Farmers’ perception, awareness and adaptation to climate change: evidence from northwest Vietnam Open Access

Vietnam is greatly affected by climate change (ISPONRE, 2009; UNDP and IMHEN, 2015). The average temperature in Vietnam increased by 0.5°C to 0.7°C from 1958 to 2007 (MONRE, 2012). Temperatures are predicted to increase in all areas by 1.1°C to 1.8°C in 2050. Annual rainfall variability is estimated to range from −16 to +36 per cent by 2050. Natural disasters are likely to become more common, and they will adversely affect agriculture. These changes will make sustainable development more difficult. The decline in crop yields under climate change is estimated to be 4.3-8.3 per cent by 2050 in the Mekong River Delta and 7.5-19.1 per cent in the Red River Delta (Yu et al., 2010). A 1 m sea-level rise is estimated to cause a 10 per cent loss in gross domestic product (Vien, 2011; CIEM, 2012). Understanding how farmers perceive climate change is necessary to implement policies promoting successful adaptation. Farmer perceptions are likely to differ among locations. These differences result from non-homogeneous factors, such as education, gender, age, culture, resource endowments and institutional factors (Mertz et al., 2009; Weber, 2010; Piya et al., 2012). Adaptation to climate change will be facilitated by perception of risks associated with climate variability, resource endowments, cultural values, social economic characteristics and the institutional and political environment (Deressa, 2007; Maddison, 2007; Gbetibouo, 2009; Ndamani and Watanabe, 2016).

Attempts have been made to enlighten Vietnamese farmers to climate change and to provide adaptive farming approaches (Le Dang et al., 2014b; Van et al., 2015; Ngo, 2016). Studies on farmers’ perception and adaptation to climate change have demonstrated favorable responses to government programs and rural communities upon successful implementation of national climate change by agriculture adaptation policy. However, no studies have been conducted in Northwest Vietnam that identify the factors that influence farmers’ perception of climate variability and choice of adaptation measures. This area is extremely vulnerable to climate change. The economy is poor and most farmers are ethnic minorities with low literacy (Van De Fliert, 2008; Schad et al., 2011; SRD, 2011). This study aims two questions: First, is it possible to study farmers’ awareness of local climate change? Second, can the adjustments that farmers have made in their agricultural practices in response to these changes be documented? This study identifies factors affecting farmer awareness and adaptation to climate change. Most of the factors affecting farmers’ awareness and adaptation measures are already known, but the impact of these factors varies among regions. Therefore, this study attempts to quantify the impacts of various explanatory factors on the adaptation probability and different household-level adaptation measures by farmers. Results of this study to identify the factors that can be used by policymakers to support agricultural adaptation to climate change.

This study was conducted in six provinces of Northwest Vietnam (Figure 1) that are geographically located between longitudes 20°39’_N and 22°49’_N and between latitudes 102°10’_E and 105°49_E (Toan and Tien, 2016). The study area is bordered by China to the north and Laos on the west. It has an area of 50.728 km² and accounts for 15 per cent of the mainland portion of the country. The population is only 4.8 per cent of the total population of Vietnam, and it has the lowest human population density (80 people/km²) (GSO, 2015). Agricultural land in this area is limited and further reduced by the fragmented terrain that is degraded by the effects of soil erosion. The region is mostly rural and the local people are largely dependent on agriculture. A large share of the population is composed of ethnic minority groups, with more than 20 different tribes living in the study area. Maize and rice are the two main crops in this area (SRD, 2011). These crops create revenues up to VND 1.4 and 1 million per person a year for rice and maize, respectively (GSO, 2012). Tea and other annual and perennial crops such as cassava, corn and cereals are additional significant sources of income and the most commonly grown cash crops (ILRI, 2014; SRD, 2010a, 2010b).

A market-based economy in Northwest Vietnam is present and economic reforms have significantly improved the rural economic structure. However, this region still faces many challenges. The economy depends heavily on agriculture and the economy has low efficiency. This area has the highest percentage of poor households in the country. Moreover, local livelihood resources are limited by the infertile cultivated land due to upland degradation, rugged terrain, steep hills, divided land, primitive cultivation practices, isolation and labor with low literacy rate. The irrigation and transportation infrastructures are poor, and there are relatively few income-generating opportunities (Jamieson et al., 1998; SRD, 2011). Therefore, the government has supported economic development projects in the Northwest provinces through policy reforms and rural action programs (ILRI, 2014).

