Introduction

Climate-variability-induced stressors adversely impact the livelihoods of smallholders and marginal communities, especially in low-income developing countries (Chatterjee and Khadka 2011). Model projections can provide a long-term view of the physical aspects of climate scenarios at the macro-scale, but fail to adequately interpret the human dimensions of climate uncertainty and risks at the micro-scale (Savo et al. 2016). Processes and dynamics vis-à-vis climate variability within particular small-scale social–ecological systems can be overlooked (McCubin et al. 2015). Climate variations alone may not adversely impact livelihoods at the micro-scale, rather, the extent of vulnerability comes from their interaction with different ecological and socio-economic factors (multiple stressors) (McDowell and Hess 2012). Understanding the ways in which these multiple stressors interact with each other to affect livelihoods is vital for developing adaptation strategies.

Evidence suggests that Southeast Asian countries are now exposed to more frequent climate extremes (Ge et al. 2019). Under such situations, understanding farmers’ perceptions about climate variability can add strength to scientific knowledge (Dang et al. 2014), while further broadening understanding of how climatic anomalies and associated stressors operate at different scales of climate and associated stressors (McCubin et al. (2015) can help to frame more inclusive adaptation strategies (McNamara et al. 2020)).

Farmers’ perceptions of climate variability have long been valued in a practical sense (Deressa et al. 2009), especially in relation to shaping adaptive practices in agriculture (Slegers 2008). Despite facing similar exposure to climate variability and other stressors, farmers from the same area may vary in their perceptions of these risks (Singh et al. 2017). For example, a rice grower lacking assured irrigation facilities will undoubtedly face greater risks of crop failure under drought or dry-spell conditions than those having access to irrigation. Risk perceptions in agriculture may nevertheless differ with climate (Simelton et al. 2013), level of technological support and the accessible resource base (Singh et al. 2012; Niles and Mueller 2016). These differences shape the adaptive capacities of different social groups (Agrawal 2008), and accordingly, their varied perceptions and adaptation responses (Smit and Wandel 2006).

Prevailing policies also influence adaptation planning. Policies formulated through top-down approaches may not always auger well to end-user needs (Donatti et al. 2019). In India, little research has been conducted to explore the links between farmers’ perceptions and policy recommendations for adaptation. Emphasis has largely been on physical and biological aspects (Asseng et al. 2013) or policy dimensions (Agrawal 2013). Efforts to record farmers’ adaptive practices were recently initiated in 122 out of 640 districts of India (NICRA 2016). However, they do not assess how farmers exposed to multiple stressors experience interacting impacts on their livelihoods (Singh et al. 2017).

While O’Brien et al. (2004) and Tripathi (2014) report on the agricultural vulnerability of different states and districts of India, respectively, scant information is available to farmers and local developmental agencies. Virtually no systematic study has been conducted at the micro-scale (village level) to assess farmers’ perceptions of multiple stressors in relation to their livelihoods in eastern Uttar Pradesh (UP). This study addresses this gap and seeks to: (i) uncover farmers’ perceptions of climate variability; and (ii) identify compounding multiple stressors, in order to (iii) understand the livelihood risks experienced over the period 2000–2015.

Conceptual Orientation

This paper considers stressors as incremental increases in a particular event, phenomenon or situation that amplifies the livelihood risks of farmers (Parry et al. 2007). Multiple stressors include climate variability (drought and flood), ecological stressors (problematic soils and depletion of fresh groundwater), socio-economic stressors (low incomes and labour shortages) and political stressors (lack of credit, infrastructure and technology) (Ribot 2014). Climate variability, alone or in combination with other stressors, may increase the risks to farmers’ livelihoods, and shape their perceptions of climate variability (Fig. 1).

Fig. 1
figure 1

Conceptual model of multiple stressors interacting with each other and causing risk on livelihoods of farmers

Full size image

Variations in knowledge of an event, process or phenomenon, are a natural corollary of differences in perception (Raymond et al. 2010). In this study, perceptions about climate variability and other stressors were defined as an individual’s ability to see, hear and experience (over the period 2000–2015) any one or combination of stressors caused by climatic phenomena alone and/or ecological, socio-economic and political factors affecting activities vital to the farmers’ subsistence. Livelihood risks are then the outcome of interactions between several stressors (Birkmann 2007). In this study, climate-caused (adverse impacts on the livelihood base) and social constructivist views (risk within society shaped by the non-climatic factors required to sustain livelihoods) were taken into account, following Scoones (1998) and Ribot (2014). Taking insights from Simelton et al. (2013), Ribot (2014) and McCubin et al. (2015), we conceptualised a framework in which climate variability was proposed as an exogenous stressor and associated risk as a consequence. By comparison, stressors relating to ecological, policy-institutional and socio-economic spheres were considered internal and compounding stressors influencing livelihoods or increasing various risks, directly or through exogenous (climate) stressors, as perceived by the studied population (Fig. 1).

Climate Variability and Other Stressors: Implications in the Context of Eastern India

The majority of small-scale and marginal Indian farmers living in fragile ecosystems depend almost entirely on local resources (Tripathi 2014), lacking access to the external resources necessary to adapt to environmental challenges (Singh et al. 2017). Lately, socio-economic and political changes, including erosion of social institutions and the continual shrinkage of common property resources (Singh et al. 2012, 2019), have dealt a further blow to livelihoods, increasing dependency on external factors to secure means of pursing a living. The situation became even more risky for resource-poor farmers, including in the study region, due to changes in climate patterns and globalisation-induced market distortions (O’Brien et al. 2004).

