1 Introduction

In the face of significant challenges posed by climate change in the current agricultural landscape, adopting adaptive measures emerges as the primary choice for farmers in responding to climate-related risks (Arbuckle et al., 2015). According to the Intergovernmental Panel on Climate Change (IPCC) report, global temperatures will continue to rise until at least the middle of the century (IPCC, 2022). Dramatic changes in the climate system lead to an increase in the frequency and intensity of extremely high temperatures, heavy precipitation, and agricultural and ecological environmental droughts (Cook et al., 2018; Orlowsky & Seneviratne, 2012). Humans and ecosystems are increasingly vulnerable to climatic hazards (IPCC, 2022). Mitigation and adaptation to climate change are two fields of constant concern (Chen & Gong, 2021). Mitigation requires scientific and technological means to expand carbon sinks or reduce greenhouse gas emissions (Panepinto et al., 2021). Adaptation refers to measures and actions to mitigate climate change impacts on natural and human systems (Aryal et al., 2020). Adaptation is adopted more widely than mitigation, especially in agriculture (Howden et al., 2007). As direct participants in agriculture, farmers’ adaptation strategies significantly affect the effectiveness of climate change adaptation in agriculture (Anwar et al., 2013).

Crop diversification stands out as one of the most favored climate change adaptation strategies among farmers, garnering extensive attention in research, particularly in regions characterized by pronounced climate variability (Khan et al., 2020; Makate et al., 2016; McCord et al., 2015; Ochieng et al., 2020). The primary goal of crop diversification is to enhance the crop diversity of agricultural lands, encompassing the genetic diversity of crop species, the number of crop planting species, and crop planting area.Footnote 1 The approach involves various agricultural behaviors, such as crop rotation, mixed cropping, intercropping, and agroforestry (Hufnagel et al., 2020; Malaiarasan et al., 2021). The practice of crop diversification can enhance the stability and productivity of agricultural systems (Maggio & Sitko, 2021; Osterholz et al., 2018), thereby alleviating issues such as water shortages, crop growth inhibition, and the spread of pests and diseases caused by climate variability (Lin, 2011). Furthermore, from an economic perspective, crop diversification can mitigate income fluctuations resulting from changes in the price of a single crop, ensuring stable incomes through diverse crop combinations to support farmers adapting to climate change (Mzyece & Ng'ombe, 2021). McCord et al. (2015) investigated the crucial role of crop diversification in semi-arid mountain agriculture in Kenya. Ponce (2020) found that farmers in the Peruvian Andes respond to increased climate variability by enhancing crop diversity. In Northern Ghana, Appiah-Twumasi and Asale (2022) observed that crop diversification can mitigate the impacts of erratic rainfall on rural households.

While crop diversification presents myriad advantages, it also poses challenges. One such challenge is the increased difficulty of mechanization compared to monoculture practices (Lin, 2011). This difficulty arises because different crops may have diverse planting requirements, making it challenging to utilize machinery efficiently. Additionally, diversified cropping systems frequently demand more labor and time inputs, as farmers must oversee multiple crops with distinct growth cycles and agronomic practices. The escalated demand for labor and time may lead to elevated production costs for farmers, thereby increasing the overall expenses associated with climate adaptation. (Cohen et al., 2020; Wu et al., 2022).

In addition to farmers’ adaptation strategies, a growing body of research underscores the significance of “Climate Change Adaptation Politics” (He et al., 2022b; Wang et al., 2024). In recent years, governments in various countries and regions have actively undertaken purposeful climate actions, particularly in the agricultural sector, with a sustained increase in government attention to climate adaptation for farmers (Aliabadi et al., 2022; Nam et al., 2022). To mitigate the impacts of climate change on agricultural production, governments worldwide have issued a series of interventional measures to help farmers adapt to climate change (Chen et al., 2014; Heath et al., 2020; Karimi et al., 2018; Mwinjaka et al., 2010). For instance, in China, policy measures addressing drought, such as providing early warning information, agricultural technical assistance, and financial or infrastructural support, have played an essential role in assisting farmers in adapting to drought (Chen et al., 2014). The Brazilian government has implemented various drought response programs to support farmers in sustaining their agricultural livelihood activities. These programs include increased welfare payments, distribution of drinking water, and longer-term agricultural risk reduction measures such as crop insurance, loans for constructing small dams or purchasing efficient irrigation systems, and subsidized rainwater cisterns (Herwehe & Scott, 2018). The Iranian government renovates and improves old irrigation canals and facilities, introduces drought-resistant varieties, and provides technical support for irrigation plans for farmers (Karimi et al., 2018). In contrast, with limited policy measures, some farmers commit suicide due to failing to cope with agricultural losses and indebtedness caused by climate change (Swami et al., 2020).

