Measurement and Spatial Differentiation of Farmers’ Livelihood Resilience Under the COVID-19 Epidemic Outbreak in Rural China

Abstract

Livelihood resilience is the ability of individuals, families or communities to withstand external shocks based on existing resources. It is an important research paradigm in sustainable development studies. The outbreak of COVID-19 and strict epidemic prevention policies have greatly impacted the production and life of rural farmers in China. The resilience of farmers’ livelihoods during the epidemic is crucial to the sustainable development of their livelihoods and regional stability. This study uses classic buffer capacity, self-organization ability, and the capacity for learning a three-dimension livelihood resilience framework using the comprehensive index, OLS, and geographical detector methods based on Hubei province and neighboring Anhui and Chongqing. Rural household survey data investigate the background of epidemic hit the livelihood of farmers resilience and its spatial distribution pattern and identify the key influencing factors. The results show that the livelihood shock faced by farmers was higher than the risk of disease, and the overall level of livelihood resilience was low after the pandemic. Financial capital and social capital can effectively help farmers to eliminate livelihood difficulties. In contrast, natural capital has a limited driving force, and physical and human capital have no obvious impact. The spatial agglomeration differentiation is obvious, indicating that the impact of COVID-19 on livelihoods was closely related to the degree of local socio-economic development and geographical location. The results of this study provide targeted recommendations for the development of epidemic prevention and livelihood resilience policies tailored to local conditions, emphasizing the importance of boosting livelihood recovery at both the government and household levels.

Introduction

Outbreaks of severe infectious diseases have accompanied the history of human development and have had a great impact, especially the sudden outbreak and rapid spread of COVID-19, which has had a huge impact on the economic development, livelihood, and well-being of countries worldwide (Hinojosa & Acosta 2021; Kelleni 2021; Van 2021). From the experience of severe acute respiratory syndrome (SARS), H7N9 influenza and other outbreaks in the past, most of which occurred in Beijing, Shanghai and other large cities, cases are clustered, so prevention and control measures and health services are mainly focused on urban community residents (Zhang 2007; Zhang et al. 2019; Boterman 2020). However, with the scale of the floating population between urban and rural areas reaching 376 million, the spread mode of “central aggregation + mobile transmission” of the novel COVID-19, coupled with the lack of primary medical resources in rural areas, makes rural areas accounting for 70% of the total population also face extreme risks (Fu & Fu 2020; Kraemer et al. 2020). For example, although the outbreak point of COVID-19 in China is Wuhan, nearly half of the cases of Xiaogan and Huanggang are distributed in towns and villages. At the same time, many sporadic rebound points of the epidemic are also located in rural areas. For example, 76.39% of the cases during the epidemic in Shijiazhuang in January 2021 were distributed in Zengcun town, Gaocheng district, among which farmers accounted for approximately 70%. From the perspective of the current world epidemic, the United States epidemic began to spread to rural areas after August 2020. In May 2021, more than 50% of the infected people in Maharashtra, India, were farmers, while the proportion was nearly one-third in the northern states, which have the largest rural population (Cuadros et al. 2021; Mumtaz et al. 2021).

During the outbreak of COVID-19, the Chinese government has taken some prevention and control measures, such as extending the Spring Festival holiday, delaying the opening of schools, shutting down businesses, and initiating an emergency response to major public health emergencies. These measures have effectively reduced mobile contact between people and the emergence of cluster infections but have also considerably impacted our economy. Due to poor public health conditions, lack of epidemic prevention materials, and inadequate information, rural areas have become critical areas for epidemic prevention. The impact of epidemic disease is bound to affect the production and life of rural residents (Fu & Fu 2020). First, the burden of epidemic prevention expenditure and the obstruction of migrant work overlap, resulting in the fluctuation and imbalance of household income and expenditure. Second, compulsory traffic control hinders seasonal agricultural activities, leading to increased agricultural costs and difficulties in agricultural production and marketing (Rahaman et al. 2021). Third, physical isolation between urban and rural areas led to the suspension of agriculture and tourism operations, and nearly 90% of family farms were forced to suspend business. From February 2020 to June 2020, the Ministry of Finance, Ministry of Agriculture and Rural Affairs, and Poverty Relief Office of the State Council issued documents demanding “a continuous stable guarantee for epidemic prevention and control of agricultural production” and “organizing cadres staying at villages in return to work as soon as possible to do a good job of helping rural households experiencing poverty caused by the epidemic.“ The infectious disease outbreak has greatly weakened farmers’ livelihood capital and livelihood capacity. Therefore, it is key to promote the post-epidemic recovery of rural areas to effectively measure the livelihood resilience of this group and explore core obstructing factors.

By analyzing the spread of COVID-19 characteristics and effective mechanisms in rural areas, this article attempts to build an evaluation index system of farmers’ livelihood resilience after a significant public health event. Then, we combine field survey data from the critical areas of Hubei Province and the surrounding areas, such as rural Anhui and Chongqing, to accurately measure farmers’ livelihood resilience index after the epidemic. At the same time, considering that COVID-19 is highly contagious and the degree of risk is closely related to the direction of population migration, the regional spatial distribution of the livelihood level of rural households after COVID-19 is further calculated. Finally, a regression model is used to determine the obstructive factors affecting the livelihood resilience of this group, and the spatial correlation of the influence of each factor is identified based on geographic detectors, which provide decision support for effectively introducing policies to prevent farmers from returning to poverty due to disease and promoting rural post-epidemic reconstruction and sustainable rural development.

Literature Review

The Impact of the Epidemic on Farmers’ Livelihoods

The epidemic has hindered farming activities, blocked sales channels for agricultural products, and stopped farmers’ migrant work activities, resulting in human, material, financial, and other capital losses. The epidemic has adversely affected farmers’ livelihood activities and livelihood capital. Previous studies on the impact of the outbreak on rural areas and farmers can be roughly divided into four aspects: First, authors realized that health poverty would deprive farmers of development opportunities and abilities, and then demonstrated the poverty-causing effects of infectious diseases, such as economic crowding, labor damage and anxiety (Lyu et al. 2020; Phukan et al. 2004); second, studies measures the livelihood risk and vulnerability of rural households during the epidemic through the impact of the outbreak on income and the accumulation of assets of rural households, using the “exposure–sensitivity–adaptive capacity” (VSD) framework proposed by the United Nations Panel on Climate Change (IPCC) (Acharya & Porwal 2020; Waqar et al. 2020); third, a combined rural medical social security system was constructed, focused on the trigger mechanism of the coping strategies of sick farmers, such as clinic behavior of heterogeneous families (Azam et al. 2020), financing behavior regarding exogenous (social assistance and government subsidies) and endogenous (out-of-pocket expenses) factors (Bhorat et al. 2021), farmer household labor supply adjustment and the change in outdoor labor in vulnerable environments; fourth, the Sustainable Livelihoods Approach (SLA) framework of the UK Department for International Development (DFID) was used to measure the path mechanism of post-epidemic livelihood capital improvement to help farmers recover their livelihoods, from the perspective of accumulation of tangible material wealth and reconstruction of intangible social networks (DFID 1999; Masanjala 2007). However, the overall goal of rural epidemic research is to resist external interference and achieve sustainable development of subsequent livelihoods after diseases affect farmers (Wijayanti et al. 2016).

