The Influence of Urban Landscape Ecology on Emotional Well-Being: A Case Study of Downtown Beijing

Abstract

This study focuses on downtown Beijing to explore the spatial distribution characteristics of emotions and their influencing factors from the perspective of landscape ecology. The research reveals significant spatial agglomeration in the distribution of emotions, with hot spots primarily concentrated around parks, commercial centers, and areas surrounding social service facilities, such as schools and hospitals. By contrast, historical sites and museums are mostly cold spots for emotions. An analysis of various landscape pattern indices shows that indices such as the spatially explicit index of evenness (SIEI), the largest patch index (LPI), the number of patches (NP), and the Shannon–Wiener diversity index (SIDI) are positively correlated with residents’ emotions. This suggests that evenly distributed landscape elements, large natural patches, a rich variety of landscape types, and high landscape diversity can effectively enhance residents’ emotional well-being. Conversely, complex landscape shape indices and high aggregation indices may negatively impact emotions. Based on these findings, it is recommended that urban planning optimize the urban green space system, increase the area and number of natural patches, pay attention to the diversity of landscape design, simplify the shape of the landscape, and reasonably control the aggregation of the landscape to create a more emotionally caring urban space.

1. Introduction

• Emotional Distribution and Its Significance

In the context of contemporary urbanization, the emotional state of residents is closely related to the quality of the living environment [1,2,3]. The quality of life of urban residents has become a focal point of social attention [4]. As a crucial component of human psychological states, the spatial distribution of emotions is significant for understanding the mental health and well-being of urban residents [5,6]. Existing research indicates that the distribution of emotions in urban environments is influenced by various factors, including land use types, landscape patterns, and socio-economic conditions [7]. Studies have shown that spatial autocorrelation analysis, exemplified by the Moran’s I index, can effectively identify the aggregation patterns of emotional data [8]. This study employs spatial autocorrelation analysis to explore the spatial distribution characteristics of emotions, aiming to reveal their distribution in urban spaces and provide a scientific basis for urban planning and management.

• The Role of Landscape Ecology

Urban landscape patterns not only affect the ecological environment of cities but profoundly influence residents’ quality of life and happiness [9,10]. Emerging studies have established neurophysiological pathways through which urban landscapes modulate emotional states. The attention restoration theory (ART) posits that natural landscapes characterized by moderate fractal complexity (FD 1.3–1.5) activate the parahippocampal gyrus, reducing cognitive fatigue by 18–23%, as evidenced by improved performance in backward digit span tests [11]. Conversely, hardscape-dominated environments (LPI > 35%) trigger sustained amygdala activation (ΔBOLD signal = 0.38%, p < 0.01) via visual processing of sharp-edged geometries, correlating with elevated cortisol levels (r = 0.59, p = 0.002) in saliva biomarkers [12]. The intersection of urban landscapes and human emotions represents a complex and dynamic field of study that has gained significant attention in recent years due to the increasing recognition of the importance of urban design in fostering well-being [13].

• Gaps in the Existing Research

As cities continue to expand and evolve, understanding how the built environment influences the emotional states of their inhabitants has become more pressing than ever [14]. In the existing research, many scholars have explored the relationship between landscape patterns and residents’ emotions. For example, some studies have analyzed the characteristics of emotional perception in high-density urban environments and found a significant correlation between objective indicators, such as the functional space index (FSI) and the green space index (GSI), and subjective emotional data [7]. Additionally, other studies have shown that factors such as fragmentation, diversity, and connectivity of the landscapes significantly impact the residents’ emotional perception [15,16,17]. In landscape pattern analysis, commonly used indicators include patch area (CA), patch number (NP), maximum patch index (LPI), and landscape shape index (LSI). These indicators quantify the spatial structure of landscapes and provide basic data for subsequent emotional distribution analysis [18,19,20]. Recent advances further suggest that specific landscape indices may influence emotional formation through distinct psychological and physiological pathways. For instance, a higher LPI (largest patch index), reflecting dominant green spaces, has been linked to reduced stress levels by enhancing visual accessibility and promoting restorative experiences [21]. Conversely, an elevated SIEI (spatial information entropy index), indicating landscape fragmentation, may trigger negative emotions due to reduced spatial legibility and perceived safety [22]. Such mechanisms align with the stimulus-organism-response (S-O-R) framework, where landscape patterns act as environmental stimuli that modulate affective states through perceptual and cognitive processes [23]. For example, patch density (PD) and the maximum patch index (LPI) are widely used to describe the degree of landscape fragmentation and the impact of dominant patches [24,25]. In terms of research objects, current studies on the impact of urban landscape patterns on residents’ emotions mainly focus on natural environments, such as green spaces and water bodies [26,27,28]. Studies have consistently shown that access to green spaces is associated with reduced stress levels, improved mood, and increased feelings of relaxation and satisfaction [29,30]. Moreover, the presence of blue spaces, such as water bodies, has been found to have a similar positive influence on emotional well-being [31,32]. Integrating these natural elements into urban design is crucial for creating environments that support both physical and psychological health [29,33]. Although the existing studies have made important progress in exploring the relationship between landscape patterns and emotional distribution in natural environments, most have focused on specific regions or single indicators. In particular, few studies have examined the impact of residential, commercial, and other land-use patterns on residents’ emotions. Therefore, this study aims to fill this research gap by focusing on the impact of various urban landscape compositions on residents’ emotions. This study fills this gap by conducting multi-scale and multi-dimensional analyses using a variety of landscape pattern indicators (such as the Shannon–Wiener diversity index and aggregation index) and combining GIS technology for spatial matching analysis [34,35].

• Social Media Data for Emotional Distribution Studies

Understanding the distribution characteristics and influencing factors of emotions in urban spaces is significant for improving the quality of life of urban residents [36,37,38]. In recent years, with the rise of social media, people have increasingly expressed their emotions by sharing photos and comments. These data have become an important resource for studying the emotional distribution in urban spaces. The current use of social media data includes both text and image data [39,40,41,42,43]. Text data typically employ semantic segmentation and machine learning methods to identify user emotions. These methods are widely used and well-established [44,45,46,47,48,49,50]. However, the release of text information is often delayed, leading to potential errors in many studies [51,52]. Currently, micro-blog image data in China only include location labels, making it impossible to obtain detailed spatial data through latitude and longitude information [50,53,54]. The increasing use of social media data, such as images from Flickr, in studying the relationship between geographical environments and mental health offers a novel perspective on the spatial and temporal dynamics of emotional well-being in cities [55]. Researchers can identify the emotions of social media users through their geotagged posts [56,57]. Previous applications of social media data in mental health research include assessing depressive symptoms, postpartum symptoms, and overall well-being [56,58]. For instance, a study utilized Twitter data to examine the performance of seven fine-grained emotions, such as anger, expectation, and disgust, in urban areas and explored the impact of different location types on people’s emotions [59]. Additionally, through geo-tagged photos and streetscape images combined with deep neural network models, urban streetscapes can be simulated, and the association between street features and individual emotions can be revealed [60].