Planned adaptation by governments and spontaneous adaptation by farmers often interact with and impact agriculture (Mersha & van Laerhoven, 2018). On the one hand, government-led planned adaptation influences farmers’ spontaneous adaptation, as effective policy measures can mitigate the impact of climate change shocks and reduce adaptation costs (Wang et al., 2019). For instance, when local governments build irrigation facilities, farmers are significantly less affected by drought and may not need to rely on drought-tolerant varieties (Zhang et al., 2021a). On the other hand, results from spontaneous adaptation can also provide feedback to the government as a basis for modifying planned adaptation policies (Bonzanigo et al., 2016). Therefore, comprehending the impact of policy measures on farmers’ adaptive behavior can better assist policymakers in adjusting policies to enhance the farmers’ livelihood.

Upon reviewing the existing literature, the predominant focus is on examining the impact of various factors related to farmers on crop diversification. For instance, Zhang et al. (2022) examined the effects of labor migration on crop diversification. Their findings revealed that the opportunity cost associated with adopting crop diversification was higher than working outside the home, resulting in a labor shortage and a subsequent reductionin crop diversification. Ge et al. (2023) identified the role of gender in adopting crop diversification, discovering that female-headed households were more inclined to maintain traditional cropping patterns and less willing to diversify their crops. Additionally, Ndip et al. (2023) observed that farmers with fragmented land were likelier to adopt crop diversification than those with less fragmented land. While these studies have extensively investigated the factors influencing crop diversification, they ignored the role of government. Government interventions are intricately interwoven with the challenges posed by the current global climate change crisis (He et al., 2022b). Hence, it is imperative to delve into the repercussions of policy measures on crop diversification. A recent study scrutinized the impact of policy measures and unearthed that grain subsidies had a detrimental effect on crop diversification at the regional level (Song et al., 2021). While this finding is a good inspiration for further work on related topics, this study was conducted on a regional level and did not focus on the farmer level. Farmers are the most vulnerable and highly exposed to climate change risks, and they are the direct and primary adopters of crop diversification, so examining the impact of policy measures at the farmer level is essential. Additionally, the urgency of adopting crop diversification may vary depending on the context of the local climate change (Wilbanks, 2003). The study area of Song et al. (2021) is Hubei Province, China, where the effects of climate change may not be as pronounced. Hence, it is imperative to scrutinize the impact of policy measures on crop diversification at the farmer level, particularly in regions with severe climate change.

In recent years, the Chinese government has implemented a series of policy measures to improve farmers’ livelihoods and mitigate the impact of climate change (He et al., 2022a). Consequently, farmers may refrain from adopting crop diversification, thus saving the costs of adapting to climate change. This study seeks to determine whether the policy measures have effectively assisted farmers in resisting the challenges posed by climate change, leading to a reduced reliance on crop diversification. The case study on key regions of climate change can provide insights for adjusting policy measures in other regions that are strongly affected by climate change, and better establish effective linkages between governments and farmers to enhance climate adaptation capabilities.

The remaining parts are as follows. Section 2 presents an analytical framework for explaining how policy measures affect crop diversification at the farmer level. Section 3 describes the study area, data sources, and the econometric model employed. Section 4 gives the results of the statistical analysis and econometric model. Section 5 provides a discussion of the results and policy implications. Section 6 concludes the main findings.

2 Analytical framework

We propose a framework for analyzing the impact of policy measures on crop diversification at the farmer level, illustrating the process through the combined effects of three dimensions (Fig. 1). A connection is established from the environmental layer to the farmer layer through perception. Climate change seriously impacts agriculture, exposing farmers to risks (Donatti et al., 2019). Crop diversification emerges as an essential response by farmers to climate change, and whether they decide to take such action depends on their subjective intentions (Arunrat et al., 2017). Therefore, farmers’ perception of climate change is crucial (Singh, 2020; Yang et al., 2021). Recent studies have underscored the importance of farmers’ perceptions of climate change in influencing crop diversification (Khan et al., 2020; Meldrum et al., 2018).

Fig. 1
figure 1

An analytical framework for the impact of policy measures on crop diversification at the farmer level

Full size image

Policy measures act as an intermediate layer, a buffer against climate change shocks. Generally, policy measures for farmers can be divided into three types: financial support, infrastructure support, and technical support. Financial support includes cash subsidies, credit, agricultural insurance, etc. (Bryan et al., 2009; Di Falco et al., 2014; Li et al., 2013). Infrastructure support includes roads, irrigation facilities, and communication networks (Alam et al., 2012; Heath et al., 2020; Sen et al., 2021). Technical support includes agricultural technique training, the provision of crop varieties, and the provision of disaster early warning information (Antwi-Agyei & Stringer, 2021; Li et al., 2017b; Pritchard, 2013). However, whether these policy measures effectively buffer farmers’ livelihood dilemmas needs testing based on their responsive behavior (Li et al., 2017b). Policy measures may impact crop diversification in two ways (Fig. 1). First, if policy measures effectively buffer climate change shocks to farmers, they could reduce farmers’ livelihood vulnerability. Consequently, farmers would not have to adopt crop diversification to cope with increased risks (Min et al., 2017). Second, in cases where the effectiveness of policy measures is not significant, farmers would still increase the level of crop diversification to cope with climate change (Wang et al., 2019).

