Impact of Forest Parkiﬁcation on Color Authenticity

: Preserving the authenticity of forest colors is essential to highlight regional characteristics and promote the sustainable development of forest landscapes. However, the factors and mechanisms inﬂuencing forest color remain unclear. We quantiﬁed 1422 forest color images from 43 parks across seven biogeographic regions in China to capture the forest color composition among regions. A generalized linear mixed-effects model was used to analyze the effects of meteorological and anthropogenic disturbance factors on forest color. Meteorological factors included accumulated sunshine hours, average temperature, accumulated precipitation, frost-free period, average wind speed, and average air quality index. Anthropogenic disturbance factors included park feature indicators (area, elevation, and perimeter-area ratio) and human activity indicators (distance to urban areas, building density, and road density). We calculated p -values and relative effect estimates to determine the sensitivity and degree of sensitivity of color to each factor. The results indicated the following: (1) forest color composition varied signiﬁcantly among different regions in China with variations observed particularly in the proportions of primary (green), secondary (yellow and yellow-green), and accent colors (orange and blue-green); (2) forest colors were sensitive to all meteorological factors; (3) orange, yellow, purple, and red were all sensitive to anthropogenic disturbance factors; and (4) forest accent colors were more strongly inﬂuenced by anthropogenic disturbance factors, particularly park features. To protect the authenticity of forest colors, it is necessary to avoid excessive borrowing of forest color schemes from different regions, control park features, reduce building area within the park buffer zones, and optimize park tourism routes.


Introduction
Forest color authenticity refers to the extent to which various colors in the forest landscape remain in their native state despite natural and human disturbances [1]. Protecting forest color landscape authenticity is essential to show regional characteristics and achieve sustainable development of forest color landscapes. It can provide better long-term and stable landscape services for people [2]. National forest parks were established to protect the authenticity and integrity of representative large-scale forest ecosystems [3]. However, with the increasing ecological and cultural needs of people, national forest parks integrate scientific research, education, and leisure functions, which inevitably involve diversified human activities, such as monitoring, protection, and recreational experiences, leading to increased anthropogenic disturbance in the forest landscape [4]. Increasing anthropogenic disturbance can weaken or even eliminate the authenticity of forest colors [5]. Therefore, factors that influence forest colors should be identified to better protect and restore authentic forest color landscapes.
Forest colors originate from changes in the growth and development of plant organs and adaptive changes resulting from environmental variability [6]. Climate is considered the primary environmental factor affecting forest color changes [7], including light, temperature, humidity, precipitation, and air quality [8]. Plant coloration depends on the relative pigment content of the plant body [9]. Previous studies have mainly explored meteorological effects on plant pigment content from the perspective of plant organs or individual levels. Generally, longer sunshine hours favor chlorophyll synthesis [10], resulting in a green appearance. When plants are subjected to stressors, such as drought and low temperatures, the chlorophyll in their bodies decomposes easily [11], while carotenoids and anthocyanins remain relatively stable [12], leading to an orange or yellow appearance. In addition, some studies have examined the impact of meteorological factors on the quantity and duration of forest colors at the landscape scale during seasons when color changes are particularly pronounced (spring and autumn) [13,14]. Furthermore, forest color landscapes are affected by a combination of temperature, photoperiod, and precipitation [15,16]. However, the impact of each meteorological factor on forest colors varies according to the season. For example, while higher temperatures favor the emergence of new leaves in spring, resulting in an increased proportion of red in the forest, they negatively affect leaf aging and coloring in autumn, leading to a reduced proportion of red [14]. Therefore, to predict the composition of forest landscape colors under changing climatic conditions, the impact of meteorological factors on the annual forest color index requires further exploration.
With the acceleration of urbanization and development of forest tourism, human activities are increasingly affecting forest color landscapes [17,18]. The expansion of human activity has reduced forest areas and increased the fragmentation of forest color landscapes [19]. Additionally, human activities have indirectly affected forest color landscapes by altering the regional climate [20] and hydrological cycles [21]. Most studies suggest that human activities have caused the start of the forest growing season to advance and become delayed by the end of the growing season [22][23][24], resulting in a decreased proportion of non-green colors in forests. Considering that they are major tourist areas, forest parks may experience excessive disturbance of forest ecosystems and degradation of tourism resources due to human activities [25]. Previous studies showed that the larger the park area and the lower the elevation, the more concentrated the visitor activity paths within the park [26,27], thus increasing the degree of interference with local forest color landscapes. The complexity of the park shape [28] and degree of land-use changes [29][30][31] lead to more severe fragmentation of forest color landscapes. The more fragmented the forest color landscape, the more threatened is the forest color authenticity [30]. In summary, human activities influence the authenticity of forest colors; however, the mechanism through which anthropogenic disturbance factors affect each forest color remains unclear.
Assuming that climate change and anthropogenic disturbances threaten the authenticity of forest color landscapes after forest park development, this study aimed to explore the key factors affecting the authenticity of forest colors, thereby offering guidance for protecting and restoring forest color landscape authenticity. Recent developments in big data technology have made it possible to conduct nationwide research on forest colors in China. We selected 43 national forest parks in seven biogeographical regions as research objects and obtained forest images of each park from Weibo, a popular social media platform in China [32]. We then quantified the color information of each image [33] to determine the characteristics of forest color composition in various regions after forest park establishment. Subsequently, we collected meteorological information (temperature, precipitation, etc.) and anthropogenic disturbance factors (park features, changes in land use inside and outside the park, etc.) for each park to investigate the impact of these factors on forest color. The main issues addressed in this study were as follows: (1) what is the forest color composition in different regions, and which colors show significant differences  (3) what is the degree of sensitivity of each forest color to meteorological and anthropogenic disturbance factors?
We hypothesized that: (1) forest color composition changes with parkification; (2) temperature, sunshine hours, and precipitation are critical meteorological factors affecting forest color, whereas lower temperatures, shorter sunshine hours, and less precipitation will decrease green percentages and increase orange and yellow percentages in the forest; (3) both park features and human activities impact forest color authenticity; and (4) different forest colors exhibit varying degrees of sensitivity to meteorological and anthropogenic disturbance factors. Validating these hypotheses could help assess the impacts of climate change and human activities on the forest color landscape composition. It can guide forest landscape planning, design, and sustainable management of parks and other urban green spaces.

