Household food security in the face of climate change in the Hindu-Kush Himalayan region

This study attempts to understand local people’s perceptions of climate change, its impacts on agriculture and household food security, and local adaptation strategies in the Hindu-Kush Himalayan (HKH) region, using data from 8083 households (HHs) from four river sub-basins (SBs), i.e. Upper Indus (Pakistan), Eastern Brahmaputra (India), Koshi (Nepal) and Salween and Mekong (China). The majority of households in SBs, in recent years, have perceived that there have been more frequent incidences of floods, landslides, droughts, livestock diseases and crop pests, and have attributed these to climate change. These changes have led to low agricultural production and income, particularly in Eastern Brahmaputra (EB) where a substantial proportion of HHs reported a decline in the production of almost all staple and cash crops, resulting in very low farm income. Consequently, households’ dependency on external food items supplied from plain areas has increased, particularly in the Upper Indus (UI) and EB. After hazards, households face transitory food insecurity owing to damage to their local food systems and livelihood sources, and constrained food supply from other areas. To cope with these, HHs in SBs make changes in their farming practices and livestock management. In EB, 11 % of HHs took on new off-farm activities within the SB and in SM, 23 % of HHs chose out-migration as an adaptation strategy. Lastly, the study proposes policy instruments for attaining sustainable food security, based on agro-ecological potential and opportunities for increasing agricultural resilience and diversity of livelihoods.


Introduction
Mountains provide 40 % of global goods and services in the form of water, hydroelectricity, timber, biodiversity and niche products, mineral resources, recreation, and flood control (Huddleston et al. 2003). However, 51 % of the almost 842 million people worldwide facing chronic hunger, are accounted for by six Hindu-Kush Himalayan (HKH) countries, i.e. Bangladesh, China, India, Myanmar, Nepal and Pakistan (FAO 2013). Moreover, recent studies (Giribabu 2013;MoHP-Nepal 2012;FSA 2009;Hussain and Routray 2012) have revealed that severity of food insecurity in mountain areas of HKH countries is significantly higher than in plain areas.
Mountain people, particularly in the HKH region, are highly vulnerable to food insecurity because of their low productivity, subsistence economies, constraints of terrain and climate, poor infrastructure, limited access to markets, physical isolation, vulnerability to natural hazards and high cost of food production and transportation (Rasul 2011;Tiwari and Joshi 2012;Ward et al. 2012, Huddleston et al. 2003, FAO 2008. The natural resource base in the HKH region, particularly soil nutrients, water and biomass, has been steadily depleted over recent years, resulting in a significant decline in food production (Tiwari 2000;Andersen et al. 2005). Increasing impacts of climate change have added to food insecurity, particularly in the Hindu-Kush Himalayan (HKH) region, where people are affected in both upstream and downstream areas (Chatterjee and Khadka 2013;Abbas 2009;World Bank 2009;ICIMOD 2008).
The livelihoods and food security of mountain communities depend heavily on the local resource base at all elevations, although the specific agro-ecological and livelihood potentials vary considerably. Subsistence agriculture, livestock and horticulture are the main sources of livelihoods, with livestock becoming more important than arable farming at higher elevations. Remittances from those who have out-migrated from these communities, small businesses, wage labor, tourism and collection of medicinal plants and other herbs also contribute to livelihoods and food security. However, in recent years, climatic and socioeconomic factors have contributed to depletion of the natural resource base across the HKH region . Climatic changes included unprecedented changes in precipitation patterns and hydrological imbalances, rises in temperature, frequent floods and degradation of the forests, rangelands and agricultural land (Nautiyal et al. 2007). Unusually heavy rainfalls, triggering Glacial Lake Outburst Floods. (GLOFs) in the mountain areas (Din et al. 2014), have also been attributed to climate change. In the Upper Indus basin, for instance, the frequency and intensity of GLOF events have increased during recent years, five occurring during 2008-2009 in the Gojal valley of Hunza alone. Analysis of GLOF events showed that they were linked to weather conditions in terms of increased temperature, rainfall and increased occurrence of heat waves. Recently (July 2015), heavy monsoon rains coupled with GLOF affected the Chitral District in Khyber Pakhtunkhwa (KPK) Province (IFRC 2015).
'Too much or too little water' is adding to the vulnerability of mountain agriculture, which is mainly rain-fed. Mountain farmers are experiencing frequent floods and prolonged droughts, resulting in low productivity of agriculture and higher prevalence of food insecurity (Hussain et al. 2016). In Nepal, for instance, the incidence of food poverty in mountain regions is 48 % compared to 18 % in the plain areas (Haslett et al. 2014: p. 28). Likewise, in mountain areas of Pakistan, including FATA, Baluchistan, Gilgit-Baltistan and KPK, nearly 60 % of the people are food insecure (FSA 2009). The past trends and the projected changes in temperature and precipitation in the HKH clearly show that temperature and precipitation patterns are changing faster in this region than neighbouring regions (Dhakal et al. 2010;Government of Nepal 2011). Unusual rainfall patterns, attributed to climate change, are also resulting in the disappearance of some crop species and varieties from local food systems owing to lack of conservation initiatives (FAO-AIPP-IWGIA 2015).
Ensuring food security has become a big challenge in the face of such changes, particularly as subsistence farmers are already facing the constraints of limited arable land, difficult terrain, unfavorable bio-physical conditions, physical isolation and limited market access . A further factor is the increasing rate of out-migration, which is causing labor shortages in agriculture, leading to underutilization of agriculture potential. The consequent reduced agricultural production has affected food security and increased the overall vulnerability of mountain people, as they have become more dependent on food from outside and thus exposed to market fluctuations.
Food security in mountain areas usually does not attract much attention from researchers and development planners owing to the physical isolation and higher cost of research and development activities (Rasul and Karki 2007). Therefore there is a dire need to understand the challenges to mountain farming systems, local adaptation strategies and food security in the face of the changing climatic conditions already described. Such understanding may provide novel adaptation strategies that allow mountain people to maintain their food security and livelihoods.
The primary purpose of this study is to understand local people's perception of climate change, its impacts on agriculture and household food security, and local adaptation strategies, using large-scale survey data collected from four river sub-basins, i.e. Upper Indus (Pakistan), Koshi (Nepal), Eastern Brahmaputra (India), and Salween and Mekong (China). It has also identified the opportunities arising out of changing climate for mountain farmers to achieve sustainable food security and livelihoods. This study is descriptive research, which was designed to rely mainly on information provided by local people on the ground. A secondary purpose of the study was to validate the findings of various scientific studies of impacts of climate change in the mountain areas of the HKH region. It is hoped that the study will provide a better understanding of the vulnerability of mountain people to food insecurity under climatic change and their ability to adapt.

