our findings are not in agreement with those of Deressa, Hassan, and Ringler [
8], who revealed that households in Benishangul Gumuz had experienced fewer numbers of droughts and floods. A possible reason for this difference could be that they did not use station-level data to study vulnerability, whereas our SPI analysis did, and it showed that the area has experienced some types of meteorological and agricultural drought in the past 35 years. The SPI for Dibatie indicated relatively high climate variability, particularly dry spells and a significant increasing temperature trend (
p < 0.001), which might have had negative impacts on crop production and livestock rearing. Late and untimely rainfall arrival in lowland agroecosystems and/or high temperatures during the crop development stage could cause a decline in yield and increase the community’s exposure and ultimately contribute to increased vulnerability [
14]. Rurinda et al. [
71] noted that increased rainfall variability together with rising temperatures reduces soil moisture availability and increases the risk of crop failure.
4.3. Sensitivity
The overall sensitivity score, which included the health, food, and water components, was lower than the overall adaptive capacity score for all three watersheds. The food component was the primary contributor in all three cases. A study carried out in Myanmar similarly revealed the food component substantially contributed to the sensitivity dimension of farm households’ climate change vulnerability [
72]. A slightly higher sensitivity score was estimated for the Aba Gerima watershed as compared to Guder and Dibatie, which was mainly attributable to its sensitivity to the food and health components. Aba Gerima farmers generally had a lower level of food source diversification, which could trigger enhanced sensitivity. In addition, in Aba Gerima, land is being utilized to expand production of khat, which reduces the amount of land available for food production [
48]. Moreover, no access to agricultural technology, a high degree of water abstraction from ground water aquifers for khat fields, and higher soil erosion severity also could have played roles in the sensitivity of households in this study site. Similarly, a higher livelihood sensitivity associated with landscape greenness, soil fertility, soil erosion, water availability, pasture availability, and plot condition was reported by Siraw, Adnew Degefu, and Bewket [
32] in the Ethiopian highlands.
Dibatie scored better on crop diversification as compared to its counterparts. Crop diversification should make household livelihoods less sensitive to climate-related adverse effects. The Benishnagul Gumuz region, particularly Dibatie, besides other crops, has a notable tradition of ground nut cultivation [
73], which is well known as a drought-resistant crop [
74]. Dibatie, however, also had a relatively higher contribution from the soil erosion and seed-saving indicators. Ebabu et al. [
75] and Abeje, Tsunekawa, Adgo, Haregeweyn, Nigussie, Ayalew, Elias, Molla, and Berihun [
48] also reported that this watershed had higher land degradation problems as compared to Aba Gerima and Guder. Guder, representing the highland area, had lower crop diversification and, consequently, is relatively more sensitive to the effects of climate-related shocks. Saving seed for the next growing season is also not often practiced in Aba Gerima and Guder watersheds. This could possibly be related to their shift to the more remunerative cash-based khat plantation and A. decurrens monocropping [
48].
Health-related problems made a greater contribution to the sensitivity dimension in Dibatie and Aba Gerima as compared to Guder. Community discussion participants at both of these sites reported that government health services in their proximity do not work properly, and they usually use private health services in the nearest town, which usually costs more than the government services. More household members were reported to miss school due to health problems in Dibatie, which is most likely related to their greater experience with chronic illness. Health problems could result in a shortage of family labour for operating agricultural lands. A similar study in Ghana revealed that the sensitivity of farmers to the impacts of climate variability was partly contributed from their higher exposure, especially in households who do not own enough livelihood capital to support agricultural labour [
24].
The water component comprised consistent water access, water resource conflict, and average time to fetch water. Guder was relatively better off than the other two watersheds, and its lower sensitivity score might be associated with recent improvements in its broader ecosystem, and the presence of
A. decurrens plantations and other sustainable land management activities [
48,
49,
68]. Excessive extraction of water for irrigating khat farmlands could serve as a potential point of conflict among downstream and upstream farmers. Farmers in Aba Gerima reported that they had no consistent access to water sources, and the time it took them to fetch water was longer as compared farmers in the other watersheds. Moreover, the distance from the river to their farms was somewhat longer, which largely limited their ability to irrigate their farms. Whereas irrigation directly minimizes the impacts of climatic stresses such as droughts, farmers at all three sites are at increased risk of water availability due their dependence on natural water sources. A similar study noted a more pronounced livelihood vulnerability to drought in rural Iran, and it was mainly associated with access to water sources [
76].
