Study on Sustainability of Shelter Forest Construction and Protection Behavior of Farmers in the Sandstorm Area of Hexi Corridor, China

Abstract

An accurate assessment of farmers’ participation in shelter forest construction and protection behavior is of great practical significance for the renewal and restoration of degraded shelter forests, improvements in the ecological environment, and the sustainable development of agroforestry. This study was based on the theory of planned behavior and structural equation modeling. The cognition and behavior of farmers in typical wind and sand areas of the Hexi Corridor were constructed to measure their participation in the construction and protection of shelter forests from the “cognition-behavior” dimension. The results showed that all three dimensions of farmers’ perceptions had a significant positive effect on farmers’ participation in the construction and protection of shelter forests. The combined path coefficients of the three dimensions were behavioral attitudes (0.337), subjective norms (0.216), and perceived behavioral control (0.170), with farmers’ behavioral attitudes having a more pronounced effect on forest construction and protection behaviors. In the behavioral response to the construction and protection of shelter forests, farmers have a positive attitude toward compensation, management, and pest control, while the response to adjustment pattern, willingness to pay, and tree species replacement is relatively insignificant. It is suggested that the government should increase the publicity of shelter forest protection, improve the ecological compensation and reward and punishment mechanism, improve the management system, and strengthen the training of farmers’ professional knowledge and skills. These measures could increase farmers’ awareness of forest construction and protection to ensure that farmers actively participate in the construction and management of shelter forests.

1. Introduction

The shelter forest plays an important role in soil and water conservation, protection of farmland benefits, reduction in wind speed, prevention of wind and sand hazards, and conservation of biodiversity [1,2,3,4]. As the world’s industrial and agricultural development causes ecological degradation, reductions in water resources, environmental deterioration, and other issues affecting the survival and development of human beings, the United Nations has proposed a decade for ecosystem restoration from 2021 to 2030. China’s Three-North Shelter Forest Project is a great ecological protection plan [5,6], and from 2020, China’s “Three North Shelter Forest Project” enters the sixth phase of the third stage. According to the direction of ecological protection and restoration in China’s 20th report and the latest national “Technical Regulations for Restoration of Degraded Shelter Forests”, and related studies, shelter forests are degrading in many places [7,8], and their renewal and restoration are urgent.

The protection system of the sand area is mainly composed of a peripheral sand-fixing forest and an inner farmland forest network of oases. Focusing on the value of constructing sheltered forests, the ecological benefit of protecting the agricultural and production living environment is still its primary role, followed by the economic benefits of the generated agroforestry. From the construction process, we found that the shelter forest moved from “collective management” to “individual management” regarding the transformation, and “private forest” and other forms of management and care also appeared [9,10]. In recent years, experts and scholars have conducted multi-level discussions from the perspective of the relationship between agroforestry development and farm households. Studies have shown that the decentralization of forest rights to individual governance not only effectively solves the problems of lack of motivation and the high cost and low quality of direct government management but also leads to the local participation of farmers in the management and active use of surplus forest products, improving management and care efficiency and farmers’ initiative. In the study of farmers’ perceptions and their behaviors, farmers’ management behaviors toward shelter forests are influenced not only by the government’s forest rights policy reform and different regional environments, but also by farmers’ individual characteristics. For example, personal satisfaction and the protection of future generations are expected in forestry, and age influences forest management [11]. In addition, food security and income diversification would enable tree-planting by farmers [12]. Finally, the government decentralization of forest rights to individuals or levying ecological service fees from private individuals would protect forest land fragmentation and promote individual motivation [13]. Thus, the relationship between farmers and forestry is inextricably linked and has value for continued research.

At present, most farmer’ perceptions and behavioral choices from the forestry perspective are focused on forestry management and reforestation, and the study found that farmers have a positive attitude toward maintaining the “sustainable development of land and people”. However, there is also a cognitive and behavioral contradiction regarding the beliefs that “forests and agriculture compete for water and nutrients, and forests are a long process from early growth to maturity, and their benefits are not as fast as those of agriculture”. As farmers are important stakeholders in the construction of shelter forests and the return of farmland to forests, it is important to clarify how farmers’ reasonable cognitive level will contribute to the construction and protection of shelter forests, which will guide the coordinated development of forestry and agriculture. In view of this, combined with the planting status of the farmers’ oasis protection system in Minqin County, Jinta County, and Guazhou County in typical wind sand areas of Hexi Corridor, China, based on the theory of planned behavior, we constructed a three-dimensional index system of farmers’ cognition of shelter forests in dry areas. One-dimensional indicators represent farmers’ behavioral attitudes, which include farmers’ perceptions of the natural environment and understanding of the policy system, and include three question options. These determine whether they believe, for example, that shelter forests can increase their income and improve their living standards. Two-dimensional indicators represent farmers’ subjective attitudes, which include farmers’ understanding of social demonstration behaviors, and reward and punishment mechanisms, and include two question options, to determine whether surrounding neighbors planting trees provide a kind of encouragement to them, etc. The three-dimensional indicator represents the farmers’ perceived behavioral control, which includes farmers’ understanding of the environment and their own ability judgment. These three question options determine the difficulty of local afforestation and their own skills, etc. In addition, we innovated six types of farmers’ behavioral responses to shelter forests, such as whether they should manage and ecologically compensate for current woodland construction and protection, or carry out tree species replacement, etc. Our study explores the mechanism of farmers’ cognitive influence on their shelter forest construction and conservation behaviors, reveals the interactions between various cognitive factors, and shows how these cognitive factors may affect farmers’ shelter forest construction and conservation behaviors. It also provides practical references for the conservation and sustainable development of regional agricultural and forest resources.

