Introduction

The concept of urban vitality was first proposed by Jane Jacobs(1961), who argued that the interaction of people with people and people with places can help regenerate and energize cities1. Scholars have studied urban vitality from different perspectives. They believe that urban vitality is a kind of diverse resident activities and complex environmental interactions affected by urban spatial patterns2,3,4,5. With the rapid development of urbanization in China, the population keeps pouring into the cities, which makes the built environment of the cities change dramatically. This has resulted in a series of urban problems, including traffic congestion, environmental pollution, and unbalanced urban-rural development, leading to a decline in urban vitality6. This decline not only affects the quality of life of residents, but also hinders economic development and sustainability. Yinchuan and many other cities are experiencing this. Urban vitality reflects the complex interplay of social, economic and cultural factors, which directly affects the attractiveness of a city and the happiness index of its residents7. Therefore, studying the impact of the built environment on urban vitality and optimizing the built environment to stimulate urban vitality is a scientific problem that needs to be solved in urban spatial planning and design, which is conducive to improving the quality of life of urban residents and providing scientific policy references for planning managers.

Research on urban vitality has mainly focused on its description and the factors that influence it8,9. The studies that have been carried out have mainly evaluated urban vitality in terms of economic, social and cultural indicators, among others10,11,12. More studies have focused on the spatial distribution of population and activities in cities6. These studies have relied heavily on traditional data sources such as questionnaires, surveys and limited Geographic Information System (GIS) data to define and measure urban vitality13. However, these methods can be subjective and limited in scope and accuracy to capture the nuanced dynamics of human activity within cities. With the popularization of big data and the development of information technology, many researchers have used multi-source big data such as mobile ___location data, mobile phone signaling data, Baidu heat maps and GPS to describe the nuanced dynamics of human activities within cities. For example, some studies have used Baidu heat map data to study the dynamic spatio-temporal relationship between human behavior and environmental studies14. Therefore, this study introduces Baidu heat map as an indicator of urban vitality. Baidu heat map provides real-time data on population distribution in urban areas. These data not only capture the population density, but also reflect the spatial and temporal patterns of human activities.

The built environment refers to the human-made environment in cities that is not naturally formed but created and constructed through human activities15. Cervero established the “3D” quantifiable system of the built environment, encompassing three dimensions: Density, Diversity, and Design16. Subsequently, Ewing supplemented it with “Destination accessibility” and “Distance to transit”, expanding it into a more comprehensive “5D” quantifiable system of the built environment, which provides a robust foundation for measuring the built environment17. Based on diverse data sources, many studies primarily utilize two-dimensional indicators, such as Point of Interest (POI) data and road networks, to describe the built environment18. For instance, some researchers have found that transportation convenience and commercial types significantly impact urban vitality19. Wang argue that POI density has the most pronounced effect on enhancing urban vitality20. While these data can provide insights into urban functions and structures, they lack descriptions of the three-dimensional characteristics of urban spaces. To address this limitation, this study incorporates three-dimensional indicators (including openness, greenness, enclosure, walkability, and imageability)21 extracted from Baidu Map Street view images, allowing for a more comprehensive understanding of the built environment and its relationship with urban vitality. This approach offers a new dimension for comprehending the impact of the built environment on urban vitality.

A significant number of scholars have investigated the relationship between the built environment and urban vitality, primarily employing regression models such as the Least Squares method from a holistic perspective22. These studies have explored the impact of built environment factors, including density, diversity, and transportation accessibility, on urban vitality23,24,25,26. To some extent, these research efforts have confirmed the correlation between urban vitality and the built environment. However, linear regression models have limitations when analyzing spatial data, as they fail to account for spatial heterogeneity and spatial autocorrelation. Given the presence of spatial heterogeneity, this study adopts the Geographically Weighted Regression (GWR) model27. By utilizing GWR, it is possible to identify regions where specific built environment factors have stronger or weaker influences on urban vitality, providing more targeted insights for urban planning and policy formulation.