Research design and study area
This study uses data which was collected through a large scale survey in four river sub-basins, i.e. Upper Indus (Pakistan), Koshi (Nepal), Eastern Brahmaputra (India), and Salween and Mekong (China) of the Hindu-Kush Himalayan region (Fig. 1). This survey tool, named 'Vulnerability and Adaptive Capacity Assessment (VACA)', was developed by the International Centre of Integrated Mountain Development (ICIMOD) in 2011 under the Himalayan Climate Change Adaptation Project (HICAP) to assess the livelihood vulnerability, adaptive capacity and responses to climatic and socioeconomic changes. VACA covers a broad range of aspects with regard to mountain agriculture, food security, livelihoods, ecosystem services, climatic and socioeconomic changes, environmental stability and local adaptation strategies. However, the present study has only used those data that were collected to investigate the linkages of climate change with agriculture and food security.
The selected four river sub-basins (SBs) are upstream parts of those large river basins, which are among the largest and the most productive ecosystems in the world. These basins provide water not only for agriculture, but also for forestry and fisheries as well as urban and industrial purposes (Sikka and Ringler 2009). Recent studies (Yu et al. 2013;Rasul & Hussain 2015;Bhatt et al. 2014;Zhen-Feng et al. 2013) have revealed that increased incidences of erratic precipitation, floods, dry spells and landslides, attributed to climate change, have led to decline in agricultural productivity and deterioration of food security. This study investigates whether these scientific findings were mirrored in local people's perception of climate change and its impacts on agriculture and household food security. Projections of climate change modeling suggest that these river basins will also impact future water availabilities in the respective countries through changes in precipitation patterns and decline in water from glaciers.