India formulated the ‘National Action Plan on Climate Change’ (NAPCC) in 2008 to accelerate the country’s adaptation to climate change. In response to the NAPCC, State Governments have also developed ‘State Action Plans on Climate Change’ (DoE 2014), though with an almost exclusive focus on top-down approaches. The study state of UP, reeling under climate variability and extreme events, has also developed policies to mitigate and adapt to climate and biophysical stressors, but has largely overlooked the socio-economic and political stressors exacerbating such climate impacts (Tripathi and Mishra 2017). As a follow-up to various national and state policies, systematic research on climate change adaptation in agriculture started with the launch of the National Initiative of Climate Resilient Agriculture project (ICAR 2011). Subsequently, contingency measures and long-term plans (MoA&FW 2016) for farmers of states including UP were drafted. Concurrently, State Governments are refining their agricultural development policies with periodic assessments of climatic variables, and issuing weekly and monthly advisories to farmers (NICRA 2016; MoA&FW 2016). However, the process continues to be led from the top-down (DoE 2014). In many cases, farmers may differ in their perceptions about climate-variability-induced stressors and associated risks, due to their localised knowledge of agriculture management practices and other contextual factors (Limantol et al. 2016). Such on-the-ground concerns, which vary with time and space, although important for sustainable adaptations, are least understood at the formal level, and fail to feature in assessments of climate-induced stressors and the development of adaptation strategies. Addressing such gaps, this study provides an insight into how farmers perceive climate variability and how their perceptions can differ from formal ways of understanding climatic events, making an essential contribution to understanding successful adaptation.

Research Design and Methodology

Study Area

Agriculture and allied enterprises are the major livelihood activity for 55% of the 1.25 billion Indian population, with agriculture contributing 14% to Indian GDP (Gopalakrishnan and Thorat 2015). Data were collected from Azamgarh district in eastern UP, India on account of increasing climate variability and extreme events in the recent past in this location (Tripathi and Mishra 2017). Azamgarh district covers 4054.0 km2 (Census 2011) with an average elevation of 64 m above mean sea level (MoEF&CC 2010). It has a dry sub-humid hot climate, average annual rainfall of 803 mm and average minimum and maximum temperatures of 5.7 °C and 41.4 °C, respectively (GoUP 2009). There are three main seasons: winter (mid-October to mid-March), summer (mid-March to mid-June) and rainy season (mid-June to mid-October). From a population of 4.613 million (Census 2011), about 65% of people are engaged in agriculture and allied activities (GoUP 2009). Rice–wheat cropping systems dominate with cash crops such as potato, sugarcane, onion and vegetables cultivated in irrigated areas to varying extents (GoUP 2009). Soils are mostly sodic in nature and over 95% of land holdings are <2.0 ha (GoUP 2009). Privately owned tube wells cover most (83%) of the net irrigated area, with the remainder irrigated by canals and other sources (Kumar 2002). About 0.15 million ha are under a common property resource system, including wetlands (GoUP 2009). The study district is second in terms of its total number of wetlands in the state (MoEF&CC 2010) and these are crucial for agricultural livelihoods (Fuys et al. 2005). The district is relatively less developed in terms of rural infrastructure and agriculture than other parts of the state (GOI 2014; Tripathi 2014).

Sampling of Study Area and Population

Azamgarh district was selected purposively, based on recent high climate sensitivity and high vulnerability levels (Tripathi and Mishra 2017). Previous studies conducted in the same localities (Singh et al. 2014) and similar situations in other places (O’Brien et al. 2004; Tripathi and Mishra 2017), indicated hidden issues of multiple stressors experienced by farmers, highlighting the need for a systematic study. Based on the considerable knowledge about issues of the selected areas, the research was undertaken in the targeted area and population with available resources (budget and time) (Truelove et al. 2015). Three villages: Sonapur, Gambhirban and Gurehtha, in the Developmental Blocks of Jahanaganj, Rani Ki Sarai and Mehnagar, respectively, were purposively selected with guidance from district agriculture department officials. All three villages have a predominance of small and marginal farmers, and two major land-use systems (rice–wheat and rice–wheat-wetland). Field data were collected during 2012–2015 in May–July and November–December, followed by verification of results with farmers during August and December 2016, and September 2017.

A total of 24 key informants (6, 8 and 10 farmers from the study villages, respectively) were interviewed to record village level information on climate variability, ecological and socio-economic information and other data. Data were also collected through PRA (participatory rural appraisal) exercises (Table 1). Four criteria: (i) small land holding (<2.0 ha) (FAO 2010), (ii) minimum 15 years of agricultural experience, (iii) permanent residence in the village and (iv) thorough knowledge of agricultural history, were used for the selection of key informants, employing a snowball technique. Selection was made with the help of Gram Panchayat members (village level first tier of democratic institution) who know most of the villagers with whom they interact in most of the village developmental plans. A list of small and marginal farmers was prepared with the help of the Gram Panchayat and key informants to select 20 farmers aged >35 years from each village (a total of 60 respondents) using stratified random sampling. These farmers acted as interview respondents. During a scoping visit, we found the majority of the younger generation were less interested in agriculture and are migrating to towns for work, so we interviewed only those >35 years, as interaction of the younger generation with agricultural activities, and thereby experiences of climate, were assumed to be lower than those who are fully dependent on agriculture and live exclusively in villages.

Table 1 Key methods and techniques applied in data collection
Full size table

Data Collection

We used a combination of qualitative and quantitative techniques to collect the primary data (Table 1).