Based on an analysis of previous research, this study clarifies its objectives and questions, using Yinchuan City as a case study. We explore the multifaceted impact of the built environment on urban vitality by employing Jane Jacobs’ theory of vitality, Robert Cervero’s 3D and 5D theories, residents’ perceptions, and big data. We quantitatively construct 5D built environment variables using Baidu Heat Map, Baidu Street View, and deep learning techniques. Initially, linear regression is utilized to analyze the global relationship between the built environment and urban vitality. Neglecting spatial heterogeneity, we then employ Moran’s I to test for spatial autocorrelation in the data. If significant spatial autocorrelation is detected, Geographically Weighted Regression (GWR) is applied to explore spatially varying relationships, as shown in Fig. 1. Through this process, we aim to quantitatively describe the spatio-temporal patterns of urban vitality and validate spatial clustering. Ultimately, this study aims to formulate urban planning and development strategies specifically tailored to enhance the urban vitality of Yinchuan, thereby enriching and advancing the existing academic discourse.

The paper is structured as follows: “Literature review” reviews pertinent literature; “Data and methodology” outlines the data and modeling methods employed; “Results and analysis” presents the findings; “Discussion” discusses key insights and proposes strategic recommendations; and “Conclusion” concludes the paper, offering avenues for future research.

Fig. 1
figure 1

Methodological framework.

Full size image

Literature review

Measurement methods and data sources for urban vitality

Urban vitality is a multidimensional concept that encompasses economic, social, cultural, and environmental elements7,12. Jane Jacobs first introduced the concept of urban vitality in 19611. Her theory of vitality emphasizes the importance of urban diversity and complexity in maintaining urban vitality, providing us with an analytical framework from a sociological perspective that highlights the interaction between the behavior of urban residents and the urban environment. Lynch (1984) suggests that urban vitality is a crucial indicator of how settlement forms support people’s lives and needs28. Montgomery and John (1998) emphasize the significance of human activities in contributing to urban vitality. Research on human activities and the built environment primarily focuses on aspects such as living spaces, health, lifestyles, and living experiences4.

The measurement of urban vitality has always been a research focus in the fields of urban planning and management7,26,29. Traditionally, urban vitality is often assessed using macro-level data such as population censuses, economic statistics, and social activity surveys30. However, these methods have limitations in terms of timeliness, granularity, and dynamism. With the development of big data technologies, the measurement of urban vitality is becoming more diversified and refined31. Researchers utilize multi-source geographic big data, such as Open Street Map, Baidu Map POI, WeChat travel data, Meituan data, and Gaode data32,33,34,35. From the dual perspectives of people and space, they construct comprehensive assessment models for urban vitality. For example, some studies select indicators from four aspects: crowd vitality, vitality diversity, activity satisfaction, and spatial interaction potential, and use the Spatial TOPSIS method to evaluate the vitality of urban neighborhoods in central Nanjing36. Additionally, measurement data for urban vitality includes the use of Weibo check-in data, POI data, and nighttime light data37. These data can reflect the spatial distribution characteristics of urban vitality from different dimensions. These studies indicate that big data provides new perspectives and tools for the quantitative analysis of urban vitality. In particular, the introduction of Baidu Heat Map data in this study, with its high spatio-temporal resolution and real-time nature, offers a new perspective for accurately capturing urban vitality38. The integrated application of these novel data sources not only enriches the dimensions of urban vitality assessment but also improves the accuracy and timeliness of the assessment results.

Indicator system of the built environment

The built environment primarily refers to the human-made environment that has been planned and modified. Research on the built environment has always been a hot topic. The term “built environment” was coined by Amos Rapoport39. Cervero and Kock Elman (1997) defined the built environment as the physical characteristics of the urban landscape and described it based on the “3D” model (density, diversity, design)40. Handy (2002) defined the built environment as a comprehensive pattern composed of human activities, urban design, land use, and transportation systems41. This includes various dimensions such as the urban built environment, regional built environment, residential built environment, and architectural built environment. Cervero expanded the “3D” model by incorporating distance to public transportation stations (destination accessibility) and overall accessibility, forming the “5D” model for describing the urban built environment17. The 5D model provides a more comprehensive framework for understanding the built environment and its impact on urban vitality.