The first stage in the study involved stratified sampling in which six provinces in Northwest Vietnam based on an index of the climate change vulnerability of the sub-national administrative areas. There are three levels of vulnerability to climate change: high, moderate and mild (Yusuf and Francisco 2010). Hoa Binh, Son La and Lai Chau were selected as representatives of the three vulnerability levels. In the second study stage, a cluster sampling technique was used to select the study sites. One district was selected from each province, and two communes were selected from each district. In the third stage, households were selected by random sampling. Table I depicts the numbers of farmers interviewed from the study areas.

The survey was conducted in January and March 2015 by ten interviewers previously trained before the questionnaire pretest. The respondents were household heads or their spouses. Participating farmers must have been engaged in agriculture, fishing, animal production or forestry for at least 10 years. A structured questionnaire was used to obtain information on socioeconomic characteristics, crops, land tenure, access to various institutional services, current and past knowledge of climate change and any adaptation measures taken. The survey was conducted in local languages.

Out of all these households, 350 were randomly selected from a geo-referenced household database of the respective communities, and, of these, 11 were not eligible for interview. Four questionnaires included incomplete or duplicated information (rate = 4.3 per cent), and the final number of completed questionnaires was 335. Approximately 90 min was required to conduct each interview.

Adaptation to climate change is a two-stage process: first perceiving the change and then deciding whether or not to adopt a particular measure. Previous studies based on perception of climate change used various models to analyze the factors determining perceptions: ordinal and nominal logistic regressions (Byg and Salick, 2009; Agossou, 2014), Heckman probit selection model (Deressa et al., 2011) and binomial probit model (Maddison, 2007; Gbetibouo, 2009; Piya et al., 2012). The logit model was used to assess farmers’ awareness about climate variability because of the essence of the decision variable; whether farmers perceived changes in temperature and/or rainfall or did not perceive changes. The logit model considers the relationship between a binary dependent variable and a set of independent variables, whether binary or continuous. The logistic model is given by Greene (2011):

where P_i is the probability of perceiving a change in the climate, and X_i is an independent variable. Therefore, the parameter β_i gives the log odds of the dependent variable, and β₀ is a constant. Odds ratio is the probability of an event happening related to not happening. It is given by Greene (2011):

The authors investigated some adaptation options. The relevant econometric model would be either a multinomial logit (MNL) (Hassan and Nhemachena, 2008; Deressa et al., 2009) or multinomial probit regression model. The drawback of these models is that one is restricted to select only one measure from a given set of adaptation measures. However, in the course of this study, the authors frequently found that farm households adopted more than one adaptation measure simultaneously. This behavior made the use of the MNL approach inappropriate. A possible remedy would be to combine similar measures into single categories (Bryan et al., 2013). However, such grouping into self-defined categories may lead to misinterpretation (Bryan et al., 2013). Furthermore, the set of explanatory variables influencing farmer decisions was also expected to be different for different adaptation measures. Therefore, the authors used the logistic regression technique to examine the factors affecting adaptation options. Table II shows the description and expected signs of explanatory variables used in this study.

Farm households will adapt only if they expect a reduction in crop production risks or an increase in net farm benefits. Consider a latent variable (⁠$Yij*$⁠), which is equal to expected benefits from the adoption of certain adaptation measures (Abid et al., 2015):

In this equation, $Yij*$ is a latent binary variable with subscript i depicting a household that adapted to climate change and j depicting eight different adaptation measures. X_k represents the vector of exogenous explanatory variables affecting farmers’ choice of adopting particular adaptation measures, and k in the subscript shows the specific explanatory variable. The α denotes the model intercept, β_k is the vector of binary regression coefficients and $εYi*≅N(0,σ2)$ is the error term that normally is distributed and homoscedastic (zero mean and constant variance) (W. Hosmer and Lemeshow, 2000).

The authors did not observe the latent variable (⁠$Yij*$⁠) directly. All authors observed the following (Abid et al., 2015):

where Y_ij is an observed variable that indicates that household i will opt for certain measures j to adapt to perceived changes in climate (Y_ij = 1) if their anticipated benefits are greater than zero (⁠$Yij*>0$⁠), and, otherwise, household i will not choose adaptation measure j if the expected benefits are equal to or less than zero (⁠$Yij*<0$⁠). Hence, equation (4) can be interpreted as:

where G (.) takes the specific binomial distribution (Abid et al., 2015).