At present, there are few analyses on the livelihood resilience of farmers after high-risk infectious diseases. For example, studies on AIDS and Ebola mainly focus on medicine and education (Bickler et al. 2015; Wu et al. 2019; Burgos-Soto et al. 2020; Kelleni et al. 2021), while studies on SARS only have sporadic discussions on income decline (Adger et al. 2009). The quantitative exploration of high-risk infectious diseases is still very weak and not systematic. Currently, scholars have combined the COVID-19 impact with the concept of resilience to study sustainability issues. For example, Rusciano & Gatto. (2022) discussed the impact of COVID-19 on the use and perception of Milan Metropolitan Agricultural Park. Found that restrictions imposed to limit the spread of COVID-19 significantly impacted the well-being of local people. Zhang et al. (2022) proposed a three-level framework of public safety resilience under public emergency, including individual, community and government levels. They argued that community resilience could improve individual resilience in the context of the COVID-19 pandemic, and government resilience positively impacts community and personal resilience. Other studies have been carried out from the perspectives of psychological resilience (Hosgor & Yaman 2022), urban resilience (Chen et al. 2021), and supply chain resilience (Vanany et al. 2021). Still, there have been no studies on the resilience of livelihoods during the epidemic from the micro subjects.

Measure of Farmers’ Livelihood Resilience

The word resilience is derived from the Latin resilio, which means the action of bouncing back to the original state. The concept of resilience was initially applied in physics and engineering to measure the ability of materials to rebound to and restore their original state when disturbed or under pressure. Holling (1973) introduced the concept of resilience into the field of ecology for the first time in the 1970s. He considered that resilience determines the persistence of relationships within systems and measures the ability of these systems to absorb changes in state variables, driving variables, and parameters and remain unchanged. Subsequently, the research field has gradually expanded to various disciplines, such as social economics and social ecology, widely used in disaster risk management, climate change, the biodiversity crisis, ecological protection and sustainable development. (Folke 2006; Gunderson 2000; Folke et al. 2010). Resilience presents the characteristics of multiple interactions as natural and social multiscale and cross-scale systems. Chambers & Conway (1992) were the first to combine resilience with livelihood research. They proposed that livelihood resilience refers to residents’ initial productivity ability to cope with and recover from external pressure events. Individuals with stronger livelihood resilience are less affected by risk shocks. Livelihood resilience is the extension of the concept of resilience and the socialization and microscopization of research objects. Resilience theory has significant advantages in analyzing the adaptive ability of livelihood systems to cope with disturbances or shocks.

Due to the differences in research perspectives and objects, the conceptual framework and evaluation system of livelihood resilience has not yet formed a unified model. DFID’s SLA framework sees livelihood resilience as an important component of a sustainable livelihoods approach (DFID 1999). Béné et al. (2012) believe that the three key attributes of resilience are absorptive Capacity, adaptive Capacity, and transformative Capacity. Speranza et al. (2014) gave a comprehensive interpretation and definition of livelihood resilience and proposed a measurement index system. According to him, livelihood resilience refers to the ability of individuals or social groups to adapt to stress and disturbances, self-organize and learn in ways that maintain or improve basic structures and functions. Its measurement system includes three dimensions: first, Buffer Capacity when facing risks; second, Self-organization through its own assets and adaptive strategies; Third, Learning Capacity that can absorb the experience and guide subsequent livelihood activities. Based on this, the subsequent research integrates livelihood capital (resource endowment) and further designs the policy evaluation tool of whether livelihood recovery can be effectively achieved. Alinovi et al. (2008, 2010) proposed a conceptual framework aimed at linking resilience to food security at the household level, whose conceptual model was later formalized as FAO’s Resilience Index Strategy and Analysis (RIMA) model (FAO 2016). The RIMA-II model views resilience as Capacity rather than an indicator of food security, measured by five indicators: access to basic services, assets, social safety net, sensitivity and adaptive Capacity. Drawing on existing research results, Quandt et al. (2017) proposed a livelihood resilience analysis framework based on livelihood capital. The livelihood resilience analysis framework proposed by Speranza et al. (2014) is most widely used in the field of livelihood research.

Currently, there are two main types of livelihood resilience research: threshold research, which measures whether the system can bounce back to the initial state or switch to a new steady state in the shortest feasible time. These are mainly based on the adaptive cycle theory of resilience. Second, the conceptual framework of livelihood resilience is used to explore the choices of livelihood strategies, barriers and resilience levels of different subjects in the context of climate disasters, institutional change and fragile ecological environment. Adger (2005) discussed the differences in livelihood resilience among different types of residents around Iranian wetlands in the context of water resource shortages. Nyamwanza (2012) defined livelihood resilience as the process in which a community or household can cope with and absorb changes, thus transforming its livelihood model to adapt to changes and challenges. Ramanath (2016) focused on global warming and attempted to analyze the action path of human and social capital enhancement in enhancing the livelihood resilience of Philippine coastal residents. Amy et al. (2017) proposed a livelihood capital-based livelihood resilience analysis framework based on existing research results. Sarker et al. (2019) discussed the resilience and influencing factors of farmers’ livelihoods in Bangladesh.

Further, Kumar et al. (2020) explored the livelihood resilience of farmers in arid areas of India under the impact of climate change. In China, a series of livelihood resilience studies have also gradually formed. Liu et al. (2020) took contiguously poor areas in southern Shaanxi as the research object. They proposed the framework of livelihood resilience and its impact on livelihood strategies under poverty alleviation resettlement. Zhao et al. (2022) evaluated the resilience of farmers’ livelihoods in key ecological function areas. They found that ecological policy, age of the head of household, diversity of livelihood, household size and environmental dependence were the key factors affecting farmers’ resilience.

Livelihood resilience research is an important driving force to achieve the sustainable livelihood of the main body, and the process of livelihood recovery helps to improve the sustainability of livelihood. The measure of livelihood level utilizing livelihood resilience analysis helps to promote the exploration of the causes of vulnerability of livelihood subjects and the importance of coping strategies to potential risks in sustainable livelihood methods. In general, livelihood resilience combines a single livelihood study with resilience, which can evaluate the micro-individual livelihood state in-depth and accurately predict farmers’ livelihood behavior (Quandt 2019; Gong et al. 2020; Pagnani et al. 2021). There are still some shortcomings in the research on livelihood resilience. First, the research framework is generally inadequate and needs further exploration and improvement. Second, most of the existing studies are static measurements and analyses based on cross-sectional data and lack research on the dynamic evolutionary process of livelihood.

