3.1 Reliability and Validity Analysis
3.1.1 Reliability Test
Reliability is used to describe the internal consistency of research variables and measurement items. Cronbach's alpha is an important criterion for testing the scientific and rationality of the items in a scale. Generally, when Cronbach’s alpha is greater than 0.7, the reliability test is considered to have passed.This study used SPSS 26.0 to conduct a reliability analysis on the scale (Table 5), and found that the overall alpha value was 0.852, proving that the questionnaire had high credibility and the scale item design was acceptable.
3.1.2 Validity Test
In the validity analysis, the KMO value was greater than 0.7 and the P value of the Bartlett test was less than 0.05, indicating that factor analysis was suitable.The results of this study showed that the KMO values of the independent variable landscape gene perception, the mediating variable local identity, and the dependent variable intention behavior were 0.870, 0.730, and 0.787, respectively, all greater than 0.7, and the Bartlett's test values were 5781.241, 921.651, and 1181.161, respectively, with a probability value P less than 0.05, indicating significant testing. The structural validity of the sample data was good, and there was a correlation between the research variables. AMOS software can be used for further factor analysis (Table 5).
Table 5 Variables and their reliability and validity(N=620)
Construct
|
Item
|
KMO
|
Bartlett’s test of sphericity
|
x2
|
df
|
Sig.
|
Landscape Gene perception
|
A1~D4
|
0.870
|
5781.241
|
91.000
|
0.000
|
Place Identity
|
E1~E3
|
0.730
|
921.651
|
3.000
|
0.000
|
Behavioral intention
|
F1~F4
|
0.787
|
1181.161
|
6.000
|
0.000
|
Note: ***P<0.001, **P<0.01, and *P<0.05
3.2 Exploratory Factor Analysis
Four principal components were extracted from the independent variables using principal component analysis to explain the total variance (Table 6). This was done according to the principle of eigenvalues greater than 1. The variance loading and contribution rates for the four components were 22.757%, 22.160%, 17.271%, and 17.079%, respectively. Together, they contributed to a cumulative variance contribution rate of 79.267%. A component serving as a mediator was extracted from the data (Table 7). The variance loading and contribution rate for this component was 79.200%, with a cumulative variance contribution rate of the same value. Another component, serving as dependent variable, was also extracted (Table 8) with a variance loading and contribution rate of 70.096%. Its cumulative variance contribution rate was 70.096%. The Kaiser normalizing method was used to perform factor rotation, resulting in a component matrix that reveals the independent variables: component 1 as environmental landscape gene, component 2 as cultural landscape gene, component 3 as layout landscape gene, and component 4 as architectural landscape gene. The mediating variable component was local identity, while intentional behavior was the dependent variable. The results indicate that the scale possesses exceptional construct validity; therefore, all components were maintained.
Table 6 Composition matrix after rotation of independent variables
|
Item
|
Component
|
1
|
2
|
3
|
4
|
A2
|
0.867
|
|
|
|
A3
|
0.856
|
|
|
|
A4
|
0.834
|
|
|
|
A1
|
0.818
|
|
|
|
D2
|
|
0.874
|
|
|
D4
|
|
0.851
|
|
|
D1
|
|
0.847
|
|
|
D3
|
|
0.815
|
|
|
B2
|
|
|
0.886
|
|
B1
|
|
|
0.844
|
|
B3
|
|
|
0.838
|
|
C1
|
|
|
|
0.877
|
C2
|
|
|
|
0.853
|
C3
|
|
|
|
0.822
|
Extraction method: Principal Component Analysis.
|
|
|
|
|
Rotation method: Kaiser normalizing maximum variance method
|
|
|
|
|
Table 7 Component matrix of mediating variables
|
Item
|
Component
|
1
|
E2
|
0.909
|
E3
|
0.891
|
E1
|
0.869
|
Extraction method: Principal Component Analysis.
|
One component was extracted.
|
Table 8 Component matrix of dependent variable
|
Item
|
Component
|
1
|
F1
|
0.879
|
F2
|
0.877
|
F3
|
0.800
|
F4
|
0.789
|
Extraction method: Principal Component Analysis.
|
|
One component was extracted.
|
|
3.3 Confirmatory Factor Analysis
To verify the correlation between potential variables, the AMOS software was used to construct a scale structural model (Figure 4), and the scale structural model was tested for structural validity, convergent validity, and discriminant validity.Structural validity reflects the fit of the model, which is determined by comparing various indicators in the research model with the indicator evaluation criteria to judge the fit of the model(Whittaker,2011).The test results (Table 9) showed that the minimum difference value to degree of freedom ratio (CMIN/DF) of the research model was 2.788, which was less than 3.The approximate root mean square error (RMSEA) was 0.039, which was less than 0.08.The goodness of fit index (GFI), relative fit index (RFI), comparative fit index (CFI), normed fit index (NFI), and other values were all greater than 0.9;the reduced fit index (PGFI), reduced normed fit index (PNFI), and other values were all greater than 0.5.This indicates that the scale model has a good fit.