Study Area
China is a vast country with diverse climates and vegetation types [34], leading to different forest color compositions among regions. To determine the influence of forest park establishment on color authenticity in each region, we divided China into seven biogeographic regions [35] (henceforth, regions) based on climate and vegetation distribution: northeast China, north China, northwest China, Mongolian Plateau, Tibetan Plateau, eastern Himalayas, and southeast China [36,37]. Using the inventory of national forest parks (henceforth, parks) in China (https://www.maigoo.com/goomai/167110.html (accessed on 17 February 2023)), we randomly selected one to two parks from each provincial administrative region for a total of 43 parks from diverse regions ( Figure 1).

Forest Color Classification and Analysis
As the primary basis for distinguishing different colors [43], hue (H) is relatively less affected by external factors, such as light and viewing distance, than saturation and value

Study Object
Forest image acquisition. We used a focused web crawler [38] to collect images between 1 December 2018 and 30 November 2019 with Weibo as the data source and the name of each park as the keyword ( Figure A1). A total of 187,149 park images were obtained.
Forest image screening. First, we removed images dominated by non-forest elements such as human portraits, streams, buildings, food, and weather. Then, we manually screened the park images based on the principle that (1) the forest part comprised at least 60% of the total image area and (2) the color situation matched the physical object [33]. Finally, 1422 high-quality images of forest landscapes from various regions were obtained (Table A1). There were 52 images from northeast China, 224 from north China, 122 from northwest China, 551 from southeast China, 118 from the eastern Himalayas, 268 from the Mongolian Plateau, and 87 from the Tibetan Plateau. There are significant seasonal differences in forest color landscapes [39]. To avoid differences between images and seasonal meteorological data, we divided the seasons according to the upload time and content of each image (Table A2).
Forest image processing. We uniformly applied adaptive gamma correction [40] and automatic color equalization [41] to the forest color images, which reduced the problem of inconsistent image quality caused by the image equipment and environment. To eliminate the impact of non-forest elements on the color extraction results, we used Photoshop to remove color components unrelated to the forest, such as buildings, sky, and water bodies [42].