Sampling design
In India, the study areas selected comprised the states of Arunachal Pradesh (districts: East Siang, Lower Dibang, Lohit) and Assam (districts: Dhemaji, Lakhimpur, Moregaon & Tinsukia). In Nepal, the selected districts were Dolakha, Kavre-Palanchowk, Khotang, Udayapur, Sunsari and Siraha. In Pakistan, the provinces Gilgit Baltistan (districts: Gilgit & Hunzanagar) and Khyber Pakhtunkhwa (district: Chitral) were selected. In China, the counties of Linang, Baosanchan, Dali, Diqing and Nujiang were selected. Within these selected states/districts, random selections of settlements were made followed by a random route procedure to choose the households. Overall, 8083 households were surveyed: 1139, 2647, 2310 and 1987 from Upper Indus, Eastern Brahmaputra, Koshi, and Salween and Mekong, respectively. Data collected for this study may not be a true representative of the sub-basins because districts in the states/provinces were selected purposively in view of their higher vulnerability to climate change impacts rather than following a systematic random sampling technique. Thus, for generalization of results at sub-basin levels, authors clearly caution that sampling design has some limitations.

Type and nature of collected data
For this study, both quantitative and qualitative data were analyzed. Quantitative data included household size, land under different cultivation practices, monthly food and non-food expenditure, agricultural income and working members of households. Qualitative data included households' reporting on cultivation of major staple and cash crops, households' response on ownership of different types of livestock, perception of climate change, reporting on out-migration, perception of climate change impacts on crop production, adaptation strategies and reporting on non-agricultural income sources. Some important points about the nature of the data are clarified below. 1) In this study, reported climatic changes (Figs. 2,3,4,5 and 6) mean 'incidence of extreme events' observed by the households in the last 10 years. 2) Some crops are categorized under both staple and cash crop categories (Tables 3 and 6) because these are being cultivated as both staple and cash crops depending on the farmers' choice in response to market access and prices. Staple crops are those that are cultivated mainly for household food consumption, whereas cash crops are cultivated with the objective of sale to earn income. 3) Changes in crop production attributable to climate change (Table 6) are average changes perceived by the households in the past 10 years. 4) Households' judgement criteria on climate change perception may vary across households but still provide some robust indications about the incidence of extreme events, attributed to climate change, and their impacts on agriculture and food security. Despite some limitations, perception based data are useful for determining if they correspond with the findings of scientific studies, which are down-scaled at least at basin level.

Findings Socioeconomic characteristics of households
The four river basins differ from each other in terms of socioeconomic characteristics (Tables 1 and 2). Household size is greatest in the Upper Indus (UI) and least in Salween and Mekong (SM) ( Table 1), perhaps reflecting China's one-child policy (Riley 2004;Ge et al. 2012). On the other hand, in Pakistan, family planning strategies, due mainly to religious and cultural factors, have not imposed a strict family size (Zulfiqar & Hussain 2014). In all sub-basins, households are mainly headed by male members (  (Keliang and Prosterman 2007). Further background details of the four SBs are given in Table 2.