The construction of a built environment indicator system is crucial for understanding and assessing urban vitality. Traditional built environment indicators have mainly focused on two-dimensional spatial aspects, such as Point of Interest (POI) density, road network structure, and land use types40. However, these indicators struggle to fully capture the three-dimensional characteristics of urban spaces and human perceptual experiences. In recent years, with the development of street view image analysis technology, three-dimensional built environment indicators such as street openness, green coverage rate, enclosure, and pedestrian friendliness have gradually gained attention42,43. These indicators not only enrich the descriptive dimensions of the built environment but also enhance the explanatory power of the relationship between the built environment and human behavior. This study will integrate two-dimensional and three-dimensional built environment indicators to construct a more comprehensive indicator system, aiming to deeply explore their impact on urban vitality.

Research progress on relationship mining methods

In studies on the relationship between urban vitality and the built environment, the choice of relationship mining methods is crucial. Traditionally, linear regression has been commonly used to examine the linear relationship between built environment factors and urban vitality indicators44. However, these traditional methods have limitations as they assume a uniform relationship across the entire study area, neglecting spatial heterogeneity. To capture this spatial heterogeneity, the Geographically Weighted Regression (GWR) model has been introduced into urban vitality research45. GWR accounts for spatial variations in the relationships between variables, enabling a more accurate analysis of how the impact of the built environment on urban vitality may vary across different locations, thus revealing the local variation characteristics of the built environment’s influence on urban vitality27. Therefore, this study selects the GWR model as the primary relationship mining method to deeply explore the spatially heterogeneous impact of the built environment on urban vitality.

Data and methodology

Study area

In order to deeply and systematically study the relationship between urban vitality and built environment, we chose Yinchuan City as the study area, the capital of Ningxia Hui Autonomous Region, China (Fig. 2). The study area of this research is the three districts of Yinchuan City. The administrative divisions include 20 towns, 6 townships, 27 neighborhood offices, 293 residents’ committees, and 282 villagers’ committees(https://www.yinchuan.gov.cn/). These three districts are not only the core areas of Yinchuan City but also possess distinct characteristics in terms of socio-economic development. Xingqing District, as a historic downtown area, boasts a thriving commercial sector and a dense population. Jinfeng District, as an emerging commercial and residential area, features modern urban planning and rapid development. Xixia District is renowned for its educational and technological resources, with a wealth of higher education institutions and research organizations. The diversity of these three districts provides rich cases for studying the relationship between urban vitality and the built environment.

Fig. 2
figure 2

Study area.

Full size image

Data sources

As shown in Table 1, the study used 8 types of data, including Geographic information data, NDVI, POI data, Baidu heat map, GDP, Nighttime Lights, Baidu Street view, Open Street map. The basic geographic information data is sourced from the National Geographic Conditions Census (http://www.ngcc.cn/), including administrative boundaries of cities and districts, hydrological data, and road network data. This paper chooses the 2020 precision 30-m NDVI data from the Institute of Geography, Chinese Academy of Sciences. The POI data was crawled from A-map at the end of 2022. The POI database contains 223,595 POIs, categorized into 14 major groups and 81 subcategories. This article utilizes Python to develop a web scraping program, fetching Baidu heat map data every 2 h, and collected a total of 139,684 street view images in the study area. Population indicators and GDP data can be obtained from the 2022 Yinchuan City Statistical Yearbook. The nighttime light data used is the 2022 annual average radiance composite image produced by the NOAA/NCEI Earth Observation Group (https://ngdc.noaa.gov/eog/viirs/).This paper refers to the officially released 2022 NPP-VIIRS nighttime light image. The Baidu Street View data is sourced from Baidu Online Maps (https://map.baidu.com). The road network data is downloaded from the Open Street Map (OSM) open-source mapping platform.

Table 1 Data sources.
Full size table

Model building

This paper further constructs a model to study the influence of urban built environment on urban vitality. Urban vitality indicators are constructed from the 5D variables (Density, Diversity, Design, Distance to transit, Destination accessibility) of the built environment. The dependent variable is urban vitality. The independent variable is the built environment of urban space.

Measurement of urban vitality (dependent variables)

The dependent variable in this paper is urban vitality. It includes NDVI, GDP, population, Baidu heat map and Nighttime Lights data. After variable data were outliers and data standardization, the five dependent variables were weighed by entropy method. The entropy value method is based on the degree of change of each index item in the system, using the information entropy tool to calculate the weight of the five dependent variables46, and finally get the value of urban vitality evaluation. This study utilizes Baidu’s heat map to quantify the density of population activity, employing a color scheme where bright red indicates higher population density and bright blue signifies lower density. Focusing on three districts in Yinchuan city, it analyzes population density data from the Baidu heat map for Wednesday, July 19, and Saturday, July 22, 2023, sampled at two-hour intervals from 8:00 a.m. to 8:00 p.m., yielding a total of 14 data points per day. Using ArcGIS 10.7 software, average population density values for these dates were computed. To assess population mobility, this paper introduces a concept of population spillover, calculated by the difference between the average density on weekdays and weekends, reflecting the dynamic movement of people across urban spaces.

