ZHENG Yang-yang ,ZHU Tie-hui ,JIA Wei

1 College of Economics and Management,Nanjing Forestry University,Nanjing 2100372,P.R.China

2 Institute of Agricultural Economics and Development,Chinese Academy of Agricultural Sciences,Beijing 100081,P.R.China

Abstract China is characterized as ‘a large country with many smallholder farmers’ whose participation in modern agriculture is key to the country’s modern agriculture development. Promoting smallholder farmers’ adoption of modern agricultural production technology is one effective way to improve the capabilities of smallholder farmers. This paper aims to explore the impact of Internet use on the adoption of agricultural production technology by smallholder farmers based on a survey of 1 449 smallholders across 14 provinces in China. The results suggest that Internet use can significantly promote technology adoption,with the probability of adopting new crop varieties,water-saving irrigation technology and straw-returning technology increasing by 0.200,0.157 and 0.155,respectively. Furthermore,the effect of Internet use is found to be heterogeneous with a greater effect on smallholder farmers having low education levels,limited training,and high incomes. To increase agricultural production technology adoption by smallholders,rural Internet infrastructure and Internet use promotion should be the focus for the Chinese government.

Keywords:Internet use,smallholder farmers,agricultural technology adoption

1.Introduction

Building the connection between smallholder farmers and modern agriculture is key to rural revitalization and an important component of agricultural modernization.Numerous studies have examined approaches that promote the connection,such as improving the organization of smallholder farmers and agricultural socialized services,and implementing specific models such as land trusteeship,land stock cooperatives and‘new agricultural business entities+smallholder farmers’(He and Wu 2019;Ye and Zhang 2020). However,existing literatures primarily focus on external drivers,that is,they pay more attention to providing services to smallholder farmers than building the capabilities of smallholder farmers. The adoption of advanced agricultural technology is an important means to build smallholder farmers’ capacity. On the one hand,adopting advanced agricultural technology can change the traditional farming methods of smallholder farmers,optimize the allocation of agricultural production essential factors,reduce agricultural production costs,and improve agricultural production efficiency (Muzari et al.2012;Wossen et al.2019). On the other hand,advanced production technologies contribute to achieving the goal of green agricultural development by reducing water use or improving soil quality (Asfaw et al.2012). Moreover,the adoption of agricultural production technology can create more market opportunities and increase income,thereby breaking the poverty trap and promoting economic growth (Asfaw et al.2010;Wossen et al.2015,2017). Schultz (1964) pointed out in Transformation of Traditional Agriculture that the key to transforming traditional agriculture is to introduce modern production technologies.

To effectively promote smallholder farmers’ adoption of advanced agricultural technologies,it is necessary to identify the key barriers of the adoption. Limited access to information is considered as the main obstacle. The probability of adopting new technology decreases significantly if farmers lack information about the benefits or how to correctly use the technology (Jack 2013). Aker(2010) believed that information supply and technology supply are equally important in the process of promoting agricultural technology. Therefore,how to enable farmers to obtain technical information conveniently and reduce information asymmetry becomes an important research question (Qiu et al.2016). The Internet,as an important means for smallholder farmers to obtain information,has a profound impact on the behavior and thinking of smallholder farmers. First,the Internet can reduce farmers’ information search cost,enabling them to obtain technical information in a timely and convenient manner.The Internet also enhances farmers’ understanding of the ‘risk-benefit aspects’ of agricultural technology,and breaks the traditional impression of production technology(Genius 2006). Second,Internet use can increase the bargaining power of smallholder farmers,expand the agricultural sales market,improve agricultural production performance,and in turn encourage farmers to adopt new technologies (Aker and Ksoll 2016;Donovan 2017).Third,Internet use can improve the management and learning capabilities of smallholder farmers,enhance their skilled human capital,and increase their income(Gao 2018;Leng et al.2020). In addition,the Internet has established an effective platform for sharing and disseminating environmental protection concepts,policies and knowledge,which has increased farmers’ awareness of environment protection literacy. Farmers with different levels of knowledge can equitably obtain and exchange environmental knowledge through the Internet,and post their comments on current environmental issues (Peng et al.2019;Gong et al.2020).