2. Materials and Methods

2.1. Study Site

The site for this study is Hexi Corridor in Gansu, located in northwestern Gansu Province, China (Figure 1). The climate of the region is a typical temperate continental arid climate, with rainfall gradually decreasing from east to west. Soils are mainly brown and gray desert soils. The desertification area of Hexi Corridor accounts for more than 90% of Gansu Province, an area with serious desertification in China. As a commercial food base in northwest China and a key construction area of the three northern shelter forests, since the start of the Three-North Shelter Forest project in 1978, a shelter forest barrier with Populus L., Elaeagnus angustifolia L., Haloxylon ammodendron, and Tamarix ramosissima was constructed in the north of the corridor within the 1200 km long wind and sand line and farmland. However, due to the constraints of natural conditions and human activities in the area, species within the area’s protection system such as mature Populus L. and Haloxylon ammodendron forests degrade and die due to drought and lack of water, and the ecological environment is still required constant attention [14].

Minqin County is located in the eastern end of the Hexi Corridor, the lower reaches of the Shiyang River. Minqin County is surrounded by two deserts, Tengger and Badan Girin, to the east, west, and north, and has a typical temperate continental climate. The area of desertification in the district accounts for 90% of the county, wind and sand disasters are serious, and shelter forests play an important role in protecting the local ecology. In 2007, Wen Jiabao, the then premier of the State Council, during his visit to Minqin, proposed that “Minqin must not become the second Lop Nor”, providing a warning regarding the deterioration of the ecological environment in the Shiyang River Basin of the Hexi Corridor. As Lop Nor was a lake in Xinjiang before 330 A.D., there was more water in the lake during this period, and the city of Loulan in the northwest was the famous “Silk Road” throat. However, due to climate change and the impact of human water conservancy projects, the water from the upper reaches decreased until it dried up, and now there is only a large salt crust. After Lop Nor dried up, the surrounding ecological environment changed dramatically, all the herbaceous plants died, the anti-sand guardian poplar trees died in pieces, and the desert advanced to Lop Nor at a rate of 3–5 m per year, and soon became one with the vast Taklamakan Desert. Since then, Lop Nor has become a place where no grass grows and is called the “Sea of Death” [15,16].

Jinta County is located in the north of the middle end of the Hexi Corridor, in the middle and lower reaches of the Hei River, at the edge of the Batangilin Desert. This has a typical temperate continental arid climate, where sandy land area accounts for 60% on the total. Sandstorms increased from 3–5 times a year in the 1980s to 20–24 times a year at present. Wind and sand disasters are serious, and the creation of a protection system is the basis of resisting wind and sand hazards and improving the habitat. Jinta County is also in the economic radiation zone of Zhangye and Jiuquan and Jiayuguan. Bearing east and west, and even south and north, Jinta oasis is also the security barrier of the Jiuquan satellite launch center and an important support for national border construction. Therefore, its geographical location is very important.

Guazhou County is located in the western end of the Hexi Corridor, the middle and lower reaches of the Shule River. With a typical temperate desert climate, the topography is dominated by mountainous areas, Gobi and flood plains, with 75 days of windy weather above level 7, every year. This area is known as the “world wind bank”. Shelter forests play an important role in maintaining the ecological security of the local oasis. Guazhou County is located in the northwest and Xinjiang Hami City, since ancient times, has been located from east to the west out of the traffic hub, holding the ancient Silk Road merchants and heavily populated towns.