Research on the Spatial Difference in Farmers’ Livelihood Level

As the most basic livelihoods of the rural system unit and the main behavior body, farmers are highly dependent on natural resources. Different levels of economic development and uneven distributions of resource endowment among regions result in significant differences in the livelihood level of farmers. The resilience of livelihoods is also significantly different when they are affected by disturbance or pressure. Existing studies on the spatial resilience of rural households’ livelihoods are mainly based on regional differences in the social economy, resources and livelihood assets. Foeken & Owuor (2001) studied the issue of multi-spatial livelihoods (households with a livelihood foothold in both urban and rural areas) based on data from an urban agriculture survey conducted in Nakuru, Kenya, in sub-Saharan Africa in 1999–2000. Bahadur (2011) evaluated rural resources and livelihood development strategies in mountainous areas of Nepal by combining socioeconomic and spatial methods. Social and economic data were collected through household surveys of randomly selected farmers, and interpolation observations were made using the families’ geographical location and spatial distribution. Wang et al. (2012) adopted the “PAR + 3S” method to analyze livelihood allocation structure assets, apartment type classification and spatial distribution characteristics by combining household livelihood asset ownership with spatial information of household management land using geographic coordinates as an identification code. To study the livelihood of farmers in the Yunnan-Guizhou Plateau region, Liu et al. (2012) used GIS and BP natural networks to simulate the spatial distribution pattern of the peasant household risk index and livelihood index emergency and response-ability index in Yunnan Province. Hirons et al. (2018) confirmed that specific characteristics of livelihood practice, knowledge, beliefs and institutions and their interactions could enhance and weaken resilience at different spatial, temporal and social scales. Augustine et al. (2020) studied farmers’ views on farm resilience to various disturbances they had to deal with in daily farm management through semi-structured interview data from the organic dairy cow (85) and sheep (43) farms in France and emphasized the factors of resilience. Awazi et al. (2021) comparatively evaluated the livelihood resilience of smallholder farmers in Isiolo County, Kenya and northwest Cameroon in the face of environmental changes. The results showed that the resilience of smallholder farmers in different regions significantly differed with the same livelihood recovery measures. Zhang et al. (2020) took Shigou town, Mizhi County, northern Shaanxi, as an example; based on the theoretical framework of livelihood resilience, the improved TOPSIS model and ArcGIS method was used to analyze the spatial differentiation of farmers’ livelihood resilience. In summary, most current studies on the resilience of farmers’ livelihoods are based on groups of a single region and a single scale. There are relatively few studies on spatial differences.

Data and Methods

Data Sources

The research data in this study include socioeconomic statistics, geographic data and social survey data. The socioeconomic statistics are mainly obtained from the 2020 Statistical Bulletin on National Economic and Social Development of each region; The geographic data comes from the geospatial data cloud platform; The questionnaire data are the questionnaire survey conducted by the research group in Hubei Province, neighboring Anhui Province and rural areas of Chongqing from May to August 2020.

Hubei Province and Wuhan Prefecture, the focus of the COVID-19 epidemic in the country, implemented the strictest quarantine policy from January 23 to April 8, 2020, with socio-economic activities almost at a standstill. The epidemic has severely impacted the economic and social development of Hubei, with an annual GDP growth rate of 5% negative, making Hubei the only province with a negative GDP growth rate in China. Hubei Province is a key area of the COVID-19 epidemic in China. By radiation of the Wuhan “1 + 8” urban circle, there were a large number of rural workers in Huanggang, Huangshi, and Xiaogan, making the proportion of confirmed cases in rural areas as high as 48–55% during the outbreak. At the same time, Yichang Prefecture and Enshi Prefecture in Hubei Province are located at the intersection of Chongqing. In addition, Anhui Province and the municipality of Chongqing are adjacent to the east and west of Hubei Province. The former is a necessary transfer station for workers in the Yangtze River Delta region. The latter is the concentrated export area of the rural labor force in the middle and upper reaches of the Yangtze River. Economic growth in Anhui Province and Chongqing municipality slowed due to the pandemic. Since the outbreak of COVID-19 in January 2020, the susceptible population has spread from Hubei Province to neighboring provinces, along with the wave of people returning home before the Spring Festival. To April, the total number of confirmed cases in the three provinces and cities has reached nearly 70,000, accounting for 55.6% of the total number of infections in China, and they are distributed in nearly 70% of the regions below the county level, which is of typical significance for epidemic control. Under the government’s strong epidemic prevention and control measures, Anhui Province and Chongqing municipality achieved a total zero in March and Hubei Province in April, respectively. According to the epidemic notification of the Joint Prevention and Control Mechanism of the State Council, districts and counties with relatively concentrated numbers of cases in townships were selected for sampling, as shown in Fig. 1.

Fig. 1.

Sampling areas and corresponding epidemic risk levels

Following the measures required by local epidemic prevention and control, investigators of the research group conducted emergency social surveys in rural areas in Hubei Province and around Anhui Province and the municipality of Chongqing from May to August 2020. According to a February 2020 outbreak period risk hierarchy based on the balance of high, moderate and low proportions of cases, a stratified random sampling survey was conducted on 49 village groups in 14 districts and counties of Hubei (Huanggang, Xiaogan, Huangshi, Yichang), Anhui (Chuzhou Lu ‘an), and Chongqing (Wanzhou, Fuling). The survey was conducted twice with an interval of one month. The questionnaire and semistructured interview investigated farmer household endowments, assets and employment, agricultural production and management. A total of 520 questionnaires were issued, and 465 valid questionnaires were returned, including 2,431 rural people.

Regarding the overall sample, 89.5% of the people were located in the urban-rural fringe, and the number of migrant workers accounted for more than 30% of the total number. Almost every household had members who had returned to their hometowns from districts and counties; the characteristics of surveyed households are shown in Table 1. The total population of rural households and the labor force is large. The number of migrant workers is more than half, and the part-time farmers and non-farmers are more than 90%, indicating that rural China’s livelihood mode has changed from agricultural to non-agricultural production. The proportion of farming income has also decreased markedly. With the implementation of the rural revitalization and common prosperity policy, the living and business area of farmers and farmers is generally higher than that of cities, and natural capital is in a dominant position, which has certain resilience in the face of the impact of the epidemic.

Table 1.

Characteristics of sample households in rural areas affected by the epidemic

Indicators		Mean value	Indicators		Mean value
Risk grade	High risk/%	40.86	Family characteristics	Household Size/(persons/household)	4.61
	Moderate risk/%	43.66		Labor force quantity/(persons/household)	3.02
	Low risk/%	15.48		Number of migrant workers per household/(persons/household)	1.73
Livelihood model	Farmers only/%	8.75		Proportion of farming income/(%)	31.42
	Part-time farmers/%	78.30		Living and operating area/(square meters/household)	136.54
	Non-farmers/%	12.95		Additional expenses incurred by households due to the epidemic/(YUAN/household)	603.76

Note: The epidemic risk levels of districts and counties were determined according to the COVID-19 Risk Assessment Report released by the Joint Prevention and Control Mechanism of the State Council in February 2020. The proportion of the nonagricultural income of farmers only, part-time farmers and nonfarmers in total income were 10% or less, 10–90%, and 90% or more, respectively.

Index Selection

Based on Speranza’s (2014) international classic livelihood resilience framework, the evaluation system of postepidemic household livelihood resilience was established, which can be divided into buffer capacity (BC), which refers to the ability of farmers to make use of their resource endowments to cope with livelihood pressure or livelihood interference under the impact of the epidemic and to increase or maintain their livelihood characteristics and functions, which mainly covers housing, land, durable goods and other assets, family member structure and health status, and employment, income and savings level (Speranza 2013; Speranza et al. 2014; Thulstrup 2015; Liu et al. 2020; Quandt 2018; Sina et al. 2019a); self-organization (SO), which reflects the ability of farmers to integrate social resources and obtain policy assistance through the network organization system and location advantage in epidemic situations, which can be represented by support from relatives and friends, advantageous channels of the cooperative and transportation infrastructure (Quandt 2017; Walsh 2016; Sina et al. 2019b; Andreas 2015); and capacity for learning (CL), which reflects farmers’ knowledge creation and skill updating and level of working experience, as well as livelihood capacity reconstruction and return to work under epidemic risk. CL generally involves farmers’ education level, post-epidemic reemployment training, the number of migrant workers, and industry (Speranza et al. 2014; Sina et al. 2019b). Based on the three dimensions of buffering capacity, self-organizing ability and learning capability, 22 indicators were selected to measure the resilience of farmers’ livelihoods after the epidemic. The corresponding indicator description and weight are shown in Table 2.

Table 2.