Moreover, according to the sustainable livelihoods framework, the asset status of farmers’ households forms the basis for their adoption of livelihood strategies (Li et al., 2017a). For this reason, livelihood assets were considered when analyzing the impact of policy measures on crop diversification. Livelihood assets consist of five components: human assets (H), physical assets (P), financial assets (F), natural assets (N), and social assets (S) (Liu & Xu, 2016).

In summary, our research focused on the impact of policy measures and farmers’ perceptions of climate change on crop diversification. We consider the crucial context of climate change and include five livelihood assets as control factors in our analysis.

3 Study area, data, and methods

3.1 Study area

The Tibetan Plateau (TP), due to its unique physical geography, unparalleled height, and fragile and sensitive environment, is not only a sensitive area to climate change but is also considered a driving machine and amplifier of global climate change. We conducted our fieldwork in the Pumqu River Basin (PRB), situated in the southwestern region of the TP (Fig. 2). The PRB is delineated by the Himalayas to the south, bordering Nepal, the Kailas Range to the north, Lake Paiku Co to the west, and the Kangcheda Snow Mountain Group to the east. It covers parts of 5 counties, including Nyalam, Tingri, Dinggye, Gamba, and Sa’gya (Zhang et al., 2023). The main river in the region is the Pumqu River, which originates from the Yebokangal Glacier on the northern slope of Mount Shishapangma Peak. The Himalayas act as a significant barrier to the warm and humid air currents from the Indian Ocean, creating a clear regional divergence in climate between its northern and southern flanks. As a result of the barrier and the “rain shadow” effect, precipitation in the PRB is low, averaging 300 mm per year. Solar radiation is intense, and sunshine hours are long, with an annual average of approximately 3,300 h of sunshine. In addition, evaporation is high in the PRB, with an annual average evaporation level of approximately 2,500 mm. The meteorological environment is harsh with a significant foehn effect (Zhang et al., 2021b).

Fig. 2
figure 2

Study area and field survey routes

Full size image

According to the observations of the meteorological stations of Nyalam and Tingri from 1987 to 2017, the climate of the PRB showed a warm and dry trend (Fig. 3). In terms of temperature, the annual average temperature is 3.2–3.8 °C, much lower than that of the plains at the same latitude. Since 1987, the annual average temperature of the PRB has shown a significant upward trend, with a warming rate of 0.41 °C 10 yr−1, and the warming rate is more significant after entering the twenty-first century. In contrast to the temperature, precipitation showed a decreasing trend. The annual change rate of precipitation from 1987 to 2017 was − 22.69 mm 10 yr−1.

Fig. 3
figure 3

Annual average temperature (a) and annual precipitation (b) change trends of the PRB from 1987 to 2017

Full size image

According to the 2019 China Statistical Yearbook (County-Level), the total population of the Tingri, Dinggye, and Nyalam Counties was 101,637 in 2018,Footnote 2 of which approximately 80% were farmers and herders, and approximately 90% were Tibetans. The total arable land area was 177.37 km2, and local crops mainly included highland barley, wheat, oilseed rape, potatoes, and beans. The livestock raised were yaks, cattle, goats, sheep, and horses. Recent studies have found that farmers in the region reclaim more farmland to adapt to climate change (He et al., 2021; Wang et al., 2020). Adaptation to climate change is critical for local farmers in a relatively arid region subject to significant climate variability and where climate change heavily influences agriculture.

3.2 Data sources

Sample selection. The local grass-roots governments in China are hierarchical, comprising counties, townships, and villages. Therefore, the sample selection involved two steps. Firstly, discussions were held with the local governments of Tingri, Dinggye, and Nyalam counties. For each county, three sample townships were selected based on the local socioeconomic, environmental, and agricultural conditions. Secondly, we employed a stratified random sampling approach to survey each sample township: villages within each township were stratified into three levels based on their socioeconomic status, and 2–3 sample villages were chosen from each level, with approximately 20–30 households surveyed in each village. In total, 684 valid questionnaires were collected, including 242 from 11 villages in Dingri County, 181 from 8 villages in Dingjie County, and 261 from 8 villages in Nyalam County. Figure 2 illustrates the sampled villages’ locations and the on-site survey routes.

