An assessment of the livelihood resilience of tourism smallholders in the upper reaches of Yihe River, China

ABSTRACT Using data from 524 households in the upper reaches of the Yihe River, China, a tourism livelihood resilience index (TLRI) was developed to explore the correlation between livelihood resilience and strategies of tourism smallholders. The smallholders were grouped into four categories: tourism oriented (TO), tourism and part-time migration work (TPM), migration work and part-time tourism (MPT) and migration work-oriented (MO). The TLRI revealed notable differences among these groups, with MPT smallholders demonstrating the lowest resilience (0.287) due to a weak buffer capacity (0.242). In contrast, MO smallholders displayed high resilience, as they were able to become successful professional tourism operators. Interestingly, households with low-diversity livelihoods (e.g. MO and TO) had higher resilience than those with high diversity (MPT and TPM). Regression findings showed that increased buffer capacity, self-organisation capacity and community protective capacity tend to motivate smallholders to engage in tourism intensification activities, while learning capacity had no measurable impact. Based on these results, three suggestions have been presented for improving the livelihood resilience of tourism smallholders. The conceptualisation of the TLRI enhances our understanding of the livelihood resilience of tourism smallholders in China and provides insights into the household operation system in tourism-guided rural transition.


Introduction
Rural tourism, as one of the important driving forces for rural revitalisation and poverty reduction has attracted increased attention from governments, NGO and academic circles (Li et al. 2020).However, impacted by many external factors (e.g.political, economic, environment and weather) (Lasso and Dahles 2018), the tourism industry is acknowledged to be vulnerable to crises (Dahles and Susilowati 2015).The full dependence of tourism-based income may put local people at risk (Tao and Wall 2009).Over the past few years, the use of the resilience approach is frequently suggested by multiple scholars (e.g.Cahyanto and Pennington-Gray 2017;Jiang et al. 2019;Orchiston, Prayag, and Brown 2016) as the best approach to respond to increasingly turbulent environment in tourism.However, limited study has been accorded to livelihood resilience of tourism smallholders in the rural transformation in developing countries.
In general, resilience refers to ability to prepare in order to mitigate, prevent and minimise losses (Aldunce et al. 2014), adapting to the change by emphasising the ability to learn and the ability to organise themselves to utilise it as a stimulus for development and improvement or innovation (Speranza C, Wiesmann, and Rist 2014).Due to its normative features and link with complex system, resilience is often criticised for being difficult to empirically measure (Carpenter et al. 2001).Many studies (c.f.Quandt 2018;Speranza 2010) indicate that for resilience measures to be meaningful, resilience needs to be contextualised.
Within the sustainable development discipline, resilience is the most germane to the Sustainable Livelihood Approach (SLA) (Adger 2006).According to Chambers and Conway (1992), a livelihood comprises the capabilities, assets and activities that contribute to a means of living.While recognising that poor people are always exposed to vulnerability context, the SLA attempt to militate against such insecurity through building up adaptive capacity, a key concept relevant to resilience (DFID 2000;Obrist, Pfeiffer, and Henley 2010).The approach has proven useful for assessing the ability of households to withstand shocks.Adaptive capacity is mainly structured by five family livelihood assets (Scoones 1998).The societal structures represented by the transforming structures and processes operating at all levels shape access to resources and opportunities (Shen, Hughey, and Simmons 2008).Thus, social structuration processes and actorperspectives are essential for building livelihood adaptive capacity (Speranza C, Wiesmann, and Rist 2014).
Based on the above, resilience thinking is actually implied in the SLA (Obrist, Pfeiffer, and Henley 2010).Livelihood resilience refers to the capacity of livelihoods to cushion stresses and disturbances while maintaining or improving essential properties and functions (Quandt et al. 2018;Speranza C, Wiesmann, and Rist 2014).Linking SLA to resilience thinking can enhance understanding of livelihood dynamics, of how households maintain and enhance their livelihoods in the face of change (Sallu, Twyman, and Stringer 2010;Scoones 2009).Recent studies attempt to enrich livelihoods system's adaptive capacity with resilience thinking (Duan, Xie, and Morrison 2022).
In recent years, there has been a growing body of literature evaluating livelihood resilience.For example, Obrist, Pfeiffer, and Henley (2010) proposed a multi-layered resilience framework.Speranza C, Wiesmann, and Rist (2014) developed a livelihood resilience framework identifying the attributes and indicators of the three resilience dimensions.Sina et al. (2019) develop a conceptual framework for measuring livelihood resilience, emphasising individual coping ability, socio-physical robustness of the local community.Unfortunately, few resilience frameworks have been tested and modified to date, especially on the micro farmer scale (Ritchie and Jiang 2019;Speranza C, Wiesmann, and Rist 2014).Recently, some studies attempt to explore the influencing factors and mechanism of community resilience (Cahyanto and Pennington-Gray 2017;Guo, Zhang, and Zhang 2018) or tourism industry itself (Hall, Prayag, and Amore 2017).Due to the complexity and endemic of livelihood system and resilience (Obrist, Pfeiffer, and Henley 2010;Shen, Hughey, and Simmons 2008), identifying critical indicators for the resilience of specific livelihood system remain to be further explored (Ritchie and Jiang 2019).
Livelihood system and its resilience are often conceptualised as locally phenomenon, and thus more local-level analyses are necessary to gain a better understanding of livelihood resilience (Ritchie and Jiang 2019).By the generally adopted "Scenic spots + communities + smallholder" tourism development model in rural revitalisation in China, Household owned and operated is the main way for farmers to participate in tourism in poverty-stricken rural areas in China (Xu and Hao 2014).In this case, Shen, Hughey, and Simmons (2008), Tao and Wall (2009) argued that sustainable tourism livelihood and the adaptability should take note of the potential trade-off between livelihood at the household level and tourism at the community level.However, little studies have been conducted on livelihood resilience of Household operation system in tourism-guided rural transformation.
Although studies of tourism crisis resilience have been conducted, they have limitations related to scope and depth.They have predominantly focused on tourism destinations without considering the livelihood system of tourism smallholders.Second, Moreover, there has been a dearth of research on the resilience of Pacific rural tourism livelihoods.This study aims to address two key questions: (1) What are the characteristics of livelihood resilience and strategies employed by tourism smallholders in poverty-stricken areas undergoing tourism-guided rural transformation?(2) What is the correlation between livelihood resilience and strategies adopted by tourism smallholders in such areas?

