1 Introduction

Drought has become a typical extreme natural phenomenon in the context of global climate change. As the world’s costliest natural disaster (Chen and Yang 2013), drought has caused an average of 6–8 billion dollars in global damages annually (Vasiliades and Loukas 2009). In recent years, the frequencies, intensities, durations, and losses of droughts are increasing due to not only the impacts of global climate change and ecological environment destruction, but also the growing water consumption caused by rising population, developing economy, and rapid urbanization process (Yuan et al. 2013).

China is a drought-prone country (He et al. 2011). Reports of droughts in China are widely scattered throughout the national meteorological literatures, constituting one of the rich observational records ranging from 1470 to present in the world (Chen and Yang 2013). Like many developing countries, rural economies in China are highly dependent on agriculture, and agriculture is more vulnerable to drought impacts (Osbahr et al. 2008). Drought can cause shortage in both surface and ground water sources and consequently has damaging effects on crop production and rearing of livestock (Ashraf et al. 2014). The statistical data in China show that the average area affected by drought is approximately 21.59 million ha annually and the annual grain losses due to drought are up to 10 billion kg (He et al. 2011). Drought is one of the major limiting factors of Chinese agriculture (Xu et al. 2013).

Adaptation is one of the policy options for reducing the potential negative impacts of drought hazard on agriculture (IPCC 2007; Ozor et al. 2012). Given the consequences and pervasiveness of drought, it is of importance to help local rural communities by promoting adaptation measures and make them more susceptible to extreme climate events such as drought and associated climatic variations (Ashraf et al. 2014; Obayelu et al. 2014).

An in-depth knowledge of local farmers’ perceptions of climate change their ongoing adaptation measures toward the drought hazard along with the factors which might influence farmers’ choices is important (Bryan et al. 2013), which could not only enhance policy measures in order to mitigate the negative impacts of drought but also increases the farm households’ capability to cope and adapt better with the challenges posed by drought to their agro-based livelihood (Ashraf et al. 2014). There are a growing number of studies of farm-level adaptation strategies and their determinants (Kabubo-Mariara 2008; Deressa et al. 2009; Solomon et al. 2014; Bryan et al. 2013; Sarker et al. 2013; Ashraf et al. 2014). However, due to different climatic conditions, farm types and other conditions such as political, economic and institutional factors (Hisali et al. 2011), adaptation choices in agriculture vary across countries and areas. Therefore, country-specific or area-specific studies of adaptation measures to the drought hazard are required (Sarker et al. 2013). The objectives of this study are to identify farmers’ ongoing adaption measures toward the drought hazard and to examine factors influencing their decisions to adapt in Yongqiao District. To the best of our knowledge, this is the first study that primarily deals with drought adaptation methods and factors affecting the choices at farm household level in China.

The remainder of this paper is organized as follows. Section 2 describes the study area. Section 3 presents data collection and the analytical framework. Section 4 reports the empirical results and discussions. The final section provides a concluding summary of our findings.

2 Study area

Yongqiao District was selected as the study area. It is the biggest county-level district in China, located in the north of Anhui Province. Geographically, district extends from 33°17′ to 34°05′ north latitudes and from 116°05′ to 117°26′ east longitudes. The total area of this district is 2868 sq km. In 2012, the population of the district was 1.88 million, among which agricultural populations accounted for about 78 %. The population density was 656 persons per sq km with an average household size of four persons. The average per capita GDP was approximately CNY 20,059 (USD 3280). Wheat and corn are the two main crops in this area.

Yongqiao is part of the drought-prone region of China (see Fig. 1). The surface of this district is generally undulating, and the soil is very porous with low capacity to retain water. This makes the area vulnerable to drought. The main problems that affect stable crop harvests are heat waves during the crop’s flowering phase in July and August and rainfall shortages. Due to the effects of climate change, farmers may face more risky situations in the future and adaptation is imperative to maintain and improve the productivity of the agricultural sector.

Fig. 1
figure 1

Distribution of agricultural drought hazards (Xu et al. 2013) and location of the study area

Full size image

3 Methods and data collection

3.1 Data collection

To identify and assess current and potential household-level adaptation strategies available to rural communities, data were collected using household surveys.

Two villages in Yongqiao District were purposively selected for this study. The reason behind this selection is that they were the only two villages that were provided a trial for wheat drought weather index insurance, which is an important adaptation measure in this region (Bryan et al. Bryan et al. 2013; Jie et al. 2013). Within each village, respondents were randomly selected. Sample size for each of the villages was proportional to the farming population residing in the particular village. In most cases, we interviewed the head of the household, who typically did the most farm work. If the head of household mainly participated in off-farm work (i.e., leaves the village for work in cities), we interviewed the family member who was the most responsible for farm work.