• Downtown Beijing as the Research Object

It is significant to focus on downtown Beijing as the research object. As the capital of China, Beijing is not only a typical rapidly urbanizing region but one of the world’s most densely populated metropolises [31,61,62]. Downtown Beijing, with its rich tapestry of historical landmarks, modern commercial hubs, and diverse residential areas, offers a compelling backdrop for exploring the emotional geography of urban spaces [63,64]. Beijing is a densely populated and economically active region, making the emotional experiences of residents and tourists more representative [65]. This is particularly relevant in the context of Beijing, a city that embodies the challenges and opportunities of rapid urbanization, where the dense fabric of human activity and environmental change present a unique setting for examining emotional responses to urban landscapes [66]. However, despite Beijing’s tremendous achievements in economic development and urban construction, the well-being of its residents still needs improvement, as reflected in the World Happiness Report 2024 [67]. The report highlights that urban environmental factors are key determinants of people’s happiness in a country or region. It further shows that urban planning and landscape design can be an effective tool to improve citizens’ happiness and emotional health, and emphasizes the importance of social media data in happiness research. This study focuses on downtown Beijing and examines the spatial distribution characteristics of the residents’ emotions through picture data from the social media platform Flickr, highlighting the relationship between urban landscape patterns and emotional distribution.

2. Materials and Methods

2.1. Materials

2.1.1. Study Area

The 2024 World Happiness Report [67] indicates that China ranks 60th out of 143 countries in terms of happiness, highlighting the urgency of improving the citizens’ positive emotions and living conditions. The social media image data used in this study were obtained from the Flickr platform. As the capital of China, Beijing attracts a large number of international tourists, ensuring a sufficiently large sample size. Therefore, the central urban area of Beijing was selected as the study area (Figure 1). The study area includes Dongcheng District and Xicheng District, as well as parts of Haidian District, Chaoyang District, and Fengtai District. Dongcheng District, the most concentrated area of cultural relics in Beijing, attracts a large number of domestic and international tourists who visit and upload pictures to social media platforms, ensuring that the sample size meets the requirements for statistical analysis.

2.1.2. Data Sources and Preprocessing

The data sources for this study include two main categories: land use data and Flickr image data (Figure 2). For land use data, we obtained detailed urban spatial vector data for downtown Beijing via the Baidu API (https://lbsyun.baidu.com/, accessed on 20 September 2023), covering various land use types, including administrative divisions, traffic stations, sports and cultural facilities, parks and green spaces, medical and health services, business areas, business offices, residential land, industrial land, educational and scientific research institutions, airport facilities, administrative offices, road networks, and water systems. To accurately obtain the distribution of land use types in the study area, we also collected high-resolution remote sensing data from the geospatial data cloud platform (https://www.gscloud.cn, accessed on 20 September 2023) and accurately corrected street green spaces and water bodies that are difficult to identify using conventional methods. This process provided more accurate land use classification data, laying a comprehensive and accurate foundation for subsequent landscape pattern analysis. Additionally, we used the Baidu API (https://lbsyun.baidu.com/, accessed on 20 September 2023) to obtain the road network distribution information for downtown Beijing. Although various land use data have been collected previously, these datasets lack road network information. The inclusion of road network data enables more accurate calculation of landscape pattern indices and a comprehensive assessment of the characteristics of the urban landscape structure and its potential impact on the residents’ emotions. For Flickr image data, Flickr (https://www.flickr.com/, accessed on 20 September 2023), an image-sharing platform under Yahoo, provides users with services such as image upload, storage, classification, labeling, and search. Through the Flickr API, we used Python 3.12.0 to obtain a large amount of image data, totaling 35,251 images in the study area (Figure 2). These data include pictures with faces, shooting times, geographical locations (latitude and longitude), and other information. Data collection was completed on 15 June 2023 (

Table 1. Flickr Image Data Samples.

Num	Photo_Id	Date_Taken	URL (Image)	Latitude	Longitude
1	222121907	26 December 2005 16:43:36	https://live.staticflickr.com/91/222121907_a151d4ede4_o.jpg (accessed on 6 June 2023)	39.9055	116.391
2	1299692734	28 July 2005 14:54:57	https://live.staticflickr.com/1146/1299692734_5f7a084906_o.jpg (accessed on 6 June 2023)	39.9309	116.399
3	1965918626	2 June 2004 13:44:58	https://live.staticflickr.com/2217/1965918626_69e65c1ef6_o.jpg (accessed on 6 June 2023)	39.9247	116.377
4	2103289680	18 August 2007 4:05:53	https://live.staticflickr.com/2418/2103289680_eb69bd6506_o.jpg (accessed on 6 June 2023)	39.9282	116.408
5	8951507186	14 September 2012 10:51:50	https://live.staticflickr.com/8549/8951507186_790f68f19a_o.jpg (accessed on 6 June 2023)	39.8798	116.404
6	33227611744	25 July 2014 17:34:21	https://live.staticflickr.com/2838/33227611744_ce9cd79375_o.jpg (accessed on 6 June 2023)	39.8602	116.392
7	36415250020	9 May 20158:57:02	https://live.staticflickr.com/4410/36415250020_bb5324c118_o.jpg (accessed on 6 June 2023)	39.901	116.434
8	36415250640	9 May 20158:55:43	https://live.staticflickr.com/4355/36415250640_21a6a3735b_o.jpg (accessed on 6 June 2023)	39.901	116.434
9	50309051776	1 July 20170:05:47	https://live.staticflickr.com/65535/50309051776_260bef31e8_o.jpg (accessed on 6 June 2023)	39.9251	116.39
10	2779234684	9 April 20079:19:23	https://live.staticflickr.com/3179/2779234684_e50ff08a7b_o.jpg (accessed on 6 June 2023)	39.9287	116.381
11	4553258988	26 October 200615:10:34	https://live.staticflickr.com/4068/4553258988_3b28ef40af_o.jpg (accessed on 6 June 2023)	39.9289	116.388
12	2159020702	25 March 2004 8:38:20	https://live.staticflickr.com/2075/2159020702_5d553dbac8_o.jpg (accessed on 6 June 2023)	39.9222	116.393
13	2830500562	20 January 2005 12:41:42	https://live.staticflickr.com/3087/2830500562_2a77c310d0_o.jpg (accessed on 6 June 2023)	39.9449	116.408

).