Quantitative Methods

Quantitative methods were used to collect data on individual perceptions about climate variability and livelihoods risks from 60 farmers using a structured interview schedule (Table 1). Day to day agricultural activities in the study area and local literature (collected during a scoping study) were used to frame statements on perceptions about climate variability and related risks or impacts on livelihoods. A questionnaire was developed with a set of positive and negative questions (e.g., positive sentence: timing of winter start is postponed; negative sentence: duration of winter is not decreased) covering different climatic variables (Tables 1 and 2, Online Resource 1). The purpose of randomly inserting negative statements was to minimise risk of bias in responses (Maddison 2006). Questions on ‘farm management’ (cost of cultivation, labour shortages, farm profits, etc.) enabled us to link farmers’ perceptions with documented trends. Questions were piloted with five farmers before final application, following edits to reduce ambiguity. The responses of farmers against each statement were measured using a 5 point scale (Likert 1932). District level rainfall (1901–2014) and temperature data (1901–2002) were accessed from the Water Portal (IWP 2017) and Indian Meteorological Department (IMD) (IMD 2017), respectively. These sources were also accessed for data on climate variability and related stressors. Policies and MSP (Minimum Support Price) data were taken from MoA&FW (2015), which complemented data on economic and market stressors.

Qualitative Methods

Using Google search engine, data on the susceptibility of the district to extreme climatic events and vulnerability were collected from the Ministry of Agriculture, Government of India and other secondary sources (details provided in online resources). Qualitative data were also obtained from 24 key informants who were asked to participate in 10 different exercises including focus group discussions (FGDs) (Table 1). Photographs and videos on floods, droughts, rainstorms and outbreaks of crop insect pests and diseases were shown to the key informants while conducting FGDs. The same strategy was followed with 60 respondents during interviews. The objective in both cases was to better capture participants’ and interviewees’ experiences of present and past events. In-depth discussions were held with key informants following FGDs to record perceptions of multiple stressors and their interactions with climate and related livelihood risks. The data on multiple stressors were collected in FGD through the finger-raising method (Mundry et al. 2000), and individual responses against multiple stressors were recorded. The weighted score was developed using these responses from across the three study villages. Questions on ‘farm management’ were further confirmed with key informants in FGDs. Supplementary field notes and photographs of specific climate events, alongside transect walks, complemented the qualitative data-set. Audio recordings were made for longer discussions on types of social groups and their socio-economic and major agricultural activities, associated risks, historical changes in land-use patterns and narratives on how changes in common property resources regimes have impacted different groups of farmers.

Triangulation of Data

After the first round of data collection, three follow-up visits were made to the sampled villages to triangulate the qualitative and quantitative data (Antwi-Agyei et al. 2013). This included discussions with key informants and available interviewees on the pattern of results of compounding stressors (collected through FGD) and climate variability, and perceived risks (collected through interviews). This exercise was a productive, iterative process, in terms of identifying circumstances surrounding the key phenomenon of climate and its contextual factors, converging central characteristics with better interpretations and omitting unacceptable points to improve the trustworthiness of results (Lambert and Loiselle 2008).

Data Analysis

Interview data were entered into spreadsheets and frequencies and scores were calculated under different categories, effectively translating some of the qualitative information into quantitative information. Using the STAR statistical packages (version 2.0.1) (IRRI 2013), score and rank values for perceptions and livelihood risks were calculated and tested for their significance applying the Wilcoxon matched-pairs sign-rank test. On the basis of highest to lowest score generated using climate and risk-related variables, ranks were estimated for the corresponding variables. First, statements (variables) relating to perceptions about climate variability and livelihood risks having the highest score were calculated and assigned rank one according to the Wilcoxon value. This highest score obtained by a statement was then applied as a base value to compare with other subsequent statements until they appeared as significant in the Wilcoxon test. The statements between two significant values were designated to a particular priority group (PG). The score of every next significant statement was considered as a base value for comparing subsequent statement(s) to find out the next PG. For example, (i) perception score of farmers on erratic rainfall was compared with statements: (ii) timing of winter onset postponed, (iii) total duration of summer increased, (iv) duration of rainy season decreased and (v) number of rainy days decreased. The statements were found to be non-significant with each other, and differed significantly from next statement: ‘duration of winter decreased’. Thus, the first set of five statements was treated as PG one.

The major multiple stressors were grouped and ranked in order of their severity by running a Mann–Whitney U-test to analyse their significance in relation to livelihood risks. Multiple stressors data were analysed using additive percentages and ranked in STAR. The compounded growth rate was calculated by fitting the exponential function as: Yt = ABt, where Log Yt = Log A + t Log B + e. Yt = minimum support prices of the crops in the year t, A = intercept, B = regression coefficient, t = time variable in years (1 to n), n = number of observation and compound growth rate (r) = (antilog of B − 1) × 100 (see details in Online Resource 1, p. 3) in a spreadsheet. Monthly mean rainfall data (mm) for the period 1901–2014 (>100 years) and 2000–2014 (15 years) were analysed in Excel (2010) considering percentages, means, standard deviations and coefficients of variation. Qualitative data were thematically categorised and cross-checked with quantitative data, particularly on perceived climate variability, multiple stressors and livelihood risks (Stringer et al. 2017). This enabled complementary patterns to be identified in the characteristics of major variables (Antwi-Agyei et al. 2013). We coded qualitative data using content analysis, and again, the major themes that emerged were analysed for patterns in major characteristics of stressors and associated risks to farmers (Antwi-Agyei et al. 2013). Doubts were discussed with key informants by phone or using social networking media such as WhatsApp. Finally, results of the study were presented to farmers in each village to gain feedback and validate the findings.