Table 9 Model ft indices of the SEM model
Statistic
|
CMIN/DF
|
RMR
|
GFI
|
AGFI
|
PGFI
|
NFI
|
RFI
|
IFI
|
TLI
|
CFI
|
PNFI
|
PCFI
|
RMSEA
|
Recommended
value
|
<3.00
|
<0.05
|
>0.80
|
>0.80
|
>0.50
|
>0.90
|
>0.90
|
>0.90
|
>0.90
|
>0.90
|
>0.50
|
>0.50
|
<0.08
|
Model calculation value
|
2.788
|
0.039
|
0.921
|
0.895
|
0.694
|
0.944
|
0.932
|
0.963
|
0.955
|
0.963
|
0.782
|
0.798
|
0.054
|
Results
|
Ideal
|
Ideal
|
Ideal
|
Ideal
|
Ideal
|
Ideal
|
Ideal
|
Ideal
|
Ideal
|
Ideal
|
Ideal
|
Ideal
|
Ideal
|
The indicators of convergent validity include factor loadings, average variance extraction (AVE), and composite reliability (CR)(Hair et al,2009). A factor loading of >0.5, AVE of >0.5, and a CR of >0.8 generally suggest a valid scale model. As shown in Table 10, all items within the layout gene perception dimension, architectural gene perception dimension, environmental gene perception dimension, cultural gene perception dimension, place identity dimension, and intentional behavior dimension of the scale have standardized factor loadings that range between 0. 676 and 0.893, which exceeds the recommended value of 0.5. 0.696, 0.674, 0.691, 0.753, 0.693, and 0.610, which all exceed the recommended value of 0.5. The AVE values for the six dimensions are as follows: The CR values for the six dimensions are as follows: 0.873, 0.861, 0.870, 0.924, 0.900, and 0.860, all of which exceed the recommended value of 0.8. This indicates that the scale model has good convergent validity.
Table 10 Aggregation validity
|
Construct
|
Item
|
Standardization factor load
|
S.E.
|
C.R.
|
P
|
AVE
|
CR
|
Layout gene
|
B3
|
0.820
|
|
|
|
0.696
|
0.873
|
B2
|
0.870
|
0.045
|
23.230
|
0.000***
|
B1
|
0.811
|
0.044
|
21.899
|
0.000***
|
Architectural gene
|
C3
|
0.763
|
|
|
|
0.674
|
0.861
|
C2
|
0.855
|
0.056
|
20.618
|
0.000***
|
C1
|
0.841
|
0.058
|
20.423
|
0.000***
|
Place identity
|
E3
|
0.836
|
|
|
|
0.691
|
0.870
|
E2
|
0.872
|
0.045
|
23.715
|
0.000***
|
E1
|
0.784
|
0.038
|
21.461
|
0.000***
|
Environmental gene
|
A3
|
0.893
|
|
|
|
0.753
|
0.924
|
A2
|
0.871
|
0.034
|
30.495
|
0.000***
|
A1
|
0.837
|
0.033
|
28.195
|
0.000***
|
A4
|
0.868
|
0.034
|
30.283
|
0.000***
|
Cultural gene
|
D3
|
0.799
|
|
|
|
0.693
|
0.900
|
D2
|
0.875
|
0.043
|
24.285
|
0.000***
|
D1
|
0.830
|
0.044
|
22.780
|
0.000***
|
D4
|
0.825
|
0.046
|
22.599
|
0.000***
|
Intentional behavior
|
F3
|
0.694
|
|
|
|
0.610
|
0.860
|
F2
|
0.865
|
0.066
|
19.083
|
0.000***
|
F1
|
0.867
|
0.066
|
19.106
|
0.000***
|
F4
|
0.676
|
0.057
|
15.414
|
0.000***
|
Note: * * * p<0.001, **p<0.01, and *p<0.05
The discriminant validity assesses distinctions among various dimensions in a structural model. The key indicator for observation is the correlation coefficient among different dimensions, with the coefficient being below the square root of the AVE to indicate the scale's discriminant validity. Results from the test (Table 11) indicate that the correlation coefficients for the perception of layout genes and architecture genes, place identity, environmental gene perception, cultural gene perception, and intention behavior dimensions are all 0. 296, 0.400, 0.471, 0.370, and 0.449 are the correlation coefficients between the dimensions of architectural gene perception and place identity, environmental gene perception, cultural gene perception, and intention behavior, respectively. The correlation coefficient between architectural gene perception and place identity dimension is 0.455, while the correlation coefficients between architectural gene perception and environmental gene perception dimension, cultural gene perception dimension, and intention behavior dimension are 0.496, 0.306, and 0.445, respectively. The correlation coefficients for the