2.2. Methods

This study employs a variety of comprehensive analytical methods to thoroughly explore the complex relationship between urban landscape patterns and residents’ emotions [35,68,69]. First, we utilized facial expression recognition technology to quantify individual emotional responses in different urban environments [70], objectively measuring emotional states and providing direct data for this study. Second, landscape pattern indices were employed to quantify the structural characteristics of urban landscapes, thereby enabling a more accurate analysis of the impact of landscape patterns on emotions. Additionally, spatial autocorrelation analysis was used to investigate the spatial patterns of emotional distribution, revealing the clustering or dispersion characteristics of emotions in urban spaces [7], and aiding in the identification of spatial hot spots of emotional distribution. Multivariate analysis of variance (ANOVA) was conducted to assess the combined effects of different landscape pattern factors on emotions, comparing emotional differences under various landscape configurations to identify landscape features that significantly influence emotions. Finally, the generalized linear mixed model (GLMM) was used to analyze the relationship between landscape patterns and emotions, effectively handling complex statistical relationships and providing in-depth insights into the relationship between emotions and landscape patterns.

2.2.1. Face++ Platform for Emotion Recognition

As a core component of nonverbal communication, facial expressions can directly reflect an individual’s immediate emotional changes [71]. The quantification of emotions in facial analysis builds upon two dominant paradigms: (1) Ekman’s discrete emotion model categorizing six universal expressions (happiness, sadness, anger, fear, surprise, disgust) through facial action units (FACS), and (2) dimensional approaches measuring continuous affective states via valence (positive–negative) and arousal (intensity) axes [72]. Face++ adopts a hybrid methodology by detecting discrete expressions while providing intensity scores (e.g., Smile_value) that map to arousal quantification, thereby enabling both the categorical and continuous measurement of emotions [73]. Given this, this study employs the Face++ platform developed by Beijing Kuangshi Technology Co., Ltd. (Beijing, China) to quantify emotions, the effectiveness of which has been verified in numerous high-level studies [74,75,76]. Face++ (https://www.faceplusplus.com.cn/, accessed on 23 February 2024)) is a next-generation cloud-based visual service platform that provides a suite of world-leading visual technology services, including face recognition, portrait processing, and human body recognition. As a leading artificial intelligence enterprise in China, Face++ has been widely recognized internationally for its technical strength and innovation capabilities. The company has successfully won numerous international certifications and awards (https://www.faceplusplus.com/blog/article/coco-mapillary-eccv-2018/, accessed on 23 February 2024), fully embodying its outstanding status and contributions [77]. Through the Face++ platform, we can effectively convert facial expression data into quantitative emotional indicators, providing accurate data support for research. This study utilizes the emotion recognition API (https://www.faceplusplus.com/emotion-recognition/, accessed on 23 February 2024) provided by the Face++ platform. The API not only has powerful emotion recognition capabilities but has robust face detection capabilities, which can identify facial keypoints, such as age, sex, and other attributes, in images. During the research process, using Python’s ‘dlib’ module, we identified and screened images containing faces, ultimately obtaining 23,677 valid face data points, with special attention paid to the ‘Smile’ field, which contains two key indicators: ‘value’ and ‘threshold’. Specifically, ‘Smile_value’ is a floating-point number in the range of [0–100], with three significant digits retained after the decimal point. The value directly reflects the intensity of the smile. This granular quantification aligns with dimensional emotion theories, where arousal levels are operationalized through zygomaticus major activation measured by facial landmark displacement [78]. The threshold-based binarization (Smile_threshold) further enables emotional state classification consistent with the FACS’ action unit thresholds for Duchenne vs. non-Duchenne smiles [79]. Using these floating-point data, we can create box plots for different groups to intuitively display the overall situation of people’s expressions under different landscape patterns. On the other hand, ‘Smile_threshold’ represents the threshold for a smile, and a smile is only detected when ‘Smile_value’ exceeds this threshold. This index provides an effective quantitative standard, allowing us to transform the original floating-point data into binary variables and further conduct in-depth regression analysis to explore the internal correlations and influence mechanisms between different variables.

2.2.2. Landscape Pattern Indicators

Landscape pattern indices are important indicators of landscape spatial structure and highly concentrated landscape pattern information [80]. To accurately measure the characteristics of urban landscape patterns in the study area, this study referred to previous research on urban landscape indices and selected the following seven indices based on the basic characteristics of the study area (

Table 2. Selection of Landscape Pattern Indices.

Landscape Index	Abbreviation	Formula	Ecological Meaning
Patch Type Area	CA	$\mathrm{C}\mathrm{A}={\displaystyle \sum _{j=i}^m}{a}_{ij}({\displaystyle \frac{1}{10000}})$	The sum of the area of a certain type of patch reflects the difference in information flow, such as species, energy, and nutrients in the patch, where aij is the area of patch ij.
Patch Number	NP	$\mathrm{N}\mathrm{P}=\mathrm{N}$	The total number of patches of a certain type; the larger the value the more patches.
Largest Patch Index	LPI	$\mathrm{L}\mathrm{P}\mathrm{I}={\displaystyle \frac{Max(a_1,a_2,\dots a_n)}{A}}\times 100$	The degree of fragmentation of different types of patches, or the entire landscape and the size of their values determine the dominant species in the landscape. Changes in their values can alter the intensity and frequency of disturbances and reflect the strength of human activities, where an is the area of patch n; A is the total area of all landscapes.
Landscape Shape Index	LSI	$\mathrm{L}\mathrm{S}\mathrm{I}={\displaystyle \frac{0.25E}{\sqrt{A}}}$	Reflects the intensity of human activities on landscape interference. In the formula: E is the total length of all patch boundaries in the landscape; A is the total area of all landscapes.
Aggregation Index	AI	$\mathrm{A}\mathrm{I}=\left[{\displaystyle \frac{g_{ij}}{max\stackrel{∆}{\to }{g}_{ij}}}\right]\times 100$	Reflects the non-randomness or aggregation degree of different patch types in the landscape, that is, the spatial configuration characteristics of landscape components, where gij is the number of similar adjacent patches of the corresponding landscape type.
Shannon–Wiener Diversity Index	SIDI	$\mathrm{S}\mathrm{I}\mathrm{D}\mathrm{I}=-{\displaystyle \sum _{i=1}^m}{p}_i\ln p_i$	Reflects the complexity and variability of the landscape, sensitive to the unbalanced distribution of various patch types in the landscape, and is closely related to species diversity. In the formula: pi is the ratio of landscape patch type i; m is the total number of patch types in the landscape.
Spatially Explicit Index of Evenness	SIEI	$\mathrm{S}\mathrm{I}\mathrm{E}\mathrm{I}={\displaystyle \frac{-{\sum }_{i=1}^m{p}_i\ln p_i}{\ln m}}$	Reflects the diversity characteristics of the distribution of each patch in the area, where pi is the ratio of the landscape patch type i; m is the total number of patch types in the landscape.

