Scenic Beauty Evaluation of Forests with Autumn-Colored Leaves from Aerial and Ground Perspectives: A Case Study in Qixia Mountain in Nanjing, China

Abstract

In recent years, on-site visitation has been strictly restricted in many scenic areas due to the global spread of the COVID-19 pandemic. “Cloud tourism”, also called online travel, uses high-resolution photographs taken by unmanned aerial vehicles (UAVs) as the dominant data source and has attracted much attention. Due to the differences between ground and aerial observation perspectives, the landscape elements that affect the beauty of colored-leaved forests are quite different. In this paper, Qixia National Forest Park in Nanjing, China, was chosen as the case study area, and the best viewpoints were selected by combining tourists’ preferred viewing routes with a field survey, followed by a scenic beauty evaluation (SBE) of the forests with autumn-colored leaves in 2021 from the aerial and ground perspectives. The results show that (1) the best viewpoints can be obtained through the spatial overlay of five landscape factors: elevation, surface runoff, slope, aspect, and distance from the road; (2) the dominant factors influencing the beauty of colored-leaved forests from the aerial perspective are terrain changes, forest coverage, landscape composition, landscape contrast, the condition of the human landscape, and recreation frequency; and (3) the beauty of the ground perspective of the colored-leaved forests is strongly influenced by the average diameter at breast height (DBH), the dominant color of the leaves, the ratio of the colored-leaved tree species, the canopy width, and the fallen leaf coverage. The research results can provide scientific reference for the creation of management measures for forests with autumn-colored leaves.

1. Introduction

Colored-leaved forests (CLFs), commonly known as “Caiye Lin” in China, has emerged recently as a popular term in the country referring to the colorful forest landscape. According to Mu, Y.X., CLF refers to the collection of forest vegetation with a good visual aesthetic [1]. Colored-leaved plants are defined as plants that exist in nature or artificially cultivated environments and have high ornamental value due to a leaf color that is significantly different from the natural green during the growing season or at certain stages of the growing season [2]. According to Zhang Z., a CLF landscape features a forest landscape composed of colorful tree-dominated plant communities [3]. In the forest landscape, color is the information element that has the greatest impact on people’s senses, and tourists’ experiences and perceptions of the forest landscape mainly come from the attractiveness of the colorful forest leaves. When building CLFs, the focus is not on such management measures as tendering, transformation, or replanting but on the selection and spatial allocation of colored-leaved trees with a high scenic beauty.

A scenic beauty evaluation (SBE) of CLFs by tourists is the main method to evaluate the aesthetic degree of a forest landscape [4]. In the mid-1960s, the concept of “landscape evaluation” was first proposed and studied in the United States, but these early landscape evaluations in the United States were mainly about the visual aesthetics of the landscape, which was a qualitative evaluation method [5]. It was not until the 1970s that quantitative research methods were developed, such as the descriptive factor method [6], the public preference method [7], and the proxy composition method [8]. Up to now, four major schools of SBE have been recognized worldwide: the expert school [9], the psychophysical school [10], the cognitive school [11], and the empirical school [12]. Among them, the psychophysical school, proposed by Daniel [13], has the most mature evaluation technology [14]. Within the psychophysical school, the SBE method was widely used in the evaluation of forest landscape quality [15,16]. The quality of an SBE is decided by both the judge’s perception of the landscape and the judge’s evaluation system. SBE, which was developed to reflect people’s perceptions and evaluations of natural beauty with a numerical magnitude, has been used in the literature to predict the aesthetic quality of the landscape and to construct a linear relationship between landscape elements and the beauty value. The process of SBE mainly includes the following three steps: a landscape scenic beauty evaluation, a landscape element analysis, and the establishment of a landscape quality evaluation model [17]. A landscape beauty score is the visual reflection of individuals or groups on photos of a landscape within a certain period of time (e.g., 10 s). Under consistent shooting conditions of the landscape photos used for the evaluation, the SBE value obtained through public evaluation has a high reliability. Meanwhile, when the SBE method is used to establish a landscape beauty evaluation model by using a “landscape beauty evaluation score”, the beauty score should be standardized, which greatly eliminates the influence of the judge’s aesthetic scale on the evaluation system, thus reflecting the beauty of the forest landscape in a more realistic way. In addition, many scholars in China and abroad have conducted research on the differences in the aesthetic attitudes of evaluators. Most scholars believe that the differences in aesthetic tendencies of people with different cultural backgrounds are not significant, since the aesthetic judgments of different evaluation groups have an obvious consistency [18]. However, some scholars believe that, compared with other groups, college students are more objective in SBE and less utilitarian [19,20]. Therefore, college students were chosen as the judges in this study.