Index measures and weights of farmers’ livelihood resilience after epidemic diseases

Dimension	Indicator	Literature citation	Description and definition	Sign	Weight	Mean value	SD
Buffer Capacity (BC) 0.182	Housing capital	Speranza (2013)	Calculation method: housing structure*0.5 + housing area*0.5 Housing structure: Civil = 1; Brick = 2; Brick and tile = 3; Steel = 4; Housing area: square meters/household	+	0.061	69.918	28.926
	Land Capital	Quandt (2018)	Calculation method: land type*0.5 + land area*0.5 Land type: dry sloping land = 1; Dry ground = 2; Irrigated = 3; Paddy field = 4 Land area: land area of each type (acres)	+	0.242	4.471	5.105
	US Durable Goods	Thulstrup (2015)	Household ownership of durable consumer goods such as cars and home appliances: 0 ~ 3 = 1; 4 ~ 6 = 2; 7 ~ 9 kinds of goods = 3; 10 or more = 4	+	0.050	2.763	0.722
	Family support ratio	Speranza et al. (2014)	Method of calculation: ratio of children aged 18 and under, elderly individuals aged ≥ 60 in families to the number of individuals participating in the labor force	-	0.016	0.506	0.315
	Family health	Sina et al.(2019a)	Calculation method: number of family members with serious or chronic diseases: 0 = 1; 1 ~ 2 = 2; 3 ~ 4 = 3; 5 or more = 4	-	0.036	0.477	0.537
	Family social insurance level	Sina et al.(2019a)	Number of people with endowment insurance*0.5 + Number of people with medical insurance*0.5 = 1; 1 ~ 2 = 2; 3 ~ 4 = 3; 5 or more = 4	+	0.044	2.499	0.600
	Income diversity	Thulstrup (2015)	Number of Sources of Household Income (types)	+	0.216	2.323	1.139
	Annual gross income	Speranza et al.(2014)	Average annual household income over the past three years: 50,000 yuan or less = 1; 50,000 ~ 100,000 yuan = 2; 100,000 ~ 200,000 yuan = 3; Over 200,000 yuan = 4	+	0.167	2.185	0.861
	Savings capacity	Speranza et al. (2014)	Household annual savings of the proportion of total income: 10% or less = 1; 10% ~ 30% = 2; 31% ~ 50% = 3; 50% and above = 4	+	0.169	2.185	0.896
Self-organization (SO) 0.514	Work opportunity	Liu et al.(2020)	The number of channels to obtain migrant work opportunities	+	0.046	1.450	0.986
	Funding support	Quandt et al.(2017)	Can you get financial help from relatives and friends (yes = 1, no = 0)	+	0.021	0.884	0.321
	Number of public officials	Liu et al.(2020)	Number of village officials and civil servants among family members and close relatives 0 = 1; 1 ~ 2 = 2; 3 ~ 4 = 3; 5 or more = 4	+	0.426	1.101	0.342
	Organizational support	Zhao et al.(2022)	Participation in agricultural co-op (Yes = 1, No = 0)	+	0.435	0.082	0.274
	Agricultural product sales channel	Speranza et al.(2014)	Types of agricultural product sales channels: No sales of agricultural products = 0; Unable to sell due to the epidemic = 1; Farmer’s market or purchase point = 2; Wholesaler on-site purchase = 3; Online sales = 4	+	0.060	1.711	1.179
	Traffic accessibility	Sina et al.(2019a)	Time required to reach the nearest health facility 15 min or less = 4; 16–30 min = 3; 31 ~ 60 min = 2; 1 h or more = 1	+	0.012	3.136	0.721
Capacity for Learning ( CL) 0.304	Level of education	Quandt (2018)	Highest level of education of family members: Primary school and below = 1; Junior high school = 2; High school/technical secondary school = 3; College/Bachelor’s = 4; Graduate = 5	+	0.068	2.742	1.052
	Skill Training frequency	Speranza et al.(2014)	Number of skill trainings in the past two years: 0 = 1; 1 ~ 3 = 2; 4 ~ 6 times = 3; 7 times or more = 4	+	0.488	1.217	0.462
	Farthest footprint	Quandt (2018)	Main working areas for migrant workers: City = 1; Cities within and outside the province = 2, outside the province = 3, no work = 4	+	0.118	2.385	1.074
	Number of migrant workers	Speranza et al.(2014)	The proportion of migrant workers in the total number of people in the household	+	0.122	0.359	0.325
	Migrant industry	Speranza et al.(2014)	Industry Category for Migrant Workers (Category)	+	0.071	1.170	0.745
	income of migrant workers	Thulstrup (2015)	The proportion of income earned by migrant workers of total income: 0 ~ 30% = 1; 31% ~ 50% = 2; 51% ~ 80% = 3; More than 80% = 4	+	0.084	2.815	1.141
	Return to work	Speranza et al.(2014)	Post-epidemic employment situation: Returned to job = 4; Blocking-up = 3; Uncertain = 2; Unemployed = 1	+	0.049	3.279	1.435

Research Methods

In measuring resilience, we first standardize the original data to eliminate the dimension and order of magnitude of the difference between indicators. Then, we use the entropy method to determine the weight coefficient and the comprehensive index method to measure the livelihood resilience index. Finally, regression analysis is used to find the farmer’s livelihood resilience and its impact factors, and a geographical detector is used to find the spatial differentiation of the causal mechanism. The empirical research framework is as Fig. 2.

Fig. 2.

Empirical Research Framework

Measurement of Livelihood Resilience

1. Data standardization: First, livelihood resilience indicators are divided into positive and negative indicators. The former indicates that the larger the data are, the greater the resilience capacity is, while the latter indicates the opposite (The positive and negative directions of each indicator have been marked in Table 2). The range method is used for data standardization and dimensionless processing, as shown in Formula (1):

1

In the formula, Z_ij is the jth index measure of the ith dimension, Z⁺ _ij and Z⁻ _ij represent the standardized values of positive and negative efficacy indicators, respectively, x_ij is survey data for each index, and x_{i, max} and x_{i, min} are the maximum value and minimum value of an indicator.

2. Index weight. The index weight is determined by the entropy weight method combining subjective and objective methods (Gautam, 2017). The entropy weight method assigns values through original indicators and data, which can profoundly reflect indicators’ discrimination ability, giving weight more objectively and with higher credibility. First, the specific gravity P of each sample index value is calculated, and the proportion of the ith peasant household sample in the jth index measure P_ij is. The entropy value e_j and weight w_j of indicator j are shown in Formula (2):

2

3. Measurement of livelihood resilience. The synthetic index method proposed by Speranza et al. (2014) is adopted. To select indicators from three dimensions of buffer capacity, self-organizing ability and Capacity for learning to measure the level of livelihood resilience. This method can clearly distinguish all dimensions and specific indicators. Based on index standardization and weight calculation, the index of each dimension is calculated separately, and then the dimension value and weight are summed by comprehensive weighting. F_i is set as the measured value of the three dimensions (BC, SO, LC), W_i is the weight of each dimension, X_i is the measured value of the sample data of the ith index, and w_i is the weight of the ith index. LRI is the livelihood resilience index of farmers, and the post-epidemic resilience level of farmers is shown in Formula (3):

3

Analysis of Spatial Differentiation

Due to the characteristics of epidemic aggregation and mobility, it is necessary to consider the spatial correlation of factors resulting from farmers’ livelihood resilience. Since the degree of risk of infectious diseases is inversely proportional to livelihood resilience, the spatial distribution of resilience factors should be similar. The geographic detector is a statistical model used to detect the spatial differentiation of objects and their driving factors (Wang & Xu 2017). Since there is no linear premise hypothesis, this method can analyze the driving force of the post-epidemic recovery of farmers’ livelihoods. Differentiation factor detection is specifically used to clarify the impact degree q of each factor on the spatial distribution of LRI values, as shown in Formula (4):

4

where h = 1, ……, L is the layer of the LRI, N_h and N are the numbers of cells of layer h and the whole region, and σ²_h and σ² are the variances in layer h and the whole region LRI, respectively. In Eq. (4) where q∈[0,1], a q value→1 indicates that the variable has a stronger driving effect on the livelihood resilience of farmers, and the LRI also presents a more significant spatial differentiation phenomenon.