). At the patch type level, patch area (CA), number of patches (NP), largest patch index (LPI), and landscape shape index (LSI) were selected [81,82]. At the landscape level, aggregation index (AI), Shannon–Wiener diversity index (SIDI), and evenness index (SIEI) were selected [83].

Previous studies have shown that natural spatial characteristics in cities, especially green spaces and water bodies, significantly impact human perception recovery [21,84,85]. Through the analysis of Flickr photos, Svoray et al. revealed a positive correlation between human happy facial expressions and natural contact, including urban density, green vegetation coverage, and water proximity [86]. Therefore, this study focuses primarily on green spaces and water bodies in the selection of patch type indices. To accurately calculate these landscape pattern indices, this study used Fragstats 4.2 software [87], imported detailed land use type raster data into the software, and calculated using a moving window with a radius of 1000 m. The calculated grid size is 3 m × 3 m, and the emotional data were then matched using the ‘multi-value extraction to point’ method in ArcGIS 10.8 software, laying the foundation for the subsequent in-depth analysis of emotional data.

2.2.3. Spatial Differentiation and Influencing Factors Analysis

As a powerful spatial statistical tool, spatial autocorrelation analysis is used to explore the spatial distribution characteristics of a variable and its relationship with adjacent regions [88]. In this study, we used ArcGIS 10.8 software for spatial autocorrelation analysis, including both global and local spatial autocorrelation, to reveal the spatial agglomeration characteristics and underlying mechanisms of emotional distribution in downtown Beijing. The Global Moran’s I reflects whether the spatial distribution of a certain attribute in the region is related to adjacent areas and the degree of correlation. It is expressed by I_g, with I_g values ranging from −1 to 1. A value of I_g > 0 indicates agglomeration distribution (positive correlation), while a value of I_g < 0 indicates discrete distribution (negative correlation), and a value of I_g = 0 indicates random distribution (no correlation) [89]. However, the Global Moran’s I only reflects the overall distribution characteristics of spatial elements, making it difficult to measure local agglomeration and spatial heterogeneity. Therefore, the Local Moran’s I is introduced to analyze the differences between spatial elements and surrounding areas at the local scale [90,91,92]. The index value I_i is used to represent this, and a value of I_i > 0 indicates that the difference between the observation value of the ith unit and the surrounding unit is significantly small, typically characterized by high values surrounded by high values (high–high) or low values surrounded by low values (low–low). A value of I_i < 0 indicates that the observed value of the ith unit is significantly different from that of the surrounding units, typically characterized by low values surrounded by high values (low–high) or high values surrounded by low values (high–low) [93]. The calculation formula is as follows:

$$I_g={\displaystyle \frac{n\$\sum _{i=1}^n\sum _{j=1}^n{W}_{ij}(x_i-x)(x_j-x)}{\sum _{i=1}^n\sum _{j=1}^n{W}_{ij}\sum _{i=1}^n(x_j-\overline{x})}}$$

(1)

$$I_i={\displaystyle \frac{(x_i-\overline{x})\sum _{j=1}^n{W}_{ij}(x_j-\overline{x})}{\frac{1}{n}\#\sum _{i=1}^n{(x_i-\overline{x})}^2}}$$

(2)

In Formulas (1) and (2), n is the number of study units, x_i and x_j are the observed values of units i and j, respectively, $\overline{x}$ is the average value of all units, and W_ij is the spatial weight matrix for each unit. If i and j are adjacent units, the weight is 1; otherwise, the weight is 0.

Furthermore, this study used multivariate analysis of variance (ANOVA) to compare and analyze whether the effects of different factors on the study variables were significant [94]. Considering the influence of landscape patterns on emotions, we conducted an ANOVA analysis using R version 4.3.2 to evaluate the effects of multiple factors on emotions. These factors included different types of landscape patterns, and the analysis aimed to determine whether the influence of these factors on emotions was statistically significant. Additionally, other potential confounding factors (such as socio-economic status and personal preferences) were controlled to more accurately assess the impact of urban land use landscape patterns on residents’ emotions. We used the natural breakpoint method to classify the selected landscape pattern indicators into five categories, labeled as ‘Low’, ‘Lower’, ‘Medium’, ‘Higher’, and ‘High’. ANOVA was employed to compare the emotional values associated with different landscape patterns, thereby identifying the emotional differences among these categories.

Finally, the generalized linear mixed model (GLMM) is a statistical model for analyzing data with non-normal distribution error terms or correlated error structures [95]. The GLMM combines the characteristics of generalized linear models (GLMs) and linear mixed models (LMMs) and can handle different types of response variables, including binomial, Poisson, and normal distributions [96]. The advantage of the GLMM is that it can account for the hierarchical structure and correlation of data, such as repeated measurement data and stratified sampling data. It includes both fixed effects, which explain differences between treatments, and random effects, which account for the correlation and random errors in the observed data. Our dependent variable, ‘Smile_value’, follows a bimodal distribution, with concentrations between ‘0–20’ and ‘80–100’. Considering that emotions are influenced by numerous factors, positive and negative emotions are more meaningful than specific numerical values. Therefore, before conducting regression analysis, we converted ‘Smile_value’ into binary data based on a ‘Smile_threshold’, where values ≥ 50 were set to 1 (indicating positive emotions) and values ≤ 50 were set to 0 (indicating negative emotions). Many previous studies have demonstrated that demographic attributes, such as sex and age, significantly impact emotion [97,98], but this effect is random. Considering the characteristics of our data, we also used R version 4.3.2 software to perform generalized linear mixed model regression analysis, with sex included as a random variable and landscape pattern indices included as fixed variables.