relationship between the place identity dimension and the dimensions of environmental gene perception, cultural gene perception, and intention behavior are 0.462, 0.371, and 0.469, respectively. The correlation coefficients for the environmental gene perception dimension and the cultural gene perception dimension, as well as the intention behavior dimension, are 0.464 and 0.503, respectively. Additionally, the correlation coefficient for the cultural gene perception dimension and the intention behavior dimension is 0.434. The square roots of the Average Variances Extracted (AVEs) for the six dimensions of layout gene perception, architectural gene perception, place identity, environmental gene perception, cultural gene perception, and intention behavior are 0.834, 0.821, 0.831, 0.867, 0.833, and 0.781, respectively. All these values are higher than the correlation coefficients between these dimensions and others, indicating excellent discriminant validity for the scale structure model.
Table 11 Scale discrimination validity test results
|
Layout gene
|
Architectural gene
|
Place identity
|
Environmental gene
|
Cultural gene
|
Intentional behavior
|
Layout gene
|
1.000
|
|
|
|
|
|
Architectural gene
|
0.296
|
1.000
|
|
|
|
|
Local identity
|
0.400
|
0.455
|
1.000
|
|
|
|
Environmental gene
|
0.471
|
0.496
|
0.462
|
1.000
|
|
|
Cultural gene
|
0.370
|
0.306
|
0.371
|
0.464
|
1.000
|
|
Intentional behavior
|
0.449
|
0.445
|
0.469
|
0.503
|
0.434
|
1.000
|
Ave
|
0.696
|
0.674
|
0.691
|
0.753
|
0.693
|
0.610
|
Square root of AVE
|
0.834
|
0.821
|
0.831
|
0.867
|
0.833
|
0.781
|
3.4 Research hypothesis test
Under the assumption of adequate structural model fit, this study tested the nine research hypotheses. It is widely accepted that latent variables are significantly correlated when the P value is below 0.05 and the CR exceeds 2. The standardized path coefficient's sign (whether positive or negative) indicates the direction of correlation. As for the results (Table 12), all nine tested hypotheses in this study present P values below 0.05, with CR above 2, verifying their validity.
Table 12 Hypothesis test
Direct path
|
Standardized path coefficient
|
S. E
|
C. R
|
P
|
Hypothetical serial number
|
Inspection results
|
Place identity
|
<---
|
Environmental gene perception
|
0.184
|
0.056
|
3.309
|
0.000***
|
H2A
|
Establish
|
Place identity
|
<---
|
Layout gene perception
|
0.190
|
0.048
|
3.926
|
0.000***
|
H2B
|
Establish
|
Place identity
|
<---
|
Architectural gene perception
|
0.327
|
0.059
|
5.587
|
0.000***
|
H2C
|
Establish
|
Place identity
|
<---
|
Cultural gene perception
|
0.150
|
0.050
|
2.990
|
0.003**
|
H2D
|
Establish
|
Intentional behavior
|
<---
|
Environmental gene perception
|
0.128
|
0.039
|
3.322
|
0.000***
|
H1a
|
Establish
|
Intentional behavior
|
<---
|
Layout gene perception
|
0.136
|
0.034
|
3.968
|
0.000***
|
H1B
|
Establish
|
Intentional behavior
|
<---
|
Architectural gene perception
|
0.153
|
0.042
|
3.662
|
0.000***
|
H1C
|
Establish
|
Intentional behavior
|
<---
|
Cultural gene perception
|
0.133
|
0.035
|
3.782
|
0.000***
|
H1d
|
Establish
|
Intentional behavior
|
<---
|
Place identity
|
0.126
|
0.035
|
3.642
|
0.000***
|
H3
|
Establish
|
Note: * * * p<0.001, **p<0.01, and *p<0.05
This study used the bootstrap method of AMOS25.0, set a sample size of 5000, and a confidence interval of 95% to test the mediating effect of local identity on the influence of landscape gene perception dimensions on intention behavior.It is generally believed that the upper and lower bounds of the Bias-corrected and Percentile do not contain 0, which proves the mediating effect to be valid.The inspection results show that (Table 13) all four mediating effect hypotheses are valid.