Exploring the use of the Internet has important theoretical and policy implications for technology adoption by smallholder farmers. Studying the role of Internet use in promoting technology adoption by smallholder farmers both enriches the smallholder production theory and expands the boundaries of the Information and Communication Technology (ICT) research. According to the Third National Agricultural Census in 2016,the number of smallholder farmers in China accounted for more than 98% of the total rural households,and the area of arable land operated by smallholder farmers accounted for 70% of the total arable land. Therefore,investigating agricultural technology adoption by smallholder farmers is of great significance to agricultural modernization in China. This paper is innovative in a number of ways.First,it focuses on how to achieve the connection between Internet use and agricultural technology choice through capacity building of smallholder farmers rather than through merely providing external supports.Second,different from other studies,this study examines agricultural technology adoption before,during and after production,respectively,as well as the heterogeneous effect of Internet use on technology adoption. Third,the study uses a unique survey data collected by the National Agricultural and Rural Development Research Institute,China Agricultural University. The sample size is large(N=1 449) and representative,covering 14 provinces/autonomous regions of China.

2.Theoretical framework

This paper draws on the theoretical framework of Yoav and Shchori-Bachrach (1973),and incorporates‘information’ as an input factor into the production function of farmers. The function is given as follows:

where Ytis the output in period t,L is land input,and Ntis other input in period t,such as new crop varieties,soil,fertilizer,etc.,Ktis the accumulated information during t,g and F are concave functions. Assuming F(·) has a constant return to scale,the function is defined as follows:

where yt=Yt/L;nt=nt/L;f=F(1:nt)/L. Assuming that f(s) has good production function characteristics. The hypothesis of Kislev and Shchori-Bachrach (1973) is applicable to the knowledge function g(·),which is the marginal contribution of knowledge is positive and shows a decreasing trend. As knowledge grows,the knowledge function converges to an upper limit,which means that with the accumulation of knowledge,the marginal productivity of information increment tends to reach zero. When knowledge is 0,the value of g(·) is very low or even 0.

When defining the cumulative information Ktand its acquisition,it is assumed that farmers obtain it through passive and active learning,and active information collection involves costs such as time or money. It is assumed that the marginal costs of information obtained at any given time are increasing.

The knowledge state (K) at time t is defined as follows:

where htdescribes the information acquired passively during this period of time. htmay increase over time. The variable Atindicates the degree of purposeful knowledge acquisition,and the information cost Ctis expressed as follows:

The profit function π of the farmer households with a scale of L is expressed as follows:

Let p denote the price of n,combine eqs.(3) and(5),and obtain the first-order derivation of n and A,respectively:

where πnand παrepresent the partial derivatives of profit with regard to n and A,respectively. Assuming that πn=0,then dn/dkt=-(g′f′/gf′′)＞0;it means that the more information farmers obtain,the more likely they are to adopt new technologies. Once a new technology is adopted,as long as the marginal productivity of information (g′) is positive,farmers with a higher level of information accumulation will adopt more new technologies. Considering two farmers with the same scale of operation,the framer acquiring more information will adopt new technologies faster than the other.Assuming that the threshold information level K＊ when g(K)=,K=K＊ is within the range of K value,the number of farmers adopting new technologies will increase until Ktexceeds K＊.

3.Data and models

3.1.Data

This study utilized the Survey of Returning Home from January to February 2019 by the National Agricultural and Rural Development Research Institute of China Agricultural University. The enumerators included undergraduate,master and doctoral students among various majors of China Agricultural University1The questionnaire mainly focuses on the national grain production survey. First,the research team determined the number of samples in each province,region,and city based on the grain production area. Secondly,according to the surveyed counties (cities)corresponding to the corresponding household registration students;the students conduct surveys on the township where the household registration is located,and select about 1-2 villages,about 15 households.. Before data collection,the enumerators were provided trainings to explain the aims,question structures,and key challenges of the questionnaire.2Since all majors of China Agricultural University are related to agriculture,the recruited enumerators have a high level of awareness of agriculture and therefore data collection is of high quality.A number of researchers were responsible for sending and receiving questionnaires and answering questions encountered during data collection,while returned questionnaires were reviewed by research leaders. The survey questionnaire included two levels,farmer questionnaire and village-level questionnaire. In each village,15-20 households were randomly selected.The survey obtained 1 952 grain grower questionnaires in total. Since this paper mainly focuses on smallholder farmers,the samples of non-smallholders with land of 30 mu (1 ha=15 mu) and above were excluded,according to the World Bank’s classification standards for smallholder farmers3The World Bank classifies rural households with an average arable land area of less than 2 hectares (30 mu) as smallholder farmers.. Inconsistent and incomplete questionnaires were also dropped,resulting in a dataset of 1 449 samples from 144 villages in 119 counties/districts of 14 provinces/autonomous regions sparsely located in eastern,central,western and northeastern China (i.e.,Inner Mongolia,Jilin,Sichuan,Anhui,Shandong,Jiangsu,Jiangxi,Hebei,Henan,Hubei,Hunan,Gansu,Liaoning and Heilongjiang),demonstrating a good regional representativeness(Table 1). Heilongjiang and Liaoning,as major agricultural production provinces,have relatively few samples. The main reason is that there are more large-scale farmers in Northeast China than in other regions.4The sample in Heilongjiang is smaller than that in Inner Mongolia. The possible reason is that there are more students participating in the Inner Mongolia survey,and the survey sample may be relatively high.