In previous studies, scholars in China and abroad mainly carried out research on the SBE of CLFs at the intra-forest scale [21,22,23], at the level of individual trees [24,25], and or at the level of view-sheds [26,27]. However, few scholars have evaluated the beauty of CLFs from the aerial perspective of an unmanned aerial vehicle (UAV). Tree species composition and the horizontal and vertical structure of a forest affect the SBE of colored-leaved forest landscapes [28,29,30]. Even for the same CLF types, the shapes and trunks of trees appear different depending on the perspective of observation, thus affecting the scenic beauty of forests [31,32]. Since the outbreak of the new coronavirus (COVID-19) in Wuhan, China, in December 2019, the virus has spread across the country, and there is still no sign of its extinction. Guided by China’s dynamic zero-COVID-19 policy, many scenic areas in China have adopted closure measures to restrict on-site visitation. As an alternative, some scenic areas have established online cloud tourism with three-dimensional, high-resolution aerial photos taken from a UAV. The factors affecting the beauty of a CLFs from the aerial perspective are quite different from those observed on the ground [33]. A comparative analysis of the factors affecting the beauty of CLFs from two different perspectives, aerial and on the ground, can provide a scientific reference for the management planning of autumn CLFs.

In this paper, Qixia Mountain National Forest Park in Nanjing, China, which has the reputation of being “the red colorful mountain with layers of dyed forest in autumn” and is listed as one of the four most famous red-leaved forest viewing areas in the country, was selected as the case study area. Through a comparative analysis of the factors affecting the beauty of CLFs from two different perspectives, the following research goals can be achieved: (1) to explore a selection method for the best viewing points of CLFs based on the tourists’ preferences; (2) to analyze the influencing factors of the SBE of CLFs from two different perspectives; and (3) to provide suggestions on improving the SBE of autumn CLFs under the situation of the COVID-19 pandemic.

2. Materials and Methods

2.1. Study Area

Qixia Mountain National Forest Park is located in the middle of Qixia District, Nanjing City, Jiangsu Province, China (E 118°94′98″~E 118°97′70″, N 32°14′80″~N 32°16′70″). The park is 20 km away from Taiping Gate, bounded by Nanjing Yangtze River Bridge in the west, Yangtze River in the north, Longtan Drum Platform in the east, and Ningzhen Avenue in the south, covering a total area of 27 km² (Figure 1). The average altitude of the park, which is high in the east and low in the west, is 286 m, with an undulating terrain composed mainly of low mountains and hills due to tectonic erosion. The park is located in the transition area between the north subtropical zone and the warm temperate zone, with a humid climate and four distinct seasons. The average annual sunshine in the park is 1628 h, the average annual temperature is 19.6 °C, and the average precipitation is 1530.1 mm.

Due to its unique climate, soil, and water conditions, there are rich plant resources in the park. There are a total of 589 species of 342 genera with 108 families, including 11 families of 14 genera with 19 species of ferns, 3 families of 4 genera with 4 species of gymnosperms, and 94 families of 324 genera with 477 species of angiosperms. The flora components are a result of the intersection of northern and southern climates, with both warm temperate deciduous broad-leaved species and subtropical evergreen and deciduous broad-leaved species, as mainly found in over-mature scenic forests. The park is covered with many huge and old trees, with a unit volume of 85 m³/hm² and a forest coverage of 94.6%, making it an important species gene bank and urban biodiversity protection base in Nanjing. With many kinds of colored-leaved species, the park in autumn has famous scenery, described as “the red mountain with layers of dyed forest” and “leaves in frost season redder than February flowers”, attracting tourists from all over the country. The park is the highlight of eco-tourism in Nanjing, and the forest is also the first choice for local people to visit and return to nature during holidays.

2.2. Research Methods

2.2.1. Viewpoint Selection

The assessment of visitor preferences was conducted by obtaining visitor comment data through Ctrip, a versatile open travel platform https://secm.ctrip.com/restapi/soa2/12530/json/viewCommentList (accessed on 18 November 2021). The structure of the obtained data included information such as user ID, check-in time (year/month/day/hour), travelogue-style comment text, pictures and ratings of scenic spots in terms of scoring the scenery, and interest and cost-effectiveness evaluation.

A convolutional neural network (CNN) was used to analyze the cognitive image preferences of visitors’ check-in images. The results were presented in the form of “adjective + noun”, where the noun describes the image content, and the adjective reflects the photographer’s emotional tendency towards a specific object. The specific technical route is shown in Figure 2, and the obtained images were parsed by DeepSentiBank, a specific technical tool, to obtain collections of APN collections taken by tourists at different attractions. Word frequency statistical analysis was applied to the adjective–noun (APN) pairs, where the noun part was regarded as the visitors’ cognitive image preference of the external landscape of Qixia mountain and the adjective part was regarded as their affective dimension. This provided data to support the assessment of visitors’ preferences. The tourists’ main viewing angle preferences, such as flat view, top view, and back view, were determined by analyzing the composition elements and the perspective deformation of the elements in the tourists’ photos. This allowed for the determination of the slope and direction preferences of tourists. The landscape texture details were also obtained to determine the visitors’ viewing distance preference. Based on the shooting destinations, the tourists’ attraction location preferences were determined. In addition, further information about tourists’ emotional preferences, route preferences, color preferences, viewing angle preferences, viewing distance preferences, and attraction (location) preferences was obtained by designing questionnaires for online and offline distribution.