The estimated parameters (β_k) of the binary logistic model only give the direction of the effect of the independent variables on the binary dependent variable and statistical significance associated with the effect of increasing an independent variable just like ordinary least squares coefficients (Peng et al., 2002). Thus, a positive coefficient β_k shows that an independent variable X_k increases the likelihood that Y_ij = 1 (which is the adoption of a particular adaptation measure in our case). But this coefficient cannot explain how much the probability of household i adopting a particular adaptation measure (Y_ij = 1) will change when we change X_k, that is, the coefficient (β_k) does not show the magnitude of the effect of a change in explanatory variable X_k on Pr(Y_ij = 1). Thus, to interpret and quantify the results, it is necessary to calculate either marginal effects or partial elasticity. Marginal effects (⁠$yij′$⁠) describe the effect of a unit change in the explanatory variable on the probability of a dependent variable, that is, Pr (Y_ij = 1). The final equation of the marginal effect (⁠$yij′$⁠) after derivation as follows (Abid et al., 2015):

Another alternative for interpreting the results of a logistic regression is to use partial elasticities, which measure the percentage change in probability of the dependent variable (adoption of certain adaptation measures to climate variability) because of a 1 per cent increase in the explanatory variable X_k. The partial elasticity of the logit model may be calculated at mean as follows (Abid et al., 2015):

To analyze the factors determining perceptions, 11 independent variables were chosen based on previous literature and specific characteristics of the study community. Some studies show that gender does not necessarily differentiate the ability to perceive climate change (Maddison, 2007). Education may influence perceptions either positively (Maddison, 2007; Deressa et al., 2011) or negatively (Gbetibouo, 2009). Among other factors, access to climate information and the factors specific to agriculture such as household extension services, soil quality and land tenure also can affect the perception of climate change (Deressa et al., 2011; Piya et al., 2012). The agro-ecological setting of farmers also influences the perception to climate change (Deressa et al., 2011).

The government of Vietnam is aware of the climate change problem and adopted the National Target Program to Respond to Climate Change in 2008, a national strategy to respond to climate change in 2011 and a national strategy on green growth in 2012. These programs integrate climate change issues into national development policies, and they become high priorities. The choice sets considered in the adaptation model included six group variables, including:

1. adjustments of crops and varieties (changing crop/livestock type, change variety, change crop structure, crop diversification);

2. adjustments to the planting calendar (change irrigation schedule, change crop rotation);

3. adjustments in planting techniques (change crop cultivation, change fertilizer/stimulus, change pesticides/herbicides, change crops quantity, change farmyard manure);

4. off-farm jobs;

5. other; and

6. no adaptation.

Following a review of literature on farmer adaptation choices, a range of household and farm characteristics, institutional factors and other factors that describe local conditions were hypothesized to influence farmer adaptation choices. The explanatory variables affecting the adaptation decision by farmers include household characteristics such as education, gender, age of the head of the household and household size. Farm characteristics include farm size and farm and nonfarm income; institutional factors include access to credit; and infrastructure includes distance to input and output markets and past climate experiences (Hassan and Nhemachena, 2008). In the empirical model, each explanatory variable is included in all six equations to test whether the impact of variables differed from one adaptation option to another. Table II presents the variables hypothesized to determine adaptation behavior.

The results of the study in Figure 2 show that about 92.8 and 91.3 per cent of the farmers in the study area perceived temperature changes in the winter (92.8 per cent) and summer (91.3 per cent) seasons (Figure 2). Many respondents perceived a slight warming for both the winter (39.1 per cent) and summer (52.8 per cent). Some farmers (23.3 per cent) perceived a slightly cooler winter and 14.3 per cent perceived a significantly cooler winter. These results showed significant differences in the perception of temperature change trends among the farmers. Temperature perception is disparate between winter and summer seasons. The main perception of summer temperatures is that they are increasing, whereas the winter temperatures tend to be decreasing (UNDP and IMHEN, 2015).

Most (89.9 per cent) of the respondents have observed winter rainfall decreases over the past 20 years. Of these, 53.7 per cent perceived a slight decrease, and 36.1 per cent perceived a significant decrease. The percentage of farmers who reported a significant decrease in summer rain was only 4.5 per cent, whereas 30.5 and 24.8 per cent perceived significant and slight increases in summer rainfall. These farmer perceptions of rainfall patterns revealed opposite rainfall trends between summer and winter.