Theoretical analysis and research hypotheses
With the continuous acceleration of China's urbanisation process and the implementation of tourism poverty alleviation policies since 1999, farmers in the study area have gradually transitioned away from traditional agricultural and forestry livelihoods to engage in agritainment operations or urban non-farming employment (Li et al. 2020).Tourism development and urbanisation have promoted the differentiation of smallholders' livelihoods.Based on the intensive level of tourism livelihood, previous studies dived the livelihood strategies into four categories (tourism-oriented types (TO), tourism and part-time migration work (TPM), migration work-oriented (MO) and migration work and part-time tourism (MPT)) leading to different livelihood outcomes (Ding et al. 2021).High-quality livelihood resilience is the basis by which rural smallholders can cope with livelihood risks and capture livelihood opportunities (Zhou et al. 2021).
In reference to previous literatures, this paper constructed a tourism livelihood resilience framework based on several contested models from multiple perspectives, to explore the correlation between household livelihood resilience and their livelihood strategies (Figure 1).These include the SLA (DFID 2000), the multi-layered social resilience framework (Obrist, Pfeiffer, and Henley 2010), as well as the indicator framework for assessing livelihood resilience (Cahyanto and Pennington-Gray 2017;Speranza C, Wiesmann, and Rist 2014).From the adopted definition of resilience, four major attributes which can further be decomposed into various proxy indicators, are usually identified, namely buffer capacity (BC), Self-organisation capacity (SOC), Learning capacity (LC) and community protective capacity (CPC) (Cahyanto and Pennington-Gray 2017;Speranza C, Wiesmann, and Rist 2014).
Buffer capacity refers to the ability of a system to absorb disturbances while remaining in the same domain of attraction (Obrist, Pfeiffer, and Henley 2010).From strategic management and sustainable livelihood perspective, buffer capacity encompasses various livelihood aspects, such as asset-focused coping strategies and short-term dynamics (Cahyanto and Pennington-Gray 2017;Obrist, Pfeiffer, and Henley 2010;Speranza C, Wiesmann, and Rist 2014).Agritainment operations require substantial financial, natural and human investments, and the availability of these assets affects the ability to engage in tourism (Su et al. 2019).Therefore, a lack of capacity to cope with risks is a major constraint on agritainment intensification or extensification.In contrast, households with high human assets but lacking high-quality natural assets may opt for migration work that depends primarily on labour force participation (Liang et al. 2010).Based on this, the study makes the following hypothesis: H1 The stronger the buffer capacity of smallholders, the more smallholders tend to tourism intensification livelihood strategies.
Self-organisation capacity (SOC) refers to the ability to renew and reorganise one's livelihood system to adapt to uncertain situations (Folke 2006).Dynamic capability theory posits that exploring new opportunities, developing new products/services and cultivating collaborative relationships are essential for building new routines (Becken and Hughey 2013;Jiang et al. 2019).Agritainment operation is prone to disruption, seasonality, consumer volatility and institutional change (Lasso and Dahles 2018).This requires innovation in household business format and Changes in the portfolio of livelihood elements to maintain and enhance their livelihoods in response to the changes (Shen, Hughey, and Simmons 2008;Su et al. 2019).Self-efficacy involves thriving in the face of significant adversity (Obrist, Pfeiffer, and Henley 2010), and is linked to overcoming potential risks within the entrepreneurial process (Pu, Zheng, and Fu 2016).Opportunity recognition and Utilisation Capabilities support families in recognising and leveraging opportunities during times of crisis to improve their livelihood systems (Zhu, Tang, and Murphy 2009).Cooperation and networks facilitate self-organisation by promoting trust, facilitating information-sharing, and providing access to resources and collaboration opportunities (Speranza 2010).By comparison, household lacking Self-organisation capacity may be subject to limitations in livelihood adaptation to dynamic context.They may be forced to turn to migration work or broaden the livelihood portfolio of activities.Based on this, the study makes the following hypothesis: H2 The stronger smallholders' self-organisation ability, the more inclined smallholders are to adopt tourism intensification livelihood strategies.
The capacity for learning implies adaptive management, indicating that a resilient livelihood system is a learning system that incorporates disseminate knowledge and previous experiences into current livelihood action (Speranza C, Wiesmann, and Rist 2014).Generally, under the background of the urban-rural dual structure, improvement in the capacity for learning, especially in education level, makes the labour force more inclined to choose non-farming employment and more advanced occupations in cities or developed Coastal Regions (Tang, Liu, and Shu 2022;Zhou et al. 2021).On the other hand, digitalisation of the tourism industry and the environmental sensitivity of e tourism industry require the smallholders to learn from past experience, access to new knowledge and skills, and identify vital knowledge for future development to enhance their adaptive capacities (Su, Wall, and Wang 2017).Failure to do so may compel smallholders to diversify their livelihood strategies from one livelihood margin to the margin of another (Lasso and Dahles 2018).Based on this, the study makes the following hypothesis: H3 Stronger the Learning capacity makes smallholders more inclined to adopt intensification or migration-work intensification livelihood strategies.
Community protective capacity (CPC) refers to the supportive factors of a social structure (Obrist, Pfeiffer, and Henley 2010).Owing to the collectiveness feature of tourism resources, tourism products, a sustainable tourism livelihood is embedded in a community tourism context (Shen, Hughey, and Simmons 2008).This leads to issues of benefit-sharing and access to tourism markets, two important forms of community participation in tourism development (Teresa et al., 2009).In developing countries, particularly in rural areas, access to the tourism market of local people is often confronted with operational and structural barriers (Tosun 2000).Thus, Institutional arrangement as well as community tourism development have been proven to be of great importance support to the tourism livelihoods development (Ashley 2000).Otherwise, if smallholders are constrained by the community transforming structures and processes, they may be obliged to undertake migration work (Su et al. 2019).Based on this, the study makes the following hypothesis: H4 Community protective capacity makes rural residents more likely to adopt tourism intensification livelihood strategies.