Farming systems in sub-basins
Farming households in SBs are mainly smallholders who practise mixed farming systems, consisting of crops, fruit and livestock (Tables 3 and 4). In SM and UI, a substantial proportion of agricultural land is under fruit and tree orchards. In UI, more than one-fourth of households' agricultural land has been turned into pasture and grassland, possibly due to frequent agricultural labor shortage (Table 1). It may also be one of the reasons for the 3-5 % fallow agricultural land in Koshi and EB basins, as reported elsewhere (Ghimire and Thakur 2014). In Koshi and EB, around 80 % of the land is under the cultivation of crops. In all four sub-basins, minor plots are transformed into kitchen gardens, e.g. in EB around 10 % of average household land are kitchen gardens. Among the SBs, the UI has the highest percentage (99.9 %) of farm households with access to irrigation systems, whereas EB has the least (16.8 %) (Table 1). Almost all farming households in UI and Koshi cultivate staple crops including cereals and vegetables (Table 3). Likewise in EB and SM, more than 90 % of farming households cultivate staple crops. In SM, nearly 68 % farming households cultivate cash crops, whereas in the other three SBs, fewer than half of the households cultivate cash crops. Choices of particular crops -both staple and cash -differ across countries owing to differences in agro-ecological potential and market factors (Pan et al. 2010). In UI, among the households who grow staple crops, 55 % grow wheat, which is the main staple crop, not only in the study area but also in other regions of Pakistan (Zulfiqar & Hussain, 2014). Farming households also prefer to cultivate summer potato and other vegetables, apple and summer maize as staple crops. Among those households who grow cash crops in UI, summer potato is the first choice due to suitability of local conditions for its cultivation and its local as well as external demand from downstream areas (Rasul & Hussain 2015). A significant proportion of households also produce fruits such as apple, cherries, apricot and walnut to generate income (Table 3).
In Koshi and EB, paddy is the main staple crop, followed by other cereals in Koshi and vegetables in EB. Similar to UI, summer potato is the main cash crop in Koshi. Instead of fruits, in Koshi vegetables are the next choices of farmers for cash crops. In Koshi, one-quarter of farming households cultivate mustard as a staple crop while one-third of EB households cultivate it as a cash crop. A significant proportion of households also prefer tea, ginger and jute as cash crops (Table 3). In SM, summer maize is the main staple crop, followed by other cereals and vegetables. In this SB, around 40 % of farming households prefer to grow walnut and tea as cash crops. Among other cash crops, garlic, tobacco and sugarcane are also very important income generating choices. Among all SBs, income from the sales of staple and cash crops in SM are significantly higher due to good market connectivity and long-run integration (Park 2008;Laping 2006). Among other SBs, UI is slightly better compared to Koshi and EB in terms of income generated income from crops, whereas EB has the least returns (Table 3).
Similar to other mountain regions, households in the study SBs raise livestock to support their food security and livelihoods. In UI and Koshi, more than 80 % households own livestock. Also in EB and SM, a substantial proportion of households raise livestock (Table 4). Among the livestock raising households, a substantial proportion of the households in UI, Koshi and EB raise cattle and goats, whereas in SM most of the households prefer to raise pigs. In Koshi more than half of the households raise buffalo, whereas in other SBs, only a small proportion of households prefer to do this. In UI, more than half of livestock raising households raise sheep (Table 4), which is in line with the higher transhumant activities in this SB (Table 1). Moreover, sheep are more resilient to water and fodder/forage-stress but can have negative longterm environmental impacts on pastures and rangelands (Shafiq & Kakar 2007). In all SBs, households also raise poultry, ducks and pigeons for home consumption as well as sale. However, bird raising is significantly higher in EB and SM. Fish catching and farming practices are prominent only in EB compared to other SBs. Households' perception of climate change The majority of households in all four SBs perceived that climate has been changing over the last 10 years (Fig. 2). In UI, among those households who perceived changes in climate, the majority reported the greater frequency of floods in their area. A significant proportion of the households also reported that they observed certain changes in rainfall patterns and temperature. Moreover, one-fifth of households faced landslides and erosion triggered by heavy rainfall and floods (Fig. 3). Manzoor et al. (2013) also revealed that after 2000-01, frequency and magnitude of floods in UI has increased due to intense rainfall in the Indus catchments and are occasionally increased by snowmelt.  In Koshi, changes in rainfall patterns were reported by more than half of the surveyed households. In this SB, households perceived that decrease in annual rainfall has resulted in prolonged dry spells and droughts. In addition, they also perceived an increase in temperature and incidence of livestock diseases over time (Fig. 4). Bharati et al. (2012) projected a 0.79-0.86°C rise in temperature in the 2030s for the Koshi Basin compared to a baseline average over 1976-2005. Increased temperature leads to greater evaporation and thus surface drying, increasing the intensity and duration of drought (Devkota and Gyawali 2015).
In the EB, incidence of drought, flood and erratic rainfall have increased, and the majority of households also observed a rise in temperature (Fig. 5). More than 55 % households reported that incidence of livestock diseases had also increased in their areas. This is consistent with other studies (Sirohi and Michaelowa 2007;Singh et al. 2000;Basu and Bandhyopadhyay 2004) where changes in temperature, rainfall patterns and humidity were directly related to increased incidences of livestock diseases.
Similar to other SBs, in SM around 47 % households reported that they observed frequent dry spells and droughts in their areas (Fig. 6). A significant proportion of households also observed a rise in the incidence of erratic rainfall, temperature rise and crop pests. SM is the only SB where households have reported an increase in crop pests over time, which was attributed to climate change (Fig. 6). This is understandable because changes in temperature, timings of seasons and rainfall patterns may lead to increased populations of weeds in grasslands, and incidences of pests and diseases of grasses and crops (Sirohi and Michaelowa 2007).  In addition to climate change, outmigration is the most prominent socioeconomic change observed in all four SBs. Outmigration from EB is relatively low compared to other SBs (Table 5). In UI and SM, migration to other parts of the country is higher than the overseas migration. However, in Koshi, overseas migration is higher than within country migration. Among the migrant sending households in Koshi, more than 74 % of households receive remittances, whereas in UI and EB, 55 % and 58 % households receive remittances, respectively. Despite the higher outmigration (40 %) in SM, only one-fifth of migrant sending households receive remittances (Table 5) due possibly to reasonable income from agriculture (Table 3), lessening the need for them.