Climate change studies in this region confirm differences in the summer and winter rainfall (Son et al., 2011; CARE, 2013; Van and Hang, 2013). The total annual precipitation decreased from 1960 to 2008 (Van and Hang, 2013). Another aspect of changing rainfall involves shorter rainfall durations but higher intensities (Son et al., 2011; UNDP and IMHEN, 2015).

One aspect of climate change is the frequency of extreme events. The Northwest mountain region is most affected by extreme weather phenomena (MONRE, 2012). Extreme events due to climate change in this area include droughts, landslides, flashfloods, untimely rains, extreme temperatures and water shortages (CARE, 2013; UNDP and IMHEN, 2015). Changes in farmer perceptions were also considered in terms of past extreme weather experiences. Farmers were asked about the frequencies of the main extreme weather they have experienced over the previous 10 years. A list of options was provided with five levels of occurrences. Changes in the annual frequencies of extreme temperatures and precipitation, recurrent droughts and flashfloods have been confirmed in the Northwest (Figure 3). From 2000 to 2009, a total of 96 flash floods occurred in the Northern mountainous region, Central Coast and Central Highlands (Schad et al., 2011). Relatively more flashfloods occurred in the provinces of Lai Chau and Son La in 1994, 1996, 2000 and 2002 (Schad et al., 2011). These weather events vary in frequency and intensity (Son et al., 2011; UNDP and IMHEN, 2015).

Household-level perception of climate change trends and extreme weather events by the majority of farmers were consistent with the actual trends (Figures 4 and 5). The mean annual temperature had a slight, but significant, increase from 1961 to 2011 (Figure 4), whereas there was a slight decrease in precipitation over the same period (Figure 5).

The correlation matrix of independent variables hypothesized to affect farmers perception of climate change is shown in Table III. Farmer age had a negative correlation with gender (ρ = −0.007 and highly positive correlation with farming experience (ρ = 0.864) at p < 0.05. This result provided a strong indication of correlation between age and farming experience variables. Thus, the age variable was dropped from the model.

A positive coefficient denotes that the factor enables correct perceptions of reality, whereas a negative sign denotes that the particular factor related to climate change trend and extreme events does not increase perceptions (Table IV). Male farmers are more likely to perceive change in rainfall than female farmers and better access to information sources, whereas females are more involved in household chores and, thus, rarely find time to access these sources of information. Similar trends have been reported in farming communities in the rural Mid-Hills of Nepal (Piya et al., 2012). The influence of education is positive for both climate change and significant for temperature, extreme temperature and drought. It is consistent with the findings of Gbetibouo (2009) and Deressa et al.,(2011) that increasing education improves the ability to perceive climate change. Farming experience will increase the probability that the farmer will perceive long-term changes in drought and abnormal temperatures. Educated farmers or the farmers with longer farming experience are more likely to perceive changes in drought or abnormal temperatures.

The results also confirm that owning a farm increases the probability of perceiving changes in droughts, flash floods and landslides. Access to extension services will increase farmer perception of changes in soil erosion and rainfall. But this is insignificant compared to farmer perceptions of other climate changes. This suggests that village-level extension services provided by government agencies are insufficient for raising awareness about climate change. Government agriculture service centers are usually distant from remote settlements and the extension agents rarely visit these villages. Extension programs place relatively little emphasis on raising farmers’ awareness of climate change. Soil fertility increases the probability of perceiving changes in abnormal temperature, soil erosion or rainfall. Access credit was found to be insignificant in our study. However, a positive sign in all perceive change indicates that access credit generally has positive effects on climate change perception.

Farmers with access to environmental information more easily perceive changes in precipitation than those without access. Area and farmer groups are two variables that increase the probability of perceiving drought changes. This result confirms earlier studies (Deressa et al., 2011; Piya et al., 2012). In the present study, climate information and farmer groups are insignificant in perception of climate change. Meanwhile these variables play important roles in helping the rural communities perceive climate changes and to make adjustments (Maddison, 2007). The level of farmer vulnerability to climate change varies by area and agro-ecological setting. Therefore, perceptions of farmers to climate change differ by region (Deressa et al., 2011). Lai Chau and Son La have high and moderate vulnerability to climate change (Yusuf and Francisco, 2010). This increases the probability of perceiving climate change in these regions compared to Hoa Binh.

Farmers who observed climate variability were asked to describe any household-level adaptations made. The adaptations used by farmers and farm households used a variety of adaptation measures in response to climate changes (Figure 6). The most common adaptations were changes in crops and varieties grown. A large number of farmers noticed changes in climate, but 17.31 per cent did not respond in any obvious way.