Construction of built environment system (independent variable)

According to the classification code of POI of Gaode map, the Diversity independent variable in this paper selects six POI data such as public service, hotel service, shopping service, food service, traffic service, and landscape service, totaling 25,270 POI points.

Design independent variables are from Baidu Street View images. Baidu Street View was obtained from Baidu map via Python. Elements were extracted from the street view images using DeepLab V3 + and the corresponding ADE20K pre-trained scene resolution dataset for image segmentation. Eventually, a total of 139,684 street view images were collected in the study area. Then, a deep learning based semantic segmentation technique (semantic segmentation) was applied to segment the street scene images into common features. Finally, based on previous studies, we mainly extracted five design metrics including Openness, Greenness, Enclosure, Walkability, and Imageability. The calculation formula is from the Rui’s article47.

In this study, the ‘Distance to transit’ variable is measured by bus station and road density, common indicators of urban accessibility and public transportation efficacy48. The road network data for Yinchuan City in 2022 was sourced from Open Street Map (OSM) and processed using Python for collection and ArcGIS 10.7 for cleaning and simplification.

‘Destination accessibility’ was analyzed with sDNA (Spatial Design Network Analysis), a tool developed by Alain Chiaradia and colleagues at Cardiff University49, operating on the ArcGIS 10.7 platform. The study utilized 0.5 km and 1.2 km radii within the Yinchuan City Road network to represent the range of pedestrian and vehicular movement, respectively. The study selected Network Quantity Penalized by Distance of Angular (NQPDA; The abbreviations “Accessibility-500m” and “Accessibility-1200m” are used in the following paper and tables.) and Two Phase Betweenness (TPBt; The abbreviations “Transportation −  500 m” and “Transportation − 1200 m” are used in the following text and tables.) as key metrics. In sDNA analysis, a higher NQPDA value indicates a greater distance, while TPBt measures the potential for movement through the urban network50. A total of 22 factors were selected as shown in Table 2.

Table 2 Variables formulas and explanations for scores of indices.
Full size table

Relationship mining methods

Regression model

This paper uses a linear regression model. The dependent variable is urban vitality, and the independent variable is the 5D built environment indicator. The global regression uses Ordinary Least Squares (OLS) regression model to measure the impact of 5D variables on urban vitality. OLS reveals the relationship between the dependent and independent variables51.

$$\:y={\beta\:}_{0}+{\sum\:}_{i=0}^{n}{X}_{i}{Y}_{i}+\epsilon\:\:\:$$
(1)

where, y is the dependent variable, and Xi Yi is the independent variable, and ε for the residual difference.

Spatial autocorrelation

Spatial autocorrelation analysis is an indicator reflecting spatial distribution, including spatial global autocorrelation and spatial local autocorrelation. Global autocorrelation analysis can reveal the spatial correlation situation of urban vitality, including three states of discrete, random and agglomeration, which is usually expressed by Moran’s I index. The formula is as follows:

$$\:{I}_{1}=\frac{n}{{\sum\:}_{i=1}^{n}{\left({X}_{i}-\overline{X}\right)}^{2}}\bullet\:\frac{{\sum\:}_{i=0}^{n}{X}_{i}{\sum\:}_{i=0}^{n}{X}_{i}{w}_{ij}\left({X}_{i}-\overline{X}\right)\left({X}_{j}-\overline{X}\right)}{{\sum\:}_{i=0}^{n}{X}_{i}{\sum\:}_{j=0}^{n}{X}_{i}{w}_{ij}}\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:$$
(2)

Local spatial autocorrelation analysis can further explain the discrete or agglomeration status of urban spatial vitality on a local spatial scale. It is usually represented by LISA agglomeration map. The formula is as follows:

$$\:{I}_{2}=\frac{n\left({X}_{i}-\overline{X}\right){\sum\:}_{j=1}^{n}\left({X}_{i}-\overline{X}\right)}{{\sum\:}_{i=1}^{n}{\left({X}_{i}-\overline{X}\right)}^{2}}$$
(3)

where: n is the number of cities; xi and xj are the spatial vitality of the ith and jth city, respectively; x is the average value of the vitality of urban tourist places; wij is the spatial weight matrix. When − 1 < I < 0, there is a negative correlation and a discrete distribution; when I = 0, there is no correlation and a random distribution; when 0 < I < 1, there is a positive correlation and a clustered distribution.