In this context, the best viewpoints for the SBE of CLFs in the park were generated using a combination of tourists’ preferred influencing factors and characterization obtained through big data from Ctrip and questionnaires, with reference to the public landscape preference method [34,35], public aesthetic psychology [36,37], and safety management code for tourist attractions (DB51/T 2312-2017). The main factors affecting the best viewing points were extracted by ArcGIS as follows: elevation (reclassification), surface runoff (depression filling, flow direction, flow rate, river network, and river network classification), slope (slope analysis and reclassification), aspect (aspect analysis and reclassification), and distance from road analysis (buffer analysis) (

Table 1. Spatial characters of the ideal viewpoint in the study area.

Influencing Factors	Characteristic Description
Elevation	Plants have vertical distribution that varies with altitude, visibility at different altitudes is different, and the visual field space expands with the increase in altitude [38].
Surface Runoff	Water is the most important element of plant life. The existence of streams can not only adjust the microclimate and improve the comfort of the environment but also enhance the aesthetic features of the landscape [39].
Slope	According to the International Geographical Union, flat slopes, slight slopes, gentle slopes, and slopes of 0–15° are regarded as the best slopes for sightseeing [40,41].
Aspect	To avoid direct sunlight and obtain the best viewing angle, the true north direction is taken as 0, and the value of 0–360° is taken in the clockwise direction; the best slope aspects are 315°–360° and 30°–45° [42,43].
Distance from Road	Tourists prefer a flat area with a certain space. The best viewing distance for plants is 1 to 2 times the height of the plants, and 20 to 30 m from the road is the best viewing distance for scenic spots and surrounding scenery [44,45]. Therefore, it is best to choose a viewing point within 25 m of the road.

). According to the description of the spatial attributes given in Table 1, the specific attributes that ideal viewpoints should have were extracted, e.g., elevation raster pixels with elevations higher than 10 m, the best water view raster pixels are within 50 m of surface runoff, the slope raster pixels should 0°–15°, the aspect raster pixels should be 30°–45° and 315°–360°, and the best visual distance raster pixels are within 25 m from the road. The ideal viewpoint distribution areas were obtained using the overlay tool in ArcGIS, after which 50 scenic forest viewpoints with good view lines (located at the summit or ridge line) in these distribution areas were generated (Figure 3). The plant species, average diameter at breast height (DBH), canopy density, leaf color type, proportion of colored-leaved trees, and other stand characteristics within a square of 10 × 10 m of each viewing point were recorded by means of on-the-spot field survey (

Table 2. Presentation of some information from the sample sire survey.

Sample	Slope	Aspect	Elevation	Depression	Average DBH	Average Tree Height	Ratio of Colored-Leaved Tree Species	Fallen Leaf Coverage
1	14	354	127	0.75	25	17.3	0.98	0.85
2	6	41	127	0.74	25	17.5	0.98	0.83
3	8	35	145	0.45	16	8.8	0.95	0.83
4	7	318	202	0.65	18	9.6	0.9	0.85
5	15	41	252	0.95	18	10.2	0.98	0.82
6	8	315	205	0.85	18	9.8	0.95	0.65
7	3	336	177	0.25	12	7.3	0.5	0.53
8	12	31	174	0.45	10	3.3	0.23	0.33
9	11	327	137	0.00	6	2.3	0	0.08
10	13	32	135	0.35	16	8.3	0.1	0.05
11	8	316	150	0.85	20	12.4	0.89	0.93
12	12	44	89	0.65	16	8.4	0.33	0.43
13	7	36	195	0.75	15	7.6	0.13	0.73
14	13	35	172	0.80	14	6.6	0.73	0.56
15	15	35	105	0.95	18	9.8	0.88	0.85
16	15	42	95	0.80	15	7.6	0.95	0.91
17	8	359	91	0.15	6	2.1	0.33	0.13
18	14	39	110	0.75	16	9.3	0.84	0.81
19	8	360	111	0.45	12	4.9	0.75	0.83
20	8	31	119	0.80	25	18.2	0.96	0.85
21	12	341	56	0.85	16	8.3	0.98	0.88
22	3	32	87	0.75	13	6.2	0.83	0.85
23	10	318	37	0.78	20	11.3	0.23	0.43
24	13	45	62	0.90	8	2.8	0.86	083
25	9	359	81	0.75	25	17.9	0.15	0.08
26	5	337	125	0.65	14	7.3	0.23	0.68
27	7	42	137	0.80	18	9.3	0.86	0.86
28	5	45	143	0.65	12	4.5	0.22	0.35
29	15	32	143	0.85	18	10.4	0.95	0.86
30	13	38	124	0.85	16	9.1	0.93	0.84
31	10	356	63	0.15	0	0	0	0
32	8	346	37	0.55	18	9.1	0.85	0.91
33	8	44	29	0.95	25	17.6	0.97	0.82
34	13	337	24	0.45	16	8.7	0.48	0.76
35	2	350	47	0.65	20	11.7	0.28	0.58
36	10	358	29	0.15	0	0	0.53	0.84
37	15	39	34	0.80	18	10.3	0.33	0.45
38	13	38	76	0.35	14	7.1	0.13	0.11
39	9	42	47	0.75	18	11.8	0.5	0.6
40	9	42	74	0.95	16	7.9	0.83	0.56
41	13	44	80	0.45	10	4.1	0.18	0.12
42	14	36	130	0.25	18	8.8	0.56	0.04
43	12	31	123	0.45	13	7.3	0.18	0.24
44	9	330	109	0.80	15	7.6	0.56	0.41
45	13	332	114	0.55	12	5.1	0.86	0.46
46	12	335	133	0.50	25	17.5	0.96	0.7
47	12	45	119	0.45	14	6.6	0.58	0.43
48	12	37	120	0.80	18	11	0.32	0.35
49	13	43	69	0.35	8	2.4	0.88	0.51
50	11	45	79	0.55	16	8.7	0.89	0.83