Adjustment of crops and varieties was the most common climate change response, whereas water management was the least used of the major adaptation methods. These results indicate that adaptation measures at the household level generally do not include advanced management technologies but are limited to simple and low expenditure measures such as changing crops or crop varieties and obtaining additional income from off-farm jobs. Very few farmers used advanced adaptation measures such as apply efficient water management technology, diversifying crop rotations, changing the crop variety.

All of the models were tested for significance and prediction accuracy. The R² for all models indicate that the statistically significant explanatory variables can explain 10 per cent to 26 per cent of the variation of farmer adaptation assessments (Table V). The χ² values for all adaptation models are positive and range from 335 and 364. The associated p values are < 0.001, and the authors conclude that the models with predictors fit significantly better than the intercept-only model. Hence, the null hypothesis (H₀) can be rejected and accept the alternative hypothesis (H1) that at least one of the regression coefficients is not zero. The classification table test results show that the overall percentage correctness for all models is > 62 per cent, which confirms the better fit of all of the models used here. Based on the results from the classification table along with global null hypothesis and pseudo-R², it can be assumed that the models in this study fit the data and accurately estimate the major factors affecting the adoption of different adaptation methods.

To quantify the impact of explanatory factors affecting farmer adaptation methods, the authors used logistic regression models for all adaptation measures. Coefficients of logistic regression explaining the direction of the effect of independent variables are presented in Table VI, and the marginal effects explaining the effect of a unit change in explanatory variables on the dependent variable are shown in Table VII. Partial elasticity calculations to illustrate the percentage impact of various factors on the probability of different adaptation measures are shown in Table VII. For continuous variables, the authors described the results in marginal form; for the binary variables, the authors used the elasticities for interpretation of results. Authors then describe the impact of various explanatory variables on the probabilities of adopting different adaptation measures in response to climate change.

There is no evidence that gender influences the probability of adaptation. Households with male heads are negatively related to almost all of the adaptation measures except for water management and nonadaptation measures were not significant. The negative correlation between adaption and gender is consistent with the findings of Maddison (2007). Years of farming experience significantly increases the probability of making adjustments of crops and varieties and taking off-farm jobs as adaptation measures. Experienced farmers are more likely to change varieties, crops, livestock and crop structure and are less likely to adjust the planting calendar. In this study, a 1 per cent increase in years of experience increased the probability of adopting changing the crop variety (1.68 per cent) and finding off-farm jobs (0.34 per cent). These results are consistent with those of Maddison (2007) and Nhemachena and Hassan (2007). These studies also found a positive relationship between farming experience and adaptation measure choices. Thus, it can be concluded that farmers with greater farming experience are likely to be more aware of past climate events and better judge how to adapt their farming to overall climate change and extreme weather events.

Education is an important factor in accessing advanced information on new improved agricultural technologies and increased agricultural productivity (Maddison, 2007; Deressa et al., 2011). In this study, the significant coefficient of education of the household head shows that the probability of adapting to climate change increases with an increase in the years of education. A 1 per cent increase in years of education leads to an increase in the probability of changing water management (1.1 per cent). Maddison (2007), Deressa et al. (2009) and Ngo (2016) also found a significant positive relationship between education of the household head and adaptation to climate change that is consistent with results of the present study.

Logistic regression results show that an increase in household size did not significantly increase the probability of adaptation though the coefficient of the adaptation options had a negative sign. Even though it is insignificant, we believe that increasing household size decreases the likelihood of most adaptation measures. This is contrary to the findings of Deressa et al. (2011), Abid et al. (2015), and Ngo (2016). Increasing the household size is not a recommended measure to increase resilience to climate change. An increase in household size increases the likelihood of poverty and thus reduces the resources available for adaptation.

“Land area” represents the total land area owned by a farm household and is a proxy for farm household wealth. The results indicate that the land area has a significant positive impact on a changed planting calendar. A 1 per cent increase in the land area increased the probability of planting calendar adjustment by 0.48 per cent. But land area had a significantly negative impact on “find off-farm jobs” as adaptation measures. This shows that larger farms are more likely to choose adaptation measures related to the farm. This is consistent with the theory that adaptation has a fixed cost element. The most likely reason for the positive relationship between adaptation and farm size is that adaptation is subject to economies of scale. Large-scale farmers are more likely to adapt because they have more capital and other resources. The result is consistent with previous studies such as Maddison (2007), Nhemachena and Hassan (2007), Ngo (2016) and Ndamani and Watanabe (2016). Land holdings in North Vietnam are highly fragmented as a result of a land allocation policy that distributed land in equitable quantities but inequitable qualities (Marsh et al., 2007). The authors believe that government should support land reform such as farmers’ cooperation for cultivating large-scale fields. Households with more land have a lower probability (0.62 per cent) of choosing “find off-farm jobs” as a response to climate change.