2.2. Theoretical Framework and Research Hypothesis

The theory of planned behavior (TPB), which is widely used to study the relationship between farmers’ cognition and behavior, summarizes the factors that influence individuals’ cognition as behavioral attitudes, subjective norms, and perceived behavioral control, and that these three cognitions influence their behavioral responses. In the subsequent descriptions, we denote behavioral attitudes, subjective norms, perceived behavioral control, and behavioral responses by BA, SN, PBC, and BA. Behavioral attitudes are the degree to which an individual likes and dislikes a particular behavior. Subjective norms are defined as social pressure from external sources (e.g., friends, government, or NGO personnel) and internal sources (e.g., family and relatives) that can determine behavior. Perceived behavioral control is defined as the individual’s perceived ability to engage in the behavior. These three structures are combined to form a positive or negative behavioral intention. Thus, when these three structures (attitudes, subjective norms, and perceived behavioral control) occur under more favorable conditions, individuals generate stronger behavioral intentions [17,18,19]. Farmers’ behavioral attitudes toward shelter forest construction and protection were understood by drawing on Shi H T et al.’s definition of economic rationality and the ecological rationality of behavioral attitudes toward returning farmland to forest. Farmers’ evaluation of their degree of awareness of shelter forest policies and the economic and ecological benefits brought by shelter forests was also considered [20], as well as subjective norms, such as guidance and pressure from society and the external environment, and their influence on individual behavioral decisions. In rural areas, people’s access to information is concentrated on communications from village committees and neighbors, and farmers’ subjective norms regarding the construction and protection of shelter forests can be understood as being formed the policy guidance and surrounding social behavior when farmers participate in the construction and protection of shelter forests. Perceived behavioral control is the perceived ease or difficulty of an individual’s performing a specific behavior. Farmers’ perceptions of shelter forests are not only including the economic benefits of increased agricultural production, but also the difficulty of afforestation and the impact of coercive land. Farmers’ perceived behavioral control of constructing and protecting forest land can be understood as their judgment of their own ability and influence on the surrounding environment when participating in shelter forest construction and protection.

According to the theory of planned behavior, individual behavioral attitudes significantly affect behavioral intentions. When applying the theory of planned behavior, although there is no research on farmers’ attitudes toward the construction and protection of shelter forests, there is more evidence on farmers’ behavior in ecological environment to support the hypothesis of this study. Maintaining a positive attitude among farmers is an important prerequisite for pro-environmental behavior, as Karppinen found that farmers’ attitudes explain their choices to produce ecological foods [21]. Wang et al. found that the more positive the farmers’ attitudes toward ecological conservation, the more willing they were to reduce the use of pesticides in agricultural production [22]. Meijer et al. showed that when farmers are aware of the benefits of tree planting and have positive attitudes towards tree planting, then they are more willing to take afforestation actions [23]. Through a study of farmers in Pingbian and Xichu counties in Yunnan Province China, Jin L S et al. found that when farmers clearly understand the policy of returning farmland to forest and recognize the necessity of ecological compensation measures, they are more willing to participate in the action of returning farmland to forest [24]. Against this research background, the first research hypothesis of this study is proposed.

Subjective norms reflect farmers’ perceptions of external social pressures when performing specific behaviors. Bossange et al. studied the characteristics of farmers’ adoption of conservation tillage and observed that researchers’ recommendations for conservation tillage may help farmers to adopt conservation tillage [25]. Mutyasira et al. used an ordered probability model and partial least squares structural equation model to test farmer’ decisions regarding the adoption of sustainable agricultural practices (SAPs) in Ethiopia and showed that subjective norms had a significant effect on the number of SAPs that were adopted. Regardless of the nature of conservation activities, and whether these were subsidized or non-subsidized, subjective norms were related to farmers’ intention to implement conservation measures [26]. Wenning et al. conducted an ongoing survey of 16 important agricultural production areas in the north and south of Xinjiang, and the results showed that farmers are influenced by the exemplary norms factors of their neighbors and friends, and that positive exemplary norms lead to more positive responses to protective forest management behavior [27]. In summary, the second hypothesis of this study is proposed.

Perceived behavioral control is related to the perceived difficulty and controllability of an individual’s performing a specific behavior. It reflects the reality that a person’s personal control over their behavior will be constrained by involuntary factors such as time, resources, and the environment. When applied to the field of shelter forest policy, a farmer’s perceived behavioral control over shelterwood conservation and construction includes their consideration of the factors that promote or hinder shelter forest construction and conservation behavior, including household endowment resources, social resources, and prior production experiences. Boz studied Turkish farmers and found that a high income, better financial support, and better communication networks were effective factors influencing farmers to participate in pro-environmental projects [28]. Lou et al. integrated 23 studies related to tea farmers through a meta-analysis, and the results showed that tea farmers were more inclined to use green technologies in tea growing when they had higher levels of perceived behavioral control [29]. Empidi and Emang studied farmers in the Cameron Highlands region and found that when farmers were more involved in sustainable agricultural management, such as attending appropriate workshops or receiving manure management programs, they engaged in more conservation practices in the Cameron Highlands forested watershed area [30]. Accordingly, the third hypothesis of this study is proposed.

Subjective norms have a significant impact on attitudes, and people tend to consider the views of important reference groups when forming their own attitudes toward particular behaviors. Farmers’ attitudes toward following state programs are represented by social factors. Petty and Cacioppo suggested that individuals’ attitudes are influenced by others and the surrounding environment [31]. Similarly, Quintal et al. asserted that individuals consider the expectations of others when forming their own personal attitudes; therefore, farmers may use subjective norms as a source of information [32]. Daxini et al. found that the stronger the subjective norms perceived by farmers, the more positive their attitude towards environmental protection and the aware farmers were of the environmental risks of nutrient loss and the need for the full utilization of natural resources [33]. Maleksaeidi et al.’s study also showed that subjective norms are predictors of attitudes and when farmers consider that untreated agricultural wastewater is often an unwelcome phenomenon, people’s attitudes towards untreated wastewater become more negative as social pressure increases [34]. Accordingly, the fourth hypothesis of this study is proposed.