Table 1 Distribution of samples by province/autonomous region

3.2.Variable selection

The dependent variables of this study are smallholder farmers’ agricultural technology adoption of three technologies:new crop varieties,water-saving irrigation technology and straw-returning technology. New crop varieties refer to those with stronger disease resistance or more obvious yield-increasing effects than general crop varieties,but the price is relatively high. Water-saving irrigation technologies refer to production technologies that can effectively save water,such as sprinkler irrigation,micro-sprinkler irrigation,subsurface irrigation and drip irrigation. Straw-returning technology refers to measures to increase fertility and increase production,whichimprove traditional incineration methods and avoid air pollution caused by incineration,including straw crushing and pressing,returning to the field with straw mulching,returning to the field by piles,returning to the field by burning,and returning to the field. Each new technology was assigned a value of 1 if the farmer adopted it,and 0 otherwise.

Internet use is the key independent variable. It is considered that even farmers have smart phones or Internet skills,they may not obtain agricultural information through the Internet. If smallholder farmers were asked directly whether they have smart phones or use the Internet,it may cause measurement errors (Ma and Zhu 2020;Nie et al.2020). The survey directly asked whether farmers use the Internet to obtain agricultural information in order to accurately measure information access through the Internet5The questionnaire asked“did you use the Internet to obtain agricultural service information?”..

This study chose the control variables from three categories:personal characteristics of farmers including age,education level,health status,and training experience (Gao et al.2018;Ma 2020);family characteristics including number of laborers,the proportion of non-agricultural income,quality of cultivated land,farm size,number of plots,household income,and subsidies (Boz 2016;Gao et al.2018);and village characteristics including poverty,economic development level,water source guarantee,and distance to the nearest trunk road (Tatlıdil et al.2009).

As shown in Table 2,farmers adopting water-saving irrigation technology,new crop varieties and strawreturning technology accounted for 28.36,63.15 and 63.35% of the total sample,respectively. Those who used the Internet to obtain agricultural production information accounted for 14.5% of the total sample. Those who received training accounted for 20.8% of the total sample.The average age of smallholder farmers was 54 years,the average number of family agricultural labor was 2.2,the average farm size was 5.7 mu,and the average annual household income was 56 000 CNY (6.5491 CNY=1 USD,according to the latest exchange rate on December 9,2020). The poverty-stricken villages and villages with guaranteed water sources accounted for 23.4 and 76.1%of the total sample,respectively.

As shown in Table 3,among the smallholder farmers who used the Internet,76.67% adopted new crop varieties,43.33% adopted water-saving irrigation technologies,and 73.81% adopted straw-returning technology. Among the smallholder farmers who did not use the Internet,60.86% adopted new crop varieties,25.83% adopted water-saving irrigation technologies,and 61.58% adopted straw-returning technology.

Table 2 Descriptive statistics of variables

Table 3 Internet use and agricultural technology adoption

3.3.Model selection

Binary probit regression modelSince adoption of agricultural technology is a binary choice,the binary probit regression model was used to analyze the influencing factors for smallholder farmers’ technology adoption. The model is defined as follows:

where Technologyiindicates whether a farmer adopts the agricultural technology;P is the probability of adoption,Internetiis Internet use,β1is the coefficient of Internet use;ziis a vector of control variables. γ is a vector of the coefficient of the control variables,regioniis a vector of dummy variables for different regions in China,and δ is a vector of coefficients of regional dummy variables.