Soil fertility status had a significant positive impact on farmer adoption of changed planting techniques. If the soil is fertile, the probability of adjustment of planting techniques is increased (0.10 per cent). These results are consistent with findings by Gbetibouo (2009).

The highly significant coefficient of proportion of non-agriculture income to total income shows that the probability of adapting to climate changes increases with increase of non-agriculture income. A 1 per cent increase in the proportion of non-agriculture income leads to an increase in the probability of adjustment of crops and varieties (0.32 per cent), “find off-farm jobs” (0.35 per cent) and leads to a decrease in the probability of adjustment of planting calendar (0.55 per cent). Availability of free extension advice related to livestock or crop production strongly increases the probability of farmer adaptation (Maddison, 2007). Access to extension services significantly increases the probability of water management and nonadaptation. Surprisingly, access to extension service decreased farmer adjustments to the planting calendar as an adaptation measure. No significant relationship was found between the extension service and other adaptation measures. This reveals that current extension services are not sufficient to support an effective adaptation process. There is a lack of updated information on adaptation to climate change available from the agricultural extension department. Hence, there is need for better collaboration at different levels of the adaptation process.

Access to credit is significantly positively related to the adaptation measures adjustment of planting techniques and water management. It also has significantly negative impacts on adjustment of the planting calendar and non-adaptation. Access to credit and loans facilitates adaptation to new technology and climate change because it allows farmers access to new technology and enables purchase of improved varieties of seeds and fertilizer. With greater financial support, farmers are able to change their management practices in response to changing climate (Nhemachena and Hassan, 2007). The positive correlation between adaption and the availability of credit observed here is consistent with the findings of Gbetibouo (2009) and Fosu-Mensah et al. (2012). If a farmer has access to credit, it increases the probability of planting techniques adjustment (0.36 per cent), ensures better water management (0.1 per cent) and decreases the propensity for planting calendar adjustment (0.12 per cent).

Farmers belonging to cooperative organizations have a higher likelihood of using adaptation practices due to the sharing of ideas and information, discussion of problems and implementation of collaborative decisions (Ndamani and Watanabe, 2015). The logistic regression results showed a significant positive association between farmer group membership and adjustments of planting techniques. The probability of adjustments of planting techniques increased by 0.17 per cent if farmers were members of a farmer group. However, this study found a significant negative relationship between farmer group membership and water management adaptation. Farmer group membership also had an insignificant relationship to other adaptation measures. These findings indicate that activities of farmer groups did not provide effective support for farmer adaptation.

The results showed that access to information on weather forecasting was positively related to the probability of adjustment of the planting calendar. Farmers with access to timely weather information are more likely to change their irrigation schedule and crop rotation. Access to information on weather forecasting also decreases the probability of no adaptation to choice 0.1 per cent. Similar findings were reported in Southern Africa by Nhemachena and Hassan (2007). However, weather forecasting decreases the probability of water management choice 0.4 per cent and had an insignificant relationship with other adaptation measures. This is likely due to the choice of adaptation measures by farmers based on experience and cost. Normally, farmers will choose simple adaptive measures that lower investment costs.

Proportion of land with long-term use right indicates farmer land tenure status as owners or tenants. The land tenure system is vital to adaptation as landowners tend to adopt new technologies more quickly than tenants. This finding has justified numerous efforts to reduce tenure insecurity (Gbetibouo, 2009; Fosu-Mensah et al., 2012). In this study, farmer land tenure status is positively associated with most of the adaptation measures except no adaptation. Even though it is not significant, the authors believe that farmers with long-term use right land are more likely to adapt their farming to perceived climate change. If the farmer is the owner or has long-term use right of land, it reduces the probability of no adaptation (0.1 per cent). Increased likelihood of adaptation for owners may due to the reason that owners are more conscious about their farm income compared to tenants as the farmer does not pay land rent and will have more capital for adaptation to climate change. This result suggests that the government should support long-term autonomy in management and use of land to farmers for their agricultural production. This is an important measure needed to support farmer adaptation.