4.4. Adaptive Capacity
Adaptive capacity plays an essential role in responding to the adverse impacts of climate change/variability, reducing livelihood vulnerability, and helping people to achieve sustainable livelihoods [
77]. Communities in all three watersheds were significantly less able to adapt to the effects of climate change. A household’s endowment of essential livelihood assets contributes to its adaptive capacity, which ultimately determines its livelihood vulnerability to certain negative consequences. Overall, communities in Aba Gerima were found to be more vulnerable, in a large part because of their low adaptive capacity, which was mainly attributable to limitations in the livelihood strategies and sociodemographic components. These components were, in turn, made up of elements, such as a relatively higher rate of illiteracy, higher age, and lower rate of engagement in livestock rearing. Possible reasons for lower livestock production may be a reduced availability of feed and restrictions on grazing in some parts of the watershed [
78]. Their lower engagement in livestock rearing may also be associated with their larger family size and their relative affluence due to their proximity to Bahir Dar city and their greater engagement in the production of cash crops (i.e., khat). In a study conducted in Kenya, livestock diversification was shown to be an important indicator in terms of lowering sensitivity to multiple stresses related to climate change [
23]. Even though the Dibatie (lowland) watershed exhibits a slightly better adaptive capacity as compared to the other two sites, in part because of its better social networking component score, households had little contact with local government offices, which ultimately constrained their institutional adaptive capacity to climate change vulnerability. Government offices are responsible for services related to response to climate-related shocks, such as issuing warnings, training residents about climate-smart technologies, providing information about markets, and delivering inputs. Hence, weaker contact with government offices may negatively affect the provision of these services. In addition, as we learned from the community discussions, most of the interviewed households are not entitled to own land, and thus cultivate land through informal renting arrangements. Land tenure insecurity has been cited as one of the constraints of production and productivity for households in Benishangul Gumuz [
79,
80]. In addition, effective, timely, and appropriate delivery of climate related warning information is of greater advantage at lower levels of administration (e.g., district level) planning and decision making for adaptation planning in the Ethiopian case [
81]. In a similar study done along the Nile basin, access to climate related information was indicated as a significant driver for household adaptation to the adverse effects of climate variability/change [
82].
A lower extent of livelihood diversification contributed to lower adaptive capacity in the Guder and Aba Gerima watersheds, possibly because the wider coverage of
A. decurrens and khat plantations in these areas reduced the amount of land that could have been used for the production of food crops [
48]. Climate variability in the form of drought could have more impact on communities with less diversified livelihood strategies [
83]. A study of weather shocks in Ethiopia indicated that off-farm livelihood diversification enhanced the capacity households to cope with climate-related shocks [
38]. A similar study in Kenya revealed that, as part of adaptive capacity, short- and long-term climate change adaptation should be supported by CBOs and enhanced social networking [
19]. Communities in Guder showed a relatively high level of vulnerability in terms of participating in farmer-based organizations and a lack of communication devices to receive climate-related information. Rural community awareness on the causes and impacts of climate calamities on their lives and livelihoods can be improved by education. In addition, the more educated community members are, it is highly likely that they will adopt climate smart technologies [
11]. Guder had a relatively higher education level, which indicates communities in this area should be more likely to adapt to the effects of climate change. A similar study carried out by Deressa et al. [
84] indicated that level of education plays an essential role in terms of choice of climate adaptation strategies and, hence, affects livelihoods. Likewise, low literacy was shown to contribute to higher vulnerability of households in Ghana [
24].
4.5. Obstruction Factors of Communities’ Adaptive Capacity
A larger number of dependents, low degree of participation in community based organizations (CBOs) and farmer-based organizations (FBOs), and lack of contact with the local government were major obstacles limiting community adaptive capacity. Moreover, being a female-headed household, not having other sources of income, and having a shortage of family labour were also major obstacles in all studied watersheds. As climate change/variability puts households under extra pressure, households with a shortage of labour could be more vulnerable by limiting their ability to diversify their income sources or go outside the community for employment. The availability and quality of human capital, including active working labour, has been shown to affect household adaptive capacity to climate change [
85]. A lower degree of participation in community- and farmer-based organizations was a substantial obstacle, primarily because these organizations provide important services to the communities. For example, CBOs include informal social networking schemes where households help each other during periods of social, as well as economic, problems. Social networking essentially reduces a community’s vulnerability to the adverse effects of climate change [
86]. Female-headed households generally have a marginal role in all walks of life and are denied many opportunities, which makes them more vulnerable to the effects of climate change [
87]. Female-headed households may not have access to social services and information because of socially constructed problems related to inequality, and this could limit their capacity to mobilize available resources to adapt to the negative effects of climate-related shocks them [
66,
84]. Similarly, a gender specific study in Ghana revealed that female headed households, due to their low sociodemographic profile, low social network, and lack of access to water and food, were indicated as vulnerable to the impacts of climate change/variability compared to their counterparts[
88]. Moreover, in Central Nepal, limited access to communication and reliable information on climate related hazards and low participation in local based organizations made female head households vulnerable [
89].