The final survey was conducted in November, 2013. A total of 220 heads of household were invited, and 200 persons participated in our study. The survey questionnaire consisted of four sections, including: (1) farmers’ perceptions regarding climate change; (2) crop planting and yield; (3) farmers’ actual adaptation measures being taken; and (4) the demographic and socioeconomic data of respondents and their households.

In this study, we define climate change as perceived changes in the average temperature and average rainfall over the last 10 years. Respondents were asked about whether they had perceived any changes in the average temperature and average precipitation over the last 10 years. Respondents were also asked to choose the top two risks to their agricultural production over the past 5 years based on a list in the questionnaire. Six common risks were provided: flood, drought, erratic rainfall, heat waves, insect attack, and bad seeds. Respondents were then asked about their opinions on whether climate change would affect their agricultural yields or not. They were asked to estimate the annual average percentage of crop losses due to bad weather (e.g., drought and high temperature) in the last 5 years and in the next 5 years, respectively.

In the second section, respondents were asked to indicate the total area of the farmland operated by each household, detailed information on the types of crops planted in 2012, and corresponding area for each type of crop.

The third section contained some questions about respondents’ choices of drought adaptation strategies, including planting new crops that are drought tolerant and/or resistant to temperature stresses, adopting new technology for water conservation that can enable more efficient use of the available water, increasing investments in irrigation infrastructure, crop diversification on the same plot or on different plots, and purchasing wheat drought weather index insurance. The adaptation measures used for this study were selected based on focus group discussions.

The last section was to collect the demographic and socioeconomic data of respondents and their households. Category format was used for age, education, and income variables to encourage reluctant respondents to provide more honest information.

3.2 The econometric model

Econometric analysis was done to examine the factors influencing farmers’ adoption of different adaptation measures. Because the adaptation decision is a binary case, to adapt or not to adapt (0,1), binary logit regression models were employed. The reason is that the logit regression model has a number of advantages compared to linear regression and log-linear regression (Ozdemir 2011). Gujarati (2004) provides an excellent theoretical exposition of the model that can be estimated as a probability. Moreover, the binary logit model has been successfully used in some recent studies (Solomon et al. 2014; Goetzke and Islam 2015; Wang et al. 2015). The specification of the empirical model or the reduced form that was estimated is as follows:

$$Y_{i} = \alpha + \beta_{i} \mathop \sum \limits_{i = 1}^{n} X_{i} + \varepsilon_{i}$$

where \(Y_{i}\) is a dichotomous dependent variable (farmer using any adaptation measure or not, specified as yes = 1, 0 = otherwise), \(\alpha\) is the Y-intercept, \(\beta_{i}\) is a set of coefficients to be estimated, and X denotes the set of explanatory variables hypothesized to influence adaptation to drought. \(\varepsilon\) is an error term.

In general, a logistic regression model can be characterized as follows (Hosmer and Lameshow 2010):

$${\text{logit}}\left( p \right) = \log \left( {\frac{p}{1 - p}} \right) = \beta_{0} + \beta_{1} X_{1} + \cdots + \beta_{i} X_{i}$$

Each estimated coefficient is the expected change in the log-odds, log(p/(1 − p)), of the dependent binary choice variable, for a unit increase in the corresponding independent variable, holding the other explanatory variables constant. In the binary logistic regression model, the interpretation of the logistic regression coefficients becomes easier after transforming of the logistic regression coefficient into odds ratios. The odds ratio is the odds of an event occurring in one group compared to the odds of it happening in another group (Goetzke and Islam 2015).

The selection of the explanatory variables in this study is based on the literature and data availability. The explanatory variables that were included in our analysis include the household head’s gender, education level, farmland owned, farming experience, household monthly income and respondents’ perception and knowledge variables. In order to test whether the effects of explanatory variables differ from one adaptation to another, each variable has been included in all equations in the empirical model (Ashraf et al. 2014).