The coefficient of distance from plot to house had a negative sign in most cases to choose adaptation. Households with plots that were a longer distance from the house had less probability of adjustments of planting calendar by 0.02 per cent and more probability of choosing “find off-farm jobs” by 0.01 per cent. This result is consistent with Ngo (2016). Improved farmer access to markets increases their ability to buy new crop varieties, implement new irrigation technologies and other important inputs they may need to change their current practices (Nhemachena and Hassan, 2007). Longer distance to the market diminishes the exchange trading embroidery products, mainly by adapting measures and to find off-farm jobs of the farmers. A 1 per cent increase in the distance of the household from the nearest local market results in a 0.008 per cent decrease in the probability of diversion to off-farm jobs and a 0.005 per cent increase in the probability of no adaptation choice.

Personal experiences effect the perceptions and responses to climate change (Niles et al., 2015; Ngo, 2016). Farmers who are aware of climate changes have a higher probability of taking adaptive measures in response to these changes. Specifically, increasing the number of drought periods increases the probability (by 0.74 per cent) that farmers will respond to changes in terms of water management but decreases the probability (by 0.28 per cent) of adjustment of planting techniques. Untimely rains increase the probability of farmers diverting to off-farm jobs by 0.14 per cent but decreases the probability of adjustment of crops and varieties by 0.29 per cent. Increase in abnormal temperature increases the probability of farmers making no adaptation by 0.34 per cent but decreases the probability of adjustment of crops and varieties by 0.58 per cent and change in water management by 0.3 per cent. Increasing landslide decreases the probability of farmers’ adjustment of crops and varieties by 0.31 per cent. Farmer risk aversion can increase after exposure to extreme climate events. This can lead to the choice of lower risk adaptive measures. In addition, adoption practices used by farmers are due, in part, to limiting factors within their system (Niles et al., 2015)

If extreme weather events are the strongest factors influencing climate change perception for farmers this has significant implications for assessing how short-term responses can affect long-term adaptation and the subsequent policies that may be needed to accompany such actions (Le Dang et al., 2014a; Park et al., 2012). In addition, climate variability involving high temperatures, drought, and water shortages require investments in irrigation systems. Public–private partnerships should be considered to allow farmers increased water control to counteract adverse impacts from climate variability and change. Improvements in accessibility and effectiveness of local services (e.g. irrigation, agricultural extension, credit and health care) are necessity for successful adaptation strategies (Le Dang et al., 2014a).

This study explored farmers’ perceptions and adaptations to climate change in Northwest Vietnam. The majority of farmers who recognized climate change had implemented at least one adaptation. This study did not deal with the perceived risks of climate change to farmers, barriers to the use of adaptation practices in agriculture or farmer assessments of private adaptive measures to climate change. However, the results do suggest policies that could improve the adaptability of farmers to climate change at the community level.

In light of the above, this study concludes that governments and development partners should integrate policies, projects and programs related to adaptation to climate change. Sources and quality of information can be an important consideration in addition to the past climate experiences of farmers and their adaptation assessments. Awareness of climate change and adaptation methods should be more focused. The government should increase the capability of research scientists and agricultural staff to develop and promote appropriate and effective technologies. Improvement in the accessibility and usefulness of local services, such as village-level extension services, flexible terms of agricultural credits and infrastructure will be needed for successful adaptation strategies in Northwest Vietnam. Other policy options include increasing farmer education, giving farmers access to new technologies and increasing irrigation investment through public – private partnerships. The government should support land reform such as farmer cooperation in large-scale production. Nongovernmental organizations and local governments, working at the grassroots level, can play an important role for disseminating the relevant information and conducting awareness-raising campaigns for farmers. They can conduct seasonal weather forecasts through village, radio and assist farm households in scheduling their crop calendars in accordance to these forecasts.

This study is part of a PhD research at Northwest A&F University, Shaanxi, China. The authors are very thankful to the College of Economics and management of Northwest A&F University; Hoa Binh Technical and Economic College, Vietnam; study team members, and farm households in HoaBinh, SonLa, LaiChau province for their cooperation and participation in the research.The authors are very thankful to anonymous reviewers for their valuable suggestions/recommendations. The authors further declare that the contents and opinions expressed in this article are exclusively of their own, and do not express or endorse any position of their employers.