Adaptation of farmers to climate-change and determinants of their adaptation decisions in South India

Agriculture, with its allied sectors, is the largest source of livelihood in India and now faces a serious challenge from climate-change. In this context, this study investigated, how farmers in India perceive climate-change, what adaptation strategies they practice, and major determinants of their adaptation decisions. Primary data forms the database collected from 400 sample farmers from southern States of India to employ Discriminant model and Multinomial Logit Model (MNLM). The results of Discriminant analysis revealed that off-farm income, income from agriculture and farming experience are the major discriminating variables with largest contributions to motivate the farmers for tackling climate-change. Findings from MNLM revealed that in addition to above three variables, access to climate-change information and education background of the farmers are also the important determinants in adoption of climate-change adaptation strategies viz., crop diversication, integrating crops with livestock, change in the planting date and adoption of soil and water conservation measures. The study highlighted increasing role of Government in the future to safeguard interests of farmers through offering a wide range of institutional, policy and technology support. Further, providing off-farm employment opportunities to the farmers is crucial to sustain their livelihoods, as such activities are less sensitive to climate-change.


Introduction
Environmental changes viz. climate change, land-use change and natural resource degradation have aggravated vulnerability of agricultural production across the countries in the world. Among these, climate change has emerged as the biggest developmental challenge especially for developing countries like India through disrupting the normal socioeconomic settings, particularly of the poor (Sunita, 2009). Its adverse effects are much severe on agricultural sector in affecting both food and nutrition security and sustainable development. So, it is imperative on the part of farmers to face climate change in agriculture through following various adaptation strategies that demand collaborative efforts from different stakeholders. Of course, the major driving force for taking up climatic adaptation strategies comes from farmers' perceptions to tackle the climate change phenomenon.
India experienced a series of droughts ( Figure 1) and the one in 1987 was one of the worst, with an overall rainfall de ciency of 19 per cent, which affected 59-60 per cent of the normal cropped area and a population of 285 million. This was repeated in 2002 when the overall rainfall de ciency for the country as a whole was again 19 per cent. Over 300 million people spread over 18 States are affected by drought along with around 150 million cattle. Food grains production registered an unprecedented steep fall of 29 m. tonnes. After 2002, the drought in 2018 is considered to be second severe one, affecting about 42 per cent of land area and 500 million people (almost 40% of the country's population). With the advent of climate change since 1990s (Sunita, 2009), failed monsoon is the primary reason for frequent droughts in India. Since it is not possible to avoid the adverse impacts of climate change (Figure 2), it is vital to promote adaptation strategies among the farmers to tackle it in their farm elds. Before this, it is essential to analyze their perceptions about climate change adaptation strategies and determinants of the same for their effective implementation. The earlier studies conducted in South Africa (Tshikororo et al, 2021), Ghana (Francis et al, 2021), Ethiopia , Uganda (Nabikolo et al, 2012), Fiji (John, 2008) etc., highlighted that farmers changed their cultivation practices as adaptation strategies in various ways viz., change in cropping calendar, crop varities, machinery for cultivation practices, crop diversi cation, integrating crops with livestock (farming systems approach), soil and water conservation practices etc. Even strategies such as System of Rice Intensi cation (SRI) and micro irrigation were adopted by the farmers to combat water scarcity situation. They also implemented strategies for coping with declining soil productivity through increasing organic manure application, compost making and application, crop rotation, crop residues retention Tshikororo et al, 2021). In India, even the Government also started promoting formation of Farmer-Producer Organizations (FPOs), when a single farmer could not afford adaptation strategies (Naveen et al, 2019). However, the study conducted by Niles et al. (2016) revealed an interesting nding that the farmers' attitude and perception towards climate change do not correlate to their actual adoption. So, it is equally important to analyze the determinants of different climatic adaptation strategies being followed by the farmers besides their perceptions to tackle climate change phenomenon. With this background, this study was focused on better understanding of perceptions and practices followed by the farmers to tackle climate change in four Southern States viz., Andhra Pradesh, Telangana, Tamil Nadu and Karnataka. As no prior research on these lines was conducted earlier in Southern parts of India, this study is certainly a contributing one to highlight farmers' perceptions as one of the major critical elements for tackling climate change and to identify major determinants for practicing various adaptation strategies.
2. Review Of Literature: Francis and Tsunemi (2015) analyzed socio-economic factors that in uence farmers' adaptation to climate change in agriculture. The empirical results of the logistic regression model showed that education, household size, annual household income, access to information, credit and membership of farmer-based organization are the most important factors that in uence farmers' adaptation to climate change. However, the researchers identi ed unpredictability of weather, high farm input cost, lack of access to timely weather information and water resources are the major constraints of farmers' adaptation to climate change.