Results and Discussion

The Level of Livelihood Resilience of Farmers After the Epidemic

General Characteristics

According to the above indicator system and measurement methods, the LRI values of farmers after the epidemic can be calculated to reflect the level at which their livelihood capital endowment provides a buffer to interference, and the farmers realize the reconstruction of sustainable livelihood capacity through self-organization and learning capability after the impact of the epidemic. The calculation results of the sample areas are shown in Table 3.

Table 3.

Household livelihood resilience and dimensions in key epidemic areas

Area	Mean value of BC	Mean value of SO	Mean value of CL	Mean value of LRI
Total	0.322	0.117	0.264	0.199
Hubei	0.304	0.116	0.237	0.187
Chongqing	0.440	0.081	0.267	0.203
Anhui	0.334	0.148	0.422	0.265

As shown in Table 3, the LRI values of rural households after the epidemic are generally low, distributed at 0.054–0.642. The average LRI values gradually decreased along Anhui, Chongqing, and Hubei regions, indicating that, similar to urban communities, epidemic risk also greatly impacted rural households. In the center of the outbreak in the Hubei area, farmers’ buffer capacity is significantly lower than the overall average. Chongqing area farmers’ livelihood capital loss is less, effectively becoming a buffer against high-risk infectious shock.

In addition, farmers in Anhui have advantages in self-organization ability and learning capability; however, although the area’s rural economic environment and surrounding areas are similar, post-epidemic LRI values were much higher than in neighboring provinces. This shows that as the Yangtze River Delta urban agglomeration increases its labor reserves, migrant workers from Anhui are concentrated in Jiangsu, Zhejiang and Shanghai, and farmers have more channels of information and training opportunities. Social cognition and improved skill levels can effectively enhance the ability to resist disease. Therefore, once the epidemic shifts from a high to a steady phase, farmers will use their livelihood ability to recover from the outbreak quickly.

Distribution Characteristics of Various Dimensions of Livelihood Resilience

Due to the spread of the epidemic along a path that directly affects the farmer’s livelihood resilience process, the natural break-point method is adopted to divide the LRI and each dimension into three levels. The overall proportion of each item is shown in Table 4. Most farmers in the sample had moderate LRI values, over 50% had low values, and only 11% had high values. The distributions of buffer capability (BC) and learning capability (CL) in each dimension were relatively balanced. Still, the former was mainly at high and moderate levels, while the latter was at low and moderate levels. In the same way, 85% of farmers with self-organization ability (SO) hovered at moderate LRI values. This shows that with continuous economic and social development in China’s rural areas, farmers’ livelihood capital accumulation is enough to withstand epidemic risk. However, major public health events still seriously restrict normal agricultural production and social activities, and the conditions and skills of rural families still need to adapt to cope with the impact of emergencies effectively.

Table 4.

Level and proportion of post-epidemic household livelihood resilience and each dimension

Recovery level	BC	SO	CL	LRI
High	0.374–0.593	0.239–0.821	0.362–0.713	0.298–0.642
	33.12%	9.25%	17.63%	10.97%
Moderate	0.243–0.374	0.119–0.239	0.221–0.362	0.181–0.298
	42.80%	6.45%	40.43%	36.34%
Low	0.037–0.243	0.004–0.119	0.0592–0.221	0.054–0.181
	24.09%	84.30%	41.94%	52.69%

As seen from the LRI level of farmers in each sample area (Fig. 3), the LRI values of farmers in Hubei Province are concentrated in the low-value area of 0.1 ~ 0.2, especially regarding the level of self-organization ability (SO), which is mostly lower than 0.1. Learning capability (CL) values are also low and moderate, 0.1 ~ 0.3. The buffering capability (BC) dispersion degree is high, distributed from 0.1 to 0.6. The LRI values of farmers in Chongqing were mostly concentrated at a moderate level of 0.2, buffer capability (BC) values fluctuated but generally had a high level of 0.4 ~ 0.6, and learning capability (CL) values were relatively balanced at a moderate level of 0.3. However, self-organization ability (SO) values were basically at a low level, below 0.1. In Anhui Province, the LRI values were slightly higher and concentrated between 0.2 and 0.3, the buffer capability (BC) values were clearly at a moderate and high level, and the self-organization ability (SO) values were mainly in the low value region below 0.1. However, there were still some moderate and high discrete values. Learning capability (CL) was widely distributed from low to high value.

Fig. 3.

Distribution of household LRI values and values of each dimension after the epidemic in each province

The livelihood capital stock of rural households in Hubei Province varies greatly, and the overall capital endowment of rural households in Chongqing is higher than that in Hubei Province, although there are slight differences. In rural areas of Anhui Province, due to the proximity of the metropolitan areas of Shanghai–Nanjing and Hangzhou, farmers’ perception, cognition and skill level are obviously heterogeneous due to migrant work. Regarding self-organization ability, various provinces and cities have a low ability. This also confirmed that the start of the new Champions League greatly impacted agricultural production and the return to employment in the city. However, farmers’ social resources make it challenging to obtain adequate capital and information support. So they badly need state and local governments to finance policy support and employment channels.

Spatial Pattern of Livelihood Resilience

To further explore the spatial association between farmers’ LRI values and key epidemic areas, the Arc-GIS 10.2 software platform was used to draw the spatial distribution trend of LRI values and all dimensions, as shown in Fig. 4. Farmers’ LRI values generally showed a “concentration of low value and dispersion of middle and high values” phenomenon, with obvious polarization differentiation. The LRI regions with low values were distributed in Huanggang in the middle of Hubei Province, Hanchuan and Xiaonan under Xiaogan. These regions are located around Wuhan. Due to the siphon effect of central cities, rural economic and social development lags behind despite a large number of migrant workers, making it difficult for farmers to eliminate the threat of the epidemic. The LRI levels of farmers in Chongqing and Anhui were higher than those in Hubei and clustered at a moderate degree of recovery, indicating a significant spatial correlation between the LRI level and epidemic risk. This shows that improving the resilience of rural communities and farmers’ awareness will greatly contribute to the sustainable recovery of farmers’ livelihoods in affected areas.

Fig. 4.

Spatial variation in LRI values and each dimension of rural households after the epidemic

At the same time, there are significant spatial differences in the recovery level of LRI dimensions, among which buffer capability and self-organization ability tend to agglomerate while learning capability has a high degree of dispersion. The areas with low buffer capacity are mainly distributed in Zigui County and Yiling District of Yichang, Hubei Province, both of which are located in the ecological barrier of the Three Gorges Reservoir area. The terrain is mainly mountainous, and these counties used to be among the nation’s poorest, with a weak economic development foundation limited by ecological and environmental protection. The dimensions of self-organization ability were generally “low in the west and high in the east.“ After the epidemic, the sales of agricultural products and migrant work activities in Fuling and Wanzhou of Chongqing and Hanchuan and Hong ‘an in Hubei stopped. In addition, the values of high learning capability in rural areas of Anhui Province are relatively concentrated, while the distribution of learning capability values of farmers in Hubei Province and Chongqing city is very scattered, with no obvious regional correlation. This indicates that the recovery level of farmers in this dimension is affected only by individual footprint experience, skill level and employment situation, and there is no direct path of action with the epidemic risk in the region.