Impacts of changes in agricultural production
In SBs, farmers' timely and adequate access to water has become a challenge. Irregular precipitation patterns, attributed to climate change, have caused severe impacts on livelihoods of millions of vulnerable people (Chen et al. 2013). Rural communities in hills and mountains are experiencing substantial impacts on water resources due to prolonged dry seasons. UI has experienced erratic rainfalls and floods (Fig. 2), whereas in the other three SBs, temperature rise, frequent dry spells and droughts have been observed (Figs. 3, 4 and 5). Irregular changes in rainfall patterns may have impacts on water availability, particularly in EB where over 80 % of agricultural land is rain-fed (Table 1).
Despite advances in agricultural technology and inputs, a significant proportion of households reported a decline in the production of their crops over the last 10 years. This decline in production may be attributed to climate change. In UI, the majority of crop growing households reported that production of main staple crops such as wheat and summer potato has decreased over time (Table 6). In Koshi, the majority of households reported that the production of summer maize, millet and mustard has decreased. In EB, changes in climatic events have caused relatively more severe impacts on production. Households perceived that production of all staple crops and most cash crops had declined (Table 6). In SM, as perceived by the majority of particular crop growers, production of some staples, i.e. wheat and maize has decreased in the last 10 years. Some opportunities are also arising for farming households. For instance, in UI, a substantial proportion of households reported an increase in the production of summer vegetables and fruits such as apple, cherries, apricots and walnuts. Likewise, a significant proportion of households reported an increase in the production of summer potato, onion and vegetables in Koshi, tea in EB, and walnut, tea, garlic, tobacco and sugarcane in SM.
Inflow of remittances from out-migrants is undoubtedly a potential source for improving local food security and livelihoods through enhancing local small businesses, transferring new technologies, and creating job opportunities for local skilled and unskilled labor. But outmigration has also added to the challenges in mountain areas. Increased outmigration and decreased interest of the youth in farming also add to the low production in agriculture . Households in all four SBs face frequent labor shortages, which together with water shortages is leading to increased amounts of fallow agricultural land (Table 1).

Adaptation to climate change
For adaptation to climate change, households have adopted various practices, which improve their resilience. Those households which perceived changes in climate were asked to report their adaptation practices during the survey. In UI, almost one-fifth of households changed their farming practices and introduced new crops (Table 7). Changes in farming practices include water conservation methods, change in sowing time and introduction of new crops, such as fruits and nuts, which are relatively more resilient to water-stress and have higher market value (IFAD 2015). Some areas such as Broghil top in Chitral and Deo Sai in Gilgit-Baltistan, which were not suitable for crop cultivation due to their harsh climates, are now under vegetable cultivation. During summer months, these areas do not have snow cover anymore. Therefore, local communities have started growing winter vegetables (i.e. potato, peas etc.) during the summer ). More than 16 % households gave up the rearing of certain livestock (Table 7). In UI, climate change has resulted in significant degradation of pastures and rangelands, which are free sources of grazing for almost 80 % of livestock in the SBs (Khan et al. 2013). Since sheep and the larger animals consume more fodder and water, as a coping strategy, livestock owners are more likely to reduce the number of the larger animals and sheep, and increase the number of local goats (Table 4). Local breeds of goats are more resilient to water and fodder/forage-stress, as revealed by a study done in Balochistan province of Paksitan (Shafiq & Kakar 2007). In  Changes in crop production are average trends perceived by households in last 10 years Source: ICIMOD Survey 2011-12 addition to adaptation practices in crops and livestock, 10 % of households also invested in disaster preparedness practices, e.g. construction of sheds and shelter for livestock and family members. In Koshi, around 18 % of households gave up planting crops which were highly vulnerable to water stress (e.g. paddy). Around 15 % households introduced new crops on their farms (Table 7). In Nepal, farmers are shifting their cropping patterns from highly water consumptive crops (i.e. paddy) to fruits and vegetables which are high value crops (Gurung and Bhandari 2009;GWP-JVS 2014: p.21, Dixit et al. 2009). A small proportion of households also changed their farming practices (Table 7). These changes included slight shifts in timing of crop cultivation, exploring improved varieties of seed and use of different agriculture practices requiring less water (Bhatta et al. 2015). Around 8-9 % of households either gave up rearing certain livestock or changed their grazing practices (Table 7). Compared to other SBs, in EB fewer households adopted new farming practices to cope with the impacts of climate change. Rather over 10 % of them started new off-farm income activities to support their food security and livelihoods (Table 7) because of high vulnerability to floods and low agricultural production (Saikia 2012;Sarkar et al. 2012). Only 5-10 % households changed their farming practices and cropping choices, whereas almost 7 % invested in some disaster preparedness practices (Table 7).
In SM, more than one-fifth of households were those who sent at least one member of the family to work either in other parts of the country or overseas as a coping strategy (Table 7). In addition to changes in farming practices and crop choices, investments in advanced irrigation technology have been made by 11-12 % households. The Chinese government is also promoting investment in the construction of water collection and utilization engineering in hill and mountain areas (GoPRC 2007).