).

2.2.2. SBE of Colored-Leaved Forests

In this study, an EOS 5D Mark IV 5D4 SLR camera and a DJI Phantom 4 Pro UAV were used to take aerial and ground photographs, respectively, of CLFs at 50 viewpoints, under the guidelines of identical external conditions, no other interference, and repeatable operation [46,47]. Clear days with sunny weather were selected for shooting photos in late November 2021, from 9:00 a.m. to 12:00 a.m. The ground photos were taken at a distance of 20 m from the viewpoint, 45° east–northwest of the oblique light, from bottom to top [48], while the aerial photos were taken 120 m above the ground (Figure 4 and Figure 5). For each viewpoint, only one representative ground photo and the corresponding aerial photo with the same geographical coordinates were selected and numbered in order to make a PowerPoint (PPT) slide for use in the SBE of the CLFs. The slides were used by 102 undergraduate and master’s students from Nanjing Forestry University, China, and Nanchang Institute of Science and Technology, China, to rank the scenic beauty of the CLFs in the study. Before evaluation, a UAV video of the overall view of the research area was shown to introduce the research background, and the students were asked to independently judge the scenic beauty of each viewpoint based on their own professional knowledge and travel experience. A 10-point system [49] was adopted for the indoor SBE of CLFs: the larger the score was, the higher the scenic beauty of the CLFs was.

2.2.3. Influencing Factors of SBE of Colored-Leaved Forests

Based on the literature review of relevant studies [50,51,52] and the landscape characteristics of a colored-leaved forest [53,54,55], the influencing factors of SBE reflected by the photos were decomposed into six landscape elements under two different perspectives: aerial and on the ground. These elements were assigned different values according to the element grading criteria. The six landscape elements observed on the ground are as follows: the average DBH, the dominant color of leaves, the trunk shape, the ratio of colored-leaved tree species, the canopy width, and the fallen leaf coverage. As for the aerial observation, the landscape elements were terrain change, forest coverage, landscape composition, landscape contrast, condition of human landscape, and recreation frequency. The decomposed landscape elements of different observation perspectives, grading criteria, and their corresponding score ranges [56,57,58,59] for the ground and aerial observation perspectives are shown in

Table 3. Landscape elements and grading criteria for the ground observation perspective.

Landscape Elements	Grading Criteria	Scores
$X_1$ Average DBH	≥20 cm 15~20 cm ≤15 cm	3 2 1
$X_2$ Dominant Color of Leaves	Brown, gray, green Yellow-green, yellow-brown Red, yellow	3 2 1
$X_3$ Trunk Shape	Straight General Bent	3 2 1
$X_4$ Ratio of Colored-Leaved Tree Species	≥80% 30%~80% ≤30%	3 2 1
$X_5$ Canopy Width	Canopy width/tree height ≥ 0.3 Canopy width/tree height 0.2~0.3 Canopy width/tree height ≤0.2	3 2 1
$X_6$ Fallen Leaf Coverage	≥80% 30%~80% ≤30%	3 2 1

and

Table 4. Landscape elements and grading criteria for the aerial observation perspective.