Conceptual model development
Four major components of livelihood resilience for tourism smallholders were identified at household and community levels by referencing existing literatures (Baffoe and Matsuda 2018;Jezeer et al. 2019;Shen, Hughey, and Simmons 2008;Sina et al. 2019;Speranza C, Wiesmann, and Rist 2014;Zhou et al. 2021): buffer capacity, self-organisation capacity, learning capacity and community protective capacity.Each sub-component was measured using a 4-point scale.
Based on previous research (Baffoe and Matsuda 2018;Jezeer et al. 2019;Kuang, Jin, and He 2020;Quandt 2018;Speranza C, Wiesmann, and Rist 2014), livelihood capital assets are used to measure buffer capacity.In this article, 11 basic indicators have been selected, which reflect the five types of livelihood assets and are based on SLA and Tourism smallholders' Livelihood System (Table 1).
Self-organisation capacity (SOC) refers to the ability to reorganise internal and external resources to adapt to rapidly changing environments (Teece et al. 2007).Based on the foregoing studies (Li et al. 2020;Quandt et al. 2018;Speranza C, Wiesmann, and Rist 2014;Zhou et al. 2021), three core variables for capturing SOC are identified: environmental insight, social cooperation qualities and the self-efficacy of smallholders.
From perspective of strategic management, three significant attributes of Learning capacity are identified: proactive learning capabilities, knowledge identification capability and commitment to learning culture (Jiang et al. 2019).In the reference literatures (Cahyanto and Pennington-Gray 2017;Jiang et al. 2019;Orchiston, Prayag, and Brown 2016;Speranza C, Wiesmann, and Rist 2014) and considering the tourism livelihood systems in rural transformation (Li et al. 2020), willingness for learning, training opportunities and information availability were identified as the proxy indicators corresponding to the three primary attributes.
Community protective capacity (CPC) refers to the enabling factors that foster resilience building by facilitating access to social, cultural and economic capital (Obrist, Pfeiffer, and Henley 2010).Communities and community tourism operation constitute the closest environment that shape farmers' livelihood (Liu 2020;Shen, Hughey, and Simmons 2008).Refer to the literature on community participation in tourism (e.g.Li et al. 2020;Quandt 2018;Shen, Hughey, and Simmons 2008), the community protective capacity (CPC) is mainly composed by factors such as community Tourism Development, Community institutional asset, and Coordination between Tourism and community, etc.
The TLRI was derived for each of the sampled households using four major components: Buffer capacity (BC), Learning capacity (LC), Self-organisation capacity (CSC) and Community protective capacity (CPC).Each major component included several sub-components.We standardised the values of the sub-components using the following equation to scale each sub-component from 0 to 1: where IndexX ij denotes the standardised value of the sub-component for household i, S ij denotes the j th raw indicator value of household i, S min and S max denote the minimum and maximum possible values, respectively.
To obtain the value of each major component, the sub-component values were averaged using the following equation: where F i denotes the value of one of the components of household i, including buffer capacity (BCI i ), self-organisation capacity (SOCI i ), learning capacity (LCI i ) and Community protective capacity (CPCI i ).IndexX ij is the indicator value of the j th indicator, and n indicates the number of sub-components in each superior component.Specifically, referring to the calculation method for the adaptation capability index (Pandey et al. 2017), livelihood resilience, and the interactions between enabling factors and capacities operate at household level (Obrist, Pfeiffer, and Henley 2010), four contributing factors of resilience were combined to calculate the TLRI i using Equation (3): where TLRI i is the value of the adaptive capacity of the i th household, BCI i is the calculated buffer capacity score, SOCI i is the calculated self-organisation capacity score, LCI i is the capacity score for learning, and CPCI i is the calculated community protective capacity score.
Adapting FANRPAN (2021) and Baca et al. (2014), the value of major component, sub-component, and the TLRI was classified into three levels: low level (0.00-0.33), moderate level (0.34-0.66) and high level (0.67-1.00).In addition to the descriptive statistics, STATA version 16 was used to perform the Kruskal-Wallis H-test for each sub-component, major component, and the overall TLRI to examine the differences across four livelihood groups.
To explore the correlation between residents' livelihood resilience and residents' livelihood strategies, we used an ordered logistic regression, as described by Zhou et al. (2021).The dependent variable was the type of smallholder livelihood adaptation strategy, which is an ordinal and multiclassified variable.The independent variables were smallholders' livelihood resilience Sub-components classified variables.The model was specified as: Y refers to the residents' livelihood strategy.X i are the sub-components of livelihood resilience (Buffer capacity, Learning capacity, Self-organisation capacity, and Community protective capacity).α j and β i , respectively, represent the parameters to be estimated of the mode.ε i is the residual terms of the model.