Farmers’ perceived behavioral control regarding shelter forest protection and construction can be understood as farmers’ perceived control ability to respond to shelter forest protection and construction. Attitudes are generally seen as mediating variables between perceived behavioral control and actual behavior, and attitudes are considered a key predictor of pro-environmental behavior. This can be seen in Rezaei et al.’s study regarding the adoption of renewable energy or support for more ecological land construction [35]. Shi H T et al. also found a significant positive effect of farmers’ perceived behavioral control on their behavioral attitudes toward fallowing [20]. Therefore, when farmers perceive they have a higher level of behavioral control, i.e., they believe that shelter forest protection and construction is less difficult and have sufficient knowledge and skills to build shelter forests, then farmers’ attitudes toward shelter forest construction are more positive and they eventually become more willing to promote specific shelter forest protection and construction behaviors. In summary, the fifth hypothesis of this study is proposed.

H1. 

Behavioral attitudes that characterize farmers’ perceptions of the protective forest environment and policy regimes positively influence farmers’ behavioral responses.

H2. 

Subjective norms that characterize farmers’ social example, and reward for forest establishment or punishment mechanisms for destruction of shelter forests positively in-fluence farmers’ behavioral responses.

H3. 

Perceived behavioral controls that characterize farmers’ understanding of the shelter forest environment and their own competence judgments positively influence farmers’ behavioral responses.

H4. 

Social example, and reward for forest establishment or punishment mechanisms for destruction of shelter forests positively influence the environmental perceptions and policy regimes of farmers.

H5. 

Farmers’ environmental understanding of shelter forests and their own capacity judgments positively influence farmers’ environmental perceptions and policy regimes.

Based on the theory of planned behavior and the existing research results, this paper proposes the above research hypotheses and constructs a theoretical framework (Figure 2).

2.3. Data Source

This study’s data were obtained from two in-depth interviews conducted by the subject team in August–September 2021 and August–September 2022 focusing on farmers in the Minqin, Jinta, and Guazhou counties of Gansu Province, which are severely affected by desertification. According to the actual situation, assessed through a stratified random sampling method and participatory rural assessment (PRA), 1–4 village farmers in each town of the three counties, which are close to the sand source and seriously affected by sanding, were selected for the questionnaire survey. A total of 713 questionnaires were obtained, and 650 valid samples were obtained by deleting the questionnaires with incomplete key information. The proportion of valid questionnaires was 91.2%. During the survey process, each household was interviewed for about 15–25 min, and photographed and recorded. The majority of the interviewed farmers were male, and their ages were mostly 40–60 and over 60 years old. The specific regional distribution and demographic characteristics of the research sample are shown in

Table 1. Regional distribution of the survey sample.

Sample Counties	Minqin	Jinta	Guazhou
Town/Village	10/22	7/16	10/26

and

Table 2. Demographic characteristics of the survey sample.

Variable	Category	N	%
Gender	Male	501	77
	Female	419	23
Age Group	0–40	48	7.4
	41–60	427	65.7
	61–75	175	26.9
Education level	Primary School	187	28.75
	Junior High School	313	48.21
	High School	150	23.04
Annual household income	≤CNY 20,000 (≤USD 2912)	48	7.36
	CNY 20,001–40,000 (USD 2912–5824)	302	46.56
	CNY 40,001–60,000 (USD 5824–8736)	218	33.49
	>CNY 60,000 (>USD 8736)	82	12.59

.

2.4. Model Construction

Structural equation modeling (SEM) is a multivariate statistical validation analysis method with the advantages of dealing with multiple dependent variables, estimating factor structure and factor relationships, and estimating the degree of fit, which usually requires theoretical or rule-of-thumb support and is widely used in the study of individual behavior in sociology and psychology [36,37]. SEM includes measurement and structural models, and in our study, the measurement model expresses the relationship between the observed variables (each questionnaire indicator) and the latent variables (BA, SN, PBC, and BR), as expressed in the form of measurement Equations in (1) and (2). The structural model expresses the relationship between each latent variable (BA, SN, PBC, and BR), as represented in the form of (3) structural equations.

$$X={\Lambda }_x\xi +\delta$$

(1)

$$Y={\Lambda }_y\eta +\epsilon$$

(2)

$$\eta =B\eta +\Gamma \xi +\zeta$$

(3)

$\eta$ is the endogenous latent variable, $\xi$ is the exogenous latent variable; $Y$ and $X$ are the observed variables of $\eta$ and $\xi$, respectively; ${\Lambda }_x$ and ${\Lambda }_y$ are the correlation coefficient matrices of the exogenous and endogenous latent variables, respectively, with their observed variables; $B$ and $\Gamma$ are the coefficient matrices of the endogenous and exogenous latent variables, respectively; $\zeta$ is the regression residual, indicating the unexplained part; $\delta$ and $\epsilon$ denote the errors between the exogenous and endogenous latent variables and the observed variables.