Landscape Elements	Grading Criteria	Scores
$X_7$ Terrain Change	No change Slight change Significant change	3 2 1
$X_8$ Forest Coverage	Forest coverage ≥ 0.8 Forest coverage 0.5~0.8 Forest coverage ≤ 0.5	3 2 1
$X_9$ Landscape Composition	Composed of trees, shrubs, and grasses; multi-layer, unevenly aged forest with dense clusters of colored-leaved trees Composed of trees and shrubs; multi-layer, unevenly aged forest with scattered distribution of colored-leaved trees Single-layer, evenly aged forest or non-forest land	3 2 1
$X_{10}$ Landscape Contrast	Distinct differences in forest color, obvious contrast of tree shades, palmate-shaped or triangular-shaped leaves, good visual effect Small differences in forest color, medium degree of shade contrast, elliptical-shaped or oval-shaped leaves, general visual effect No difference in forest color, no shade contrast, needle-shaped or lance-shaped leaves, poor visual effect	3 2 1
$X_{11}$ Condition of Human Landscape	National or provincial cultural relics protection units, no dead trees or bushes around, no weeds, no debris and rotten bricks, etc. General monuments or buildings, surroundings without dead trees and bushes, weeds, debris, or rotten bricks Buildings in disrepair, surroundings with dead bushes, weeds, debris, and rotten bricks	3 2 1
$X_{12}$ Recreation Frequency	Low recreation frequency, no runners, picnic campers, scenery viewers, etc. Medium recreation frequency, scattered distribution of runners, picnic campers, scenery viewers, etc. High recreation frequency, dense distribution of runners, picnic campers, scenery viewers, etc.	3 2 1

, respectively.

2.2.4. Data Processing and SBE Model Building

(1)

Data normalization

There are differences in the aesthetic scales between different individuals, and it is necessary to standardize the SBE scores in order to effectively reduce such differences [60]. The formula is as follows:

$$Z_{ij}=\left(R_{ij}-{\bar{R}}_j\right)/S_j$$

(1)

$${\bar{R}}_j=\frac{1}{n}{\displaystyle \sum }_{i=1}^n{R}_{ij}$$

(2)

$$S_j=\sqrt{\frac{1}{n-1}{\displaystyle \sum }_{i=1}^n{\left(R_{ij}-{\bar{R}}_j\right)}^2}$$

(3)

where $Z_{ij}$ refers to the standardized value of jth judge’s evaluation of the CLFs in the ith slide; ${\bar{R}}_j$ refers to the average value of the jth judge’s SBE of all 50 slides; $R_{ij}$ refers to the jth judge’s evaluation of the CLFs in the ith slide; and $S_j$ refers to the jth judge’s standardized value of all 50 slides.

(2)

SBE model building.

The standardized score of the SBE of the CLFs was taken as the dependent variable of the evaluation model, and the influencing factors of the SBE of the CLFs were divided into different levels and assigned different values. Then, a Pearson correlation analysis was conducted. The factors with weak correlation with the dependent variable were eliminated, so only the factors with strong correlation were retained as the independent variables. Finally, the SBE models of multiple linear regression were established using SPSS. In the process of model building, the following principles were applied: (1) prioritizing the elimination of factors with low correlation coefficients; (2) considering the multi-collinearity among independent variables; (3) prioritizing the retention of factors with better determinability and interpretability.

3. Results and Analysis

3.1. Relationship between Landscape Elements and Beauty Values from the Ground Perspective

One-way ANOVAs between the landscape elements and SBE values (beauty values) from the ground perspective showed that there were significant or highly significant differences between $X_1$ (p = 0.033), $X_2$ (p = 0.001), $X_4$ (p = 0.001), $X_5$ (p = 0.000), $X_6$ (p = 0.000), and the SBE values, while $X_3$ (p = 0.123) was not significantly different (Figure 6). Therefore, $X_3$ could be excluded in the subsequent analysis.

The relationship between the evaluation rank of the landscape elements and SBE values was that the average DBH evaluation rank was I, and the SBE values were the lowest at 4. As the $X_4$ evaluation rank increases, the dominant color of the leaves becomes less beautiful, and the SBE values become smaller. The degrees of beauty for the number of colored-leaved tree species, canopy width, and fallen leaf coverage were the lowest when the grade rank was I. All three landscape elements gradually increased with the increase in evaluation rank. The results of multiple comparisons showed that the differences in average DBH evaluation rank were not significant at rank II or III, and there was a tendency for the SBE values to decrease, indicating that the SBE values might decrease to some extent with the increase in average DBH level.

3.2. Relationship between Landscape Elements and Beauty Values from the Aerial Perspective

The results of the one-way ANOVA of each landscape element and the value of the SBE of CLFs from the aerial perspective showed that there were significant or highly significant differences between $X_7$(p = 0.006),$X_8$ (p = 0.036), $X_9$ (p = 0.002), $X_{10}$ (p = 0.000), $X_{11}$ (p = 0.000), and the SBE values, while $X_{12}$ (p = 0.053) was not significantly different (Figure 7). Therefore, $X_{12}$ could be excluded in the subsequent analysis.