Forest Color Classification and Analysis
As the primary basis for distinguishing different colors [43], hue (H) is relatively less affected by external factors, such as light and viewing distance, than saturation and value [44]. Therefore, in this study, color classification was performed based on a range of hue values. We used a secondary K-means clustering approach for color classification [33] with the following steps: (1) determine the optimal k-value using the elbow method [45] to cluster the color information of each forest image, and (2) set the k-value to eight to cluster the H-value of the clustered colors of each image [46]. The secondary clustering results were then adjusted according to human eye color sensitivity [42,47]. Finally, we quantified the forest colors as orange [0, 25] or (345, 360], yellow (25,55], yellow-green (55,75], green (75,140], blue-green (140, 165], blue (165, 220], purple (220, 290], and red (290, 345] (Figure 2), and performed the color index calculation: where A H i is the total number of pixels occupied by hue i, and A n is the total number of pixels in the plant part of that image.   Meteorological conditions are the main factors causing changes in plant color [7]. We obtained meteorological data for each park from 1 December 2018 to 30 November 2019 from the Chinese historical weather website (Table A3). From the weather website (https://rp5.ru (accessed on 25 February 2023)), we collected meteorological data for the closest meteorological station (Table A4) for each park, including daily average temperature (T), daily average wind speed (F), and daily precipitation (R). The China Meteorological Information Service Center (https://data.cma.cn/ (accessed on 25 February 2023)) and the air quality historical data website (http://www.aqistudy.cn/historydata/ (accessed on 25 February 2023)) provided the monthly cumulative sunshine hours (SH) and average air quality index (AQI) for each park's prefecture-level city (Table A4). We calculated the average or cumulative values of the meteorological variables for each season (Table A2). The frost-free period (FFP) was obtained from the official website of each park (accessed on 10 March 2023).

Anthropogenic Disturbance Factors
Forest parkification refers to the process of transforming a natural forest area into a designated forest park through anthropogenic disturbances. This involves modifying the landscape to create recreational spaces while aiming to preserve the natural resources within the park's boundaries. The primary objectives of forest parkification include providing areas for leisure activities, cultural events, scientific research, and educational opportunities within a controlled and managed environment [3]. Anthropogenic disturbances affect the natural environment and ecosystems owing to human products, presence, and other social activities [48,49]. To measure the extent to which parkification impacts forest color landscapes, we selected six anthropogenic disturbance variables from park features and human activities.
Park features, including park size, shape, and location, significantly affect vegetation diversity [50]. Therefore, we considered that park features also affected the authenticity of forest color and selected the following three variables (Tables 1 and A3). (1) The area of the park (AREA), which reflects the horizontal extent of human activities, was obtained from the official website of each park (accessed on 10 March 2023). (2) The perimeter area ratio of the park (PAR) describes the complexity of the park shape; the more complex the park shape, the stronger the edge effect [28]. This was calculated using park perimeter/park area × 100. The boundaries of each park were obtained from the Baidu map (https://map.baidu.com (accessed on 25 March 2023)). (3) Elevation of the park (ELEV), which conveys the vertical extent of human activity, was obtained from each park's official website (accessed on 10 March 2023).
Human activities refer to the development, utilization, and protection of the natural environment during human social development [51]. Buildings, roads, and towns, as products of human social development, have increased human influence on forest color landscapes [52,53]. Three human activity variables were selected for this study (Table 1).
(1) The distance to the nearest city center (D) reflects the impact of urbanization on forest color [54]. We identified the city center nearest to the park using ArcGIS 10.6 and obtained the minimum linear distance to the park. (2) Building density (BD), which reflects the intensity of human modification of nature [53], was calculated as the building area/3 km buffer area of the park × 100. We obtained building data from the RiverMap 4.1 software.
(3) The road network density (RD) reflects park [55]. The higher the road network density, the more significant the impact of human activities on the forest color landscape [52]. It was calculated as the road length/3 km buffer area of the park × 100. We obtained road network data from the RiverMap 4.1 software. Average air quality index for the corresponding season of the forest image Frost-free period (FFP, day) Number of days between final frost day and first frost day in a year [56] Park feature Area of the park (AREA, ha) The maximum horizontal extent of human activity in the park The perimeter area ratio of the park (PAR, km/km 2 ) The perimeter of the park divided by its area Elevation of the park (ELEV, m) The maximum vertical extent of human activity in the park