Endogenous switching modelInternet use does not satisfy random sampling,that is,there are systemic differences in the initial conditions before Internet use,and there is a problem of ‘self-selection’ in Internet use by smallholder farmers. Internet use decision-making is the result of a combination of many factors,and some factors are unobservable,such as farmers’ cognition of the Internet. Taking into account the selection bias caused by observable and unobservable factors and being inspired by Abdulai and Huffman (2014) and Ma et al.(2018),this study adopted the Endogenous Switching

Model to empirically analyze the impact of Internet use on the adoption of agricultural production technology by smallholder farmers.

In the first stage,the probit model was used to estimate the probability of smallholder farmers’ Internet use:

where Interneti＊ is an unobservable latent variable. When Interneti＊＞0,Interneti=1,that is,farmers use the Internet.Siis the control variable. In addition to including the variables in Zi,it also includes the variable ‘proportion of households using a computer’. α and β are parameters to be estimated,and μiis a random disturbance term.

The second stage estimated the impact of Internet use on technology adoption by smallholder farmers,

where Technology1iand Technology0irespectively represent the adoption of agricultural production technology by small farmers who use the Internet and those who do not use the Internet. α1,α0,γ1,γ0represent parameters to be estimated,and ε1i,ε0irepresent random disturbance items.

Based on the eqs.(11) and (12),the agricultural production technology adoption of small farmers using the Internet can be written as eq.(13),and the counterfactual model can be expressed as eq.(14).

The average treatment effect on the treated (ATT) of the agricultural production technology adopted by small farmers using the Internet can be expressed as the difference between eq.(13) and eq.(14).

4.Results and discussion

Firstly,using a probit model,this paper estimated the baseline impact of Internet use on the adoption of agricultural production technology of smallholder farmers.Secondly,based on farmers’ education level,training and income,the heterogeneous influence of Internet use on farmers’ agricultural production technology adoption was estimated. Finally,the Endogenous Switching Model was used to address the self-selection problems of Internet use.

4.1.Baseline regression

As shown in Table 4,multi-collinearity check was conducted to obtain a variance inflation factor of 1.28,indicating that there is no serious multi-collinearity among the independent variables. The regression coefficients showed that Internet use increased the probability of adopting new crop varieties,water-saving irrigation technology and straw-returning technology by 0.200,0.157 and 0.155,respectively. As an important means for information acquisition,Internet use can help farmers obtain more market information,technical information and policy information. In neoclassical economics,the market is perfect and fully competitive. However,in reality,farmers face information asymmetry and market price fluctuations as sellers. Farmers must consider where the agricultural products are sold,how they are sold,and to whom they sell in order to maximize profit.According to Transaction Cost Theory,information search cost is an important part of transaction costs,and reducing information search costs can increase sales in farmers’ markets performance,promoting the adoption of agricultural technology for smallholder farmers (Tadesse and Bahiigwa 2015). For example,smallholder farmers can obtain price information timely through the Internet and make judgments on future price trends,which reduces information asymmetry and market risk,and consequently the risk of agricultural technology adoption.Similarly,through the Internet,smallholder farmers can timely and accurately understand the potential risks and benefits of new technologies,and master the application and operation of new technologies,thereby reducing the risks and uncertainties of new technologies. In addition,smallholder farmers can obtain more agricultural policy information through the Internet. Especially under the background of Rural Vitalization Strategy,Internet use can enhance farmers’ policy awareness and explore market opportunities based on the latest agricultural policies,thus stimulating smallholder farmers’ agricultural technology adoption. In addition,Internet use had a greater effect on the adoption of new crop varieties than on that of the other two technologies,which may be due to the more obvious effect of new crop varieties on crop yield and income. Specifically,compared with the other two technologies,smallholder farmers pay more attention to the relevant information about new crop varieties and can obtain more information on new crop varieties by using the Internet. For control variables at the individual level,the effect of age on the adoption of water-saving irrigation technology was significantly negative. This may be due to the fact that older people are more reluctant to accept new information and adopt new technology than their younger counterparts. The effect of education on the adoption of water-saving irrigation technology was significantly negative,which may be due to the fact that farmers with higher education levels are more engaged in non-agricultural employment,and thus show less willingness to invest in agricultural productions.Training had a significant and positive impact on the adoption of water-saving irrigation technology and strawreturning technology. This is expected as training can increase farmers’ understanding of agricultural production technologies and increase their capacity of implementing these technologies (Jia et al.2013).