Study area
Yihe River is an important tributary on the south bank of the Yellow River, with a total length of 264.88 km and a drainage area of more than 6100 square kilometres.The upper reaches of the Yihe River are located in Luanchuan County, Funiu Mountain Area in Central China (Figure 2).The upper reaches of the Yihe River have a length of 11 km and a drainage area of more than 1053 square kilometres, with an elevation of 450 m to 2212.5 m.It is the core area of a national nature reserve that provides biodiversity protection, soil and water conservation along the middle route of the South-to-North Water Transfer Project.The per capita cultivated land area here is close to the warning line of 0.05-hectare per capita cultivated land determined by FAO.It is one of 14 contiguous poverty-stricken areas in China.
The "project plans for poverty alleviation of Henan Province (2014-2020)" regards tourism as an important way to accelerate the development of rural revitalisation in this area.The development of rural tourism is centred on natural ecological resources and follows a "Scenic spots + communities + smallholder" pattern (Chen et al. 2020;Wu, Yang, and Chen 2020).At present, 45 villages prioritise rural tourism as their primary industry, with 1205 agritainments operated by smallholders and employing 3880 workers.However, tourism smallholders in the region are exposed to several livelihood risks, including competition, demand change, seasonal fluctuations, knowledge and policy risks (Cao, Wang, and Duan 2016).As a result of the fluctuations in tourist visits, many smallholders engage in migrant work during low season to diversify their income.
Based on the tourism destination life cycle (TDLC), Lao Junshan, Chong Dugou and Zhuangzi have reached the mature stage of tourism development.Smallholders in these villages have been involved in tourism for more than six years.Yangzigou's tourism economy and smallholder participation are significantly impacted by imperfect community governance and frequent replacement of community tourism operators.Beixiang, Xiexin and Yangshu are poor villages with limited resources.
However, the government's anti-poverty and rural revitalisation strategies promote tourism development in Xiexin and Beixiang, and the rural tourism system is in its early stages.Yangshu is a seasonal village located on top of a mountain.Villagers migrate to nearby towns during winter due to limited water supply, power supply and heating.In summer, they return to the countryside for agritainment operation.