Questionnaire collection process. The data collection period for the questionnaire was from July to August 2018. Seven pre-trained Tibetan students acted as translators in the survey to overcome language barriers and get accurate information. Before the formal survey, a 7-day pre-survey of farmers was conducted in the survey area, and the questionnaire content was improved based on feedback from respondents. A semi-structured interview approach was used to administer the questionnaire during the formal survey. Most respondents were household heads, and other family members provided additional information. Each household underwent a 1–2 h interview. There was no subjective involvement or bias in the entire investigation process. The questionnaire primarily included closed-ended questions, where respondents selected or provided answers based on their circumstances (e.g., “How many types of crops do your household grow, and what are their respective areas? How do you perceive the overall changes in precipitation in the past decade? (Decrease = 1, No Change = 2, Increase = 3)”). Additionally, there were some open-ended questions on specific topics (e.g., “What aspects of livelihood do you think the government has assisted your family (if any)?”). The questionnaire covered basic household information (age, education level, health status, etc.), household assets (crop varieties, planting areas, agricultural equipment, income, etc.), policy measures (subsidies, low-interest loans, skills training, etc.), and perceptions of climate change.

The questionnaires were administered to seek farmers’ perception of the state of climatic events over the years spanning over a decade, with a primary focus on broad trends in climate variability. Despite the dynamic nature of climate change, the types of agricultural risks brought about by climate change have not changed, and farmers’ adaptation strategies have not undergone significant alterations. Crop diversification has consistently been an essential adaptive strategy for farmers. Therefore, our research focuses on a prevalent phenomenon, and time may not significantly impact farmers’ behavior.

3.3 Methods

3.3.1 Dependent variable

The explanatory variable is crop diversification. This study used the Shannon index to measure the degree of crop diversification (Smale et al., 2003). Following Martey (2022), the index is calculated as follows:

$$Y = - \sum P_{i} \ln P_{i}$$
(1)

where Y is the crop diversification index, and the larger its value is, the more types of crops and the more evenly distributed the planting area of various crops is; Pi is the ratio of the planting area of the ith crop to the total planting area of the farmer, and when Pi = 0, we stipulate that ln(Pi) = 0; that is when a farmer does not plant any crops, the crop diversification index is 0.

3.3.2 Independent variables

According to the analytical framework, the study considered policy measures, farmers’ perceptions of climate change, and five livelihood assets. The independent variables are described below.

Policy measures: Regarding financial support, we considered crop planting subsidies, non-agricultural subsidies, and the total number of household loans. In terms of technical support, agricultural technique training was considered. Local governments often organize community-based activities for farmers to enrich their agricultural knowledge. Regarding infrastructure support, we selected the distance from the house to the nearest county-level or higher road. The Chinese central government has been promoting the construction of a new socialist countryside since 2006Footnote 3 and explicitly proposed strengthening the construction of rural roads.

Climate change: Crops are susceptible to changes in temperature and precipitation (Mondal et al., 2015). Therefore, when farmers perceive an increase in temperature or a decrease in precipitation, they will adopt crop diversification to reduce the climate change risks. Consequently, we selected farmers’ perceptions of temperature change and precipitation change as indicators of climate change perception.

Human assets: The “total household population” reflects the size of a household. It is generally believed that a large household can significantly increase crop diversification (Goswami et al., 2017). Since the household head is typically the critical decision-maker of a household and has absolute rights (Zhou et al., 2020), the “age of the household head” and “education level of the household head” were selected. It is assumed that the younger the household head is, the lower the education level is, and the more likely the household is to reduce the type and area of crops grown.

Financial assets: When households earn higher off-farm employment income, they may prefer to rely on off-farm employment rather than remain in agriculture (Lu, 2020).

Social assets: Farmers’ participation in community organizations can expand their social networks, achieve more risk awareness and risk coping ability (Li et al., 2018), and more easily improve crop diversification.

Natural assets: The “average area of cultivated plots” reflects the natural capital of a household. The fragmentation of cultivated land can meet crop diversification requirements and reduce farmers’ natural and market risks (Di Falco et al., 2010). Therefore, it is assumed that the larger the average area of cultivated plots is, the lower crop diversification levels are.

Physical assets: Physical assets are measured by agricultural equipment and livestock numbers. The input of agricultural equipment can improve production efficiency. The more agricultural equipment is used, the easier it is to diversify planting. On the TP, livestock is a measure of wealth and can be used to guarantee the ability to cope with risk (Nyima, 2014). Livestock breeding reduces household labor input into agriculture and may be negatively related to crop diversification. The descriptive statistics of each variable are shown in Table 1.