GWR model

The spatial heterogeneity of the individual variables of the Ordinary Linear Regression (OLS) model cannot be captured. To make the results obtained more realistic, this paper uses Geographically Weighted Regression (GWR) to analyze the results. This model is a local regression model in which the geographic coordinates of the sample to be studied are added to the regression parameters52. It can capture the spatial heterogeneity of the impact of the respective variables on urban vitality. It has a better effect than the OLS model. The specific expression of the GWR model is shown in Eq:

$$\:y={\beta\:}_{0}\left({a}_{i}{\:,b}_{i}\right)+{\sum\:}_{i=1}^{n}{\beta\:}_{k}\left({a}_{i}{\:,b}_{i}\right)+\:{\epsilon\:}_{i};i=\text{1,2},\dots\:n$$
(4)

Results and analysis

Measurement results of urban vitality

The calculation of urban vitality value is based on the entropy weight method, which can objectively determine the weight of each indicator, and then reflect the degree of contribution of each indicator to the overall urban vitality. Next, the entropy weight method is used, as shown in Table 3, to determine the weight of each dependent variable. Finally, according to the results of the entropy weighting method, the weight of NDVI is 3.243%, the weight of GDP is 26.699%, the weight of population is 41.784%, the weight of nighttime lights is 26.61%, and the weight of Baidu’s calorific value is 1.663%. These weights reflect the relative importance of each indicator in measuring urban vitality.

Table 3 Entropy weighting.
Full size table

The Urban Vitality Distribution map in Fig. 3a indicates a concentration of vitality in the central districts, with the highest levels in Xingqing District, particularly in central areas like Yuhuangge North Street and Zhongshan South Street. Jinfeng District follows, with its vitality centered around West Shanghai Road and Middle Beijing Road. Xixia District exhibits the least urban vitality. In this paper, we use the difference between the average heat value on weekends and weekdays to represent the distribution of people flow density. The results from the distribution of Baidu heat map show in Fig. 3b–d. From the spatial distribution, the flow of people is mainly concentrated in the center of Xingqing District, Jinfeng District and Xixia District, and there is an obvious trend of high frequency and high-density distribution. On weekdays, at 08:00, the flow of people is mainly concentrated in the center of the city where office buildings are concentrated and reach its peak at 10:00 and 16:00. And after 18:00, it starts to gather in the residential area. On weekends, the flow of people is mainly concentrated in the urban area at 8:00 p.m. At 10:00 p.m., the flow of people starts to flow out of the center of the urban area and is distributed in all areas. According to the analysis of population overflow, compared with weekdays, it is obvious that on weekends, the flow of people starts to evacuate outside the center of the city. Between 12:00 and 18:00, the flow of people is mainly concentrated in the center of the city where there is a commercial area and the surrounding rural areas with scenic spots. And after 18:00, the flow of people from the center of the city is dispersed to various areas.

Fig. 3
figure 3

Urban vitality distribution.

Full size image

Results analysis of the global regression model

Linear regression is used to study the linear relationship between independent variables and dependent variables. The independent variables are Diversity, Design, Distance to transit, Destination accessibility, which includes 17 independent variables. The dependent variable is Urban vitality. The dependent and independent variables are regressed stepwise and there are four models. The first model was a linear regression of the dependent variable and the Diversity variable. The second model is a linear regression of the dependent variable and the Design variable. The third model is a linear regression of the dependent variable and is Distance to transit. The fourth model is a linear regression of the dependent variable and the Destination accessibility variable.