The relationship between the level of landscape elements and the SBE values analysis was as follows: the SBE values were the largest when the level of terrain change was I, which was 7.2, and the SBE values gradually decreased with the increase in the evaluation rank. The degree of beauty of the forest coverage, landscape composition, landscape contrast, and condition of the human landscape were the lowest when the evaluation rank was I. All four landscape elements gradually increased with the increase in evaluation rank. Multiple further comparisons showed that the difference in the SBE values was not significant when the landscape’s combination rank was at II and III, which indicated that the SBE values of the two landscape elements did not affect the beauty values of CLFs when they reached a certain level.

3.3. SBE Model of CLF Building

3.3.1. Variable Selection of Landscape Elements for CLFs

Pearson correlation analysis was conducted for the selection of SBE-influencing variables. By eliminating some factors with a weak correlation with the SBE, only the factors with a strong correlation with the SBE were retained as the independent variables of the multiple linear regression model (

Table 5. Pearson correlation between landscape elements and SBE from the ground perspective.

Landscape Elements		${\mathit{X}}_1$ Average DBH	${\mathit{X}}_2$ Dominant Color of Leaves	${\mathit{X}}_3$ Trunk Shape	${\mathit{X}}_4$ Ratio of Colored-Leaved Tree Species	${\mathit{X}}_5$ Canopy Width	${\mathit{X}}_6$ Fallen Leaf Coverage
${\mathrm{SBE}}_1$(Ground Perspective)	Person Correlation	0.405 *	−0.681 *	0.233	0.771 **	0.702 **	0.721 **

* at the p < 0.05 level (two-tailed), significant correlation; ** at the p < 0.01 level (two-tailed), highly significant correlation.

and

Table 6. Pearson correlation between landscape elements and SBE from the aerial perspective.

Landscape Elements		${\mathit{X}}_7$ Terrain Change	${\mathit{X}}_8$ Forest Coverage	${\mathit{X}}_9$ Landscape Composition	${\mathit{X}}_{10}$ Landscape Contrast	${\mathit{X}}_{11}$ Condition of Human Landscape	${\mathit{X}}_{12}$ Recreation Frequency
${\mathrm{SBE}}_2$(Aerial Perspective)	Person Correlation	−0.475 *	0.468 *	0.534 **	0.724 **	0.775 **	0.319

* at the p < 0.05 level (two-tailed), significant correlation; ** at the p < 0.01 level (two-tailed), highly significant correlation.

).

As shown in Table 5, the correlation between each index and the beauty of the trees showed that $X_2$ (dominant color of leaves), $X_4$ (ratio of color-leaved tree species),$X_5$ (canopy width), and $X_6$ (fallen leaf coverage) were highly significantly correlated with SBE₁ (p < 0.01), and $X_1$ (average DBH) was significantly correlated with SBE₁ (p < 0.05). The $X_3$ (trunk shape) was not significantly correlated with SBE₁ (p > 0.05), so it was excluded from the SBE₁ model for the ground perspective of CLFs. The correlation coefficients showed that $X_1$, $X_3$, $X_4$, $X_5$, and $X_6$ were positively correlated with SBE₁, but $X_2$ was negatively correlated with SBE₁. From Table 6, $X_9$ (landscape composition), $X_{10}$ (landscape contrast), and $X_{11}$ (condition of the human landscape) were highly significantly correlated with SBE₂ (p < 0.01), and $X_7$ (terrain change) and $X_8$ (forest coverage) were significantly correlated with SBE₂ (p < 0.05). $X_{12}$ (recreation frequency) was not significantly correlated with SBE₂ (p > 0.05), so it was excluded from the SBE₂ model for the aerial perspective of CLFs. The correlation coefficients showed that $X_8$,$X_9$, $X_{10}$, $X_{11}$, and $X_{12}$ were positively correlated with SBE₂, but $X_7$ was negatively correlated with SBE₂.

3.3.2. SBE Model of CLFs

With $X_1$, $X_2$, $X_4$, $X_5$, and $X_6$ as the independent variables and the corresponding degree of beauty values as the dependent variables, the SBE₁ model of the ground perspective of CLFs was established; with $X_7$,$X_8$,$X_9$, $X_{10}$, and $X_{11}$ as the independent variables and the corresponding degree of beauty values as the dependent variables, the SBE₂ model of the aerial perspective of CLFs was established, as shown in

Table 7. Evaluation model of the degree of beauty of the CLF landscapes.