Results

Perceptions of Climate Variability

The Wilcoxon test indicated that farmers within PG 1 (deep black, Table 2) perceived that rainfall had become erratic over the period 2000–2015; that there were alterations in the onset and duration of different seasons and a decrease in the number of rainy days. Farmers within PG 2 (light black colour) perceived that while the duration of winter had significantly (p = 0.05) decreased, there were visible changes in local weather as evidenced by early onset of summers and increasing frequency of drought events. In particular, farmers experienced extended dry-spells and droughts in 2000–2002, 2009 and 2012, while severe droughts were experienced in 2013 and 2014 (Table 3, Online Resource 1), but these were not recorded by planning and developmental agencies. Farmers within PG 3 opined that the frequency of rainstorms, flash floods and extended dry-spells had increased. Farmers also perceived rainstorms Phailin (2013) and Hudhud (2014), which were not reported in secondary data (Table 3, Online Resource 1), and farmers found it increasingly difficult to predict the weather using traditional indicators.

Table 2 Prioritisation of farmers’ perception about climate variability with Wilcoxon matched-pairs signed ranks test
Full size table

Despite increased uncertainty, farmers still depend on 22 bio-meteorological indicators (Table 4, Online Resource 1), and knowledge of clouds and winds to predict local weather and rainfall patterns. For example, unseasonal/untimely appearance of insect pests is considered to be an indication of higher atmospheric humidity in otherwise dry months (Fig. 1a–c, Online Resource 2). Poor access to weather forecasts from formal sources further enhances uncertainty and greatly reduces farmers’ choices in deciding on adaptation strategies, as they first need to know to what they are adapting.

Farmers within PG 4 and 5 (indicated by the lightest black colour) perceived alterations in the occurrence of ‘loo’ (hot winds blowing during May–June), excess rains (but without any adverse impacts) and drizzling rains over the 30-year period (Table 2). Farmers considered ‘loo’ to be a reliable indicator of a ‘good monsoon’ (i.e., sufficient and evenly distributed rainfall). Drizzling rains (sawan ki jhadi, low intensity rains during August), perceived to be critical for the growth and productivity of rice, and in field preparation for Rabi (winter) season crops, water harvesting for irrigation and weed decomposition, have now become rare. The importance of drizzling rain is reflected in a local folktale

‘Yadi purva aur uttara baras jati hai, to kisan ko pure saal khushal kar jati hai’….

[Drizzling rains in ashlesha (rainwater is considered to be of average quality) and magha (good quality) constellations (in August) are important for the year-round happiness of the farmers] (Key informant: Pujari and Mahajan, April 2014).

The distribution and amount of monsoonal rainfall during mid-June to the end of September determine the success of year-round agricultural activities. The majority (62.4%) of farmers perceived that rainfall had become ‘erratic’, while 26.6% perceived ‘less rainfall’ and 5.0% perceived ‘excess’ rainfall between 2000 and 2015 (Fig. 2, Online Resource 2). A good harvest of two local millets is virtually synonymous with a good monsoon as summarised in the following folktale:

‘Sanwa, sathi (bhandai) 60 din, barkha pawe raat din’…

[If sanwa (Echinochloa frumentacea) and bhandai (rainfed paddy variety) receive even modest but continuous rains during July–August, they will mature within 60 days (Key informant: Mahajan, August 2014)].

Evenly distributed rainfall is considered to be a boon for rice and other Kharif season crops. Failing to receive such rainfall at critical stages of the crop cycle for rice (especially at transplantation to grain setting stage), is perceived to be a drought. Consequently, we analysed secondary climate data and found that an erratic trend in monsoonal rainfall was observed during 2005–2014, a period that witnessed 5 excess rainfall years (33.33%, e.g., year 2005 and 2007 received heavy rains) and 9 deficient rainfall years (60.0%) (Table 5, Online Resource 1). Rainfall received during this period varied from very low (381.2 mm in 2014) to very high (1351.6 mm in 2007). The coefficient of variation (%) also increased during this period as compared with 1975–2014, particularly in June (84.95%) and September (66.33%). The critical months, during which normal and even distribution of rains is a precondition for the better growth of rice, have witnessed decreases in rainfall; e.g., 20.47% in July (period of initial rice growth), 9.07% in August (vegetative growth and tillering) and 10.88% in September (panicle initiation) (Table 5.1, Online Resource 1). This trend, and even recent systematic efforts to record monsoonal anomalies (for issuing agroadvisories to farmers) (Table 5.2, Online Resource 1) were similar to farmers’ perceptions.

Perceived Climate Variability and Related Livelihoods Risks

Wilcoxon signed-rank test results revealed that with the highest rank score and PG 1, farmers perceived that overall livelihood risks had increased due to climate variability, and that this risk was further compounded by ecological (e.g., high soil pH) and anthropogenic stressors (market, institutional and policy, technological and social factors) (Table 3). Incidences of diseases and insect pests were perceived to have increased with the use of agrochemicals over the past 20 years resulting in higher costs of cultivation. As a consequence, farmers’ dependence on external resources had increased.

Table 3 Prioritisation of climate variability led agriculture and livelihoods risks as perceived by farmers with Wilcoxon matched-pairs signed ranks test
Full size table

Poor groundwater recharge and physical changes in aquatic bodies ascribed to reduced drizzling rains were perceived as direct consequences of climate variability. Drying of surface water bodies had increased the livelihood risks of the Bhar community (who mostly reside around the wetlands and possess sodic lands known to be less productive than normal soils). The Bhar community was traditionally dependent on a biodiverse fish catch of about 25 species in the 1980s but this reduced to 4-5 species at the time of data collection, while the Yadav community was primarily dependent on community pond water for livestock. Farmers with PG 2 opined that frequent anomalies in weather (seasonal cycles, rains and heat stress) had necessitated (significant, p = 0.05) frequent seed replacement, as home grown seeds were more susceptible to diseases and insect pest incidence and did not yield well when used successively for more than 3 years. Consequently, farmers perceive the genetic vigour of rice and wheat varieties, which have been reduced, necessitating their replacement every 3–4 years. Erratic rainfall and increases in the maximum (by 0.10 °C; CV 1.14%) and minimum temperatures (by 0.15 °C; CV 2.13%) (Table 6, Online Resource 1) over 30 years (1972–2002) seem to have altered the micro-climate.