Human activity
Distance to the nearest city center (D, km) The minimum linear distance between the park and the nearest prefecture-level city center Building density (BD, %) Building area within the 3 km buffer of the park divided by its buffer area Road network density (RD, km/km 2 ) The total length of roads within the 3 km buffer of the park divided by its buffer area

Data Analysis
All statistical analyses and mapping were performed using R, version 4.1.3 (R Core Team, Vienna, Austria).
The forest color index is the ratio of the data converted from count data. The Kruskal-Wallis test was used to analyze the differences in forest color indices among regions with p < 0.05 considered statistically significant. In addition, we divided forest color into three categories: primary, secondary, and accent colors (color indices of 30%-100%, 10%-30%, and 0%-10%, respectively) [57]. In this study, we used the color composition of unparked forests in the same biogeographic region as baseline data for color authenticity. The color composition of unparked forests was obtained from previous studies.
A generalized linear mixed model with a binomial error structure was used to investigate the influence of environmental factors on forest color indices. We fitted the model to each of the seven color indices (except the green index) using meteorological and human activity indicators as fixed effects and parks as random effects. The model was constructed as follows: where Hi is the color index of hue i, park is the code of each park, and A H i is the total number of pixels occupied by hue i. We normalized the z-scores of all explanatory variables for fixed effects to compare the model parameter estimates. Collinearity among variables was tested using the variance inflation factor (VIF), ensuring that all explanatory variables satisfied the criterion of VIF < 10 [58] (Table A5). The heterogeneity of the residuals was tested using the Durbin-Watson test [59]. R 2 m (marginal R 2 , accounting for fixed effects) and R 2 c (conditional R 2 , accounting for full model effects) were calculated to estimate the goodness of fit [60]. The significance of each explanatory variable was assessed using a χ 2 test [61], taking p < 0.05 as statistically significant. The p-value was used to determine whether each color was sensitive or insensitive to the variable. The sensitivity or insensitivity of the forest color was employed to determine whether its authenticity was affected.
To evaluate the relative importance of the predictors as drivers of forest color indices, we calculated the effect of the parameter estimates for each explanatory variable relative to all parameter estimates in the model [62]. The following three identifiable variance fractions were examined: (1) meteorological variables, (2) park feature variables, and (3) human activity variables. Relative effect estimates were used to determine the sensitivity of each forest color to meteorological and anthropogenic disturbance factors (park features and human activities). The sensitivity of the forest color was used to measure the degree to which its authenticity was affected.