Table 13 Hypothesis testing of mediating effect
Standardized mediation path
|
Mediation effect value
|
SE
|
Bias-corrected percentile 95%
|
Inspection results
|
Lower
|
upper
|
P
|
Environmental gene perception→Place identity→Intentional behavior
|
0.031
|
0.012
|
0.012
|
0.060
|
0.000***
|
Establish
|
Layout gene perception→Place identity→Intentional behavior
|
0.033
|
0.013
|
0.013
|
0.064
|
0.000***
|
Establish
|
Architectural gene perception→Place identity→Intentional behavior
|
0.047
|
0.016
|
0.023
|
0.087
|
0.000***
|
Establish
|
Cultural gene perception→Place identity→Intentional behavior
|
0.024
|
0.010
|
0.008
|
0.048
|
0.001**
|
Establish
|
Note: * * * p<0.001, **p<0.01, and *p<0.05
3.5Analysis Based on Demographic Characteristics
The independent samples t-test is used to analyze discrepancies in direct quantitative data between two separate groups, such as gender and birthplace. It is a method for testing differences. On the other hand, the one-way ANOVA analysis is a statistical technique that is utilized to ascertain whether certain demographic characteristics, such as age, income, and education level, have an effect on an outcome. Age, length of residency, level of education, monthly income, and occupation have a significant impact on test results, as demonstrated in Table 4. Male gender has a higher correlation than female gender with environmental genes, layout genes, and behavioral intentions, as evidenced in the study results (Table 14) with a significance level of p<0.05. Furthermore, birthplace has a statistically significant positive correlation with each latent variable (p<0.001). Age is significantly and positively correlated with all latent variables (p < 0.001). As residents' age increases, their perception of landscape genes, local identity, and willingness to protect also increase. Additionally, the correlation between these variables is stronger among local populations than non-local populations. Additionally, residence time is also positively correlated with all latent variables (p < 0.001). The longer residents stay in an area, the more they perceive and value landscape features and local identity, as well as their willingness to protect them. Each latent variable shows a significant positive correlation with occupation (p<0.001).As occupation increases, residents' perception of landscape features, local identity, and willingness to protect them also increase. Additionally, monthly income shows a significantly positive correlation with each latent variable (p<0.001). There is also a positive correlation between monthly income and residents' perception of landscape elements, local identity, and willingness to protect them (p<0.001). Increased education levels correlate positively with every latent variable, leading to increased appreciation from residents for local identity, landscape genes, and willingness to protect.
Table 14 characteristics of different population differences
Variable
|
Independent sample t-test
Significance (bilateral)
|
One-way ANOVA/p
|
Gender
|
Birthplace
|
Age
|
Residence time
|
Occupation
|
Education level
|
Monthly income
|
Environmental gene
|
0.002**
|
0.000***
|
0.000***
|
0.000***
|
0.000***
|
0.000***
|
0.000***
|
Layout gene
|
0.005**
|
0.000***
|
0.000***
|
0.000***
|
0.000***
|
0.000***
|
0.000***
|
Architectural gene
|
0.063
|
0.000***
|
0.000***
|
0.000***
|
0.000***
|
0.000***
|
0.000***
|
Cultural gene
|
0.160
|
0.000***
|
0.000***
|
0.000***
|
0.000***
|
0.000***
|
0.000***
|
Place identity
|
0.838
|
0.000***
|
0.000***
|
0.000***
|
0.000***
|
0.000***
|
0.000***
|
Intentional behavior
|
0.009**
|
0.000***
|
0.000***
|
0.000***
|
0.000***
|
0.000***
|
0.000***
|
Note: * * * p<0.001, **p<0.01, and *p<0.05