Resource use efficiency had also decreased, compelling many small and poor farmers to migrate to cities for more reliable income-generating jobs (Table 3). As noted with PG 3, frequent alternate wetting and drying (i.e., flash floods followed by droughts/extended dry-spells) had increased the risk of adverse impact from sodic soils (significance p = 0.05) naturally found in the study area. Migration of agricultural labourers to cities had further compounded farmers’ risks as they are now required to pay higher wages than 10–15 years previously. Heat stress had also lowered the efficiency of farm labourers per day (PG 3, Table 3). For example, during the 1990s, one labourer would uproot six panji (1 panji = 5 bundles of rice seedlings) per day; this had decreased to three at the time of data collection. Currently, about 44 labourers are required to transplant rice in 1.0 ha, which is almost twofold higher than in the early 2000s when only 24 labourers were needed. Results also indicated that climate variability, along with socio-economic and policy factors, had significantly reduced crop yields and had accelerated the loss of local agrobiodiversity (p = 0.05).

Multiple Stressors

Climate Variability

The additive percentage with rank analysis indicated that climatic stressors were predominant among the multiple stressors impacting farmers’ livelihoods (Table 4). Results on individual sub-climatic factors revealed that reduced numbers of rainy days (23.08% response; significant at Mann–Whitney U p = 0.01) and other rainfall-related anomalies were perceived as sub-stressors by 35.58% farmers. Changes in seasonal cycles (21.15% response; significant at p = 0.01) and sudden changes in weather patterns (20.19% response) were further observed (Table 4; see also Table 7, Online Resource 1).

Table 4 Multiple stressors impacting agriculture-based livelihoods as perceived by farmers
Full size table

Compounding Socio-economic and Market Stressors

Rising costs of cultivation were perceived (23.23% response) (significant) as the most substantial economic stressor (Table 5). For example, during 2014 the cost of rice cultivation ranged from Rs. 30,000 to 35,000 ha−1 and that of wheat from Rs. 20,000 to 25,000 ha−1 as compared with Rs. 10,000–12,000 and Rs. 7000–10,000 ha−1, respectively in 2000. A decadal (2005–2015) data trend indicated costs of cultivation increased by about 3.2 times in rice and wheat crops, while MSP increased by only two and half times in both the crops with a compounded annual growth rate of 9.82 in rice and 8.36 in wheat (MoA&FW 2015) (Fig. 3, Online Resource 2).

Table 5 Results from Mann–Whitney U statistics on multiple stressors impacting farmers’ livelihoods
Full size table

Uncertain and volatile market prices (perceived by 20.20%), unorganised markets and limited financial support from the government (19.19% response for each) (significant at p = 0.01) have reduced profit margins. Farmers reported that during extended dry-spells/drought years, as in 2012, crop productivity declined by 30–40%. Inadequate arrangements by the state government for rice procurement, coupled with poor market infrastructure, compelled farmers to distress sale produce through middlemen at considerably lower prices (e.g., Rs. 600–700 and Rs. 1000–1200 q−1 for coarse grained rice against the fixed minimum support price of Rs. 1250). Low income from rice crops also adversely affects the sowing of ensuing Rabi crops as farmers usually purchase seeds, fertilisers and other inputs after rice sales.

Farmers perceived that labour crises (25.58%) and low income (22.09%) (district average annual income Rs. 9859 year−1, Tripathi 2014) have increased their difficulties in undertaking their cropping activities in a timely way (Table 4). During rice transplanting and harvest seasons, reduced labour availability compels farmers to spend more money (Rs. 140–150 per half day compared with average wages of Rs. 100) for the same duration of work during other times of the year. Some farmers leased out their lands to those households with more family labourers, and more financial capital, to reduce risks caused by labour scarcity. Social marginalisation (22.09%), declining interest of youth in farming (20.93%) and unproductive education (9.5%) (all significant at p = 0.01) reduce the prospects of gainful employment.

Compounding with Institutional- and Policy-Related Stressors

Farmers lacked access to subsidised inputs such as micro-irrigation equipment, while electricity and labour supply were limited (perceived by 24.2%). Despite contingency plans being developed by the relevant institutions, improved seeds and other inputs did not reach farmers, revealing a major institutional stressor. Most farmers had no knowledge of government plans and schemes. The Mahatma Gandhi National Rural Employment Guarantee Act (MGNREGA) and Public Distribution System (PDS) have further increased the risks as perceived by 22.7% of farmers (policy stress) (Table 4). MGNREGA has led to labour shortages affecting agricultural activities over the last 10 years (2005–2015) (Fig. 4, Online Resource 2). Those with extended family, having more family labourers and practicing adhiya and rai (lease systems), had relatively fewer perceived risks associated with labour availability. It also emerged that labour demands were especially high during rice transplanting and erratic rainfall/unreliable water supplies tended to further accentuate the problem. In dry years, farmers were compelled to purchase irrigation water (Rs. 50 h−1). As most of the tube wells were electrically operated, extended power cuts strengthened the bargaining power of labourers who usually demanded relatively more wages. Under the PDS scheme, economically weaker families are entitled to 35 kg of rice and wheat each per month at nominal prices (Rs. 3 and 2 kg−1), respectively, and which has also increased the lack of interest among labourers to work on farmers’ fields (policy stress), as reported by 90.0% of farmers.