Forest Color Composition by Region after the Establishment of Forest Parks
From a national perspective, green (36.30%) was the primary color of forest landscapes, yellow (26.94%) and yellow-green (19.04%) were secondary colors, and the remaining colors were accent colors ( Figure 3). Orange (H1), yellow (H2), yellow-green (H3), green (H4), and blue-green (H5) indices differed significantly among regions (p < 0.01). Orange was a secondary color in the forest color landscape in northeast China (20.86%) and north China (11.04%) and an accent color in other regions. Yellow was the primary color in the forest landscape of the Mongolian Plateau (45.96%) and northeast China (38.10%), and a secondary color in other regions. Yellow-green was the primary color in the forest landscape of the Tibetan Plateau (34.89%), an accent color in northeast China (8.66%), and a secondary color in other regions. Green was a secondary color in the forest landscape of the Mongolian Plateau (25.02%) and northeast China (20.90%), and a primary color in other regions. Blue-green was a secondary color in the forest landscape of north China (10.20%), and an accent color in other regions. The blue (H6), purple (H7), and red (H8) indices were not significantly different among regions (p > 0.05), and all showed accent colors in the forest landscape (Table 2). significance of each explanatory variable was assessed using a χ 2 test [61], taking p < 0.05 as statistically significant. The p-value was used to determine whether each color was sensitive or insensitive to the variable. The sensitivity or insensitivity of the forest color was employed to determine whether its authenticity was affected.
To evaluate the relative importance of the predictors as drivers of forest color indices, we calculated the effect of the parameter estimates for each explanatory variable relative to all parameter estimates in the model [62]. The following three identifiable variance fractions were examined: (1) meteorological variables, (2) park feature variables, and (3) human activity variables. Relative effect estimates were used to determine the sensitivity of each forest color to meteorological and anthropogenic disturbance factors (park features and human activities). The sensitivity of the forest color was used to measure the degree to which its authenticity was affected.

Forest Color Composition by Region after the Establishment of Forest Parks
From a national perspective, green (36.30%) was the primary color of forest landscapes, yellow (26.94%) and yellow-green (19.04%) were secondary colors, and the remaining colors were accent colors (

Effect of Meteorological Factors on the Forest Color Authenticity
The sensitivity of different colors to meteorological factors varied ( Figure 4, Table A6). T, R, F, and FFP had significant negative effects (p < 0.05) on H1, indicating that orange was sensitive to these four meteorological factors. The proportion of orange in the forest decreased with higher temperatures, more precipitation, higher wind speed, and a longer frost-free period. However, SH and AQI had no significant effects on H1 (p > 0.05) ( Figure 4A). SH, T, R, F, and AQI had significant effects on H2 and H3 (p < 0.01), indicating that these indices were sensitive to these five meteorological factors. The proportion of yellow in the forest increased with longer sunshine hours, lower temperatures, less precipitation, lower wind speeds, and better air quality ( Figure 4B). The proportion of yellow-green in the forest increased with shorter sunshine hours, higher temperatures, lower precipitation, higher wind speeds, and better air quality ( Figure 4C). FFP had no significant effects on H2 and H3 (p > 0.05). SH, R, and AQI had significant effects on H5 (p < 0.01), indicating that blue-green was sensitive to these three meteorological factors. The proportion of blue-green in the forest increased with shorter sunshine hours, increased precipitation, and poorer air quality. Nevertheless, T, F, and FFP had no significant effects on H5 (p > 0.05) ( Figure 4D). SH, T, F, and FFP had significant effects (p < 0.05) on H6, indicating that blue was sensitive to these four meteorological factors. The proportion of blue in the forest increased with longer sunshine hours, lower temperatures, higher wind speeds, and longer frost-free periods. R and AQI had no significant effect on H6 (p > 0.05) ( Figure 4E). SH, R, F, AQI, and FFP had significant effects on H7 (p < 0.05), indicating that purple was sensitive to these five meteorological factors. The proportion of purple in the forest increased with shorter sunshine hours, less precipitation, higher wind speed, poorer air quality, and shorter frost-free periods. However, T had no significant effect on H7 (p > 0.05) ( Figure 4F). All meteorological factors had significant effects (p < 0.01) on H8, indicating that red was sensitive to SH, T, R, F, AQI, and FFP. The proportion of red in the forest increased with shorter sunshine hours, higher temperatures, less precipitation, higher wind speeds, poorer air quality, and shorter frost-free periods ( Figure 4G).  Figure 4F). AREA, ELEV, PAR, BD, and RD had significant effects on H8 (p < 0.01), indicating that red was sensitive to these five artificial disturbance factors. The proportion of red in the forest increased with larger area, lower elevation, more complex shape of the park, less building density, and more road network density within the park buffer. D had no significant effect on H8 (p > 0.05) ( Figure 4G).