Table 1 Descriptive statistics of variables (n = 684)
Full size table

3.3.3 Model selection

The dependent variable is the crop diversification index. Suppose the farmer does not diversify planting, which is 0. Based on this, we used the Tobit model for econometric analysis. The model can effectively estimate the dependent variable continuously distributed in positive values and contain a portion of observations that take a value of 0. The econometric model is set as follows:

$$Y = \left\{ {\begin{array}{*{20}l} {0 , \quad y^{*} \le 0 } \\ {y^{*} = \alpha_{1} + \alpha_{2} \left( {X_{i} } \right) + \alpha_{3} \left( {Z_{j} } \right) + \alpha_{4} \left( {X_{i} } \right) \times \left( {Z_{j} } \right) + \alpha_{5} \left( {A_{K} } \right) + \varepsilon , \quad y^{*} > 0} \\ \end{array} } \right.$$
(2)

Equation (2) explores the interactive effects of climate change and policy measures on crop diversification. Where Y is the value of the crop diversification index of households, y* is the potential explanatory variable, Xi is the climate change indicator, Zj is the policy measures indicator, and Ak is the livelihood asset indicator, including human assets, financial assets, social assets, natural assets, and physical assets. \({\alpha }_{1}\) is the constant term. \({\alpha }_{2},\;{\alpha }_{3}, \; {\alpha }_{4},\; {\alpha }_{5}\) are the estimated coefficients. \(\varepsilon\) is a random disturbance term with normal distribution characteristics.

STATA 12.0 was applied for the quantitative analysis. Considering the possibility of multicollinearity between different independent variables, the Pearson correlation coefficient (PCC), tolerance, and variance inflation factor (VIF) were chosen for testing. The results showed that the absolute value of the PCC between household size and livestock was the largest at 0.558 < 0.8, the tolerance of household size was the lowest at 0.468 > 0.1, and its VIF was the largest at 2.14 < 10. All independent variables passed the test, indicating no multicollinearity between independent variables.

4 Results

4.1 Descriptive statistics

In a sample of 684 households, the average household size was approximately 6 members, with a SD of 2.619. The mean age of the household heads was 48.81 years, with a SD of 12.554, indicating that most households likely possess rich agricultural practical experience, enabling them to adopt appropriate adaptation strategies. The quantified indicator of the educational level of the household head had an average value of 1.781, with a SD of 1.086, indicating that the majority were illiterate or had only attended preschool. The overall low level of education may result in difficulties in applying modern agricultural techniques. The average area of cultivated plots per family was tiny, at only 0.07 hectares, with a SD of 0.063, indicating the land was fragmented, which was conducive to the adoption of crop diversification. Additionally, the mean value of the agricultural equipment index was 4.41, with a SD of 2.190, suggesting that these households primarily rely on horses, cattle, and small-scale agricultural equipment, reflecting a relatively low level of agricultural modernization. Detailed analyses of the results regarding climate change, crop cultivation, and policy measures are in the following two sections.

4.2 Climatic perceptions and crop planting of farmers

4.2.1 Farmers’ perceptions of climate change

Farmers’ perceptions of temperature were consistent with the data reported by the meteorological station (Figs. 3, 4). 87.1% of the farmers perceived an increase in temperature, indicating more significant warming. Among the farmers who perceived changes in precipitation, 68.6% perceived an increase, 11.7% perceived a decrease, and 19.7% perceived no change. Actual precipitation in the PRB fluctuates, with an overall decreasing trend. However, the perceptions of most farmers did not consistently match actual conditions (Fig. 4). This is similar to the findings of He et al. (2021) and Wang et al. (2020).

Fig. 4
figure 4

Perceptions of climate change of sample farmers in the PRB

Full size image

4.2.2 Crop planting

According to our survey, households grow 1–7 kinds of crops. Most households grow 3 or 4 types of crops, accounting for 34.36 and 32.46%, respectively (Fig. 5). An average of 3.12 crops were planted per household. The main crops grown were highland barley, spring wheat, beans, potatoes, oilseed rape, and vegetables (Table 2). Grain crops are mainly planting types. Highland barley and rape were grown by the majority of farmers, with 650 and 516 households, respectively, followed by potatoes (306 households), beans (262 households), spring wheat (42 households), and vegetables (34 households). Highland barley covered the largest planting area at 1,700.69 hm2 and was followed by rape (1,078.16 hm2), beans (156.04 hm2), and potatoes (95.95 hm2), while spring wheat (7.62 hm2) and vegetables (1.55 hm2) were planted in relatively small areas.