Table 4 shows the detailed results of the linear regression. The significance p-value of the F-test is 0.000, which presents significance at the level and rejects the original hypothesis that the regression coefficient is 0. Therefore, the model basically meets the requirements. For the covariate covariance performance, the VIF value of variable Public is greater than 7.5, there is a covariate relationship, and the covariate independent variable is removed. All other variables have VIF less than 7. The adjusted R² for the 4D variables are 0.64, 0.15, 0.514, 0.52. The linear regression results show that Landscape, Hotel, Shopping, Openness, Greenness, Imageability, Road, Bus station, Accessibility − 500 m, Transportation-500, and Accessibility − 1200 m have a total of 11 independent variables have a significant effect on urban vitality.

Table 4 Regression model of built environment and urban vitality.
Full size table

Results analysis of the spatial autocorrelation

In this paper, we use the “Moran’s I” tool in ArcGIS software to calculate the Moran’s I index of each variable. As shown in Table 5, the Moran’s I index of Walkability variable is − 0.023, and the Z score is − 0.577, and the P-value is more than 0.5. Therefore, this variable should be removed from the GWR model. The Moran’s I index, Z score, and P value of all other independent variables are greater than zero, and P is less than 0.05, indicating that the variables are more spatially clustered and qualify for the GWR model. The Moran’s I value for most of the variables ranged from 0. 063 ~ 0.828, indicating positive spatial autocorrelation at the p = 0.01 level of significance.

Table 5 Global Moran’s I Index.
Full size table

This study employs Geoda software to delve into the local spatial autocorrelation of urban vitality distribution, producing a LISA cluster map to visualize the data. According to Luc (2010), the LISA cluster map is instrumental in pinpointing clusters of various values, such as high-high, low-low, high-low, and low-high value clusters. The analysis reveals that the local Moran’s I for urban vitality stands at 0.880, indicating a strong degree of local spatial correlation. As illustrated in Fig. 4, high-high value clusters predominantly occupy the central urban areas of Xingqing, Jinfeng, and Xixia Districts, while low-low value clusters are generally found in rural locales. During weekdays, the local Moran’s I for average heat value is 0.786, with high-high value clusters concentrated in the central urban zones of the districts. On rest days, this value slightly decreases to 0.733, yet high-high value clusters remain prevalent in both the central urban and rural areas of Xixia District, along with central Xingqing and Jinfeng Districts.

Fig. 4
figure 4

LISA clustering distribution map.

Full size image

Results analysis of GWR

Moran’s I statistics indicate that there is a significant and strong positive spatial autocorrelation present among most of the dependent and independent variables in the study. In the process of GWR, variables that had a p-value greater than 0.001 were omitted from the analysis. This exclusion resulted in the removal of four variables, after which the GWR was applied to assess spatial heterogeneity. The results show that the GWR model has significant improvements compared to OLS regression. The adjusted R2 for the Diversity variable is 0.909. The adjusted R2 for the Distance variable is 0.833, the adjusted R2 for the Destination variable is 0.746, and the adjusted R2 for the Design variable is 0.166. The R2 and adjusted R2are significantly larger and the residual squares are significantly smaller, indicating that the GWR model has a better goodness of fit. Finally, the Moran’s I value of the residuals in the GWR model is low, indicating high spatial independence. Compared with the OLS regression results, the parameter estimates of GWR showed significant spatial differences. The presence of positive and negative parameters with different proportions of significant values for all variables indicates that the spatial distribution of GWR is uniform.

The analysis of the GWR model sheds light on the spatial relationship between road density and urban vitality. In the central urban areas of Xingqing, Jinfeng, and Xixia Districts, the ‘Road’ variable demonstrates a positive correlation, implying that higher road density correlates with increased urban vitality. Conversely, this correlation turns negative in most of the eastern parts of Xingqing District and the northern and southern regions of both Jinfeng and Xixia Districts, as depicted in Fig. 5a,b. Regarding the ‘Bus station’ variable, a positive correlation with urban vitality is observed predominantly in the central areas of Xingqing, Jinfeng, and Xixia. However, areas with a high concentration of bus stations and heavy pedestrian traffic, typically commercial zones, display a negative correlation. As shown in Fig. 5c–f, the Landscape variable is negatively correlated with urban vitality in Xixia District as a whole. The main reason is that the residential density in this area is not high, and the north side of West Helanshan Road is mainly a university town. Food and shopping variables are strongly positively correlated with Shengli Street in Xingqing District, and negatively correlated with Zhongshan North Street and Yuhuangge North Street. In Jinfeng District, there is a positive correlation in West Shanghai Road and Middle Beijing Road, and a negative correlation in Yuehai Community. This area has a high residential density. The center of Xixia District is positively correlated. The center of Xingqing and Jinfeng districts show negative correlation in areas where people gather densely. Figure 5h,i illustrates that the indicators of Openness, Greenness, Enclosure, Walkability, and Imageability within Design exhibit a positive correlation in the city’s center, whereas this correlation turns negative in the peripheral areas. Figure 5g,j–l demonstrates that in the central area of Yinchuan City, there is a positive correlation between the Accessibility − 500 m and Transportation-1200 variables, while areas with high pedestrian density exhibit a negative correlation.