Effect of Anthropogenic Disturbances on Forest Color Authenticity
Anthropogenic disturbance factors from forest parkification only showed significant effects (p < 0.05) on the H1, H2, H7, and H8 ( Figure 4, Table A6). AREA, ELEV, and BD had significant effects (p < 0.05) on H1, indicating that orange was sensitive to these three anthropogenic disturbance factors. The proportion of orange in the forest increased with larger park area, lower park elevation, and higher building density. However, PAR, RD, and D had significant effects (p > 0.05) on H1 ( Figure 4A). ELEV had a significant negative effect on H2 (p < 0.05), indicating that yellow was sensitive to the elevation of the park. The proportion of yellow in the forest decreased with an increase in park elevation. AREA, PAR, BD, RD, and D had no significant effects (p > 0.05) on H2 ( Figure 4B). Park feature factors (AREA, ELEV, and PAR) had significant effects (p < 0.05) on H7, indicating that purple was sensitive to park feature factors. The proportion of purple in the forest increased with a larger area, lower elevation, and simpler shape of the park. Human activity factors (BD, RD, and D) had no significant effects (p > 0.05) on H7 ( Figure 4F). AREA, ELEV, PAR, BD, and RD had significant effects on H8 (p < 0.01), indicating that red was sensitive to these five artificial disturbance factors. The proportion of red in the forest increased with larger area, lower elevation, more complex shape of the park, less building density, and more road network density within the park buffer. D had no significant effect on H8 (p > 0.05) ( Figure 4G).

Differences in Forest Color Authenticity among Regions
Different regions have developed different forest color landscapes under the combined influence of the natural environment and vegetation types [63]. Previous studies have indicated that green is the primary forest color in all regions, owing to the limitations of plant pigments [64]. By contrast, this study found that the primary forest color in the Mongolian Plateau and northeastern China was yellow, whereas green was only a secondary color. This can be attributed to the establishment of forest parks that have increased human influence on the forest color landscape [25], resulting in a change in the color composition of these two regions. In addition, the vegetation of the Mongolian Plateau and northeastern China is dominated by temperate grasslands and temperate mixed coniferous forests with a cold, dry climate and long autumn and winter, which make plants prone to yellow, resulting in a greater proportion of yellow than green in the forests [65,66]. Consistent with the findings of previous studies, the predominance of green was particularly prominent in southeastern China and the eastern Himalayas, probably because subtropical evergreen broad-leaved forests dominate the vegetation in these two regions [67] where plants exhibit green all year round. Yellow-green was expressed as the primary color in the Tibetan Plateau, which may be because alpine grasslands, and scrub are the dominant vegetation types in this region with a large proportion of herbaceous plants [68] comprising a large proportion of yellow-green in the forest. Orange and bluegreen had large proportions in northern China, and both appeared as secondary colors, which may be because temperate deciduous broad-leaved mixed forests dominate the vegetation in this region and seasonal changes in forest color are apparent [69], resulting in diverse secondary colors. Blue, purple, and red were not significantly different among regions and were all accent colors, probably because these three colors mainly appear in flowers in the forest [36], and flower colors account for a minor proportion of the forest color landscape.
In summary, forest parkification may change the forest color composition. The differences in primary and secondary forest colors among different regions mainly depend on the vegetation type. Vegetation types result from joint selection by humans and nature in a region [70]. In forest color landscape creation, it is necessary to avoid excessive borrowing among different regions and increase the proportion of native tree species, which will protect the forest color authenticity in each region to highlight regional characteristics.