Fig. 5
figure 5

Number of households growing different quantities of crop varieties in the sample farmers of the PRB

Full size image
Table 2 Household number and planting areas of different crop varieties in the sample farmers of the PRB
Full size table

4.2.3 Crop diversification index

The average crop diversification index was 0.732, with a standard deviation of 0.390 and a maximum value of 1.692 (Table 1). Figure 6 shows a normal distribution of the crop diversification index (greater than 0). 87.57% of households have adopted crop diversification, the largest number had a crop diversification index between 0.8 and 1.0, and 76.61% had a crop diversification index greater than 0.5. A recent study showed that the average crop diversification index of households in different regions of Ghana was around 0.760 (Martey, 2022). Dessie et al. (2019) calculated the crop diversification index of 0.769 for 385 rural households in Northwest Ethiopia and pointed out this is a high level. These studies indicated that crop diversification in PRB was high.

Fig. 6
figure 6

Household number of different crop diversification indexes in the sample farmers of the PRB

Full size image

4.3 Policy measures by the government

4.3.1 Governmental subsidies

Households received diverse and high-volume subsidies from the government (Table 3). The government paid crop planting subsidies based on the amount of a given crop grown by a household. It mainly targets barley, wheat, rice, potatoes, etc., and the subsidy amount varies slightly between crops. The subsidies aim to increase the coverage of good seeds, enhance the ability of crops to adapt to climate change, and improve crop yields. Animal husbandry subsidies mainly include fine breed subsidies for livestock and livestock death subsidies (Yu et al., 2021). Livestock subsidies encourage farmers to breed improved varieties to improve their livestock’s ability to adapt to the environment and grow; the government also provides subsidies for livestock that die unexpectedly. In Tibet, weather-related disasters such as extremely low temperatures, snowstorms, and increased diseases caused by high temperatures are serious causes of livestock mortality. Livestock subsidies are therefore essential to alleviate the pressure of climate change on households’ animal husbandry. Forestry subsidies refer to subsidies for the Conversion of Farmland to Forests (Cao et al., 2009). Living subsidies mainly include minimum subsistence allowances, social security, and old-age allowances.Footnote 4 The PRB is located at the border between China and Nepal. To maintain border stability and improve production and living conditions, the Tibet local government has taken the lead in establishing an allowance system for border residents.Footnote 5 Statistics on the subsidies of 684 sample farmers are shown in Table 3. Among the 684 households, the subsidy coverage rate reached 100%, and the average subsidy amount per household was 18,203.92 yuan. The subsidies alleviate the pressures of living and provide financial support for farmers to cope with climate change.

Table 3 Subsidies obtained by sample farmers of the PRB
Full size table

4.3.2 Low-interest loans

Five hundred and fifty one households have obtained low-interest loans from government-related financial institutions (such as the Agricultural Bank of China and Rural Credit Cooperatives) with a coverage rate of 80.56% and an average loan amount of 54,181.29 yuan. With the support of the financial policies, banks, and financial institutions provide low-interest loans to farmers based on their household credit ratings, with no collateral or guarantee, to encourage them to develop industries such as special planting and breeding, family hotels, and ethnic handicrafts, and to promote specialization and off-farm employment. The loan amounts of most households range from 40,000 to 80,000 yuan (Fig. 7). The result showed that low-interest loans provide adequate financial support.

Fig. 7
figure 7

Household numbers with different low-interest loan amounts in the sample farmers of the PRB

Full size image

4.3.3 Agricultural technique training

Agricultural technique training encompassed instruction on using agricultural machinery, techniques for preventing and controlling crop pests and diseases, scientific crop planting techniques, and disseminating agricultural management knowledge. In the context of climate change, traditional farming methods face challenges due to an over-reliance on natural conditions. Farmers participated in agricultural training organized by the local government to systematically learn agricultural knowledge and skills. This training can improve agricultural productivity and enhance the intensification of agricultural production, thereby reducing dependence on natural conditions and bolstering the capacity to cope with climate change risks. Of the households surveyed, 338 did not receive agricultural technology training, while 346 participated in agricultural technique training. They reported that the training explained climate change, the essentials of agricultural production, and the detailed application of new technologies. By acquiring these skills, it may be possible to diminish households’ reliance on traditional climate adaptation methods—crop diversification.

4.3.4 Road construction

Road construction serves as fundamental, pioneering, and service-oriented infrastructure in rural areasakin to the “capillaries” that play a crucial role in economic development. It facilitates the growth of local special industries, ecological tourism, commodity processing, logistics and express delivery, and other emerging sectors. Furthermore, it contributes to the shift from conventional farming practices to modern agricultural methods. There is a Chinese saying that “To get rich, first build a road (yao xiangfu, xian xiulu); if there have roads, all livelihoods will prosper (gonglu tong, baiye xing)”. Roads are the key to opening the door for material and information communication with other places, helping farmers get rich and indirectly improving their ability to adapt to climate change. We also counted the household numbers at different distances from their houses to the nearest county-level or higher road. A total of 440 houses were within a distance greater than 0 km and less than or equal to 1 km. However, only 86 houses were located more than 1 km and less than or equal to 5 km away. Surprisingly, 158 houses were located beyond 5 km. The statistics indicated that more than 20% of the households were still inconvenient in terms of transportation and had difficulty contacting the outside.