Abrham et al (2017) analysed the smallholder farmers' climate change adaptation strategies and determinants of their adaptation decisions in the Central Rift Valley of Ethiopia. Findings of the study revealed that the farmers' capacity to choose effective adaptation options is in uenced by household demography, as well as positively by farm size, income, access to markets, access to climate information and extension, and livestock production. So, the researchers suggested the need to support the indigenous adaptation strategies with a wide range of institutional, policy, and technology support. Creating opportunities for non-farm income sources is equally important, as such activities are less sensitive to climate change. Furthermore, providing climate change information, extension services, and creating access to markets are crucial.
Alex et al (2017) analysed the household determinants that contribute to climate change adaptation strategies in the Mount Rwenzori area of South Western Uganda. A Multinomial Logistic Model (MNLM) was used to assess the drivers of farmers' choice for adaptation practices, factors in uencing the choice of adaptation, and barriers. The study concluded use of different crop varieties, tree planting, soil and water conservation, early and late planting, and furrow irrigation are the major adaptation practices. The ndings of Discrete choice model indicated that age of the household head, experience in farming, household size, climate change shocks, land size, use of agricultural inputs, landscape position (location), and crop yield varied signi cantly (p > 0.05) and in uenced farmers' choice of climate change adaptation practices. Researchers concluded that inadequate information on adaptation methods and nancial constraints are the major barriers for adaptation and hence, suggested increasing support from government and other stakeholders. Analysis. The study concluded that formal education, agricultural education, age group, farming experience and off-farm occupation signi cantly contributed towards farmers' perception regarding tackling of climate change.
3. Methodology 3.1. Sampling procedure: This study was conducted in four States of Southern India viz., Andhra Pradesh, Telangana, Tamil Nadu and Karnataka, as they occupy prominent positions in the cultivation of major crops like paddy, maize, groundnut, cotton, chillies, sun ower, tobacco, tomato, banana, cashew, coconut and cardamom. More than half of the Gross Area Sown (GAS) across these states viz., Andhra Pradesh (52%), Telangana (63%), Tamil Nadu (54%) and Karnataka (75%) is under rainfed condition. Further, in all these States, the share of marginal and small farmers in the total number of holdings was more than 80 per cent (2011 census). In Andhra Pradesh (Ananthapuramu, (540 mm), Telangana (Jogulamba Gadwal, 533 mm), Tamil Nadu (Tiruppur, 600.3 mm) and Karnataka (Chitradurga, 507.4 mm) are the major drought prone districts due to scanty (normal) rainfall (India Meteorological Department, 2019).
According to Yamane (1967), the minimum sample size in the study should be: So, this study involved a cross-sectional survey of 400 sample farmers @ 100 random sample from each of the above four districts during 2019-20. Data are collected relating to perceptions of them towards tackling climate change and for identifying the major determinants for climate change adaptation (drought coping) strategies followed by them in the study area. A structured questionnaire was employed among the sample farmers with the assistance from local Agricultural O cers, who interacted directly with the farmers at the local level.
In the context of present study, two groups of farmers were made viz., farmers willing to tackle climate change (Yes = 1) and farmers not willing to tackle climate change (No = 0). As per the survey, 256 farmers are willing and practicing climate change adaptation strategies and remaining 144 farmers are not willing to tackle climate change. Socioeconomic characteristics of sample farmers (Table 1) were hypothesized to contribute to discriminating between the two categories of farmers. it is used to classify farmers into two (or more) mutually exclusive and exhaustive categories or groups based on a set of independent variables. That is, discriminant model is used to distinguish between two categories of farmers for tackling climate change: willing to tackle climate change and non-willing to tackle climate change coded as 1 and 0 respectively.

Empirical
These two possible categories are de ned by number of factors, which simultaneously in uence the farmers' willingness to tackle climate change. In this study, information related to independent variables (Table 1) are used to calculate discriminant score Z for a given farmer as follows: Z i = β 0 + β 1 *X 1 + β 2 *X 2 + β 3 *X 3 + β 4 *X 4 + β 5 *X 5 + + β 6 *X 6 + β 7 *X 7 + ε where, Z is the discriminant score that maximizes the distinction between the two categories. Before running discriminant analysis, it is important that data used must be independent and normally distributed (Khemakhem and Boujelbene, 2015). So, Kolomogorov-Smirnov test was employed to prove the data are normally distributed. Further, multicollinearity among the independent variables was also tested through computing Pearson's correlation matrix. As the highest absolute value of correlation coe cient between each of variable is less than 0.7, multicollinearity problem was ruled out in this study. In the next step, discriminant analysis (direct method) is applied to the sample data.  Table 2 shows the description and expected signs of explanatory variables used in this study. The estimation of MNLM was conducted by normalizing one category, which is named as 'base category'. The adaptation measures were grouped into above four major categories because, farmers used more than one strategy, and the base category was 'No adaptation strategy.' That is, Climate-change adaptation strategy -the dependent variable (Dummy), 4 = crop diversi cation, 3 = integrating crop with livestock, 2 = change planting date, 1 = adoption of soil and water conservation practices and 0 = No adaptation strategy.   To know, whether a signi cant difference exists between the means of two groups, a one way ANOVA is carried out. Each of the predictor variable is treated as a dependent variable and the category of dependent variable as an independent variable and the results are present in Table 5. It is found that the signi cant difference is observed in case of FE, OFFI and AI for which the P-values are less than 5 per cent level of signi cance. However, in case of other predictors, no signi cant difference was observed (P-value > 0.05). Another way of evaluating the performance of the discriminant function is to investigate the eigenvalue and the canonical correlation coe cient (Table 6). An eigenvalue (0.113) indicates the proportion of variance explained ie., a large eigenvalue is associated with a strong function and hence, 100 per cent of the variance was explained. This shows that the function is highly signi cant and potential enough in classifying the categories. The canonical relation is a correlation between the discriminant scores and the levels of the dependent variable. A high canonical correlation coe cient (0.948) indicates a function that discriminates well between two categories of dependent variable and also infers no overlapping among them. Squaring the canonical correlation suggested that 89.8 per cent of the variation in the grouping variable was explained (Halagundegowda, 2017; Nguyen, 2017). As shown through Table 4, unexplained error is 10.2 per cent (Wilks' λ: 0.102). So, out of total variation, 89.8 per cent of the variability is explained by this model (canon corr 2 = 0.898).