Data collection
The questionnaire and "face-to-face" method were used to obtain data at rural households and community level.A survey was conducted and 550 household from seven tourism villages (Table 2) were interviewed in October 2019, to obtain relevant data of communities and farmers in 2019.At the community level, the survey of village committee members was conducted mainly through structural questionnaires supplemented by open-ended interviews.From this, we obtained the community tourism development data and the list of community tourism participants.In terms of  smallholders, 7 tourism villages were selected according to the distance from the county seat and the level of economic development by using judgmental sampling.Then, according to the list of tourism participants provided by the village committee, randomly select farmers to conduct household survey.Generally, the head of the household was interviewed, but if she/he was not available, the spouse was interviewed.A total of 550 questionnaires were distributed to tourism smallholders, and 524 effective questionnaires were collected, with a recovery rate of 95.27%.By the reliability and validity test, the value of corrected item-total correlation coefficient f value, Cronbach's alpha coefficient, Kaiser-Meyer-Olkin value and the Bartlett sphere test p-value are 0.631, 0.875, 0.857 and 0.00, respectively.

Livelihood strategies
By using k-means clustering method, the smallholders were divided into four livelihood types based on income from tourism and migration work (Ding et al. 2021;Li et al. 2020) TO, TPM, MO and MPT.
The livelihood characteristics of the four livelihood groups are shown in Table 3.The p-values for the Kruskal-Wallis H-test showed significant differences among the four livelihood types in livelihood assets (Figure 3).TO and TPM small farmers are the major groups of agritainment operation embedded in the rural area, who take root in rural areas and take rural tourism as the main livelihood (Ding et al. 2021;Wen et al. 2020).TO smallholders are often rural elite groups who have information, resources and other advantages brought by structural holes (Wu, Li, and Qiao 2014).They transform their resources advantages into entrepreneurial advantages, and form a demonstration effect on local residents through local network connection (Liu 2020).TPM smallholders are followers of TO farmers in agriculture operation.Their adaptability to the new development environment, ability to capture development opportunities and participation in rural tourism are lower than that of TO smallholders.MPT smallholders are rural marginal groups, who generally have low ability to participate in urban nonagricultural employment and rural tourism, and often face the pressure of family income and expenditure (Jia et al. 2018;Li et al. 2020).MO farmers are rural-urban floating population who are engaged in labour-intensive industries in developed eastern regions or cities in the late twentieth century.Nowadays, with the upgrading of coastal industries and the implementation of a rural revitalisation strategy in central and western China, the return of floating populations is essential to rural development (Tang, Liu, and Shu 2022).MO smallholders use county-level central cities as their primary living space (Luo, Cao, and Gu 2020).Faced with the pressure of high expenditure, especially in housing and education, they draw upon their material, financial and human capital accumulation and social connection with the countryside to diversify their livelihoods by participating in agritainment operation (Ding et al. 2021).