Influencing Factors of Livelihood Resilience Level and Spatial Differentiation of Farmers

The results of the previous study show that there are significant differences in the level of livelihood resilience of farmers in different regions. In order to further clarify the influencing factors, multiple linear regression and geographic detector methods are respectively used to analyze the level and spatial differentiation of livelihood resilience and explore the key impacts.

Driving Factors of the post-epidemic Livelihood Resilience of Farmers

Previous studies have shown that the livelihood capital owned by peasant households is the core of resisting all kinds of risk shocks (Singh et al. 2022; Pour et al. 2018), which involves the collection of resources and capabilities owned by peasant households that can be used to improve production and life. Livelihood capital is the core of the theory of sustainable livelihoods. The stock and structure of livelihood capital play a decisive role in the income-generating activities of individuals and households. Five livelihood capital definitions from the DFID of the United Kingdom are selected here. The following influencing factors are selected based on the actual impact of the epidemic on farmers in China: natural capital refers to the natural resources owned by farmers, as farmers in Chongqing, Hubei and Anhui mainly use land farming, with very little capital stock in forestland and water areas (Uddin et al. 2020); physical capital refers to the place where farmers live and produce and the tools they use, most of which rural households invest in housing and durable goods; human capital refers to the knowledge and skills possessed by family members and individuals, including household size and labor force, education of the head of household, and health status (Vesely et al. 2017); financial capital refers to the cash and loans that people have at their disposal in the process of production and living, reflected by household income, savings and borrowing (Wen & Hanley 2015); and social capital refers to the interpersonal trust and social network built by farmers, which depends on the geographical, kinship and friendly relationships among farmers, including human contact, participation in management organizations and the number of relatives in social groupsof cadres (Taylor & Distelberg 2016; Ning 2017). In summary, a total of 13 variables in 5 dimensions are obtained. Specific explanations and descriptions are shown in Table 5.

Table 5.

Explanation of variable definitions

Types of capital	Literature citation	Explanatory variable	Variable definition
Natural capital	Quandt (2018)	Cultivated land area	Area of arable land owned by a family (acre)
	Quandt (2018)	Cultivated land quality	Dry slope = 1; Dry land = 2; Irrigated land = 3; Paddy field = 4
Physical capital	Speranza et al.(2014)	Housing structure	Structure type: Civil = 1; Brick = 2; Brick and tile = 3; Steel = 4
	Thulstrup (2015)	Number of durable goods	Household ownership of durable consumer goods such as cars and home appliances 0 ~ 3 = 1; 4 ~ 6 = 2; 7 ~ 9 kinds of durable goods = 3; 10 or more = 4
Human capital	Speranza et al.(2014)	Family size	Total Family size (persons)
	Speranza et al.(2014)	The householder age	Age of head of household (years)
	Thulstrup (2015)	Labor force ratio	Numebers of labor force/total numbers of household population (%)
	Speranza et al.(2014)	Frequency of medical visits	Number of medical visits (times) for all family members in the previous year
Financial capital	Speranza et al.(2014)	Total household income	50,000 yuan or less = 1; 50,000 ~ 100,000 yuan = 2; 100,000 ~ 200,000 yuan = 3; Over 200,000 yuan = 4
	Speranza et al.(2014)	Ability to save	10% or less = 1; 10%~30% = 2; 31%~50% = 3; 50% and above = 4
	Speranza et al.(2014)	Loan demand	Whether the household borrowed during and after the epidemic (yes = 1, no = 0)
Social capital	Sina et al.(2019a)	Social consumption	Amount of Social Consumption of family in the last year (Yuan)
	Zhao et al.(2022)	Organizational support	Participation in agricultural co-op (Yes = 1, No = 0)
	Liu et al.(2020)	Number of public officials	Number of village officials and civil servants among family members and close relatives: 0 = 1; 1 ~ 2 = 2; 3 ~ 4 = 3; 5 or more = 4

The post-epidemic farmer LRI was taken as the dependent variable, livelihood capital as the independent variable, and the OLS model was used to analyze the influencing factors of the LRI. In the multicollinearity test, the coefficient of variance inflation (VIF) was less than 10, the adjusted R² = 0.873 > 0.6, and the statistic corresponding to the F value had a P = 0.000, proving that the model had high goodness of fit. The results are shown in Table 6.

Table 6.

Regression analysis of influencing factors of farmer LRI values after the epidemic

Types of capital	Explanatory variable	Mean value	Standard error	Standardized Coefficients (Beta)	VIF
Natural capital	Cultivated land area	4.471	5.105	0.089***	1.414
	Cultivated land quality	2.531	1.275	0.067***	1.462
Physical capital	Housing structure	3.297	0.671	0.111***	1.371
	Number of durable goods	2.763	0.722	0.016	1.421
Human capital	Family size	4.611	1.837	0.074***	1.587
	The householder age	56.241	10.104	-0.012	1.196
	Labor force ratio	0.676	0.223	0.045**	1.319
	Frequency of medical visits	1.938	0.865	-0.002	1.142
Financial capital	Total household income	2.185	0.861	0.136***	1.657
	Ability to save	2.185	0.896	0.162***	1.302
	Loan demand	0.262	0.440	0.126***	1.172
Social capital	Social consumption	2.918	0.925	-0.027	1.176
	Organizational support	0.082	0.274	0.690***	1.166
	Number of public officials	1.101	0.342	0.285***	1.075

Note: ***, ** and * are significant at the levels of 1%, 5% and 10%, respectively.

As seen from the regression results in Table 6, all indicators of natural capital have a significant positive impact on the LRI, which on the one hand, indicates that agricultural income is still an important source of material consumption for rural households. However, the proportion of agricultural income in total income decreases yearly. On the other hand, most of the support from the country’s post-epidemic resilience is directly linked to land cultivation, such as increased subsidies for agricultural production activities such as alternative farming and planting for pest control. However, only housing in the material capital category significantly affects post-epidemic farmers’ LRI values, while the stock of durable goods has little promoting effect. This phenomenon stems from the fact that housing is still the main investment of rural households, which can directly reflect their economic strength. However, in recent years, with the improvement of the downward channel of industrial products, durable goods such as household appliances are not scarce, so the proportion of expenditure in total consumption has been continuously reduced, which can hardly reflect the asset level of rural households.

Regarding human capital, household members and the number of individuals in the labor force who can effectively promote livelihood resilience after disease, age and medical conditions that cause adverse effects are insignificant. The possible reason is as follows: First, more family members can give inter-generational support to allow young farmers to have more opportunities as migrant workers, which is more conducive to the diversification of peasant household livelihoods and alleviates the impact of risk. Second, although aging and long-term illness have hindered epidemic prevention and treatment among families, with the implementation of the targeted poverty alleviation medical assistance policy of the new rural cooperative medical care system and the new rural insurance system, the situation of “elderly individuals without shelter to support them” and “returning to poverty due to illness” has dramatically improved.