Fig. 5
figure 5

Spatial distribution of regression coefficients of influencing factors of GWR model.

Full size image

Discussion

The spatiotemporal distribution characteristics of urban vitality

This study utilizes the entropy weight method to integrate multi-source information, including the Normalized Difference Vegetation Index (NDVI), Gross Domestic Product (GDP), population, Baidu Heat Map, and nighttime light data, thereby establishing a multidimensional evaluation system. This approach comprehensively considers urban vitality from various aspects such as the environment, economy, and population activities. Notably, population holds the highest weight of 41.784%, suggesting that areas with higher population densities tend to exhibit stronger vitality. The analysis of Baidu Heat Map provides a detailed spatiotemporal perspective. Spatially, it clearly depicts the high concentration of pedestrian flow in the central areas of Xingqing, Jinfeng, and Xixia districts. This centralized distribution pattern not only reflects the attractive force of urban functional layouts on population activities but also implies the high agglomeration and attractiveness of these areas in terms of economics, culture, and commerce. From a temporal dimension, the dynamic changes in the heat map precisely capture the periodic rhythms of urban residents’ activities6. During weekdays, the peak heat in office areas aligns closely with working hours, followed by an orderly shift to residential areas, vividly showcasing the dual structure of production and life in urban living and its orderly transition. The surge in heat in commercial and entertainment areas during weekends reflects residents’ strong demand and active participation in consumption, socializing, and leisure activities during their leisure time, indirectly reflecting the diverse release and dynamic balance of urban vitality under different time patterns. This aligns generally with the research conducted by Chen (2023)53. Building on this foundation, this paper further deepens our understanding of the spatial interaction mechanisms of urban vitality through population spillover analysis, which reveals the population flow trajectories between different regions and shapes the overall pattern of urban vitality at a macro level.

The impact of the built environment on urban vitality

Linear regression analysis reveals the overall relationship between built environment factors (independent variables) and urban vitality (dependent variable). Through stepwise regression of 17 independent variables within the 5D model (Diversity, Design, Distance to transit stops, Destination accessibility), it was found that Landscape, Hotels, Shopping, Openness, Greenness, Imageability, Roads, Transit stops, 500-m Accessibility, 500-m Traffic, and 1200-m Accessibility have significant impacts on urban vitality. This indicates that the presence of diverse landscapes and shopping facilities enhances the attractiveness of a region, while good road and transit accessibility improves travel convenience. The positive coefficients of variables such as hotels and shopping suggest that an increase in these facilities is associated with higher urban vitality. Studies by Wang found that the density of road intersections has a positive impact on urban vitality20. However, this study found a negative impact of 500-m accessibility. Variables with negative coefficients, like this one, imply that there may be an optimal level of accessibility, beyond which negative effects may arise, possibly due to issues such as overcrowding or traffic congestion.

The Geographically Weighted Regression (GWR) analysis reveals significant spatial variations in the impact of different factors. In the central urban areas of Xingqing, Jinfeng, and Xixia districts, the “Road” variable is positively correlated with urban vitality, indicating that a denser road network in these areas contributes to enhanced vitality. However, in the eastern part of Xingqing district and the northern and southern regions of Jinfeng and Xixia districts, the correlation turns negative, suggesting that the impact of road density on vitality depends on the specific context. The “Bus Station” variable also exhibits complex spatial relationships. While it is positively correlated with urban vitality in central areas, it shows a negative correlation in areas with a high concentration of transit stops and heavy pedestrian flow53,54. This implies that an excessive number of transit stops may lead to congestion, hindering the free flow of people and thus reducing urban vitality. In Xixia district, the overall “Landscape” variable is negatively correlated with urban vitality, which may be attributed to the relatively low residential density and the presence of a university town in the area. The “Food and Shopping” variables have different correlations across different regions. In Xingqing district, they are strongly positively correlated with Shengli Street but negatively correlated with Zhongshan North Street and Yuhuangge North Street. In Jinfeng district, they are positively correlated with West Shanghai Road and Beijing Middle Road but negatively correlated with Yuehai Community. These results emphasize the importance of considering local contexts and spatial heterogeneity when formulating urban planning and development strategies.