In addition, we grouped the farmers under each policy measure according to the amount or type and counted the level of crop diversification among the different groups, as detailed in the supplementary materials.

4.4 Regression results

4.4.1 Baseline results

The impacts of climate change on crop diversification are shown in Table 4. In conformity with the final results (column 4), climate change significantly affected crop diversification. The results showed that farmers will improve crop diversification when they perceive an increase in temperature or a decrease in precipitation. Among the livelihood asset indicators, the age of the household head, education level of the household head, off-farm employment income, and agricultural equipment were significantly and positively associated with crop diversification, and the livestock number was significantly and negatively associated with crop diversification.

Table 4 The impacts of climate change on crop diversification in the sample farmers of the PRB
Full size table

4.4.2 Interaction effects

Different policy measures interacted with farmers’ perceptions of temperature and precipitation as independent variables and crop diversification index as the dependent variable to explore farmers’ crop diversification decisions under policy measures (Tables 5 and 6).

Table 5 The interactive effect of policy measures and temperature changes on crop diversification in the sample farmers of the PRB
Full size table
Table 6 The interactive effect of policy measures and precipitation changes on crop diversification in the sample farmers of the PRB
Full size table

Specifically, as shown in columns (3) and (4) of Table 5, the estimated coefficients of the interaction term between low-interest loans and temperature perception, agricultural technique training, and temperature perception were significantly negative. Comparing the results in Table 4, where the coefficient on temperature perception was significantly positive, the results suggested that low-interest loans and agricultural technique training buffered the negative effects of climate change and reduced farmers’ willingness to adopt crop diversification. Farmers with more loans and attended training had less crop diversification when they perceived the same degree of temperature changes.

As shown in column (2) of Table 6, the estimated coefficient of the interaction between non-agricultural subsidies and temperature perceptions is significantly positive, contrary to the significantly negative coefficient of precipitation perceptions in Table 4, indicating that non-agricultural subsidies also changed farmers’ attitudes towards adopting crop diversification under climate change. Combining the results in Tables 5 and 6, non-agricultural subsidies, low-interest loans, and agricultural technique training all significantly changed the direction of impact when interacting with climate change perceptions, suggesting that policy measures led to different adaptive behaviors of farmers under the same climate perception scenario.

5 Discussion

5.1 Climate change and crop diversification

The results indicated that farmers perceive an increase in temperature as promoting crop diversification, while an increase in perceived precipitation tends to decrease crop diversification. According to our survey (Fig. 4), farmers’ perceptions of temperature change in the PRB generally aligned with actual conditions, and an increase in temperature significantly correlated with higher levels of crop diversification. In Kenya, Ochieng et al. (2020) observed that farmers would adopt crop diversification in the presence of high temperatures. Similar findings are found for farmers’ perceptions of precipitation change and crop diversification. In relatively dry areas, farmers perceive precipitation decrease, prompting them to adopt crop diversification to cope with possible drought. For example, Zhang et al. (2022) found that farmers increase their level of crop diversification if they perceive precipitation decreases in the Hehuang Valley of the TP, an arid and semi-arid region. Similarly, Sertse et al. (2021) found that in the arid Raya Azebo district of Ethiopia, farmers were aware of precipitation decreases, with 99% responding through crop diversification. More than 60% of farmers perceived an increase in precipitation, which was inconsistent with actual precipitation trends and might affect their ability to adapt to climate change. Farmers who perceive an increase in precipitation due to a bias in perception tend to have lower crop diversification. However, meteorological stations indicated that the region would be drier, posing a severe threat to these farmers. Therefore, we propose to provide accurate weather information to farmers to improve their accuracy in awareness of climate change.

5.2 The moderation effects of policy measures on climate adaptation behavior by farmers

Non-agricultural subsidies had a significant negative effect on crop diversification. Non-agricultural subsidies provided by the government increased farmers’ household assets and improved their ability to resist risks. The average amount of non-agricultural subsidies accounted for 35% of the average annual gross income of the 684 households we surveyed, indicating that these subsidies were important to these households. Non-agricultural subsidies attenuated the risk of agricultural disasters to which farmers may be exposed due to climate change and reduced their livelihood vulnerability so that they do not have to adopt crop diversification to cope with risks, thus negatively impacting crop diversification.