2.5. Questionnaire Design Description and Software Analysis

The study set four latent variables based on the theory of planned behavior: behavioral attitude, subjective norm, perceived behavioral control, and behavioral response. The questionnaire contained 21 questions, which included farmers’ basic information, such as survey location, survey time, name, gender, and age. The farmers’ responses about shelter forests include BA1 to BR6; each latent variable contained 2–3 questions, and each question had five options for farmers. For example, BA1, Knowing the policy of forest construction and protection, had the five options of strongly disagree, disagree, neutral, agree, and strongly agree. The behavioral response consisted of two parts, farmers’ willingness to pay for forestry and forestry protection, and farmers’ suggestions for forestry and forestry protection. Willingness to pay included not willing to pay, less than ￥10, ￥10–50, ￥51–100, and more than￥100, and it reflects the amount people are willing to donate for protective forests as an environmental item [38]. Against the background of arid and semi-arid areas, farmers’ suggestions regarding the construction and protection of shelter forests were put forward by consulting agroforestry experts and organizing the responses into five categories: adjusting the pattern (adjusting the spacing, harvesting, etc.); replacing tree species (replacing and increasing drought-resistant tree species); compensating (increasing the proportion of ecological water use, returning cultivated land to equivalent acres of forest); managing (managing nurturing, replanting); and managing pests and disease. The farmers’ options also included five responses ranging from strongly disagree to strongly agree. The software we used for the analysis was IBM SPSS and AMOS, including a variable description, tests of reliability and validity, and graphing. A description of its operation can be found in the cited literature [39,40]. The specific measurement items and descriptive statistics are shown in

Table 3. Description of main test items and variables.

Latent Variable	No.	Test Items	Average	Standard Deviation
Behavioral attitude	BA1	Knowing the policy of forest construction and protection	3.625	0.805
	BA2	Shelter forests increase income and improve living standards	4.018	0.724
	BA3	Shelter forest to improve the ecological environment	4.285	0.846
Subjective norms	SN1	Participation of family and neighbors in the management of forest land	3.762	0.715
	SN2	Government policy rewards forestry and conservation practices	3.577	0.973
Perceptual behavior control	PBC1	Little difficulty in constructing forests and protecting them	2.903	0.908
	PBC2	Acquire professional knowledge and skills in forestry and forest protection	3.283	0.902
	PBC3	Able to assume the impact of forestation and conservation on agriculture	3.192	0.962
Behavioral Response	BR1	Willingness to pay for participation in construction and protection	3.780	0.960
	BR2	Compensation	3.926	0.749
	BR3	Replacement tree species	3.420	0.907
	BR4	Management of pests and diseases	4.143	0.829
	BR5	Management	4.185	0.860
	BR6	Adjusting the pattern	4.038	0.904

; the questionnaire is available in the Appendix A.

3. Results

3.1. Reliability and Validity Tests

To judge the reliability of the results, four latent variables, BA, SN, PBC, and BR, need to be tested for reliability before empirical testing (

Table 4. Reliability and validity test results of the variable data.

Latent Variable	Observed Variables	Crombach’s α	KMO	Bartlett’s Test	Factor Load
BA	BA1	0.882	0.856	7791.371 (0.000)	0.830
	BA2				0.872
	BA3				0.866
SN	SN1	0.852			0.906
	SN2				0.892
PBC	PBC1	0.942			0.919
	PBC2				0.885
	PBC3				0.916
BR	BR1	0.915			0.731
	BR2				0.781
	BR3				0.700
	BR4				0.875
	BR5				0.867
	BR6				0.789

). Cronbach’s a value of 0.882, 0.852, 0.942, and 0.915 were tested for reliability for BA, SN, PBC, and BR, respectively. Crombach’s α is a statistic that is the average of the discounted half reliability coefficients obtained from all possible methods of item division of the scale, and is the most commonly used reliability measure. Crombach’s α coefficient was considered to be more reliable above 0.7 [32], indicating that the scale passed the reliability test. The validity was tested by factor analysis, and the test result KMO value was 0.856, while Bartlett’s spherical test significance was 0.000. The Kaiser–Meyer–Olkin measure of sampling adequacy compares the relative magnitude of simple correlation coefficients and partial correlation coefficients between variables. The KMO value was required to be above 0.6, and a value of 0.8–0.9 was obtained; Bartlett’s test required p < 0.05, and the above test indicated that the scale passed the validity test. Factor loadings indicate the degree of correlation between the factor and the topic, with higher loadings indicating that the topic is more representative of the factor; a result usually higher than 0.45 is better. The potential variables in the structural equation (SEM) are measured with multiple measures, and if all potential variables with multiple indicators are used in the structural equation analysis, there is a risk that the model fit will be reduced because the mathematical and theoretical operations of the structural equation are too complex and the estimated parameters are too many. From the results of the previous analyses, the results of the validation factor analysis for each variable indicate that the construct validity meets acceptable standards; therefore, it is reasonable to replace multiple measures with a single measure. In structural equation models, the likelihood ratio test can be used for models with nested relationships, i.e., the change in the chi-square test value of the goodness-of-fit of the model and its degrees of freedom are calculated as the difference required to obtain the chi-square statistic and its degrees of freedom (called chi-square difference test). If the change in the chi-square value is greater than the change in its degrees of freedom, this means that the change in the model is indeed an improvement. The root mean squared error of asymptotic (RMSEA), which is less affected by the sample size, is a better absolute fit indicator: the smaller the value of this indicator, the better the model fit. The goodness-of-fit index (GFI), conventional fit index (NFI), and comparative fit index (CFI) were also used. The data values of these three fit indices are limited to between 0 and 1, and the closer they are to 1, the better the fit of the model is. It is generally considered that values above 0.8 are considered acceptable for the fit of the data to the theoretical model. In our study the validated factorial and SEM nested models were used to test for convergent and discriminant validity, and the results of the four-factor model test (