4 Results and discussion

4.1 Demographic profile of the sample

The typical participant in our sample was 40.81 years, with a range from 18 to 77 years. Ninety-five percent of the respondents was married. Approximately 42 % of the sample was men. This could be due in part to the fact that a large proportion of Yongqiao’s young single and male people choose to seek jobs in urban areas instead of farming. The average farming experience was approximately 20 years, with 28 % of the respondents having farming experience of more than or equal to 25 years. Data on education indicate that 2 % of the respondents was illiterate, 22 % had completed their primary education, and 58 and 18 % had attained junior high school and higher education, respectively. The average household size of the respondents was five persons as compared to the average household size (four persons) of the district. The mean farm size was 2.36 acres. The average monthly income of the sampled households was 569 US dollars, which is comparable to the Yongqiao census data.

4.2 Perceptions on climate change

The survey results indicate that 95 % of the respondents had heard about climate change. An overwhelming majority of the farmers perceived an increase in the average temperatures (82 %) and a decrease in the average precipitation (68 %) over the last 10 years. These results are consistent with the real change trend of the average temperature and average rainfall in this region (De et al. 2012).

The results indicate that the first risk to the respondent’s agricultural production in the past 5 years was drought and the second risk was heat waves, which are in line with the real situation. Approximately 97 % of the respondent thought that climate change would affect their agricultural yields. Respondents thought that the probability of incurring crop losses during the upcoming growing season was 29 %. Roughly 96 % said that bad weather (e.g., heavy rain, drought, or high temperature) caused damage to their crops in 2012. The annual average crop loss due to bad weather in the last 5 years was estimated to be 34 %. These results are comparable to the average loss caused by natural disasters in China (Balzer and Hess 2009).

4.3 Household actual adaptation strategies

The actual adaptation strategies adopted by farmers in the study area in cushioning the effects/impacts of the drought hazard are shown in Table 1. It can be seen that the most commonly used adaptation practices were planting new drought-tolerant crop varieties and increasing investments in irrigation infrastructure. More than half of the respondents purchased wheat drought weather index insurance. Approximately half of the respondents stated they had adopted new technologies on water conservation.

Table 1 Farmers’ actual adaptation strategies
Full size table

4.4 Determinants influencing farmers’ adaptation measures

The dependent variable in the empirical estimation is the choice of an adaptation option from the set of adaptation measures listed in Table 1. Table 2 shows the definitions of the explanatory variables used for our estimation as well as their major statistical values. The explanatory variables for this study include the respondent’s gender, farm size, education level, household income, knowledge on adaptation such as awareness on climate change impacts and estimations on future losses caused by climate change. The results of the regression analysis on the determinants of adaptation are shown in Table 3.

Table 2 List of explanatory variables and descriptive statistics
Full size table
Table 3 Regression results of the binary logit models
Full size table

4.4.1 Planting new crops

From Table 3, the results show that the coefficient on farmers’ household income was positive and significant as expected. These results imply that the farmer with more household income was more likely to plant new drought-tolerant crop varieties than poorer farmers. This finding is understandable because richer farmers would have a higher capacity to buy and plant new crops. Moreover, higher income farmers may have a longer term planning horizon (Deressa et al. 2009).

The coefficient on farm size is negative and significant, suggesting that a farmer with more land is less likely to adopt the adaptation measure of planting new drought-tolerant crop varieties. This is to be expected because planting new drought-tolerant crop varieties would cost more if a farmer has more farmlands.

The two variables associated with the individuals’ perceptions of climate change impacts are positively and significantly related to planting new drought-resistant crop practices. These results suggest that if a farmer perceives his future crop yield would be affected by climate change, or that he has a high estimation on crop yield loss caused by climate change in the future, he would be more likely to plant new drought-resistant crop varieties. An odds ratio corresponding to a particular variable is interpreted as the likelihood of adopting the adaptation measure compared to the omitted category. In this model, for ‘perception,’ the omitted category is the group who did not perceive climate change can affect crop yield. The odds ratio for ‘perception’ is 15.16, suggesting the group who perceived climate change can affect crop yield 15.16 times as likely to be planting new drought-tolerant crop varieties compared to the omitted group. Therefore, the perception of climate problems and the potential benefits of taking action is an important determinant in the adoption of new adaptation measures (Hassan and Nhemachena 2008).

4.4.2 Adopting new technology

Drought not only affects the water supply to crops, but it also makes the crops more vulnerable to pest attacks (Ashraf et al. 2014). Table 3 shows that there are a limited number of factors influencing whether farmers adopted new technologies on water conservation or not. However, there are some noteworthy findings with respect to using new technology decisions.

The variable, male, has results showing that the coefficient is positive and highly significant, suggesting that male-headed households are more likely to adopt new technologies. This is to be expected because men are more informed about new technologies than women (Deressa et al. 2009). In rural China, men generally listen to the radio and television, where new technologies are usually broadcast.