Household loans had a significant negative effect, implying that the more low-interest loans a farmer receives, the less crop diversification there is. The statistical results showed that among households that have received low-interest loans, 92.38% have loans of less than 100,000 yuan, suggesting that microloans may have been invested in agriculture to diversify the crop portfolio. However, as loan amounts increase, farmers may choose to invest them in another field (e.g., animal husbandry, developing rural tourism, and leisure agriculture). This finding echoes Kemboi et al. (2020). Farmers’ access to credit in Elgeyo Marakwet County, Kenya, significantly negatively affected crop diversification, as most farmers use loans for purposes other than crop production. However, in Northeast India, Kumar et al. (2022) found credit to increase farmers’ investment and risk-taking ability and improve the level of crop diversification. This difference may be explained by the fact that farmers in Northeast India are more inclined to diversify their agriculture due to differentiated policy and geographical contexts. In the PRB, higher amounts of low-interest loans and diversified subsidies have reduced crop diversification. Such financial policy measures have reduced farmers’ vulnerability and improved farmers’ ability to cope with climate change.

Agricultural technique training had a significantly negative effect on crop diversification. Makate et al. (2016) found that farmers who obtain agricultural extension services are 38.4% more likely to adopt crop diversification than those who do not. The governmental agricultural extension organization has technical knowledge of crop production and agricultural management that can help farmers implement crop diversification. However, the impacts of this variable in the PRB were the opposite, as the fact that agricultural experts promoted drought-resistant crops at the agricultural technique training in the PRB, which were greatly adapted to the arid climatic environment so that farmers did not need to rely too much on diversified planting to resist climatic risks. Moreover, Wang et al. (2020) pointed out that households in PRB generally have irrigation facilities, thus the agricultural technique training may be more focused on agricultural intensification and specialization.

For the climate change adaptation of farmers in developing countries, a large number of studies have recommended that governments take policy measures to help farmers (Bhatta & Aggarwal, 2016; Challinor et al., 2007; Eshetu et al., 2021; Fahad & Wang, 2018; Jamshidi et al., 2019). Although governments in most developing countries have provided policy assistance to farmers, the effects have not always been satisfactory (Chinse et al., 2019; Waldman et al., 2017). Because government policy support in these countries is limited (Wilk et al., 2013), policy measures do not cover most farmers (Mwinjaka et al., 2010), or policy measures are not well embedded in rural households (Udmale et al., 2014), most farmers continue to face livelihood dilemmas caused by climate change. This study found that the policy measures adopted by the Chinese government perform well, enabling farmers to cope with climate change without increasing their crop diversification levels and lowering the adaptation threshold.

5.3 Policy implications

The results indicated that crop diversification is merely an adaptation strategy that farmers are forced to adopt. Farmers were more willing to take on certain risks with cash or credit support and may not lean towards diversified cultivation, as specialized practices often yield higher efficiency.

While increasing the level of crop diversification or augmenting government policy support contributes to enhancing farmers’ climate change adaptation capabilities, the heightened crop diversification demands additional management time and labor investment. Simultaneously, the governmental financial resources are not inexhaustible, posing constraints on the potential expansion of support. Thus, it becomes imperative to strike a suitable balance between crop diversification and government assistance, tailoring diverse improvement measures for different types of farmers.

The government should prioritize farmers with higher levels of crop diversification, providing them with increased cash subsidies, accessible credit channels, and more targeted technical assistance to enhance their climate change resilience. Furthermore, continuous monitoring of these farmers is crucial. The government could help them expand production scale, elevate specialization levels, and facilitate the transition towards agricultural intensification and modernization when suitable.

For farmers with lower levels of crop diversification, the government may consider offering policy-oriented agricultural insurance if they are heavily reliant on agriculture. Given their vulnerability to climate variability, this measure can help mitigate the impacts of climate fluctuations. Alternatively, if these farmers primarily engage in non-agricultural activities, the government could provide more training for non-agricultural occupations, enhancing their employment skills and offering support regarding non-agricultural employment information and opportunities. This approach aims to diversify farmers’ income sources, reduce their dependency on agricultural activities, and enhance their overall livelihoods.

6 Conclusions

Based on a survey of 684 households in the PRB of the Tibetan Plateau, this study explored the impact of policy measures on crop diversification under climate change. The econometric results showed that non-agricultural subsidies, low-interest loans, and agricultural technique training significantly negatively affected crop diversification. The findings suggested that effective policy measures reduce crop diversification as they reduce farmers’ exposure to climate shocks, allowing them not to adopt additional adaptation strategies. However, it should also be noted that the reduction of crop diversification may also lead to a resurgence of agricultural risks. Therefore, the government needs to further implement precise policies and adopt differentiated policies for farmers with different levels of crop diversification.

This study also has limitations. Due to the lack of a long-time series of follow-up questionnaire data, we cannot know the differences in crop diversification across seasons and whether farmers pursue the development of agricultural specialization and intensification afterward, which needs follow-up in the future.