The impact of blue-green spatial landscape pattern on population distribution pattern in Chinese cities

Abstract

With urbanization accelerating, the impact of blue-green spaces on population distribution has gained increasing importance. This study uses the Chinese Annual Land Cover Dataset (CLCD) and ArcGIS to examine urban blue-green space patterns from 2000 to 2020. A spatial Durbin model (SDM) is applied to assess how these spaces influence population distribution across Chinese cities. The findings indicate that: (1) Urban blue-green spaces significantly affect population distribution, with disparities in public service resources amplifying the impact; (2) The effects of blue and green spaces vary, with integrated blue-green planning fostering population growth; (3) The influence of these spaces exhibits geographical heterogeneity, attracting more population in eastern regions, while drawing people from neighboring areas in central and western regions; (4) The impact also varies by city size, with larger cities experiencing negative effects from the independent increase of blue or green spaces, but positive effects from combined blue-green spaces, while smaller cities see positive effects from water bodies but negative impacts from green spaces.

Introduction

Understanding the spatial distribution of populations is essential for grasping the dynamics of human-environment relationships¹ and is a central concern in geography². In China, as the working-age population peaked in 2015, urban centers face intensifying competition for talent, a trend expected to accelerate as the overall population peaks³. While many factors influencing population distribution have been identified, the role of urban blue-green spaces, which combine natural and artificial water bodies (“blue spaces”) and green areas (“green spaces”), has not been thoroughly explored⁴. These spaces, including lakes, rivers, and parks^5,6, provide crucial ecological services and recreational opportunities, which can influence population patterns by making urban areas more attractive to residents^7,8. This study aims to fill this gap by investigating how blue-green spaces, integral to the urban ecosystem⁹, impact urban population distribution — a factor that has largely been overlooked in past research^10,11.

Urbanization and its impact on blue-green spaces

Urbanization, closely linked with population migration and aggregation, has rapidly progressed in recent decades¹². This urban expansion has been remarkably fast in China since the new century, leading to significant changes in population distribution patterns. However, this urban growth often consumes substantial land resources, compromising the quality and extent of urban blue-green spaces, vital to ecosystems and resident welfare¹³. These spaces are crucial for environmental health, quality of life, and community welfare¹⁴, and their neglect can adversely affect urban ecological environments and sustainable development¹⁵. The changes in urban blue-green spaces influence population distribution, underscoring the need for enhanced focus on their conservation and planning in global urbanization¹⁶.

Related research, research gaps, and research question

The study of urban blue-green spaces, emerging from traditional ecological spatial exploration¹⁷, initially focused on assessing environmental benefits such as air purification, mitigating urban heat island effects, and biodiversity protection. Research by Vieira et al., 2018¹⁸ highlighted variations in the effectiveness of urban parks in air purification and climate regulation, underlining the importance of thoughtful urban planning and green space design. Studies have consistently shown the critical role of blue-green spaces in mitigating urban heat islands¹⁹ and, despite recognising their importance for biodiversity, research on optimizing these spaces for more effective ecological services is still developing^20,21. Recent years have seen a shift towards examining the socio-economic impacts of blue-green spaces²², including health benefits for residents²³, enhancement of city image²⁴, and influence on real estate development²⁵. Although urbanization and environmental factors have been extensively examined in the literature, the specific role of urban blue-green spaces in shaping population distribution remains insufficiently explored. Therefore, this study seeks to address this gap by investigating the central research question: How do urban blue-green spaces influence the spatial distribution of populations in China, and what underlying mechanisms drive this relationship?

Historically, population distribution patterns have been studied as part of urban modeling literature, with early works dating back to the mid-20th century. One of the foundational models in this area was the Lowry Model²⁶, which used econometric methods to analyze urban systems and population distributions. These initial models were largely based on econometrics and spatial interactions. Over time, however, the field has evolved by introducing more sophisticated tools such as cellular automata and GIS. Cellular automata models, such as those proposed by Batty²⁷ and Waddell²⁸ allowed for the simulation of urban growth dynamics, considering both spatial and temporal factors in population distribution.