Impact of Climate Change on Forest Color Authenticity
Previous studies have reported that forest color landscapes are influenced by temperature, precipitation, and sunshine hours [15,16]. Higher temperatures lead to increased yellow-green and red indices in forest areas, possibly because of the promotion of new leaf growth (yellow-green and red) and flowering (red) [14]. Increased precipitation provides ample water for plant growth and promotes flourishing vegetation [71], leading to an increase in blue-green indices in forested areas. Previous studies have indicated that longer sunshine hours promote chlorophyll formation in plants, resulting in a greener appearance [10]. However, this study found that shorter sunshine hours result in higher yellow-green and blue-green indices in forests. This can be attributed to various factors such as latitude, weather, and elevation [72], with areas experiencing shorter sunshine hours tending toward evergreen tree species, such as those found in southeastern China. In addition, the results of this study indicate that changes in forest color are associated with frost-free periods, wind speed, and air quality. The longer the frost-free period, the longer the plant growth period, and the more likely they are to appear green [73], resulting in decreased orange, purple, and red indices. Increased wind speeds within the tolerance range of plants promote flower unfolding [74], resulting in an increase in purple and red indices in forests. However, increased wind speeds can also cause the color-changing leaves of deciduous trees to detach from the plant body [75], reducing the orange and yellow indices. As air quality worsens, blue-green, purple, and red indices in forests increase. This is likely owing to increased haze, which causes more mixed colors in the forest [76].
Climate change poses a significant threat to the authenticity and protection of forest color landscapes. Increasing human activity and emissions of greenhouse gases and air pollutants have resulted in problems such as local temperature increases and poorer air quality in forest parks [20,77], causing decreased orange and yellow indices. First, it is necessary to adopt eco-friendly travel methods and strictly adhere to tourism regulations when visiting forest parks. To address the issue of decreased orange and yellow indices caused by human-induced climate change, appropriate deciduous native tree species and flowering plants with orange and yellow coloring should be planted in forest parks, thereby increasing the likelihood and proportion of orange and yellow.

Impact of Anthropogenic Disturbances on Forest Color Authenticity
This study found that park area, elevation, perimeter area ratio, building density, and road network density all significantly affect forest color indices. Park elevation and area often have opposite effects on forest color indices, which may be related to varying levels of human activity. Larger parks and lower elevations tend to have higher levels of human activity [26,27]. Deciduous trees with colorful leaves are highly appreciated for their long display periods and contribution to park landscapes [78]. Accordingly, areas with high human activity tend to have a larger proportion of colorful deciduous trees, leading to an increase in color indices (orange, yellow, purple, and red). The perimeter area ratio of parks reflects the extent of interaction between human societies and forest environments [79] and has a significant impact on the purple and red color indices found in forests. Furthermore, increasing building density leads to an increase in the orange index but a decrease in the red index. This could be because plants in regions of high-building density are more susceptible to environmental stressors, such as soil pollution and water scarcity [80], resulting in a shortened growing season [81] and an increased proportion of orange in the forest. In forests, red is mostly present in the form of flowers with light being the primary regulatory factor of plant flowering [82]. An increase in road network density results in an increase in streetlights, which provide favorable conditions for flower blooming. Accordingly, road network density has a significant positive influence on the red index.
In summary, balancing the scope, manner, and intensity of human activities is key to preserving the authenticity of forest color [4]. The size of park visiting areas should be controlled to maintain the natural properties of forest landscapes. Higher elevations have relatively fragile plant habitats and are more vulnerable to human activities [83]. Accordingly, close-range human visits to these areas should be minimized. In highly scenic forest landscapes with rich colors, artificial constructions should be limited; moreover, it is necessary to follow the principle of "controlling and reducing community and resident numbers as much as possible" [84]. Optimizing human travel paths within parks [85] and minimizing road network density can significantly reduce the impact of human activities on the authenticity of forest colors.

Different Degrees of Sensitivity to Forest Color Authenticity
As we hypothesized, a combination of climate change and anthropogenic disturbances influences the authenticity of forest color. Orange, yellow, purple, and red were more sensitive to the external environment than yellow-green and blue-green. This may be because orange, yellow, purple, and red are associated with plant phenological changes (ecological responses by plants to changes in the external environment, such as flowering and leaf discoloration) [14]. Simultaneously, yellow-green and blue-green depend more on other factors, such as plant growth status. Furthermore, the results of this study rejected the conclusion that meteorological factors are the primary regulators of forest color [7]. We found that meteorological factors were the key factors affecting secondary forest colors (yellow and yellow-green), and human disturbances were the key factors affecting forest accent colors (orange, blue-green, blue, purple, and red). Among the anthropogenic disturbance factors, park features had the most significant influence on forest accent color. This may be because forest accent colors mainly appear in flowers and are less sensitive to meteorological factors [36] but are susceptible to human activities and anthropogenic selection. Although accent colors account for a minor proportion of the forest color landscape, they often serve as color harmonizers, making the landscape more colorful and layered [57]. Therefore, it is necessary to stringently control the influence of human interference factors on forest color. Before a park is established, we must fully grasp the scope and intensity of local human activities to determine the park's scale, location, and shape.