Agricultural income and household food consumption
Traditionally, agriculture is assumed to contribute to the food security and livelihoods of households in the mountains through providing diverse foods and contributing to household income. However, due to increasing climatic vulnerabilities and market uncertainties, the contribution of agriculture to household income has significantly decreased over time (ISET 2008). In all SBs, the majority of households reported that agriculture and livestock were sources of their income. However, they are main sources of income for only a small proportion of households (Table 8) due mainly to declining productivity resulting from climatic hazards and labor shortages.
Although farm production contributes to household food consumption, yet the local people have to buy several other food items from the market due to declining productivity and diversity in mountain agriculture. In Koshi, farm production is the main source of food for almost half of the households (Fig. 7). In EB and SM, home production is the main contributor to food consumption for 23 % and 38 % households, respectively. In the UI, only 10 % households reported that home production is the main source of their food requirements (Fig. 7). In all four SBs, households are heavily dependent on external sources of food such as stores and markets where food items are mostly supplied from downstream plain areas. It is also evident from food expenditure data in three SBs, i.e. UI, Koshi and EB, more than half of household monthly expenditure was spent on food items, whereas in SM, the share of food expenditure was 40 % (Fig. 8). The supply of food items from downstream areas is likely to be limited, particularly after hazards such as landslides blocking roads (Andersen et al. 2005;MoHP-Nepal 2012).

Food security strategies in the time of environmental shocks
While agriculture remains an important contributor to household food security in mountainous areas, non-agriculture sources such as daily wage, salaried employment, small businesses and remittances, are becoming increasingly important to sustaining livelihoods, especially for households with small landholdings Bhandari and Grant 2007). When environmental shocks occur such as floods, prolonged dry spells, drought or erratic rainfall, most farming households in the SBs face transitory food insecurity due to damage to their farming systems and other livelihood sources. In the UI and SM, 86 % and 91 % households, respectively, reported that they had faced at least one shock in the 12 months prior to the survey, whereas in Koshi and EB, almost all households faced this situation (Table 9). Households adopt various strategies to reduce the impact of shocks on household food security. In the UI, one-fifth of households reported that they replaced expensive food items with the cheaper ones and about the same proportion borrowed money from relatives. Almost 14 % of households reduced their spending on clothing, whereas 10 % made changes in their farming practices, growing crops which take less time to harvest.
In Koshi and EB four out of five and three out of five strategies, respectively, to combat shocks threatening food supply involved borrowing money from different sources (Table 9). Those who did not borrow money bought food on credit from local shopkeepers or stores, sold livestock or reduced spending on clothing. Incurring debts as a coping strategy to obtain food or non-food items may be effective in the short term but debts may increase in the long term, negatively affecting community resilience and sustainable food security (Milbert 2009). In the SM, in addition to borrowing money, 21 % sought work outside the community, 18 % within the community, 19 % of adults restricted their consumption and 16 % spent savings on food.