At the same time, loan financing, household income, and savings deposits in terms of family financial capital have caused an increase in LRI values, and the impact of these factors increased in turn. Further confirmed that the rural family could not only be content with self-marketing; a certain amount of funds deposited to withstand emergencies and “microloans to help farmers” in rural areas will become the key driving force of revival after the epidemic. In addition, in terms of social capital, the introduction of channels and information sharing by agricultural cooperatives makes it easier for participating farmers to break through the epidemic blockade and commence agricultural product sales as soon as possible. The social resources brought by close relatives who are public officials ca. help by providing policy information and production and living materials. This shows that compared with the social structure of urban communities, the rural social network is still predominantly structured face-to-face, and farmers’ geographical, kinship and friendship relations are significant for sustainable livelihood development. No significant influence on social consumption confirmed the isolation management during outbreaks in rural communities, which put social and personnel gathered behavior under control.

In summary, the dimensions of financial capital, social capital and natural capital are significant driving factors of farmer households’ LRI values. However, the normalized beta coefficient of natural capital is mostly less than 0.1, which proves that the driving effect of this dimension is not obvious in economic statistics. Therefore, after the epidemic, it is crucial to provide farmers with sufficient startup funds for production resumption, dredge channels for agricultural product sales, and provide opportunities for migrant workers. However, as the share of agricultural income continues to decline, the stock of agricultural resources such as land can help, but the effect is minimal. In addition, the human capital mentioned in previous studies has no significant impact. On the one hand, the proportion of disease in rural residents during the epidemic is generally not high, with little restriction on labor ability. On the other hand, rural primary health care and security systems have improved, and the bottleneck effect is less prominent.

In order to ensure the reliability of the estimation results, this article carried out the robustness test of the regression results by replacing part of the explanatory variables and deleting samples. Among them, the replacement of explanatory variables is respectively “durable number,“ “labor,“ “family income,“ and “family number of public servants” replacement for the types of “agricultural machinery,“ “male labor share,“ “income of migrant workers,“ “the number of close relatives,“ cut 20% of the sample is a random cut sample to return, the results are shown in Table 7. Compared with Table 6, the significance and direction of action are consistent, which proves the robustness of the results.

Table 7.

Results of the robustness test

Types of capital	Explanatory variable	Change the explanatory variables	Cut sample (randomly cut 20%)
Natural capital	Cultivated land area	0.150***	0.077***
	Cultivated land quality	0.031	0.064***
Physical capital	Housing structure	0.146***	0.094***
	Number of durable goods	0.033	0.014
Human capital	Family size	0.083***	0.077***
	The householder age	-0.042*	0.000
	Labor force ratio	0.002	0.052**
	Frequency of medical visits	-0.012	-0.015
Financial capital	Total household income	0.050**	0.147***
	Ability to save	0.208***	0.149***
	Loan demand	0.131***	0.113***
Social capital	Social consumption	-0.034	-0.030
	Organizational support	0.759***	0.710***
	Number of public officials	0.005	0.292***

Note: ***, ** and * are significant at the levels of 1%, 5% and 10%, respectively.

Influencing Factors of the Spatial Differentiation of Livelihood Resilience

Individual livelihoods of farmers are closely related to local society, economy and geographical location. Different levels of economic development and uneven distribution of resource endowments between regions restrict the resilience of farmers after the epidemic. Zhejiang City, Dangyang City, Daye City, Hanchuan City in Hubei Province, Fuling District in Chongqing City, and Tianchang City in Anhui Province have relatively high per capita GDP and rural residents’ income. A high level of social and economic development foundation is conducive to the livelihood recovery of farmers. Yichang Zigui County, Chongqing Wanzhou District, and Anhui Yuan District are full of mountainous gullies, and low convenience geographical location limits the livelihood recovery of farmers. At the same time, Outbreaks of infectious diseases have a strong ability to spread, so there is a corresponding spatial pattern of farmers’ LRI values. To further explore the cause mechanism, combined with the classic literature and COVID-19 characteristics (Zhang et al. 2020) from various provinces and cities’ natural, geographical, economic, and social aspects, choosing six indicators “resource terrain, agricultural development degree, a distance of towns, regional economic situation, farmers’ income, and population density,“ for the analysis of the influencing factors of the spatial distribution of LRI values, as shown in Table 8.

Table 8.

Description of geographical detection indicators

Dimensions	Indicators	Definition
Natural environment	Resource terrain (X1)	Average elevation (meter)
	Agricultural development degree (X2)	The output value of primary industry (Yuan)
Geographical location	Distance of towns (X3)	Drive to nearest town (minutes)
Economic level	Regional economic situation (X4)	Per capita GDP (Yuan)
	Farmers’ household income (X5)	Per capita net income of rural residents (Yuan)
Social factor	Population density (X6)	Population per square kilometer

First, the indicators of each detection dimension are discretized to realize the transformation between numerical variables and type variables, and then the dependent and independent variables are matched. Finally, the geographical detector model is imported, and the q value of the effect of each factor on the LRI is calculated according to Formula (4), as shown in Table 9.

Table 9.

Detection results of influencing factors of the spatial layout of farmers’ LRI values

q-Value	X 1	X 2	X 3	X 4	X 5	X 6
BC	0.225	0.259	0.448	0.225	0.330	0.211
SO	0.221	0.317	0.195	0.416	0.028	0.276
CL	0.114	0.125	0.249	0.042	0.024	0.118
LRI	0.180	0.217	0.367	0.245	0.026	0.332

In Table 8, the villages’ location and the farmers’ net income are the key factors in determining the distribution of buffer capacity (BC), which shows that remote areas are also susceptible to epidemic disease. The lack of family livelihood capital, products, and the transportation of epidemic prevention materials is an inconvenience and may lead to livelihood difficulties for farmers with a low risk of disease. Farmers with strong self-organization ability (SO) are mainly concentrated in regions with high economic and agricultural development levels. The possible reasons are as follows: on the one hand, local financial strength directly impacts the strength of funding projects, which will play a crucial role in the rural recovery after the epidemic. On the other hand, strict agricultural organizations (professional cooperatives or e-commerce platforms) bring high output and effectively provide farmers with production and marketing channels to solve issues during the epidemic. The distance from cities and towns significantly contributes to farmers’ learning capability (CL), which proves that the proximity of central cities with abundant employment opportunities can increase the human capital of farmers and promote the non-agricultural transformation of livelihood to increase household income. The contrast is that the resource terrain and the population density of the LRI effect are not significant. First, regional resource endowment is not the main factor that enhances the ability to resist disease; compared to farmers in high-altitude mountains and plains, there is no obvious advantage in regional resource endowment for epidemic disease recovery. Second, although a larger population density leads to more strict epidemic prevention measures, the mandatory physical barrier quickly controls the spread of the disease. The temporary closure of villages and roads does not lead to the complete stagnation of farmers’ livelihood activities.

Discussion

The sudden outbreak of COVID-19 has brought a series of adverse impacts on rural areas with relatively weak medical and health systems and inconvenient transportation networks. The livelihoods of rural households are inevitably impacted by the outbreak of the epidemic, which affects the ownership and access of rural households’ livelihood capital and thus affects the resilience of their livelihoods. Previous studies on the epidemic and rural areas and farmers mainly involved public services (Torres et al. 2020), rural e-commerce economy (Mastronardi et al. 2022), mental health (Gundogan 2021), food consumption (Ebn et al. 2021), etc. Little research has been done on the relationship between household livelihood resilience and public health events. Based on data from rural economic and social emergency surveys after the outbreak of COVID-19 from May to August 2020, this study calculated the LRI level of rural households in the epidemic area by constructing a measurement system for their post-epidemic livelihood resilience and identified the core influencing factors through measurement methods. At the same time, the natural breakpoint method was used to grade LRI values. Then, Arc-GIS 10.2 software and a geographic detector were used to analyze the spatial distribution characteristics and critical driving forces of rural household livelihood recovery at the county scale of the sampling area. A multi-scale study was conducted on the spatial distribution of individual household livelihood resilience, regional overall resilience and resilience, which expanded the depth and breadth of livelihood research under the impact of public health events. It made a certain contribution to the existing research on livelihood resilience.