Evaluation Model	R2	F	p
$SB{E}_1=1.881+0.213X_1-0.336X_2+1.196X_4+0.33X_5+0.469X_6$	0.629	6.448	0.000
$SB{E}_2=0.29-0.146X_7+1.118X_8+0.162X_9+0.405X_{10}+1.364X_{11}$	0.740	10.811	0.000

. As can be seen from Table 7, the R² of the SBE₁ model for the ground perspective was equal to 0.629, indicating that the five framing elements $X_1$, $X_2$, $X_4$, $X_5$, and $X_6$ were able to explain 62.9% of the variability in the SBE₁ values of the plant landscape of CLFs from the ground perspective. Moreover, p < 0.01, and the F-value was equal to 6.448, indicating that the association between these five landscape factors and the scenic beauty was very obvious and the established regression model was valid. From the ground perspective, according to the absolute value of the regression coefficient of the SBE₁ model, the degree of influence of each landscape element on the SBE₁ is ranked as follows: ratio of colored-leaved tree species > fallen leaf coverage > dominant color of the leaves > canopy width > average DBH. The R² of the SBE₂ model for the aerial perspective was equal to 0.74, indicating that the five landscape elements $X_7$, $X_8$,$X_9$, $X_{10}$, and $X_{11}$ could explain 74% of the variability of the SBE₂ values of the plant landscape in CLFs from the aerial perspective. Moreover, p < 0.01, and the F value was equal to 10.811, indicating that these five landscape factors were significantly associated with the degree of beauty and the regression model was established. From the aerial perspective, according to the absolute value of the regression coefficients of the regression model, the degree of influence of each landscape element on the SBE₂ is ranked as follows: condition of the human landscape > forest coverage > landscape contrast > terrain change > landscape composition.

4. Discussion and Limitations

4.1. Discussion

This study on the relationship between the degree of beauty of CLFs and landscape elements is helpful to understand and grasp the influencing factors of forest tourism and provide scientific reference for the management and planning of CLF landscapes. The main contributions are as follows: (1) providing technical paths for forest tourism behavior measurement based on big data and deep learning methods and constructing a big data-based landscape preference method for tourists, which provided technical support for the selection of the best viewpoints in CLFs and helped the application of forest tourism planning from the tourists’ perspective; (2) introducing two perspectives, aerial and ground, to analyze the influencing factors of the SBE in CLFs and providing data supporting the study of the spatial optimization of CLF landscapes.

By examining the relationship between the individual elements of a CLF landscape and the SBE from both aerial and ground perspectives and comparing them with previous studies [61,62], the following more consistent relationships can be drawn: (1) The leaves of most tree species are green in color, with few yellow and red trees. Compared with brown and green leaves, yellow and red leaves are more unique, more beautiful, and more dazzling in color. (2) A big colored-leaved tree with a large DBH is more visually attractive than a colored-leaved tree with a small DBH [63]. (3) The ratio of colored-leaved species has a significant impact on the beauty of a forest landscape. When the ratio of color-leaved species is up to 80% or more, the visual impact formed by the clustered color-leaved tree species is more attractive to tourists. (4) A landscape with a large forest canopy has a higher aesthetic value than a landscape with a small forest canopy. Being in a CLF landscape with a large canopy gives tourists a more tranquil feeling, while a CLF with a small canopy only gives tourists a dull and monotonous feeling. (5) When the coverage of fallen leaves is equal to or greater than 80%, a secluded space is formed within the CLFs, which improves the scenic beauty of the CLFs to a certain extent. (6) The scenic beauty of a CLF landscape with significant terrain changes is greater than that of a forest landscape with no changes [64]. (7) In terms of forest coverage, the higher the forest coverage is, the higher the scenic beauty value is. When the forest coverage is greater than 80%, the SBE value of the CLF is the highest. (8) The differences in the SBE value caused by different compositions are more obvious. For landscapes with a higher SBE value, the landscape is composed of unevenly aged trees, there are dense clusters of colored-leaved species, and the community structure is complex [65]. (9) Landscape contrast has a great influence on the scenic beauty of the landscape. Contrast in a forest landscape includes not only the framing elements but also the contrast of color and shape among the framing elements. For example, palm-shaped leaves are more eye-catching than oval leaves [66]. (10) The condition of the human landscape has a significant impact on the scenic beauty of the landscape from the aerial perspective. An inferior human landscape with dead wood and rotten bricks destroys the overall visual beauty and reduces the beauty of the landscape.

In China, with the rapid development of the national economy and the heating up of the domestic tourism industry, hundreds of millions of people go out for sightseeing and recreation. However, when tourists go out to travel, they often go to an unfamiliar environment where it is difficult to quickly find a viewpoint that can meet their own aesthetic needs within a short time. The optimal viewpoints selection method developed in this paper can deal with this problem more successfully and, therefore, can be applied to other national forest parks and scenic areas in China and abroad.