Resilience differentiation among livelihood strategies
The values of the TLRI and its components of four livelihood groups are presented in Table 4.The average TLRI, four major components, and 18 sub-components ranged widely from 0.000 to 1.000.The p-values for the Kruskal-Wallis H-test showed significant differences among the four livelihood types for all major and sub-components, except for the Learning opportunities (p = 0.113).The result is consistent with Su et al. (2018), who show that may the rise of livelihood freedom brought by tourism participation varies among community members due to the tension between requirements of initial investments and the assets constraint (Su et al. 2019).
The mean value of the major indicator on Buffer capacity for the four livelihood strategies ranged from 0.264 (low resilience) to 0.706 (high resilience).The order of the mean value of Buffer capacity of four livelihood strategies is as follows: MPT < TPM < MO < TO.MPT smallholders had the lowest Buffer capacity (0.264) and lowest five livelihood assets.Especially, the sub-indicators on Financial assets, Social assets, Physical assets and Natural assets of MPT smallholders were 0.242, 0.277, 0.150 and 0.267, respectively.These findings are consistent with the actual situation in upper reaches of the Yihe River that most rural households in deep mountain communities are confronted with low adaptation capacity due to the remoteness, low accessibility and availability of resources (Cao, Wang, and Duan 2016).On the other hand, previous studies found that participation in tourism sometimes enlarges the gaps in livelihood asset endowments among smallholders (Su et al. 2018;Su et al. 2019).In our study, TO smallholders had the highest livelihood resilience with respect to Financial assets (0.917), Physical assets (0.819) and Natural assets (0.792), while MO smallholders had highest Social assets (0.511) and Human assets (0.556).Additionally, Compared with MO smallholders, TPM smallholders had higher Physical assets (0.496) and Natural assets (0.504).The result indicated that in the context of urban-rural dual structure, farmers with high human assets tend to urban non-agricultural employment (Liang et al. 2010).By contrast, natural assets and material assets are the basis for farmers to participate in agritainment operation (Kuang, Jin, and He 2020), which in turn contributes to the accumulation of material assets, especially in the context of the upgrading of tourism industry driven by the upgrading of tourism demand (Ding et al. 2021;Lu et al. 2020).
The mean value of the major indicator on Learning capacity of the four livelihood strategies ranged from 0.607 (moderate resilience) to 0.856 (high resilience), indicating that the resilience in terms of Learning capacity was not fairly low.The sub-components of TPM smallholders, MO smallholders and TO smallholders, are higher than 0.66 except for the Learning motivation of MO smallholders and Knowledge acceptance ability of TPM smallholders.The results could be attributed to the fact that supported by the rural revitalisation policy, local government provides a lot of training and information for smallholders in tourist destinations through the "Touyan project entrepreneurship training for people returning to the countryside" and "digital village project", etc. (Li et al. 2020;Su et al. 2019).On the other hand, as tourism development replaces traditional livelihood activities and the knowledge and skills that accompany them, livelihood transformation rising from tourism development and the dynamic context drive smallholders to learn and to access the information (Chen et al. 2020).TO smallholders showed the highest Learning capacity (0.856) among the four livelihood groups flowed by MO smallholders (0.760) and TPM smallholders (0.716), while the MPT had the lowest Capacity for learning.Previous studies found that MPT smallholders are generally the disadvantaged groups who are living in marginal situations in rural transformation, especially in terms of new livelihood knowledge and information (Jia et al. 2018;Pandey et al. 2017).On the other hand, earlier studies showed that the MO smallholders and TO smallholders are often rural elite groups (Wu, Li, and Qiao 2014) who are at the structural hole nodes which brings them information and resource advantages (Li et al. 2020;Liu 2020).
The mean value of the major indicator on self-organisation capacity of the four livelihood groups ranged from 0.458 (moderate resilience) to 0.894 (high resilience).TO smallholders had the highest self-organisation capacity (0.894) due to the high self-efficacy (0.875), high Opportunity recognition capabilities (0.958) and high Opportunity utilisation capabilities (0.847).By our survey, as the rural elite in the rural transformation, TO smallholders had successfully transformed into professional tourism operators with entrepreneurial attributes and been better integrated into the modern industrial system (Li et al. 2020).MO smallholders had high self-organisation capacity (0.674) following TO smallholders.According to the survey, MO smallholders are a group with higher education, which enables them to obtain high-income employment opportunities in towns and become the middle class (Liang et al. 2010).Good education and Broad vision made them insightful (0.800), while the High housing prices, high education and medical expenditure in cities in China make them have moderate self-efficacy (0.578) and Opportunity utilisation capabilities (0.644).TPM smallholders showed high Capacity for Self-organisation (0.672), moderate self-efficacy (0.620), high Opportunity recognition capabilities (0.710) and moderate opportunity utilisation capabilities (0.627).Previous studies found that middle-income group in rural areas are willing to follow TO smallholders to participate in limited innovation rather than take risks, due to the relaxation of income constraints (Wang 2009).As the marginal group in mountain rural area, MPT smallholders were generally subject to survival constraints.The family's task is to pursue income flow and avoid risks (Gao and Fan 2019;Wang and Wang 2019).Correspondingly, they showed moderate self-efficacy (0.389), Opportunity recognition capabilities (0.519) and opportunity utilisation capabilities (0.466) in the rural transformation and uncertain context.These findings are consistent with MARA (2019) which showed that imperfection of public services and social security leads to insufficient endogenous development momentum of rural households.
The mean value of the major indicator on Community protective capacity of TO smallholders and TPM smallholders ranged from 0.718 to 0.750 (both highly resilient), indicating that TO smallholders and TPM smallholders receive high support from community tourism development.Referring to the sub-components for the two livelihood types, the Development vitality of community tourism is the only factor that is moderate.The result indicated that Lack of Development vitality seemed to be one of the major constraints for livelihood development of TO and TPM smallholders (Pang and Bao 2022).Previous studies found that rural tourism destinations with mature tourism development and perfect community governance system provide more support for tourism participation and livelihood development of farmers (Li and He 2021;Zhao, Yang, and Zhang 2015).By contrast to TO and TPM smallholders, the mean value of the major indicator on community protective capacity of MPT and MO smallholders ranged from 0.450 to 0.594 (both moderate resilience), indicating that the MPT and MO smallholders received fewer support from community tourism system.Interestingly, similar results are found for four sub-components.According to our survey, MPT and MO smallholders are generally located in tourism villages with low tourism development level or incomplete tourism governance system.This is in line with prior research showing that access to the tourism market is constrained by the overall community tourism economic development (Zhao, Yang, and Zhang 2015).Furthermore, local people are commonly denied any significant opportunity to participate in the tourism market due to the community institutional assets (Pang and Bao 2022;Shen, Hughey, and Simmons 2008).