Table 5. Discriminant validity among study variables.

Measurement Models	Model Inclusion Factors	X2	df	RMSEA	CFI	TLI	∆X2	Note
a. Single factor	BA + PBC + SN + BR	3536.351	77	0.263	0.555	0.474
b. Two factors	BA + PBC, SN + BR	2165.313	76	0.206	0.731	0.678	1371.038 ***	Compared to a
c. Three factors	BA + PBC, SN, BR	1648.243	74	0.181	0.797	0.751	517.07 ***	Compared to b
d. Four Factors	BA, PBC, SN, BR	588.688	71	0.105	0.933	0.915	1059.555 ***	Compared to c

Note: “+” indicates that the two factors of before and after were combined into one factor; *** Significant at p < 0.05.

), with fit indices of X² = 588.688, df = 71, RMSEA = 0.105, CFI = 0.933, and TLI = 0.915, showed a good fit of the study model.

3.2. Model Fitness Check

The theory of planned behavior has a more restrictive scope of application, so further model testing was required. The theory of planned behavior has good explanatory and predictive power and provide a good theoretical basis for many studies in the areas of applicable behavior, including eating behavior, motor behavior, and social and learned behavior, including BA, SN, PBC, and BR. In this paper, the fit test of the indicators was performed using the AMOS 24.0 tool, and since the initial model fit did not meet the criteria, this criterion is shown in the recommended values of relevant scholars in

Table 6. Results of the modified structural model fit index.

Overall Model Suitability Index	Statistical Test Value	Estimated Value	Recommended Value	Fitting Results
Absolute Index	x2/df	4.803	<5.00	Qualified
	GFI	0.938	>0.90	Ideal
	RMSEA	0.077	<0.10	Ideal
Value Added Index	NFI	0.961	> 0.90	Ideal
	RFI	0.944	>0.90	Ideal
	CFI	0.969	>0.90	Ideal
Simplicity index	PGFI	0.572	>0.50	Ideal
	PNFI	0.676	>0.50	Ideal

, the model needed to be corrected according to the correction index. The results were ideal, indicating a good model fit (Table 6) [41].

3.3. Analysis of Empirical Results on the Influence of Farmers’ Perceptions on Shelter Forest Construction and Protection Behavior

The farmers’ responses for latent variables in the three counties are shown in Figure 3. Farmers in Minqin County had scores higher than or equal to the results of farmers in Jinta and Guazhou counties except for the low mean score of PBC1, which indicates that afforestation is more difficult in Minqin County as far as PBC1 is concerned. This is because Minqin is formed by the merging of two major deserts, as shown in Figure 1, and the desert’s influence make afforestation more difficult. From the latent variable responses from the overall regional perspective, farmers’ perceptions regarding shelter forests were greater in Minqin County than in Guazhou County and Jinta County. In addition, answers regarding behavioral responses revealed high scores for BR2 and BR5 and low scores for BR3 and BR6 in all three counties, indicating that farmers responded strongly to compensation and management, and less to tree replacement and the adjustment of patterns.

The results of model testing are shown in

Table 7. Standardized estimation results of SEM path coefficients of farmers’ perceptions and responses.

Hypothetical	Path	Estimate	S.E.	P	Results
H1	BA→BR	0.337	0.04	***	Establish
H2	SN→BR	0.216	0.028	***	Establish
H3	PBC→BR	0.170	0.029	***	Establish
H4	SN→BA	0.164	0.031	***	Establish
H5	PBC→BA	0.242	0.034	***	Establish

*** Significant at p < 0.001.

and Figure 4. The concept of path coefficient is equivalent to the regression coefficient shown in the regression analysis, which is used to illustrate the relationship between the model elements. A regression coefficient > 0 indicates that the two elements are positively correlated: the larger the value, the stronger the correlation. In addition to the path coefficient, the p-value of the path coefficient test is usually used to determine whether the path coefficient is significant. In our study the paths of all latent variables in the Table 7 model are significant, indicating that farmers’ behavioral attitudes, subjective norms, and perceived behavioral control all have significant positive effects on farmers’ behavioral responses to shelter forest construction and protection, confirming the previous hypotheses H1–H5.