Table 4 Regression results:The impact of Internet use on the of agricultural technology choice1)

Regarding the household-level controls,the more labor a household had,the greater the probability of it adopting straw-returning technology. One possible reason is that a certain amount of labor is needed to implement the technology. The proportion of non-agricultural income had a significantly negative impact on the adoption of new crop varieties and straw-returning technology. The reason may be that the higher the proportion of non-agricultural income,the less time,labor and capital smallholder farmers spend on agriculture. This is contrary to thefinding of Issahaku and Rahaman (2019) that farmers’increased income through non-agricultural employment can promote their adoption of sustainable production measures. It is also a common expectation that the more high-quality cultivated land the households have,the more likely they are to adopt new technologies. However,this study found that larger farm size would reduce the probability of adopting new crop varieties and water-saving irrigation technology. This is in line with Hu et al.(2019)that higher cost and risk associated with larger land may reduce farmers’ willingness of adopting new technologies.A few family-level controls had mixed influences on the adoption of new technologies. For example,the number of plots had a positive impact on the adoption of new crop varieties while having a significantly negative effect on the adoption of straw-returning technology. The possible reason is that the more scattered the number of land parcels,the lower the land income;therefore,smallholder farmers with a large number of plots are more likely to adopt new crop varieties to improve land productivity.Meanwhile,the more dispersed the land,the higher the cost of straw returning. Similarly,household income had a significantly negative effect on the adoption of water-saving irrigation technology but a positive effect on the adoption of straw-returning technology. It is assumed that high family income is largely due to off-farm employment. This means that high income families have less labor and time to be spent on agriculture,and thus are lacking the amount of labor required by water-saving irrigation technology. However,higher income promotes the adoption of straw-returning technology because it improves the risk resilience of smallholder farmers (Carter et al.2016). The impact of subsidy on the adoption of water-saving irrigation technology was significantly negative,while that on the adoption of straw-returning technology was positive. The possible reason is that the water-saving irrigation technology requires sufficient funds while the straw-returning technology does not.

Regarding the village-level controls,living in a poor village had a significantly negative impact on the adoption of straw-returning technology,which may be due to their financial constraints. It was also found that the higher the level of economic development,the higher the probability of adopting water-saving irrigation technology and strawreturning technology. It is intuitive that farmers living in villages with better economic conditions have a higher level of awareness of new technologies than those in villages with worse economic conditions. Compared with smallholders in the eastern region,the probability of adopting new crop varieties in the middle region was reduced by 0.098,the probability of adopting strawreturning technology in the western region was reduced by 0.253,and the probability of adopting new crop varieties,water-saving irrigation technology and straw-returning technology in the northeastern region was decreased by 0.087,0.191,and 0.240,respectively. The possible reason is that the eastern region has higher income,more advanced production concepts,and more inclined to adopt production technology than other regions. The land output in the western region is relatively low. In order to reduce production costs,farmers are unwilling to adopt the technology of straw-returning technology.Smallholders may plant corn,wheat or rice at the same time,which makes it difficult to distinguish them by crop type. Therefore,this paper does not divide samples by crop types.

4.2.Heterogeneity analysis

The heterogeneous impact of Internet use on adoption of agricultural production technology was investigated from three aspects:smallholder farmers’ education,training and income (Tables 5-7). In terms of education,this paper classified primary schooling and below as a low education level,and junior high schooling and above as a high education level. The marginal effects of Internet use for the low education group were significantly higher than that of the high education group. This indicated that farmers with low education level had limited access to information and lacked agricultural production technology information. The use of the Internet can significantly enrich farmers’ information resources and promote the adoption of modern production technology. As Hojo (2002) put forward,agricultural production technology adoption varied across education levels. Farmers with an average or low education level had a higher probability of technology adoption than those with a higher education level.

Table 5 Heterogeneity analysis on farmers’ education characteristics

Table 6 Heterogeneity analysis on farmers’ training characteristics

In terms of training,the marginal effects of Internet use among the trained smallholder farmers were significantly higher than that of the untrained ones. One explanationmay be that the trained smallholder farmers can have higher agricultural production skills and obtain more information through Internet use,which promotes the adoption of agricultural production technology. This is consistent with the conclusion of Jia et al.(2013)who found that training could reduce farmers’ fertilizer consumption to a certain extent.