Testing of theoretical assumptions
Table 5 shows the results of the regression of the four dimensions of smallholders' livelihood resilience and their livelihood strategy choices.Test of parallel lines (0.796), test of Model fitting (0.000) and the value of cox & snell r 2 (0.617) indicated that the ordinal multi-classification logistic regression model can be applied to explore and perform quite well in explaining the influence of livelihood resilience on livelihood strategies.Regression results show that Buffer Capacity, Self-organisation capacity and Community protective capacity have a significant negative impact on the choice of tourism intensification livelihood strategies.By contrast, the correlation between Learning capacity and livelihood strategy choice is not significant.

Discussion
Based on survey data of tourism smallholders in the upper reaches of the Yihe River, this study empirically examined the livelihood resilience of tourism smallholders by the TLRI and analysed the correlations between smallholders' livelihood resilience and their livelihood strategies.Based on our survey, most households in tourism-guided rural areas engage in multiple livelihood activities, which aligns with existing literature on the subject.Studies such as those by Ellis (2000) and Zhang et al. (2019) indicate that smallholders favour income diversity to sustain their families during rural transitions.Our results reveal that overall, households with less diversified livelihoods have higher scores in the major and sub-components of the TLRI, such as MO and TO, compared to households with more diversified livelihoods like TPM or MPT.This finding is surprising and contradicts conventional observations that higher livelihood diversity leads to improved sustainability (Su et al. 2019).The results could be attributed to the fact that a less diversified livelihood approach may be more helpful in accumulating income and assets, which boosts a household's adaptability to cope with and adjust to change (Tao and Wall 2009).Conversely, high-diversified livelihoods are often passive selections that maintain household income and expenditures in a dynamic context (Lasso and Dahles 2018).From an evolutionary standpoint, households adopt either low or high diversification levels reactively or proactively to respond to changes in the tourism industry (Duan, Xie, and Morrison 2022;Ritchie and Jiang 2019).
Our regression result and descriptive statistic are consistent with research hypothesis H1, Su et al. (2019), andLasso andDahles (2018), indicating that smallholders with better buffer capacity are more likely to engage in tourism intensification livelihood.The reason for this could be that central and local governments have proposed to cultivate tourism as a strategic pillar industry of the national economy and taken it as an Important Approach for Rural Revitalisation (Li et al. 2020).This prompts smallholders with high Buffer capacity to strengthen investment in agritainment operation, especially rural elites, despite the risks and uncertainties (Wu and Shi 2021).Consistent with hypothesis H2, Pu, Zheng, and Fu (2016), Zhu, Tang, and Murphy (2009), and Speranza (2010), we found significant correlation between smallholders' Self-organisation capacity and their livelihood strategies.The possible reason is that the arrival of the information society and China's economic and social transformation have led to rapid changes in the tourism supply and demand (Qin 2016).This requires dynamic adjustment of tourism smallholders' livelihood systems to adapt to environmental changes, otherwise, they will be eliminated by the new development environment (Chen and Liu 2019).Consistent with hypothesis H3, Liu et al. (2020), andSu et al. (2019), suggests that learning capacity facilitates adaptation to dynamic contexts and promotes smallholders' engagement in tourism intensification livelihood by utilising their family endowments.Consistent with hypothesis H4 and Shen, Hughey, and Simmons (2008), our study found a significant positive correlation between community protective capacity and smallholders' tourism intensification livelihood strategies.This may be because high community protective capacity leads to fair distribution of tourism benefits and access to markets, thereby encouraging smallholders to participate in tourism development (Xu and Hao 2014).