The survey found that, with the economy’s and society’s development, farmers have already had a certain amount of livelihood capital in the face of risk shocks. However, the epidemic still had a specific impact on the production and life of farmers in the affected areas, and the LRI level of farmers after the epidemic was generally low, showing a trend of successive decline in the provinces of Anhui, Chongqing and Hubei (Nath et al. 2020). The cushioning and self-organization ability of farmers in Hubei was weak, the livelihood capital loss of farmers in Chongqing was small, and rural families in Anhui had the advantage in learning capability. On the whole, the buffer capability of farmers’ LRI levels is evenly distributed at the high level, but the learning capability is evenly distributed at the low level. There are also significant differences between provinces and cities. Hubei farmers’ buffer capability dispersion degree is high, while Anhui farmers’ learning capability distribution is uneven. Although the LRI dimensions of farmers in Chongqing are relatively concentrated, their self-organization ability is extremely poor.

Starting from the five types of livelihood capital of the DFID, the resumption of farmers’ production and labor is the top priority of livelihood resilience, so financial and social capital have the most significant contribution to the LRI. In addition, although agricultural benefit policies are often linked to land, natural capital also positively impacts rural post-epidemic recovery, but the non-agricultural livelihood of farmers greatly reduces this effect (Melvani et al. 2020). In the context of the decline of industrial products and the improvement of social security, material and human capital are no longer the key to reconstructing farmer LRI levels after the epidemic.

Regarding spatial location, rural households with low resilience are relatively concentrated, while rural households with high resilience are scattered as a whole. The low-value LRI area in Hubei is located in the surrounding countryside of Wuhan. At the same time, there is little difference between Chongqing and Anhui, and the LRI fluctuates in the median area. In addition, although remote village groups were free from the virus, the economic impact was more obvious, and areas with a high population density did not hinder the recovery of farmers’ livelihoods due to the control of the epidemic. The regions with high agricultural, industrial development and high household income have obvious advantages in buffer capacity and self-organization ability, and the corresponding agricultural resources have little influence on the comprehensive livelihood resilience level.

The resilience of rural communities is not directly related to the distribution of the epidemic, unlike the recovery of urban communities that are directly affected by the transmission path of the virus. The necessary spatial, physical isolation of rural areas for a short time can effectively block the spread of the epidemic and hardly hinder the recovery of farmers’ livelihoods. In contrast, economic pressures outweigh disease risks in areas where economic development is lagging, and rural households have inadequate capital stocks. Therefore, after the epidemic, consolidating the results of out -of -poverty at the same time, clinging to the new opportunities for rural “homesickness” economic development, contemporary living space reconstruction, and ecological agricultural product value transformation brought by the epidemic, and making full use of disease resistance against the background of agricultural industry policies can create a better “camp farming environment” to achieve common prosperity for rural development and farmers.

Post-epidemic measures can no longer focus on agricultural production but need to adjust policies dynamically based on different stages of household livelihood resilience. When the epidemic has just been contained, government organizations and public institutions should organize to buy instead of donations, logistics enterprises and e-commerce cooperatives should place door-to-door orders, agricultural product sales channels should be expanded, and micro-credit should be used to help seasonal agricultural production. At the release stage, when the epidemic is slowing down, we should focus on providing one-off subsidies or reducing or exempting social security contributions to small and micro-enterprises that absorb migrant workers from affected areas and organize migrant workers to return to work and find employment in an orderly manner through work-for-work training and unified labor export. In the regression stage of basic control of the epidemic, we should consider improving the rural primary medical system and the diversified livelihood capacity of farmers.

Conclusion

In this study, an assessment framework for livelihood resilience under the impact of COVID-19 was established, three dimensions of buffer capacity, self-organization capacity and learning capacity were identified, and their influencing mechanisms were explored from different scales. Under the impact of COVID-19, the overall resilience of farmers’ livelihoods was low, and there were great differences in the resilience of farmers’ livelihoods in different regions. Meanwhile, the levels of resilience in each dimension also showed significant heterogeneity. Therefore, for post-COVID-19 assistance, targeted policies should be introduced according to local conditions. In Hubei Province, where the epidemic has most had an effect, and the buffer capacity is weak, special attention should be given to elderly, sick, and disabled individuals in rural areas and areas of relocation. Through real-time monitoring and basic living security, we must resolutely prevent people from returning to poverty due to the epidemic. For Anhui farmers who have advantages in learning capability but are temporarily blocked, it is necessary to actively promote local labor channels, such as community factories and poverty alleviation workshops, due to the restrictions on migrating for labor. The main bottleneck faced by farmers in Chongqing lies in their self-organization ability, so it is very important to ensure the smooth transportation of products and solve the difficulties of the farming industry.

The judgment of farmers’ livelihood resilience under the impact of the epidemic helps determine the work focus of production and life resilience after the epidemic, mastering the individual farmers’ livelihood resilience, the overall regional resilience and the spatial distribution characteristics of each dimension, and provide prediction and analysis support for the government to judge the affected degree of farmers, the resilience after the epidemic and the direction of livelihood development. Although the risk of the spread of the epidemic is lower in rural and township groups than in urban communities, the resilience of livelihoods of rural households is generally weaker. Therefore, in the follow-up prevention and control of COVID-19, on the one hand, we should break the paralysis mentality after the relatively stable epidemic situation in cities, focus on blocking the transmission path of the virus to rural areas, strengthen prevention and control information, education and guidance, and enhance the awareness and ability of farmers. On the other hand, we need to work out contingency plans for rural health emergencies, promote capacity building of community-level medical and health service facilities, and strengthen the role of village-level clinics and pharmacies as outposts and inspection points.

Due to subjective and objective reasons, this article inevitably has some limitations. First of all, due to the requirements of the government’s prevention and control policies, the randomness of the questionnaire respondents is strong, and the sample size is insufficient. Second, the cross-sectional data can only reflect the current livelihood status of farmers and make a reasonable prediction of the future development direction but cannot reveal the dynamic change process of farmers’ livelihood elasticity and livelihood development. Thirdly, socioeconomic variables and indicators are not considered in the analysis of overall regional resilience and spatial pattern, which is not accurate enough. In the study of the combination of public health emergencies and livelihood resilience, more dynamic studies are needed to compare the changes in livelihoods before and after to judge the sustainability of livelihood development better. It is necessary to pay attention to systematic research, combine the theory of resilience and the theory of sustainable development, judge the dynamic development stage of the livelihood system, and provide a theoretical basis for formulating a livelihood strategy.

Authors’ Contributions

All authors substantially contributed to this research work. Hengxing Xiang and Xu Zhao were responsible for conceptualizing the study and writing the initial draft of this manuscript. Xu Zhao and Feifei Zhao contributed to the editing and multiple reviews of the manuscript. Hengxing Xiang performed the data processing and analysis. All authors read and approved the final manuscript.

Declarations

Competing Interests

The authors declare that they have no competing interests.