Recent academic research on population distribution patterns and their influencing factors provides a foundation for this study. Studies have explored population density²⁹, population centers³⁰, internal population structure³¹, and trends in population agglomeration or dispersion³². Researchers analyzing population distribution often consider natural geographical elements like topography³³, hydrology³⁴, vegetation³⁵, and climate¹⁹, alongside socio-economic factors such as economic development³⁶, transportation infrastructure³⁷, and public services³⁸. Analytical tools used in these studies include multiple regression models^39,40, structural equation models⁴¹, and geographic detectors^42,43.

Despite these advancements, the influence of urban environments, particularly urban blue-green spaces, on population distribution has remained relatively under-explored. While some research addresses natural environmental factors, comprehensive studies integrating remote sensing data for long-term analysis of urban blue-green spaces and their impact on population distribution are scarce. Recent research, however, has begun to examine the role of urban green spaces and water bodies. Studies have suggested that blue-green infrastructure, including parks, rivers, and lakes, improves environmental quality, influencing population density and spatial distribution^44,45. These studies have integrated advanced remote sensing techniques and GIS to map and quantify the blue-green spaces in urban areas, providing a spatially explicit analysis of their impact on population trends^46,47.

Research contribution and significance

This study introduces two main innovations in response to existing gaps in research. Firstly, it systematically analyzes China’s urban blue-green space landscape from 2000 to 2020 using the annual land cover dataset and refined spatial analysis methods. Secondly, it uses the spatial Durbin model to explore the under-researched relationship between urban blue-green landscapes and population distribution. The spatial Durbin Model is a spatial econometric model that includes direct and indirect spatial effects in spatial data analysis. It extends the classic spatial autoregressive model by incorporating spatially lagged variables of both the dependent and independent variables. This model is particularly useful when analyzing the interdependencies between regions or cities. These approaches enhance our understanding of how urban blue-green spaces interact with population distribution, offering valuable insights for urban planning and sustainable development.

Additionally, the study leverages remote sensing technology and novel datasets to examine the interaction between China’s urban blue-green space patterns and population distribution from 2000 to 2020. This macro and in-depth analysis aims to comprehensively understand the mutual influence between blue-green spaces and population distribution. The research has a two-fold value: theoretically, it addresses gaps in the current understanding of the relationship between blue-green spaces and urban population distribution, offering new perspectives and methodologies. Practically, it provides data support and strategic insights for urban planning and management, aiming to optimize the allocation of urban blue-green spaces, enhance urban livability, and promote sustainable urban development.

Theoretical framework

Elucidating spatial attributes, particularly population distribution, is a cornerstone of geographic inquiry and is crucial to understanding the intricate human-environment relationship⁴⁸. This study focuses on China’s increasingly prominent urban landscape, where the interplay between rapid urbanization and spatial restructuring of blue-green spaces profoundly reshapes population distribution patterns.

The conceptual underpinning of this research is rooted in the notion that urban blue-green spaces - a composite of natural and artificial water bodies (blue spaces) and green areas (green spaces) - are not mere passive elements of the urban landscape but active determinants of environmental quality and urban livability. By offering leisure, entertainment, and ecological services, these spaces significantly enhance cities’ attractiveness, thereby influencing urban dwellers’ decision-making process regarding their living spaces⁵.

Central to our exploration is the hypothesis that the configuration and extent of these blue-green spaces exert a discernible influence on the population distribution, both within the cities and in their contiguous regions. This hypothesis dovetails with the broader theoretical premise of spatial effects, which argues that characteristics or phenomena in one area are influenced by and, in turn, influence neighboring areas⁴⁹. This interplay is believed to be mediated by various factors, including but not limited to the natural environment, human activities, and socioeconomic processes⁵⁰.

To substantiate this hypothesis, the study employs a methodological approach that harnesses the SDM’s analytical prowess. This choice predates the model’s ability to capture the spatial interdependencies inherent in the urban landscape and population dynamics⁵¹. The SDM’s utility is further augmented by its integration with data sourced from the CLCD and city population censuses, analyzed through geographic information systems technology, notably ArcGIS.

Moreover, the research goes beyond mere correlation, delving into the causative mechanisms that link urban blue-green spaces with population distribution. It considers various control variables - environmental quality, economic development, and public services - which are pivotal in shaping urban living conditions and, by extension, population distribution patterns⁵². These variables reflect the quality of urban life and serve as key attractors or detractors for urban populations, further complicating the relationship between urban landscapes and their inhabitants⁵³.