Regarding the income level,smallholder farmers with less than the average income were grouped as lowincome farmers,and the rest were grouped as highincome farmers. The marginal effects of Internet use for high-income farmers were significantly higher than that for the low-income group. This may be due to the fact that the high-income group can obtain more information about agricultural production technology through the Internet,and are more familiar with the risks and benefits of agricultural production technology;while high-income smallholder farmers have more capital to invest in agriculture and bear the risks of agricultural production technology.

4.3.Robustness test

This paper used endogenous switching model (ESM)method to address endogenous problem,and the results are shown in Table 8. The robustness of results showed that Internet use had a significant positive impact on agricultural production technology adoption. First,Internet use could significantly promote smallholder farmers’adoption of new crop varieties by 27.6% (ATT is 0.276,t-value is 16.08). Second,Internet use could significantly promote the adoption of water-saving irrigation technology by 38.4% (ATT is 0.384,t-value is 20.17). Third,Internet use could significantly promote the adoption of strawreturning technology by 9.4% (ATT is 0.094,t-value is 5.40). These results combine to show that Internet use had different effects on the adoption of technologies by smallholder farmers. Specifically,Internet use had a greater effect on the adoption of new crop varieties(27.6%) and water-saving irrigation technologies (38.4%)than on the adoption of straw-returning technology (9.4%).

Table 7 Heterogeneity analysis on farmers’ income levels

Table 8 The average treatment effect of the impact of Internet use on the adoption of agricultural production technologies

5.Conclusion and policy recommendations

5.1.Conclusion

Using data of 1 449 farmers from 14 provinces/autonomous regions collected by National Agricultural and Rural Development Research Institute of China Agricultural University,from January to February 2019,and employing probit model,IV-probit model and PSM method,this paper explored the impact of Internet use on smallholder farmers’ adoption of agricultural production technologies. The results showed that Internet use could significantly promote the adoption of agricultural production technologies by smallholder farmers,increasing the probability of adopting new crop varieties,water-saving irrigation technology and straw-returning technology by 0.200,0.157 and 0.155,respectively.Heterogeneity analysis showed that the impact of Internet use on smallholder farmers’ technology adoption had a greater impact on the farmers with low education level,training experience and high income. After addressing the self-selection and endogenous problems of Internet use,the conclusion was still robust. This study nevertheless has some limitations. First,surveyed households were not randomly selected and were more based on the social relationship of the investigators rather than the strict random sampling approach;Second,there were more students participating in the survey in some provincesthan in other provinces,which may also lead to more samples being recovered in certain provinces. Third,the usage of dichotomous dependent variables may not be optimal,missing some important variations such as the degree and time of technology adoption.

5.2.Policy recommendations

This paper puts forward three policy recommendations.First,it is necessary to strengthen the construction of rural Internet infrastructure. At present,the infrastructure has not fully met the Internet use needs of smallholder farmers,primarily because of the problems of high network cost and slow network speed in many rural areas. It is in urgent need to strengthen the construction of agricultural information,disseminate the information to villages and households,and encourage the digital transformation of agricultural production. Second,it is important to provide training to smallholder farmers,assisting them in acquiring or improving Internet use skills. By December 2018,there were 222 million Internet users in rural areas of China,accounting for 26.7% of total Chinese internet users,and a 12.91 million increase from 2017. The rural Internet popularization rate was 38.4% in 2018,0.3 percentage point higher than that at the end of 2017 (China Internet Network Information Center 2019).More attention should be paid to the training of Internet use skills for smallholder farmers. Taking into account the awareness and acceptance of smallholder farmers,the training content and training methods about smallholder farmers should be different from those for large-scale farmers. Finally,policy formulation should take into consideration the heterogeneity of smallholder farmers.The effectiveness of Internet use varies significantly by characteristics of farmer individuals and households.Therefore,the government should formulate targeted policies according to the needs of different smallholder farmers and provide tailored agricultural production methods to farmers of different knowledge,skill and income levels.

Acknowledgements

This work was supported by the Agricultural Science and Technology Innovation Program of Chinese Academy of Agricultural Sciences (CAAS-ASTIP-IAED-2020-06;CAAS-ASTIP-IAED-2021-SR-02;CAAS-ASTIPIAED-2021-06).

Declaration of competing interest

The authors declare that they have no conflict of interest.