Behavioral attitudes. The path coefficient of farmers’ behavioral attitudes on their behavioral responses was 0.337, which was significant and positively correlated at the 1% level, indicating a strong association in the positive direction compared to SN and BA. This shows that the more positive farmers’ behavioral attitudes, the more positive farmers’ behavioral responses to the construction and protection of shelter forests, confirming research hypothesis 1. The standardized factor loadings of 0.80, 0.90, and 0.85 for the three variables observed for behavioral attitudes were not significantly different, indicating that farmers’ knowledge of the shelter forests policy, their perception that constructing shelter forests can increase their income and improve their living standards, and their agreement that shelter forests improve the ecological environment can all increase their behavioral response to constructing and protecting shelter forests.

Subjective norms. The path coefficient of the influence of farmers’ subjective norms on their behavioral responses was 0.216, which was significant and positively correlated at the 1% level. This indicates that farmers’ behavioral responses are more positive when they are influenced by the supervision and demonstration of the community regarding forest establishment and protection, and research hypothesis 2 is confirmed. The standardized factor loadings of 0.81 and 0.96 for the two variables observed in the subjective norm indicate that the behavioral response of farmers’ participation in shelter forest construction and protection is encouraged and guided by their relatives, neighbors, and government policies, with government policies rewarding forest construction and protection behavior proving to be the more significant factors affecting their behavioral response. The path coefficient of the influence of farmers’ subjective norms on their behavioral attitudes is 0.164, p < 0.001, which is significantly positive, indicating that farmers’ subjective norms will influence their behavioral responses by mediating their behavioral attitudes. Therefore, hypothesis H4 holds.

Perceived behavioral control. The path coefficient of farmers’ perceived behavioral control on their behavioral responses was 0.170, which was significant and positively correlated at the 1% level. This indicates that the level of difficulty and knowledge perceived by farmers in the construction of shelter forests affects their behavioral responses; therefore, research hypothesis 3 holds. From Figure 2, it is clear the assumptions regarding the impact of forest construction and protection on agriculture is the most important factor determining the control of their perceived behavior (0.97). Through the survey, this is closely related to the shelter forest’s yield increasing effect on agriculture and the impact of coercive land on the district. Secondly, the difficulty of constructing and caring for forests (0.89) and the knowledge and skills required (0.89) was presumed to be greater by farmers in the three counties, due to the water and soil restrictions in the arid zone itself, as well as the competition for water between living, agriculture and forest land. Preliminary planting, later management and an assured water supply are key to the healthy development of shelter forests. The path coefficient regarding the influence of farmers’ perceived behavioral control on their behavioral attitudes was 0.242, which was significant and positively correlated at the 1% level, indicating that farmers’ perceived behavioral control would influence their behavioral responses by influencing their behavioral attitudes, confirming that hypothesis H5 holds.

In the test variables regarding farmers’ willingness to pay and related suggestions, the path coefficients of each variable were ranked as management (0.98) = compensation (0.98) > pest management (0.96) > adjustment pattern (0.73) > willingness to pay (0.71) > tree species replacement (0.60), indicating that farmers’ management of shelter forests (management, replanting) and compensation measures (increasing the proportion of ecological water, replanting) are the main concerns during the process of forest construction and protection. In the process of forest protection, the management of shelter forests and compensation measures were shown to be the key concerns of farmers. The second major concern was the management of pests and diseases. According to farmers’ reflections and view of the field, embroidery disease, rotten skin, and damage by aspen are problems for the overall farmland shelter forest in the study area, in addition to problems of powdery mildew, dead branches and rodent damage in the wind and sand control forest. Adjusting the pattern so that the three counties undertaking oasis shelter forest construction are concentrated at the periphery of the wind and sand mouth, with surrounding roads, canals, and fields around, formed a more stable pattern distribution for the shelter forest system. Behavioral adjustment regarding this were less obvious to other behavioral intentions, while farmers reflect that the current development of a “small grid, narrow forest belt” and low coverage of the forest belt configuration, is conducive to the protective benefits of shelter forests. Finally, willingness to pay and tree species replacement were studied. As the three counties are located in the arid and semi-arid areas, farmers are willing to pay a certain amount to strengthen the construction and protection of shelter forests and improve the local environment. The payment amount is low but the proportion is high. In the region, the drought-tolerant vegetation system of farmland shelter forests with Populus L., jujube, Ulmus pumila, and land wind and sand-fixing forests with sorrel and tamarisk has matured. Farmers think that replacing tree species is not realistic; instead, to prevent the degradation of shelter forests in the region, farmers are focused on strengthening the harvesting of degraded trees, replanting drought-resistant species and management.