Data and methods

Spatial measurement model construction

In geographic spatial research, the spatial effect is an essential consideration⁵⁴. In simple terms, the spatial effect refers to how neighbouring areas influence specific characteristics or phenomena in one area⁵⁵. This influence originates from various causes, including but not limited to the natural environment, human activities, and socioeconomic processes⁵⁶. In this study, the spatial effect is a critical theoretical assumption. We posit that the pattern of a city’s blue-green space landscape impacts its neighboring regions’ population distribution. To test this hypothesis, we have chosen to analyse using the spatial Durbin model. In spatial econometrics, assessing the presence of spatial autocorrelation is a crucial first step in understanding spatial effects. Moran’s Index is a commonly used statistic for this purpose. It measures the degree to which a variable is spatially correlated with its neighbors, helping to determine whether the spatial distribution of a variable, such as blue-green spaces, exhibits patterns of clustering or dispersion. A positive Moran’s Index suggests that cities with similar blue-green space proportions are clustered together, while a negative value indicates a dispersed pattern. This study uses Moran’s Index to assess the spatial autocorrelation of the blue-green space distribution across the 352 cities. The results of this analysis inform the specification of the Spatial Durbin Model, which incorporates spatially lagged dependent and independent variables to explore spatial relationships among cities further. The spatial Durbin model helps us examine the spatial influence of the geographical distribution of urban blue-green spaces on population distribution. The primary model expression is as follows:

$$\:\begin{array}{*{20}c} {y_{{it}} = \alpha \: + k\sum {\:_{{j = 1}}^{n} } w_{{ij}} y_{{jt}} + \beta \:x_{{it}} + \theta \:\sum {\:_{{j = 1}}^{n} } w_{{ij}} x_{{jt}} + u_{t} + \lambda \:_{i} + \varepsilon \:_{{it}} } \\ \end{array}$$

(1)

In this formula, $\:{y}_{it}$ and $\:{x}_{it}$ represent the dependent and independent variables of city $\:i$ in year $\:t$, respectively; $\:{y}_{it}$ represents the total urban population of city $\:i$ at time $\:t$; $\:{y}_{jt}$ is the total urban population of city $\:j$ at time $\:t$; while $\:{x}_{jt}$ is the corresponding independent variable of city $\:j$ in year $\:t$; $\:\alpha\:$ is the intercept term; $\:{\epsilon\:}_{it}$ is the error term; $\:k$ and $\:\theta\:$ respectively represent the coefficients of the spatial lag terms of the dependent and independent variables; $\:{\lambda\:}_{i}$ and $\:{u}_{t}$ represent individual and time-fixed effects, respectively; $\:{w}_{ij}$ is an element in the spatial weight matrix, indicates the weight that reflects the geographical relationship between city $\:i\:$and its neighboring city $\:j$; and $\:n$ represents the number of cities.

To ensure the robustness of the results and reduce potential randomness, we conduct diagnostic tests for multicollinearity, such as the VIF test, to assess the potential correlation between the independent variables. Additionally, we perform significance tests for the spatial lag terms to confirm the spatial interdependencies between cities.

It should be added that “neighbouring areas” refers to other cities geographically adjacent to the city. Conversely, “local” pertains specifically to the city itself. Given the scale of this study, which involves 352 cities across China, “local” consistently refers to the individual city under investigation.

To avoid the randomness that might result from the regression of a single weight matrix and to test whether the compression of space and time changes the impact of the blue-green space pattern on the population distribution pattern, this study sets up three spatial weight matrices: geographic distance weight $\:{W}_{1}=\raisebox{1ex}{$1$}\!\left/\:\!\raisebox{-1ex}{${d}_{ijgeo}$}\right.$, where $\:{d}_{ijgeo}$ represents the geographic distance between cities $\:i$ and $\:j$; economic distance weight $\:{W}_{2}=\raisebox{1ex}{$1$}\!\left/\:\!\raisebox{-1ex}{${d}_{ijeco}$}\right.$, where $\:{d}_{ijeco}$ represents the absolute value of the difference in actual per capita GDP between cities $\:i$ and $\:j$; and service distance weight $\:{W}_{3}=\raisebox{1ex}{$1$}\!\left/\:\!\raisebox{-1ex}{${d}_{ijser}$}\right.$, where $\:{d}_{ijser}$ represents the absolute value of the difference in actual per capita public service resources within the cities $\:i$ and $\:j$. The per capita public service resources are obtained by taking a weighted average of the three library, teacher, and doctor indicators we construct in this study.