Conclusions
The proposed TLRI, composed of the SLA and resilience, can identify multiscale livelihood resilience contributing factors for households' livelihood system in rural transition.It emphasises the combination of social structuration processes and actor-perspectives.The proposed framework and its empirical results enable stakeholders to identify components of livelihood resilience and help them to build future resilience to rural tourism crises.
Rural residents' livelihood strategies are divided into tourism-oriented types (TO), tourism and part-time migration work (TPM), migration work-oriented (MO), and migration work and part-time tourism (MPT), and portfolio of multiple activities is commonly adopted by rural residents due to dynamic context.The Kruskal-Wallis test indicates a significant difference among four groups in livelihood resilience.MPT smallholders have the lowest resilience (0.287) with the low Buffer capacity (0.242) as the greatest constraint.By contrast, MO smallholders showed high resilient because they had successfully transformed into professional tourism operators and better integrated into the modern industrial system.Low-diversity livelihoods were more resilient than high-diversity livelihood.
In terms of the relationship between livelihood resilience and strategies, smallholders tend to engage more in tourism intensification activities when they have stronger buffer capacity, selforganisation capacity and community protective capacity in their livelihood resilience.However, at the current stage of the case site, learning capacity does not significantly impact tourism intensification activities.The statistical analysis indicates that smallholders with high learning capacity usually choose either tourism intensification or migration-work intensification.
Based on our findings, we propose the following recommendations.Firstly, the Rural Revitalization Bureau should prioritise providing entrepreneurship training and information to MPT smallholders to enhance their self-efficacy and their ability to identify and capitalise on opportunities.Secondly, since the tourism industry is susceptible to crises, encouraging cross-urban and rural part-time businesses can create diversified employment opportunities beyond relying solely on tourism.Thirdly, although TO smallholders make up a small percentage of tourism smallholders, they are important innovators in agritainment operation and serve as role models for other groups.Therefore, it is crucial to provide TO smallholders with knowledge and information to foster continuous innovation and entrepreneurship.
In addition, there are still some deficiencies in this study.For instance, it only evaluates the resilience level of farmers' livelihood at a specific time, whereas the livelihood of farmers is a dynamic process.Future studies should adopt a longitudinal approach with multiple data collections to assess the evolutionary process of livelihood resilience.Also, the livelihood system's resilience is often viewed as a localised phenomenon, and this paper only examined the livelihood resilience in the upper reaches of the Yihe River in China, discussing the relationship between residents' livelihood resilience and livelihood strategies.Conducting comparative research on the resilience of tourism smallholders in various geographical settings could broaden the understanding of livelihood resilience.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Figure 2 .
Figure 2. Location of the study area and the survey sites.

Table 1 .
List of indicators selected for each dimension of resilience.

Table 2 .
Sampling distribution of sample farmers.

Table 3 .
General characteristics of the four livelihood groups.
Figure3.The livelihood assets difference of four livelihood groups.

Table 4 .
Value and Kruskal-Wallis test of TLRI, major-and Sub-components.

Table 5 .
Correlation between livelihood resilience and tourism smallholders' livelihood strategy.