4. Discussion and Conclusions

4.1. Discussion

Behavioral attitudes, subjective norms and perceived behavioral control were found to characterize farmers’ perceptions and had a significant positive effect on their behavioral responses to shelter forest construction and conservation. This is largely similar to the views of related scholars studying farmers shelter forest perceptions and behavioral intentions [20,24]. However, when comparing factor of the perceived behavioral control, which played the dominant role in farmers’ cognition in the Loess Plateau and Yunnan-Guizhou areas, this study found that farmers’ behavioral attitudes in the typical wind and sand areas of the Hexi Corridor had a more pronounced effect on their forestation and protection behaviors. This also indicates that farmers’ perceptions of the ecological and economic roles of shelter forests in the arid zone represent the most direct perception of their afforestation activities [42,43]. Therefore, the government and relevant authorities need to continuously promote shelter forest policy propaganda, enhance the popularity and distribution of this propaganda, emphasize the positive impact of shelter forest construction on farmers’ economic income and ecological benefits, and guide farmers increase their knowledge of shelter forest construction and protection. Secondly, with in farmers’ social network, the protection of forest construction and examples of this object should become a focus. A model household should be established, with friends and neighbors providing exemplary leadership, to fully cultivate farmers’ consciousness and initiate management. Finally, the policy incentive should be carried out in a timely fashion, for the farmers who have achieved outstanding shelter forest protection and construction work should be rewarded and publicized, and clear reward and punishment standards should be established. It is recommended to improve the “area to people, responsibility to people” shelter forest reward and punishment system.

Expectations regarding management and forestry ecological compensation are key factors in farmers’ behavioral responses and could serve to promote protection behavior [44,45]. For farmers in Minqin, Jinta, and Guazhou, the demand for ecological water in the region is the key to healthy development of shelter forests. At present, during the construction and subsequent development of shelter forests in the three counties, farmers’ behavioral responses all show that strengthening management, replanting, and managing pests and diseases are still the focus regarding the future restoration of degraded shelter forests and sustainable development in the region. The latest “Technical Regulations for Restoration of Degraded Protection Forests” promulgated by the state in 2020, also highlight the adoption of replanting and management approaches to restore the current degraded shelter forests. Therefore, we can encourage farmers to set up private shelter forest protection organizations under the guidance of the government by establishing agroforestry cooperatives, and we also encourage social funding to improve farmers’ skills via training on protective forest cultivation, renewal, pest control, and fire prevention. In addition, previous studies focusing on shelter forest conservation were mainly limited to studying direct financing for shelter forest construction and conservation, or focused on farmers’ willingness to plant and operate shelter forests. However, the results of this study show that when farmers have high subjective norms and perceived behavioral control over shelter forest construction and protection, they display a variety of responses, such as actively participating in pest and disease control and management of protection forests and degradation restoration. Therefore, in the concrete practice of shelter forest construction and protection, the relevant departments should also encourage and publicize that farmers’ can participation in the construction and protection of shelter forest through various ways, such as “using labor and donating funds”, to enhance the participation rate and popularity of shelter forest construction and protection.

This study on the construction and protection behaviors of farmers in Minqin, Jinta, and Guazhou counties of the dryland corridor aims to provide some reference for the restoration of degradation and the sustainable development of shelter forests in dry sand areas. There are still shortcomings in the study, such as the quantification and description of farmers’ perceptions of shelter forests with respect to changes in government policies and the economic benefits brought by shelter forests require long-term dynamic tracking. In addition, knowledge of the trade-offs between the adverse effects of agricultural income enhancement, such as coercive land, and competition for water between humans and land, and the ecological and economic benefits achieved during the construction of shelter forests could be enriched by subsequent studies.

4.2. Conclusions

The study investigated the influence of farmers’ behavioral attitudes, subjective norms and perceived behavioral control on their behavioral responses to shelter forest construction and conservation through a survey of farmers in Minqin, Jinta, and Guazhou counties, typical wind and sand areas of the Hexi Corridor, China. This study was based on the theory of planned behavior and structural equation modeling, with the following main findings.

Farmers’ cognitive influences on shelter forest construction and conservation behaviors, farmers’ behavioral attitudes, subjective norms, and perceived behavioral control all have significant effects on shelter forest construction and conservation behavioral responses. Farmers’ subjective norms and perceived behavioral control have a significant positive influence on their behavioral responses to shelter forest construction and conservation through the mediation of farmers’ behavioral attitudes, in which farmers’ behavioral attitudes were shown to have more influence on farmers’ behavioral responses than subjective norms and perceived behavioral control.

The magnitude of the combined path coefficients of the three dimensions of farmers’ perceived responses to shelter forest construction and conservation behaviors were behavioral attitudes (0.337) > subjective norms (0.216) > perceived behavioral control (0.170).

Among the five tested variables of farmers’ behavioral responses, farmers considered compensation and management to be the current priorities regarding shelter forest construction and protection, followed by pest management. Compared to the other three variables, farmers’ responses regarding adjustment pattern, willingness to pay for forest construction and protection, and tree species replacement were lower.

The study suggests that farmers need clear, direct information and various forms of education (workshops, field trips, and even personal mentoring) to implement protected forest conservation and creation as much as possible. Furthermore, it is not enough to emphasize the long-term ecological benefits: farmers need financial compensation or incentives to ensure the long-term protection of shelter forests and provide a more secure future for rural populations.