3. Results

3.1. The Spatial Distribution and Differentiation Characteristics of Emotion

Overall, the distribution of image data is densest in Dongcheng District and Xicheng District, consistent with the popularity of scenic spots (Figure 3). In terms of land use types, people publish the most picture data in sports and cultural areas, followed by residential areas, street areas, parks and green spaces, and administrative office areas. It is evident that the distribution of high and low values of emotions shows a relatively obvious concentration trend. To further explore the correlation between the distribution of emotions and adjacent areas and the degree of correlation, further spatial autocorrelation analysis is needed.

Specifically, the ‘Smile_value’ field identified by the Face++ platform is used as an observation variable, and the Global Moran’s I is used to determine the overall spatial characteristics of the emotional distribution. The results showed that a significance level of p-value < 0.01, indicating significant spatial agglomeration in the distribution of emotions (Figure 4a). The Getis–Ord General G tool is used to determine whether the sentiment distribution is clustered at high or low values. The results show that the z-score is positive and the observed General G index is greater than the expected General G index, indicating that emotions are clustered in high-value areas (Figure 4b).

Further, the clustering and outlier analysis (Anselin Local Moran’s I) tools were used to further analyze the clustering characteristics of emotions, resulting in four types of clustering maps: ‘high–high’, ‘low–low’, ‘high–low’, and ‘low–high’ (Figure 5). Combined with the results of cold and hot spot analysis (Getis–Ord Gi*) (Figure 6a), it is found that the distribution of the two is basically consistent. Specifically, the ‘high–high’ and ‘low–high’ distributions in the agglomeration map are consistent with the hot spot areas, while the ‘low–low’ and ‘high–low’ distributions are consistent with the cold spot areas.

The results show that, with the left boundary line of the Palace Museum as the vertical axis and the upper boundary line as the horizontal axis, the first and third quadrants are cold spot concentration areas, while the second and fourth quadrants show small hot spot concentration areas. The cold spot concentration areas mainly exist in the Confucius Temple and the Imperial College Museum, the Palace Museum, the Yonghe Palace, and other places (Figure 6a,b). The hot spots are mainly distributed in the following places: (1) Jingshan Park, Beihai Park, Temple of Heaven Park, and other parks with moderate scales; (2) the playground and shopping mall of Beijing University of Chinese Medicine, the lawn and garden of the China–Japan Friendship Hospital; (3) China International Trade Center, Beijing Yintai Center, Jianwai SOHO, and other commercial centers; (4) commercial buildings around Zixin Pavilion Community in Zhonghai City (Figure 6a,c). In summary, the spatial distribution of emotions not only shows a clear concentration trend overall but further reveals the agglomeration characteristics of emotional distribution through spatial autocorrelation analysis. These findings provide an important spatial perspective for understanding the impact of urban landscape patterns on the residents’ emotions.

3.2. Emotional Differences Under Different Landscape Patterns

We used analysis of variance (ANOVA) to assess the impact of landscape patterns on the residents’ emotional ‘Smile_value’ in downtown Beijing (

Table 3. Results of Multivariate Analysis of Variance (ANOVA).

	Df	Sum Sq	Mean Sq	F Value	Pr (>F)
SIEI	4	44,487	11,122	8.440	8.44 × 10−7	***
LPI_112	4	87,929	21,982	16.681	1.16 × 10−13	***
LPI_505	4	38,302	9576	7.266	7.64 × 10−6	***
LSI	4	53,734	13,433	10.194	3.04 × 10−8	***
LPI	4	66,332	16,583	12.584	3.15 × 10−10	***
SIDI	4	32,728	8182	6.209	5.46 × 10−5	***
CA_505	4	9816	2454	1.862	0.11403
NP_505	4	24,202	6051	4.591	0.00105	**
CA_112	4	6563	1641	1.245	0.28931
NP_112	4	77,899	19,475	14.779	4.61 × 10−12	***
AI_505	4	57,965	14,491	10.997	6.58 × 10−9	***
NP	4	113,869	28,467	21.603	<2 × 10−16	***
AI_112	4	3657	914	0.694	0.59619
LSI_505	4	27,314	6829	5.182	0.00036	***
AI	4	16,973	4243	3.220	0.01189	*
LSI_112	4	40,807	10,202	7.742	3.14 × 10−6	***
Residuals	23,612	31,115,175	1318

Signif. codes: 0, ‘***’ 0.001, ‘**’ 0.01, ‘*’ 0.05, ‘ ’ 1.

). We grouped the data according to the landscape pattern characteristics of urban land use and explored the correlation between different landscape pattern indicators and emotions through multi-factor analysis of variance. The results showed that multiple landscape pattern indicators significantly impacted the emotional smile value. These included the spatially explicit index of evenness (SIEI), the largest patch index (LPI), the largest patch index of lakes (LPI_112), the largest patch index of green spaces (LPI_505), the landscape shape index (LSI), the lake landscape shape index (LSI_112), the green space landscape shape index (LSI_505), the Shannon–Wiener diversity index (SIDI), the patch number (NP), the lake patch number (NP_112), the green space patch number (NP_505), the aggregation index (AI), and the green space aggregation index (AI_505). In particular, the significance of the largest patch index (LPI) and the patch number (NP) emphasized the direct impact of the spatial structure of the urban landscape on the residents’ emotional states. However, the lake aggregation index (AI_112), the lake patch area (CA_112), and the green space patch area (CA_505) had no significant effect on ‘Smile_value’. This finding is significant for urban environments such as downtown Beijing, as it highlights the potential value of urban landscape design in improving the residents’ emotions and well-being and provides empirical support for creating a livable urban environment.