Conclusions
In four river sub-basins (SBs) of the Hindu-Kush Himalayan region (HKH), i.e. Upper Indus (UI), Koshi, Eastern Brahmaputra (EB), and Salween and Mekong (SM), households perceived that increased incidences of natural hazards such as floods, droughts, landslides, livestock diseases, crop pests, erratic rainfall and temperature extremes, attributed to climate change, were significantly influencing agricultural production, income and household food security. Increased rates of out-migration have also resulted in labor shortages in agriculture, possibly adding to a decline in productivity, food availability and farm income. Traditionally, agriculture is assumed to contribute to the food security and livelihoods of households in the mountains through providing diverse food and contributing to household income. However, due to increasing vulnerabilities, the contribution of agriculture in household food consumption and household income has significantly decreased over time. Particularly in EB, the majority of farm households reported a notable decline in the production of almost all staple and cash crops, resulting in the least farm income compared to other SBs. Due to decreased food production, households' dependency on external food items supplied from downstream plain areas is increasing. Although agriculture and livestock are contributing to the income of a substantial proportion of households in SBs, yet these are no longer the main sources of income for the majority of households. Generally, in SBs, households have to rely on nonagricultural income sources to buy expensive food and nonfood items, supplied from plain areas.
To cope with the climate change risks, households have adopted various strategies. These include changes in farming practices such as the introduction of new resilient crops, abandoning certain highly water consuming crops and giving up rearing certain livestock, which are vulnerable to water and fodder stress. In addition, households in UI and EB invested in preparedness for hazards such as floods and landslides, and in Koshi they made changes in livestock grazing practices to avoid excessive degradation of pastures and rangelands. In EB, a notable proportion of households took on new offfarm activities due to the increased vulnerability of agriculture. One-fourth of households in SM decided to migrate as an adaptation measure, either to other areas of China or overseas to find off-farm income opportunities. Households in this SB also invested in irrigation to cope with water stress. Although SBs have invested in adaptation practices to mitigate extreme climate events attributed to climate change, households in these areas still face transitory food insecurity when hazards, such as floods, landslides and droughts strike, owing to their effects on farming systems and other livelihood activities.
In view of this study's findings, the following policies are suggested in order to achieve sustainable food security in mountainous areas subject to extremes of climate.
& Governments need to establish separate food security policies for mountains and plains because mountains are different from plains in terms of nature, type and magnitude of vulnerabilities. There is also a need to re-evaluate agricultural policies under the projected changes in climatic conditions (Lu et al. 2012). Some current policies appear to be inappropriate. For example, the Nepalese government promotes cultivation of rice and pulses in mountainous areas, although these crops are resource-intensive and very sensitive to water stress. & Climate change has also brought some opportunities, which are not adequately capitalized. For instance, In UI, a substantial proportion of households observed an increase in the production of summer vegetables and fruits such as apple, cherries, apricots and walnuts over the 10 years. Likewise, a significant proportion of households reported an increase in the production of summer potato, onion and vegetables in Koshi, tea in EB, and walnut, tea, garlic, tobacco and sugarcane in SM. National and subnational planning processes should take into account such rising opportunities while preparing strategies to achieve sustainable food security in the SBs. & Government and non-government experts in HKH countries need to identify the specific zones within the SBs with higher agro-ecological potential for specific high value crops such as fruits, nuts, vegetables, tea, tobacco and other cash crops. Strategies may focus on exploiting the existing potential through land use intensification, efficient water use, integration of livestock and crop diversification. & Areas having less agro-ecological potential and that are highly vulnerable to hazards may be encouraged not to pursue agricultural activities. In such areas, strategies may focus on the subsistence use of resources, ecotourism and non-agricultural enterprises to reduce the dependence on local resources and ensure food security. & Governments should encourage private investment in production and post-harvest facilities. There is also a need to improve accessibility to institutional services (i.e. roads, markets, extension services and technology . Capitalizing on local potential and opportunities will also help to control out-migration, excessive switching to the non-farm sector and frequent labor shortages in agriculture.

Future research
This is a descriptive study of local perceptions of climate change, its attributable impacts on agriculture and food security, and local adaptive measures. In short, this study is a situation analysis of food security and climate change linkages in the study areas. It is hoped that it will provide a good platform for researchers to design and conduct empirical studies in the same areas to further understand the key factors of vulnerability and adaptation to climate change, and to establish different adaptive measures to achieve food security in the face of climate change.