Limitations
To the best of our knowledge, this was the first study to explore the impact of parkification on forest color landscape authenticity from a nationwide perspective. We clarified the forest color composition of each region after parkification and provided new ideas for forest color landscape planning, design, and sustainable management. However, this study had limitations.
Due to data availability and geographical limitations, we used the Sina Weibo platform as the data source for forest imagery rather than collecting primary data on-site. However, the forest images obtained from the Sina Weibo platform exhibit inconsistencies in quality, which manifest as follows: (1) image quality is affected by human factors such as equipment and shooting modes; and (2) forest image screening is inevitably affected by the preferences of image uploaders and researchers [33]. In future research, it is advisable to utilize consistent equipment and shooting modes (such as the same camera and accessories, same aperture, and shutter speed) to conduct regular on-site sampling of forest color landscapes from different biogeographic regions. Sampling should occur at the exact locations, in the same directions, angles, and observation distances, while documenting on-site conditions such as illumination and color temperature. Furthermore, adopting an automatic image selection algorithm minimizes the influence of individual preference on study results.
Regarding the baseline determination of forest color authenticity, forest color data before parkification was unavailable, making it impossible to quantify forest color authenticity accurately. However, compared with previous studies, we found that forest color composition changed after parkification. Therefore, we suggest continuously monitoring forest color landscapes and adopting the color data from the initial monitoring period as baseline data to explore the degree of influence of forest parkification on color authenticity.

Conclusions
This national-scale study quantified forest colors using Weibo images as the data source. We found that the proportion of primary and secondary colors in forest composition was affected by factors such as vegetation type, resulting in significant differences among regions. However, there was no significant variation in the proportion of accent colors (blue, purple, and red) across regions. We further explored the impact of environmental factors in parks, including meteorological and anthropogenic disturbances, on the forest color landscape. The results showed that forest color was sensitive to all meteorological factors with the proportion of orange and yellow decreasing under anthropogenic climate change. Some forest colors (orange, yellow, purple, and red) were sensitive to anthropogenic disturbances in the following ways: (1) larger park areas, lower elevations, and larger building areas led to a higher proportion of orange; (2) lower park elevations resulted in a higher proportion of yellow; (3) larger park areas, simpler shapes, and lower elevations resulted in a higher proportion of purple; and (4) larger park areas, more complex shapes, lower elevations, lower building densities, and higher road network densities led to a higher proportion of red. The sensitivity of forest secondary colors to meteorological factors was higher than that to anthropogenic disturbances. Forest accent colors were more sensitive to anthropogenic disturbances than to meteorological factors with the greatest sensitivity being to park features. In conclusion, to protect the authenticity of forest main and accent colors, excessive borrowing among different regions should be avoided during forest color landscape construction, and the proportion of native tree species should be increased. To protect the authenticity of forest accent colors, it is necessary to prioritize park area, shape, and location control; reduce the construction of land in park buffer zones; and optimize park tour routes.
This study analyzed the impact and degree of influence of meteorological factors and anthropogenic disturbances on different forest colors, which provided new ideas for authentic protection and sustainable management of forest color landscapes. However, some limitations need to be addressed. In future studies, we will enhance image quality by improving image acquisition and screening. We can continuously monitor forest color landscapes to grasp the dynamic changes and anthropogenic disturbance degree of forest color, which will enhance protection of forest color landscapes' authenticity. Moreover, we will investigate the effects of climate change and human interference on the spatial patterns of forest color landscapes to provide in-depth guidance for planning and constructing sustainable forest color landscapes in parks and other urban green spaces.        OR is odds ratio and CI is 95% confidence interval of OR.