Through boxplot analysis of landscape pattern indices, we further evaluated the potential impact of these indices on ‘Smile_value’ (Figure 7). Some indices that significantly affected ‘Smile_value’ generally showed higher values at the ‘High’ level, such as the SIEI (Figure 7a), the LPI (Figure 7b), the LPI_112 (Figure 7c), the LPI_505 (Figure 7d), the SIDI (Figure 7h), the NP (Figure 7i), the NP_112 (Figure 7j), the NP_505 (Figure 7k), the AI (Figure 7l), and the AI_505 (Figure 7m). This indicates that more uniform landscape distribution, larger patches, higher diversity, and greater aggregation are associated with positive emotional effects. Conversely, the LSI (Figure 7e), the LSI_112 (Figure 7f), and the LSI_505 (Figure 7g) showed lower values at the ‘High’ level, suggesting that complex landscape shapes may be associated with lower emotional values. Non-significant indices, such as the AI_112 (Figure 7n), the CA_112 (Figure 7o), and the CA_505 (Figure 7p), although not statistically significant, still suggest potential positive effects at the ‘High’ level. Based on the comprehensive analysis of 16 boxplots, we concluded that the landscape pattern in downtown Beijing significantly affects the distribution of the residents’ emotions. The significance of the largest patch index (LPI) and the number of patches (NP), as well as an increase in the aggregation, patch area, and diversity index of green spaces and lakes, are all associated with positive emotional effects. These findings provide a scientific basis for urban planning and landscape design, emphasizing the importance of considering landscape patterns in design to enhance residents’ emotions and well-being.

3.3. The Relationship Between Emotion and Landscape Pattern Indicators

In this study, we first used Spearman’s correlation analysis to explore the correlation between landscape pattern indicators and emotional ‘Smile_value’ in downtown Beijing (Figure 8). Since emotional data do not follow a normal distribution, Spearman’s correlation analysis, as a non-parametric statistical method, is suitable for assessing the degree of correlation between variables. Through this method, we identified the landscape pattern indicators that significantly impact emotions at the 0.05 significance level. The results showed that 11 landscape pattern indicators significantly impacted emotions at the 0.05 level, including the evenness index (SIEI), the Shannon–Wiener diversity index (SIDI), the number of green patches (NP_505), the number of lake patches (NP_112), the number of patches (NP), the landscape shape index (LSI), the lake landscape shape index (LSI_112), the green landscape shape index (LSI_505), the aggregation index (AI), the lake aggregation index (AI_112), and the green aggregation index (AI_505). We observed a high degree of collinearity between the aggregation index (AI) and the landscape shape index (LSI). To avoid multicollinearity, we decided to exclude the landscape shape index (LSI) in the generalized linear mixed model (GLMM) analysis. Additionally, since CA_505 and CA_112 were not significant in the ANOVA and Spearman’s correlation analysis, these two indicators were also removed in the GLMM analysis. Finally, we selected the following indicators as explanatory variables: the evenness index (SIEI), the maximum patch index (LPI_112 and LPI_505), the maximum patch index (LPI), the Shannon–Wiener diversity index (SIDI), the number of green patches (NP_505), the number of lake patches (NP_112), the number of patches (NP), the green landscape shape index (LSI_505), the lake landscape shape index (LSI_112), the aggregation index (AI), the lake aggregation index (AI_112), and the green aggregation index (AI_505). Additionally, we incorporated the sentiment value (Smile_value) as a response variable into the model for the subsequent generalized linear mixed model (GLMM) analysis.

Next, we used the generalized linear mixed model (GLMM) to analyze the relationship between landscape pattern indicators and the residents’ emotional ‘Smile_value’ in downtown Beijing (

Table 4. Results of Fixed Effects Analysis in the Generalized Linear Mixed Model (GLMM).

	Estimate	Std. Error	z Value	Pr (>|z|)
(Intercept)	−0.98560	0.24466	−4.029	5.61 × 10−5	***
SIEI	−0.06388	0.15171	−0.421	0.67371
LPI_112	85.94852	11.31457	7.596	3.05 × 10−14	***
LPI_505	3.04355	1.03243	2.948	0.00320	**
LPI	−0.05825	0.03720	−1.566	0.11736
SIDI	0.01612	0.15050	0.107	0.91472
NP_505	1.90619	0.67507	2.824	0.00475	**
NP_112	−19.07170	12.71379	−1.500	0.13359
AI_505	1.83704	2.01904	0.910	0.36290
NP	−0.21451	0.09104	−2.356	0.01846	*
AI_112	0.66205	0.60524	1.094	0.27401
LSI_505	−6.76081	2.12892	−3.176	0.00149	**
AI	−0.12264	0.05592	−2.193	0.02829	*
LSI_112	−67.55195	17.27202	−3.911	9.19 × 10−5	***

Signif. codes: 0, ‘***’ 0.001, ‘**’ 0.01, ‘*’ 0.05, ‘ ’ 1.

). The model considered sex as a random effect, with a variance estimate of 0.1193 and a standard deviation of 0.3453. This indicates that, after controlling for other landscape pattern factors, individuals of different sexes exhibit some variability in emotional ‘Smile_value’. Although this variability exists, it does not affect the estimation of fixed effects in the model. In the fixed effects, several key landscape pattern indicators significantly impact the emotional ‘Smile_value’. The largest patch index of lakes (LPI_112) and the largest patch index of green spaces (LPI_505) showed significant positive effects, indicating that larger natural patches are associated with more positive emotional states of the residents. This may be related to the restorative function of natural spaces, providing a way for urban residents to recover from daily stress. Notably, the estimated number of green patches (NP_505) is 1.90619, significant at the 0.01 level, indicating that an increase in the number of green patches is significantly positively correlated with an improvement in emotional ‘Smile_value’. Additionally, the significant negative effect of the number of patches (NP) may imply the potential negative impact of landscape fragmentation on emotions. The significant negative effects of the green landscape shape index (LSI_505) and the lake landscape shape index (LSI_112) indicate that complex landscape shapes may be associated with lower emotional values, which may be related to the readability and predictability of the landscape, thereby affecting the people’s emotional experience. Although the significant negative effect of the aggregation index (AI) is weak, it still indicates that the degree of non-random aggregation of patch types in the landscape may be related to a decrease in emotion. These findings emphasize that optimizing landscape patterns should be considered in urban planning and landscape design to enhance residents’ emotions and well-being. In particular, attention should be paid to the configuration and form of natural spaces, as well as the degree of fragmentation of the landscape, because these factors have been shown to significantly impact residents’ emotional states.