According to Table 3, we find that the coefficient on the variable, years of farming experience of the household head, is negative (−0.03). This variable also shows high statistical significance, at the 1 percent level or better. These results imply that if farmers have more experience in farming activities, they are less likely to adopt new technologies on water conservation.

4.4.3 Increasing investments in irrigation systems

From Table 3, it has been observed that farming experience and household income have positively insignificant influence on farmers’ decisions over increasing investments in irrigation systems during the drought.

Our results show that the coefficient on male is positively and significantly related to the dependent variable. As in the case of technological practices, males can access more agricultural adaptation practice information than females, making them more likely to invest in irrigation systems.

According to Table 3, we find that the variable of the education level of the farmer shows a positive and significant coefficient. This finding may be interpreted as farmers with more education are more likely to increase investments in irrigation infrastructure. This is to be expected, as better education can help farmers to understand information regarding the impact of climate change and adaptation strategies, thus increasing their investments in irrigation systems. Evidence from various sources also indicates that there is a positive relationship between the education level of the household head and his adaptation to climate change (Maddison 2006; Deressa et al. 2009).

4.4.4 Crop diversification

Crop diversification is one of the well-accepted adaptive strategies used by farmers to diffuse risk (Adger et al. 2003; Ashraf et al. 2014). From Table 3, it has been observed that farmers’ farm size, farming experience, education level, and estimations of future crop loss caused by climate change were positively and statistically related to their adaptation choices over crop diversification. This finding suggests that these variables have positive influence on the likelihood of practicing multiple cropping as an adaptation measure to the drought hazard.

4.4.5 Purchasing insurance

Crop insurance programs play an important role in agriculture policy (Wang et al. 2015). Table 3 shows that more factors influence farmers’ decisions on purchasing wheat drought weather index insurance.

First, farm size is positive and significant, showing that a farmer with more farm land would be more willing to buy insurance. This is to be expected, as larger farms will generally face more risks in the future, which should drive greater demand for insurance.

The results of Table 3 show that the number of years on farming activities is positively and significantly related to the insurance adaptation practice, suggesting that a farmer with more farming experience would be more likely to buy insurance.

The regression results show that the coefficient on the level of education of the farmer is positive and significant. According to the estimated odds ratio of ‘Higheduc,’ the high educated group is 2.68 times as likely to buy the weather index insurance compared to the (omitted) low educated group. This is expected as education equips one with a better understanding of the insurance policy details and of how to process information provided by different sources regarding the weather index insurance. Thus, better-educated household heads are more responsive to the crop insurance program.

The household income variable shows a negative and significant coefficient (−0.99). This is a surprising finding, and the expectation is that a richer farmer would be more likely to buy insurance. However, the demand for buying insurance may be influenced by many factors (Wang et al. 2015).

The awareness of future loss caused by climate change is positively and significantly related to the decision of purchasing weather index insurance, showing that farmers who have a high crop loss estimate are more likely to purchase wheat drought weather index insurance. Several studies have found that farmers’ awareness and perceptions positively and significantly affected their decisions to adopt adaptation measures (Araya and Adjaye 2001; Tenge and Hella 2004; Hassan and Nhemachena 2008).

5 Conclusions

This study analyzed actual adaptation choices toward the drought hazard made by farmers in China along with factors influencing their choices of adaptation measures. We found that most farmers in our study area were aware of climate change and they were taking measures to reduce the potential negative impact of drought. The main practices actually followed by farmers were mostly taken in combination with other measures and not alone. The major adaptation options included planting drought-resistant crop varieties, increasing investments in irrigation infrastructure, and purchasing weather index insurance.

The regression results show that factors influencing farmers’ adoption of different adaptation strategies varied. Men were more likely to adopt new technologies on water conservation and to increase investments in irrigation infrastructure. Farmers’ educational levels were positively and significantly related to increasing investments in irrigation infrastructure, crop diversification, and purchasing weather index crop insurance. This implies that better-educated farmers are more likely to adapt better to the drought hazard. Farmers’ perceptions of climate change and its impact were positively associated with their adoption of adaptation measures such as planting new drought-resistant crop varieties and increasing investments in irrigation infrastructure. Other factors explaining farmers’ choices of adaptation measures include farming experience, farm size, and household income. Government policy should target improving the significant determinants outlined above to boost farmers’ adaptation and hence to reduce vulnerability (Sarker et al. 2013).