During adverse situations, small and marginal communities sustain their livelihoods by accessing common property resources. In the past, marginal farmers, like the Bhar community, freely accessed the Badaila lake and other surface water bodies to harvest wild rice, as well as to grow local rice varieties and catch fish for consumption and local sale. The natural drainage water, flowing from Badaila lake, was utilised for irrigation. From the 1990s onwards, a top-down management regime for common property resources led to changes in structures and functioning of these resources. Currently, the Village Panchayat controls such resources (following norms prescribed by the State Government), which are auctioned to private contractor(s) to generate higher revenues. Badaila lake and village ponds were leased out to the highest bidder (private contractors) for a 1-year period. This created a ‘tragedy of the commons’ situation. The Bhar community faced hardships (policy stress) as they could only access certain parts of the lake for fishing and the collection of other aquatic resources (wild rice, lotus, vegetables, etc.); the remainder was under control of the contractor, and any encroachment may have invited legal action.

The top-down approach to policy implementation by the State Agriculture Department, often overlooking on-the-ground realities and farmers’ needs, had also significantly (p = 0.01) enhanced farmers’ risks (21.7%). Most of the intended policies/inputs did not reach farmers in time. For example, district KVK (KrishiVigyan Kendra) conducting Front Line Demonstrations (FLDs) in wheat, rice and other crops had limited reach. Such FLDs mostly benefit only a few large scale farmers who are in regular contact with KVK and able to manage the required inputs to adopt the new varieties.

Compounding with Ecological Stressors

Due to erratic rainfall, groundwater had declined (22.8% response) between 2000 and 2014 (Fig. 5, Online Resource 2). Abandonment of conventional open-wells and changes in structures and functioning of community ponds (perceived by 21.7%) which recharge groundwater (Table 4) further aggravated the problem. For example, in Sonapur village, we recorded almost no water storage in seven community ponds and natural water courses connected to the Badaila lake due to low rainfall and inappropriate anthropogenic activities (e.g., encroachment, inappropriate modifications in the form and shape of village ponds by Panchayat without considering traditional ecological knowledge). Landscape modifications and blocking of natural drains had altered the hydrological balance resulting in rapid run-off to low lying areas, depriving upland crops of water. Farmers experienced that during rice transplanting, the water table often drops below 10.8 m, especially during drought years or delayed monsoon. Low rainfall and water pumping for irrigation not only exacerbate groundwater depletion, but also increase energy use (increased irrigation hours) and irrigation costs (cf. Redfern et al. 2012).

Changes in land-use patterns, i.e., replacement of millets, local vegetables, oilseeds and pulses with a rice–wheat system, has led to over-withdrawal of groundwater and severe erosion of local agrobiodiversity, compounding risks (Table 4). Rich agrobiodiversity is considered key to the well-being of economically weaker and socially marginal groups who opined that although climate variability (e.g., reduced rainfall) has led to agrobiodiversity loss, land-use changes from a rainfed system (e.g., short duration upland paddy and local pearl millet, maize and redgram and black gram) to an irrigated rice–wheat system, seemed to have inflicted more damage. Due to assured availability of irrigation water, the majority of farmers, with relatively bigger land holdings and more resources, had switched to improved varieties of wheat and paddy crops, which also need more inputs, increasing dependence on external resources. Farmers perceived that land-use change (19.6%) and soil sodicity (soil pH 8.8–9.2) (18.5%) (both significant at p = 0.01) caused a decrease in soil quality and fertility, respectively. Sodic soils also have a degraded structure and low infiltration capacity resulting in prolonged submergence of and damage to crops even during normal rainfall years (Table 4). Farmers perceived that with concomitant impact of green revolution practices that started in the 1960s (e.g. mono-cropping with intensive use of inputs), mixed and diversified cropping systems, sheep and cow herding in crop fields, collective cropping and resource sharing mechanisms are now become virtually non-existent. Such changes caused many traditional adaptive practices and related social institutions, helpful in degraded sodic lands, to disappear.

Compounding with Technological Stressors

Farmers had limited access to improved agricultural technologies (27.7%) attributed to poor extension (26.5%) (significant at p = 0.01). Technologies in rice, wheat and other crops being adopted by farmers are less suited to farmers’ needs, and specific recommendations for niche problems are unavailable. For example, most farmers are unaware of salt-tolerant varieties of rice (CSR-43 and CSR-36) and wheat (KRL-210 and KRL-213) (24.0%, significant at p = 0.05). Private input dealers often provide genetically impure seeds and misleading recommendations as to appropriate practices according to farmers. Further, gypsum- and pyrite-based technologies implemented by the state government to reclaim sodic soils rarely reached farmers; yet, technologies to circumvent soil-related stressors like sodicity and erratic rainfall are perceived to be absolutely essential (significant) (21.7%).

A summary of analysis of pooled stressor data indicated that although overall climate variability was found to be one of the major significant stress (Table 5), institutional and policy stress (p < 0.000), social and technological stress (p < 0.000), economic and market stress (p < 0.01) and ecological stress (p < 0.02), in that order, were other stressors that have significantly influenced farmers’ risk perceptions.