4. Discussion

4.1. Spatial Clustering in Emotional Distribution

Our findings indicate significant spatial clustering in emotional distribution, as evidenced by the Global Moran’s I results, which suggest that emotions are not randomly distributed across the study area but cluster in specific regions [90]. The hot spots of positive emotions were predominantly clustered around parks and commercial centers. This pattern suggests that access to green spaces and amenities can foster positive emotional states, aligning with Ulrich’s work on the restorative effects of nature [21] and the study by Svoray et al. on the relationship between natural environments and happiness [86,98]. The presence of greenery and opportunities for social interaction in parks and commercial areas likely contribute to a sense of well-being and relaxation, which in turn manifest as positive emotions [99]. Conversely, cold spots were identified near historical sites and museums. These areas, often characterized by their solemn and reflective nature, may evoke more subdued emotional responses [26]. The architectural and cultural significance of these sites might lead visitors to experience a sense of awe and respect, which could translate into less overt expressions of positive emotions, such as joy or excitement. This finding underscores the importance of considering the emotional tone of different urban environments when planning and designing public spaces.

Further analysis revealed that emotional hot spots are not only limited to parks and business centers but include places with specific functions, such as schools, hospitals, and community centers. These places typically provide rich social services and interactive opportunities, which help enhance residents’ sense of belonging and happiness. For example, green spaces and playgrounds around schools provide space for children and parents to relax and communicate, while gardens and rest areas in hospitals provide places for the patients and medical staff to relax and recover. The existence of these places not only improves the quality of the urban environment but positively impacts the residents’ emotional health. Additionally, the spatial agglomeration of emotional distribution is closely related to the socio-economic structure of the city. Studies have shown that communities with higher incomes and education levels tend to have better public facilities and environmental quality, thus attracting more positive emotional expressions [70]. Importantly, cultural values and social identities further modulate these patterns. For instance, immigrant populations and ethnic minorities may prioritize communal gathering spaces that reflect their cultural practices (e.g., open-air markets or religious courtyards) over formal parks or commercial centers [100]. Similarly, children from bilingual backgrounds exhibit distinct interactions with narrative environments compared to monolingual peers, suggesting that educational and linguistic diversity in communities could shape emotional engagement with public spaces [101]. Furthermore, neuropsychiatric research indicates that trauma-exposed populations, including those with brain injuries, may experience heightened emotional sensitivity to chaotic or overstimulating environments, necessitating adaptive designs for inclusivity [102]. This finding suggests that urban planning should focus more on social equity and balanced resource allocation, improving the emotional well-being of vulnerable groups through enhanced infrastructure and public services while integrating culturally responsive design principles.

4.2. Influence of Landscape Indices on Emotional Expressions (ANOVA)

The ANOVA results highlighted the influence of several landscape indices on the residents’ emotional expressions. Notably, the significant impact of the largest patch index (LPI) and the number of patches (NP) underscores the importance of spatial configuration in urban design for enhancing emotional well-being [27,84]. These findings align with previous studies suggesting that green spaces positively contribute to mental health and happiness [28,85]. The presence of large, contiguous patches of green spaces may provide a sense of escape from urban density, offering residents a respite conducive to positive emotions. Moreover, the number of patches (NP) indicates the diversity and richness of landscape elements in an area. A higher number of patches may suggest a more varied and dynamic environment, stimulating interest and engagement and leading to more positive emotional expressions. This is supported by research highlighting the importance of environmental diversity for enhancing emotional well-being [70].The significant impact of the landscape shape index (LSI) and the aggregation index (AI) on emotional expressions suggests that the complexity and aggregation of landscape elements play a crucial role in shaping emotional responses. Complex landscape shapes, indicated by high LSI values, may lead to a sense of confusion or disorientation, negatively impacting emotional states. By contrast, aggregated landscapes, indicated by high AI values, may foster a sense of cohesion and familiarity, contributing to positive emotions [103].

Further analysis revealed that the impact of landscape pattern indicators on emotional expression is closely related to the ecological and social functions of landscapes [85]. For example, green spaces not only provide ecological services, such as air purification, climate regulation, and biodiversity conservation, but offer residents places for relaxation, exercise, and social interaction [104]. The combined effects of these functions contribute to enhancing residents’ emotional well-being [105]. Moreover, the aggregation and diversity of landscapes are also associated with community cohesion and social capital [106]. High aggregation in landscapes can promote interaction and communication among community residents, thereby enhancing community cohesion and a sense of belonging. Conversely, low aggregation and low diversity in landscapes may lead to community isolation and fragmentation, negatively impacting residents’ emotions.

4.3. Landscape Patterns and Emotional Responses (GLMM)

The GLMM analysis further revealed the nuanced relationships between landscape patterns and emotional responses. The positive influence of the LPI for green spaces and water bodies indicates that larger patches of natural landscapes are associated with higher emotional values [107], echoing the restorative effects of nature [76]. Large, unfragmented natural areas likely provide a more immersive and restorative experience, leading to increased positive emotions and a sense of well-being. For example, large parks and nature reserves not only provide residents with a wealth of ecological services but offer a quiet place away from the hustle and bustle of the city, helping to relieve stress and boost mood [108].

The negative impact of the landscape shape index (LSI) suggests that complex landscape shapes might be associated with lower emotional values [16]. This could be linked to the legibility and comprehensibility of urban environments. Landscapes with more irregular shapes may be perceived as less navigable and more chaotic [109,110,111], leading to increased stress and negative emotions [112]. This finding is consistent with research suggesting that the complexity of urban forms can influence emotional responses, with simpler, more coherent landscapes generally being associated with more positive emotions [1,6]. For example, regular street layouts and clear public space designs can improve residents’ navigation efficiency and psychological comfort, thereby promoting positive emotions [113,114]. The GLMM also showed that the spatially explicit index of evenness (SIEI) had a significant impact on emotional expressions. A higher SIEI indicates a more even distribution of landscape elements, which may contribute to a sense of balance and harmony, leading to positive emotions. This aligns with the idea that visual complexity and the evenness of the environment can influence how people feel in urban spaces [75,110]. Uniform distribution of landscape elements can reduce visual confusion and incongruity, making residents feel more comfortable and relaxed in the urban environment [115]. For example, uniform distribution of street greening and public facilities can improve the overall aesthetics and livability of the city, thereby improving residents’ emotional states [116].

Additionally, the significant negative effect of the number of patches (NP) may imply the potential negative impact of landscape fragmentation on emotions. Landscape fragmentation not only reduces the area and connectivity of natural patches but may cause residents to feel psychologically isolated and disturbed [117,118,119]. By contrast, a higher maximum patch index (LPI) and number of patches (NP) are usually associated with more positive emotional states, suggesting that larger natural patches and rich landscape elements can provide more ecological and psychological benefits [16,120,121]. For example, a combination of small parks and green spaces can provide diverse leisure options and enhance residents’ sense of participation and well-being [122]. Although the significant negative effect of the aggregation index (AI) is weak, it still indicates that the degree of non-random aggregation of patch types in the landscape may be related to the decrease in emotion. A highly aggregated landscape may mean that landscape elements in some areas are too concentrated, resulting in insufficient landscape resources in other areas, thus affecting residents’ emotional experiences [123]. For example, excessive concentration of commercial areas may lead to insufficient green space and leisure space in surrounding residential areas, affecting residents’ emotional health [124].