Since the end of 2019, a new form of online tourism called “cloud tourism” has emerged, which is the active response of the tourism industry to the new situation of COVID-19. In the context of the normalization of the dynamic zero-COVID-19 policy, the number of scenic areas adopting “cloud tourism” has increased rapidly. The demand from subjects who like cloud tourism is showing a trend of diversification. Although the younger generation is still the mainstay, people of other age groups are also gradually participating in online sightseeing. At the same time, the supply subjects of “cloud tourism” are also becoming diversified. Local governments, tourism companies and platforms, online social platforms, and self-media guides have played important roles in promoting online travel programs. Different supply entities have begun to cooperate with each other to enhance the marketing effect of online tourism projects. For example, the Yunnan Provincial Cultural and Tourism Department used the official tourism platform APP “Travel Yunnan Online” to “move” more than 900 scenic areas online with local cultural and tourism departments and enterprises. The Dunhuang Research Institute, Gansu, China, and Tencent jointly developed a cell phone application named “Travel Dunhuang Online”. Therefore, our research results can provide a new approach to the SBE of cloud-tourism-oriented CLFs.

4.2. Limitations of the Study

It should be pointed out that only the landscape elements of tree morphology and biological characteristics were considered in this study, while other elements such as animals, sunlight, microorganisms, and humidity were not studied. Studies in China and abroad have shown that these factors mentioned above also have a certain influence on the scenic beauty of CLFs [67,68]. For example, light is the main reason for the scenic beauty of color; the clarity of a landscape under the front-facing light is higher, but the shadow effect generated under backlit conditions can also improve the visual effect of a CLF landscape to a certain extent. In addition, this study decomposed landscape elements such as the trunk shape, landscape combination, landscape contrast, and human landscape when evaluating the landscape quality of a colored-leaved forest; these elements were subjective in nature, and this subjectivity can affect the final evaluation results [69,70]. Secondly, the study lacked information on color indices and lacked a more detailed reference on color management and the management of colored-leaved forests. In addition, the indices obtained through the principles and methods of forestry were more objective [71]. Therefore, other non-landscape-influencing factors and color index information should be effectively introduced in the subsequent research on the evaluation method of a CLF landscape’s beauty, which can provide a scientific reference for the development of more detailed management measures for autumn CLFs.

5. Conclusions

In this paper, based on the big data landscape-preference mining technology of tourists and the GIS spatial overlay analysis function, the best viewing spots of CLFs for tourists were screened out. The theories of botany, ecology, and aesthetics were combined together to evaluate the SBE of autumn CLFs at 50 viewpoints in Qixia Mountain National Forest Park in 2021 from the aerial and ground perspectives. Our research results showed that the SBE of CLFs from different perspectives was affected by different landscape elements. The major elements affecting the beauty of the ground perspective were the average DBH, the dominant color of the leaves, the ratio of colored-leaved tree species, the canopy width, and the fallen leaf coverage. It should be noted that the tree trunk shape had little influence on the scenic beauty of the SBE from the ground perspective and was not considered to be the dominant affecting factor in the SBE from the ground perspective. In Qixia Mountain Forest Park, most of the colored-leaved trees, which include Acer palmatum, Prunus cerasifera, Cotinus coggygria, Zelkova serrata, Liquidambar formosana, and Sapium sebiferum, belong to semi-natural forests. The management measures for CLFs are mainly the sanitary cutting of dead trees, thinning, and light-releasing felling, with a little artificial pruning or twisting of tree trunks. Therefore, there is no significant difference in the shapes of tree trunks in many CLFs. The major elements affecting the SBE of the aerial perspective of CLFs in the park were terrain changes, forest coverage, landscape composition, landscape contrast, and the condition of the human landscape. The reason why recreational behavior was not considered to be the dominant factor of the SBE model may be that people’s recreational activities in the forest look very small from the aerial view of a UAV, so they can be neglected. In general, the dominant factors that affect the scenic beauty of CLFs from different viewing perspectives are different. Therefore, our research results put forward the following suggestions for the creation of management planning for autumn CLFs:

(1)

Regional micro-topography method: When creating landscape planning for flat terrain, an undulating micro-topography should be created. In areas with large terrain changes, planting belts, terrains, steps, and cultural walls on the slope should be established to create a multi-level changing landscape space.

(2)

Dominant color substitution method within a small woodland: reasonably control the forest density by thinning, replant red- and yellow-leaved trees on tourists’ preferred viewing routes, increase the coverage of densely distributed and patchy colored-leaved forest in a small woodland, plant colored-leaved trees on bare land, and promote the succession of forests to mature and over-mature forest stands.

(3)

Landscape retouching method: Through the combination of trees, shrubs, and grass and by adding water bodies, the forest landscape can be enriched. The forest landscape should be kept clean through the selective cutting of cluttered trees. The scenic beauty could be further improved by updating inferior landscape buildings and planting colored-leaved trees with forest cultural value.

(4)

Mutual complementary method. Since obvious contrasting colors and appreciable leaf shapes can improve the beauty of CLFs, trees with the dominant color of the landscape as their complementary color should be planted, and the ratio of trees with irregular leaf shapes should be increased.