Improvement strategy

Based on the distinct characteristics of Xingqing, Jinfeng, and Xixia districts, this paper proposes targeted and multidimensional comprehensive improvement strategies. In the central area of Xingqing District, a multidimensional approach should be implemented to enhance urban vitality. Regarding transportation, it is essential to ensure smooth long-distance connectivity, optimize the layout of transportation facilities, and introduce intelligent traffic management systems to improve operational efficiency. At the same time, historical and cultural relics should be preserved and integrated with modern architectural styles to enhance the city’s image and identifiability. For areas where food and shopping negatively correlate with vitality, commercial formats should be differentiated and upgraded, traffic flow should be optimized, and parking facilities as well as public transportation networks should be improved to boost regional vitality.

In Jinfeng District, vibrant areas such as Shanghai West Road and Beijing Middle Road should continue to pursue high-end and international development strategies. By strengthening cooperation and exchanges with advanced cities both domestically and internationally, high-end commercial, business, and cultural resources should be attracted to enhance the district’s internationalization level and integrated service functions. For less vibrant areas like Yuehai Community, residents’ needs should be thoroughly investigated to optimize public service facilities such as education, healthcare, and culture, thereby enhancing the convenience and attractiveness of the community and boosting overall vitality.

Xixia District should fully leverage the knowledge and talent advantages of the university town to promote the integration of industry, academia, and research, as well as innovative development. Universities and enterprises should be encouraged to collaborate in establishing innovation and entrepreneurship bases and platforms for the transformation of scientific and technological achievements, thereby facilitating knowledge spillovers and industrial innovation and injecting new development momentum. Interaction between the university town and surrounding communities should be strengthened by sharing sports facilities and cultural venues, organizing joint cultural activities, and promoting population exchanges and cultural sharing to create a vibrant community atmosphere. In the central area, commercial service functions should be further enhanced, and public space layouts should be optimized to attract diversified commercial formats and cultural entertainment facilities, thereby creating an integrated service center that combines shopping, dining, leisure, and cultural experiences to enhance the region’s attractiveness and radiation capacity.

Conclusion

This study comprehensively employed various advanced methodologies and abundant data to deeply explore the relationship between the built environment and urban vitality in Yinchuan City, achieving significant results. By integrating multi-source data and applying the entropy weight method, a comprehensive assessment framework for urban vitality was established. Combining linear regression with geographically weighted regression analysis, the study analyzed the impact of the built environment on urban vitality from both macro and micro perspectives, revealing overall dependencies and spatial heterogeneities. The global linear regression provided macro directions for urban planning, while the local GWR analysis offered scientific bases for refined planning. The application of multi-scale analysis effectively compensated for the deficiencies of traditional research in spatial analysis, providing a new paradigm for urban research. These recommendations aim to optimize the built environment through precise policy implementation, stimulate the endogenous driving force of urban vitality, and achieve sustainable urban development. These suggestions hold strong practical significance and operability.

This study acknowledges the limitations of using Baidu heat maps, POI data, and street view images as proxies for urban vitality, given their inherent biases and potential errors. Baidu heat maps may be influenced by data availability and accuracy, POI data may not fully capture urban complexity, and street view images can face issues like occlusion and quality variations. Future research should integrate additional data sources and apply rigorous validation to enhance reliability. Exploring urban vitality from diverse perspectives—such as impacts on industries or demographics—and analyzing long-term trends can deepen understanding. Multidisciplinary approaches combining sociology, economics, and geography are recommended to capture cities’ multidimensional nature. Addressing sustainability and adaptability amid environmental changes is crucial for offering forward-looking urban planning and policy guidance.