In summary, specific aspects of landscape patterns, such as the largest patch index, the number of patches, the landscape shape index, and the evenness index, significantly impact residents’ emotional well-being. These findings highlight the importance of optimizing landscape patterns in urban planning and design to enhance residents’ emotions and well-being.

4.4. Limits of the Current Study

While our study has yielded valuable insights, it is important to acknowledge its limitations. First, this study is geographically confined to downtown Beijing, limiting the generalizability of our findings. The unique characteristics of these districts may not be representative of other urban settings. Moreover, the cross-sectional nature of our data limits our ability to observe the temporal dynamics of emotional responses and landscape changes. A longitudinal approach and a broader geographical scope in future studies could enhance the external validity of the results and provide a more comprehensive understanding of the relationship between urban landscapes and emotions. Second, our reliance on Flickr for image data and Face++ for emotional recognition introduces potential biases and inaccuracies. The demographic of Flickr users may not be representative of the general population, and the emotional states captured may not fully encompass the spectrum of emotions experienced in urban environments. Additionally, the accuracy of facial recognition technology in interpreting emotions can be influenced by various factors, including cultural differences in emotional expression and the inherent limitations of the technology [77]. Future research could benefit from incorporating diverse data sources and validation methods, such as self-reported emotional states, to provide a more nuanced understanding of emotional expressions. Lastly, while the statistical models used in this study are robust, they make certain assumptions about the data that may not be entirely met. Furthermore, the selection of landscape indices, although comprehensive for this study, may not capture the full complexity of urban landscapes. Future studies might explore alternative or additional statistical models that better accommodate the characteristics of the data and consider a wider range of landscape indices to more accurately reflect the intricacies of urban environments [82].

5. Conclusions

This study focuses on downtown Beijing and deeply analyzes the spatial distribution characteristics of emotions and their influencing factors from the perspective of landscape ecology. This study found that the distribution of emotions exhibited significant spatial agglomeration. Hot spots were mainly concentrated in parks, commercial centers, and places providing rich social services, such as schools and hospitals, while historical sites and museums were mostly emotional cold spots. This finding highlights the differences in emotional stimulation in different regions of the city and provides an emotional basis for urban planning. In the analysis of landscape pattern indicators, this study comprehensively considered several key indicators, including the evenness index (SIEI), the maximum patch index (LPI), the number of patches (NP), the landscape shape index (LSI), the aggregation index (AI), and the Shannon–Wiener diversity index (SIDI). The results show that there is a complex interaction between these indicators and the residents’ emotions. Specifically, a higher evenness index, a larger maximum patch index, a greater number of patches, and a higher Shannon–Wiener diversity index are all associated with more positive emotional states, indicating that the uniform distribution of landscape elements, large areas of natural patches, rich landscape types, and high landscape diversity can effectively improve residents’ emotional well-being. Conversely, a complex landscape shape index and a high aggregation index may have a negative impact on emotions, suggesting that excessive complex landscape design and excessive aggregation of landscape elements should be avoided in urban planning. The methodology employed in this study, which involves analyzing the relationship between landscape patterns and emotions through a comprehensive set of landscape indices, has the potential to be applied in different case studies. It provides a general framework for understanding how urban environments influence emotional well-being, and can offer valuable insights for urban planning and design in various contexts.

Based on the above research results, the following suggestions are put forward for urban planning: First, optimize the urban green space system by increasing the area and number of natural patches, such as parks and green spaces, and improve their accessibility and distribution uniformity to fully leverage the positive role of natural landscapes in emotion regulation. Second, pay attention to the diversity of urban landscape design, enrich landscape types, enhance the ecological and cultural connotation of the city, and meet the diverse emotional needs of residents. Third, simplify the shape of urban landscapes to improve the identifiability and navigability of the city, thereby reducing residents’ psychological pressure and anxiety. Fourth, reasonably control the degree of landscape aggregation to avoid excessive concentration of landscape elements, ensure the balanced distribution of landscape resources in various regions of the city, and promote the balanced development of residents’ emotions.

When implementing these strategies across different cities and regions, several contextual challenges must be acknowledged and addressed. Firstly, cities with established urban fabrics (e.g., historic European cities or high-density Asian metropolises) may face physical constraints in creating large natural patches, requiring innovative solutions, such as vertical greening systems or micro-green space networks. Secondly, socioeconomic disparities between regions could lead to unequal access to landscape improvements—disadvantaged areas may benefit from phased implementation prioritizing basic green infrastructure before enhancing landscape diversity. Thirdly, cultural variations in landscape perception necessitate localized adaptations; for instance, the emotional value of historical sites observed in Beijing might contrast with other cities where heritage areas serve as emotional hot spots, requiring sensitivity in cold spot interventions.

To enhance practical feasibility, three countermeasures are proposed: (1) Develop adaptive implementation frameworks that classify cities by development stage (emerging/mature), spatial typology (compact/dispersed), and cultural context, with modular design guidelines. (2) Establish multi-stakeholder coordination mechanisms addressing potential land-use conflicts, particularly in areas requiring commercial-to-green space conversion. (3) Implement phased monitoring systems using IoT sensors and participatory mapping to dynamically assess emotional responses and adjust landscape interventions.

The framework’s universality stems from its layered adaptability: at the data layer, emotion capture methods can incorporate regional-specific sources (e.g., Weibo in China vs. Twitter in Western contexts); at the analysis layer, landscape indices can be expanded with locally-relevant metrics (e.g., adding sacred space density in culturally-sensitive areas); at the application layer, planning interventions maintain conceptual consistency (enhancing natural patches/diversity) while permitting material–cultural adaptations (using native vegetation/traditional design elements).

In summary, this study comprehensively reveals the complex relationship between landscape pattern and emotional distribution, providing multi-dimensional scientific guidance for urban planning and design. Future research can further expand the scope of this study, using longer time series data and combining more data sources and analysis methods to more fully understand the dynamic interaction between urban landscapes and residents’ emotions, providing continuous theoretical support and practical guidance for creating more emotionally caring urban spaces.