So, strong association is detected (canonical correlation = 0.948) between two groups namely, set of all independent variables and two categories of farmers (dependent variable).  Table 7 presents the summary data for the discriminant analysis and the analysis yielded one discriminant function for two categories of climate-change adaptation. The ndings include both unstandardized and standardized discriminant (canonical) function coe cients and they are meant for evaluating the relative contribution of each of the predictor variables as discriminators between two categories. When predictors are measured in different units, the magnitude of an unstandardized coe cient provides little indication of its relative contribution to the discriminant function. So, standardizing the coe cients is necessary, so as to have a common scale of measurement for comparative purposes as all the predictor variables (Kumari, 2017). In the derived function, the sign indicates the direction of the relationship and magnitude indicate the extent of contribution to the group discrimination. It is important to note that the larger the standardized coe cient (b), the larger is the respective variable's unique contribution to the group discrimination (irrespective of the sign of the coe cient). All the predictors except TRG are positively in uencing the discrimination of groups. It is further apparent from the analysis that OFFI (b5 = 0.658), AI (b7 = 0.558) and FE (b1 = 0.517) are the highest discriminating variables with largest contributions.
This result means that appropriate attention should be given towards promoting off-farm employment opportunities, pro tability of agriculture and as well as due recognition to the FE in order to motivate them to practice/implement climate-change adaptation strategies. So, by using the variables and the standardized coe cients, the required discriminant equation (discriminator) is shown below:   loaded on the discriminant function. That is, these predictors were, therefore, not signi cantly associated with climatechange adaptation strategies. Clearly, EC is the weakest predictor and suggests that it is not associated with adaptation strategies, but a function of other un-assessed factors. The group centroids are the averages of the Z values calculated by the estimated model, which can use to evaluate the expected position of the concerned farmers' categories. (Uddin, 2013). As can be seen in Table 10, the centroid of nonwilling category is -0.447 and the centroid of the 'willing' category is 0.252. This implies that if someone's score on the discriminant function is positive (closer to 0.252), then that respondent is probably willing to tackle climate change. On the contrary, if a person's score on the discriminant function is negative (closer to -0.447), then the data probably came from the 'non-willing' category. On calculating the cut score (halfway between the two centroids) ie., -0.097 and if an individual person's score on the discriminant function (calculated by plugging in their scores on predictor variables) is above -0.097, then the respondent is probably from the 'willing' category. On the contrary, if the discriminant function score is below -0.097, then the respondent is probably from the 'non-willing' category.   84%)), FS (on integrating crop with livestock (2.67%)) and TRG (on change in planting date (3.02%)) have exerted signi cant positive in uence on the adoption of climate-change adaptation strategies.

Summary and Conclusions:
Climate-change is considered as one of the biggest threats to global agriculture and India, in particular. So, it is essential for the farmers to plan and resort for different climate-change adaptation strategies to stabilize their annual income and sustain in the farm business. Discriminant analysis revealed that OFFI, AI and FE are the highest discriminating variables with largest contributions. So, these variables are to be given due attention in the study area to motivate the farmers for practicing climate-change adaptation strategies. The ndings of MNLM revealed that FE and AI are the major determinants that contribute towards adoption of selected climate-change adaptation strategies followed by AC, EDU and OFFI. These ndings highlight that the role of Government is crucial in the ensuing years to safeguard the interests of farmers through a wide-range of institutional, policy and technology support. Among the above determinants, creating off-farm employment (income) opportunities to the farmers deserves special mention, as those activities are less sensitive to climate-change Tshikororo, 2021). To conclude, the aspects like linking farmers to markets, improving access to climate-change information, knowledge about various climate-change adaptation strategies (longterm drought proo ng measures) etc., should be included in the existing formal agricultural extension system of the Ministry of Agriculture and Farmers' Welfare and other line Ministries to bene t the farming community.   ROC curve