Discussion

Climatic Stressors

The results revealed that farmers perceived an increase in erratic rainfall, decreases in the duration of rainy season and number of rainy days, and changes to seasonal cycles (Table 2), similar to the meteorological data. However, in the case of extreme events including recent rainstorms (MoA&FW 2016), the different knowledges diverged. Such differences may be due to meteorological data collection approaches, and farmers’ experiences being spatially distant (cf: Simelton et al. 2013). Sometimes intense rain or rainstorms are a localised phenomenon, not captured in observations due to methodological issues (Seneviratne et al 2012), and therefore farmers respond to unrecorded events (Callo-Concha 2018). Although the study district is assessed to be occasionally drought prone by developmental agencies (Table 3, Online Resource 1) (MoA&FW 2016), farmers perceived a significant increase in the frequency of extended dry-spells and droughts during the study period. This may be based on their past and recent exposure to moderate and severe droughts. These micro-level perspectives tend not to be incorporated into policy making at higher levels (O’Brien et al. 2004; Antwi-Agyei et al. 2017), resulting in climate policy and planning that do not reflect the realities experienced (England et al. 2018).

Interestingly, the farmers’ perceptions were entirely different from formal institutions (MoA&FW 2016) regarding normal rainfall. Farmers considered rainfall to be normal when it was evenly distributed throughout the season to support the optimum water requirements of the rice crop during its different growth stages; thus they contextualised climatic phenomena in reference to agronomic attributes. The IMD considers a normal rainfall year to be when rain is 98–104% of long-period average (LPA, based on 50 years of data); below normal rain if LPA is 90–96%; and deficient if LPA is below 90%. With these definitions, the droughts experienced may be masked. Previous studies have also shown that Indian farmers (including those of the study region) are increasingly facing drought impacts (Das 2010; Nambiar 2016). These observations point to a mismatch between micro-level realities and formal definition of climatic phenomena; distorting social learning trajectories (Antwi-Agyei et al. 2017) and prompting non-holistic and maladaptive adaptation strategies that eventually put the farmers in a ‘lock-in condition’, making it more difficult for them to move onto alternative adaptation pathways (Antwi-Agyei et al. 2018).

Other Stressors Compounding Climatic Risks

Overall, climate variability risks were compounded by economic, market and institutional stressors. Farmers’ perceptions of weather events are often shaped by the crops they grow and changes in resource and land-use patterns (Slegers 2008; Adimassu and Kessler 2016), infrastructure support (Niles and Mueller 2016), and socio-economic and political factors (cf: Meze-Hausken 2004), which combine to shape uncertainty over agricultural management and associated livelihoods. For example, although groundwater is a reliable source of irrigation for farmers in the study area, poor electricity supply and higher diesel costs, may further amplify perceived risks, especially during extended dry-spells and/or drought (Udmale et al. 2014). In addition, market prices for rice tend to be lower after harvest (20–30%), irrespective of climatic conditions. But, to meet multifarious needs, farmers (particularly Bhar and Yadav), have to sell their produce at these prices because they lack storage capacity, and need to repay debts to input dealers. A recent assessment has indicated that climate change can impact farmers’ livelihoods by reducing income (15–20%), exacerbating an already difficult situation for many farmers. Consequently, institutions need to provide more cost effective and market-smart adaptive strategies (GOI 2018).

Risk perceptions of the farming communities often vary according to their resource endowments and the local agro-ecological conditions. Despite that some of the large landholders had recently switched from mixed farming to rice–wheat mono-cropping systems, they seem less prone to risks from escalating input costs and market volatilities. Contrarily, small landholders depending more on livestock (e.g., Yadav and Bhar) were likely to be adversely affected by such changes. We also noticed that farmers with relatively productive soils (i.e., less sodic) and assured access to tube-well water for the rice crop, did not perceive drought or delayed rainfall to be a major limiting factor to crop production. However, poor access to or virtual lack of these enablers remained a serious concern to the livelihood security of marginal famers (i.e., Bhar community) who rely more on common property resources. There is convincing evidence that social–ecological and institutional factors (Callo-Concha 2018) play a critical role in shaping farmers’ perceptions of stressors. Two of our study communities who rely more on common property resources, seemed to possess better collective knowledge and community institutions for informed decision making and adaptation to the aforementioned stressors (Singh et al. 2016; McNamara et al. 2020; Wilson et al. 2020).

Another key finding of this study was that institutional policies and recommendations prescribed for adapting to the stressors were hardly suited to local needs. For example, policy thrusts on promoting farm diversification, growers’ associations, and improving agri-market infrastructure (MoA&FW 2019) did not elicit an enthusiastic response from farmers. Similarly, generic indicator-based crop insurance policies overlooked the real local needs and did not gain traction (OECD 2018). These discrepancies prevented farmers from accessing and employing external resources and recommendations as means to adaptation. This problem can be ascribed to poor coherence between farmers’ actual needs and institutional policies (Birthal et al. 2007) and insufficient attention being paid to the local socio-cultural and biophysical conditions (cf: Nelson 2011; Donatti et al. 2019). Consequently, farmers continued to rely heavily on local resources, including the use of proven local indicators for weather prediction.

Similarly, MGNREGA compounded farmers’ risks by creating agricultural labour shortages (policy led social stress) (RBI 2018), leading farmers to depend on costly technologies funded by increasing debt (cf: Bhargava 2014). Such debts extend forward (Antwi-Agyei et al. 2018), affecting both current and future livelihood risks (Agrawal 2008). Relevant developmental agencies could execute their policy-centric activities through Village Panchayat instead, improving coherence with farmers’ activities, while scheduling rural development activities outside agricultural peak activities may provide win-wins here (FICCI 2015). The PDS policy increased the area under rice–wheat cropping due to assured returns, decreased conservation and cultivation of local varieties (Sahai 2011; Pingali 2012) and increased dependence of farmers on external resources and other essential food items like pulses, edible oil, fish, etc., and can be treated as a policy stressor. Notwithstanding, the PDS and MSP policies inculcate profit-seeking attitudes among farmers, simultaneously eroding their risk buffering capacity (Agrawal 2008; Pingali 2012). Although the NAPCC (NAPCC 2008) and Sustainable Agriculture Policy (MoA&FW 2010) have attempted to incorporate smallholder farmers’ perspectives (MoA 2007), multiple-stress-led risks were poorly mainstreamed in policy (Dev 2012).

Risks are magnified when natural and/or anthropogenic factors diminish the functional ecosystem services of common property resources. Decreases in total annual rainfall (Table 5.1 and 5.2, Online Resource 1) and increased frequency of droughts have reduced availability of food resources from common property resources with wide-ranging ramifications for the livelihood security (Singh et al. 2019). While structural changes in the management of and access to the common property resources were considered prior to implementing the National Food Security Mission 2007 in India, knowledge, technology and capacity building supports under NFSM seem to have benefitted the resource rich farmers most (Manjunatha and Kumar 2015). Therefore, more needs to be done to reframe current policies to support management of common property resources such that risks for farmers depending heavily on such resources for subsistence are not amplified (Agrawal 2003; Singh et al. 2012).

Way Forward for Social Learning and Collective Action

Our key findings provide many relevant cues to develop efficient and locally compatible adaptation policies to deal with the various stressors facing farming communities. Future policies should consider that both individual wisdom and collective experiences are equally important and together constitute a social construct that ultimately shapes farmers’ perceptions of multiple risks (Riedlinger and Berkes 2001). Such constructs, in addition to providing key insights to better capture the relationships among stressors, risks and adaptation measures (Ison et al. 2013), hold enormous utility to formulating shared responsibilities and action plans for effective adaptation (Baud et al. 2011). This would also enable mainstreaming of local knowledge into future adaptation policies on risk management (Ison et al. 2013). However, in order to achieve this goal, collaborative learning (Paschen, Ison 2014) remains a prerequisite to narrowing down asymmetrical balances in the entire knowledge generation process (Klenk and Meehan 2015). This can be pursued via a participatory approach right from the beginning: identifying the major stressors, prioritising the adaptation interventions and monitoring the intended outcomes (Clements et al. 2011). From a practical utility standpoint, the generic weather advisory issued to farmers barely reduces the local-specific uncertainties and risks in crop production (MoA 2007), and local communities with poor adaptive capacity remain reluctant to switch to externally imposed adaptation strategies (McNamara et al. 2020). These inadequacies can be addressed by designing and translating knowledge into more tangible and usable forms, and ensuring shared decisions (Clements et al. 2011). Such inclusive and collective strategies would better enable smallholder farmers’ to cope with multiple risks in a holistic manner (Kanji and Greenwood 2001; McNamara et al. 2020).

Conclusion and Lessons

This research aimed to understand farmers’ perceptions of climate variability and other stressors in relation to the livelihood risks. In rural eastern UP we found that climate variability-induced risks have increased over time and are compounded by ecological, socio-economic and techno-political stressors that remain unaddressed (Fig. 2). Although farmers’ perceptions of rainfall patterns were similar to the meteorological data from formal sources, they differed considerably from the formal definition of droughts and rainstorms, due partly to their rather localised impacts, and partly to the compounding effects of non-climatic stressors. Poor access to weather forecasts compels farmers to rely on local indicators for agricultural planning but is increasingly becoming unpredictable, and enhancing risks. The adverse impacts of climate-induced anomalies, changes in common property resource management and land/water degradation were more pronounced among the most marginalised farmers.

Fig. 2
figure 2

Identified significant climate and other multiple stressors (in order of severity) impacting farmers’ livelihood. Note. Identification of stressors 1–6 is based on the pooled data from the Wilcoxon matched-pairs signedand Mann-Whitney U test

Full size image

We found that differential risk perception among stakeholders can create difficulties in distinguishing ‘what is’ and ‘what ought to be’ with regard to climate-induced risk assessment, and designing micro-scale adaptation strategies. We highlight the need for participatory dialogue between farmers and policy makers to reconcile differences and to frame commonly agreed adaptation pathways that avoid lock-ins to maladaptive practices. Most adaptation policies for Indian farmers focussed on adaptation to climatic stressors are still largely based on top-down approaches, and lack integration of farmers’ perspectives. While several policies have been launched in India to enhance the adaptive capacity of smallholder farmers, a wide gap still exists, resulting in ineffective risk reduction. Robust coherence between agriculture and rural development policies, as well as infusion of a plural perspective into both top-down and bottom-up approaches are needed if farming livelihoods are to become more resilient.

Further research could seek to classify the study farmers into different categories based on their resource endowments, considering e.g., land holding, income and ability to access external resources to understand how farmers’ perceptions of and responses to multiple risks varies across these groups. This could help to further nuance the different supports that different kinds of farmers need in order to adapt (cf. Stringer et al. 2020). Based on the key results of the study, we propose the following policy recommendations:

  1. (1)

    Integrated studies on the perceptions of farmers, experts and policy makers on multiple stressors are needed, juxtaposing them with farmers’ livelihood risks, to better capture micro-level insights to inform more nuanced policy actions.

  2. (2)

    Future policies on agricultural adaptation in general, and those aimed at improving adaptation in marginal areas facing multiple risks in particular, should integrate local situations into formal frameworks for early and wide adoption by the target farming communities.

  3. (3)

    Policies dealing specifically with climate change adaptation should consider the community’s knowledge of climatic changes, and their resource endowments. Integration of locally proven indicators, after further participatory validation, for weather prediction with formal knowledge could enhance farmers’ adaptive capacity in dealing with climate-related stressors.

  4. (4)

    Climate-related information communicated to the farmers should be in a usable form to enable effective adaptation decisions for livelihood security.