Spatial effect decomposition

We employ a partial derivative method to decompose the total effect of blue-green space patterns on population distribution to reduce potential bias from point estimation in the spatial regression model. This approach helps us isolate the direct and indirect effects, ensuring a more precise understanding of the relationships between the variables and mitigating the risk of randomness in the regression results⁵⁷. Consequently, the above SDM can be rewritten as:

$$\:\begin{array}{*{20}c} {Y = \left( {I - kW} \right)^{{ - 1}} \left( {X\beta \: + WX\theta \:} \right) + R} \\ \end{array}$$

(2)

In this equation, $\:Y$ represents the dependent variable matrix, which contains the values of the population distribution across cities; $\:I$ is the identity matrix of dimension $\:n\:\times\:\:n$, where $\:n$ is the number of cities in the study. It is used as part of the transformation applied to the spatial weight matrix; $\:k$ is a scalar coefficient that represents the spatial lag coefficient of the dependent variable, capturing the effect of neighboring cities on the city’s population; $\:X$ represents the matrix of independent variables that influence the population distribution, such as urban green space, blue-green space, and other socio-economic factors; $\:\beta\:$ is the vector of coefficients corresponding to the independent variables in matrix $\:X$. Each $\:\beta\:$ represents the impact of the corresponding independent variable on the population distribution; $\:WX$ represents the spatially lagged independent variables, accounting for the effect of neighboring cities’ independent variables on the population distribution of the target city; $\:\theta\:$ is the vector of coefficients corresponding to the spatially lagged independent variables in WX. These coefficients measure the extent to which the characteristics of neighboring cities influence the city of interest; $\:R$ is the residual term, including the intercept and error terms. For the independent variable $\:X$, the partial derivative matrix of its corresponding dependent variable $\:Y$ can be written as:

$$\:\begin{array}{*{20}c} {\left[ {\frac{{\partial \:E\left( Y \right)}}{{\partial \:x_{{1k}} }} \ldots \:\frac{{\partial \:\left( Y \right)}}{{\partial \:x_{{Nk}} }}} \right] = \left[ {\begin{array}{*{20}c} {\frac{{\partial \:E\left( {y_{1} } \right)}}{{\partial \:x_{{1k}} }}} & { \cdots \:} & {\frac{{\partial \:E\left( {y_{1} } \right)}}{{\partial \:x_{{Nk}} }}} \\ {\: \vdots } & {\: \ddots \:} & \vdots \\ {\:\frac{{\partial \:E\left( {y_{N} } \right)}}{{\partial \:x_{{1k}} }}} & { \cdots \:} & {\frac{{\partial \:E\left( {y_{N} } \right)}}{{\partial \:x_{{Nk}} }}} \\ \end{array} } \right] = \left( {I - kW} \right)^{{ - 1}} \left[ {\begin{array}{*{20}c} {\beta \:_{k} } & { \cdots \:} & {\omega \:_{{1N}} \theta \:_{k} } \\ \vdots & { \ddots \:} & \vdots \\ {\:\omega \:_{{N1}} \theta \:_{k} } & {\: \cdots \:} & {\:\beta \:_{k} } \\ \end{array} } \right]} \\ \end{array}$$

(3)

In this equation, $\:E\left(Y\right)$ represents the expected value of the dependent variable matrix $\:Y$, which is the anticipated population distribution across cities; $\:{x}_{1k}$ to $\:{x}_{Nk}$, these represent the individual elements of the $\:k$-th independent variable in matrix $\:X$, where $\:k$ refers to the specific variable being analyzed and 1 to $\:N$ are the individual values for each city or observation; $\:{\beta\:}_{k}$ is the coefficient associated with the $\:k$-th independent variable in matrix $\:X$, measuring the effect of that independent variable on the dependent variable, $\:Y$; $\:{\theta\:}_{k}$ is the coefficient associated with the spatially lagged $\:k$-th independent variable, indicating the influence of $\:k$-th variable’s spatially lagged version on the dependent variable; $\:{\omega\:}_{1N}$ and $\:{\omega\:}_{N1}$ represent the off-diagonal elements of the matrix, capturing the interaction between the $\:k$-th independent variable and the spatial lag terms for different cities. These terms indicate how the spatially lagged independent variables interact with each other across cities; $\:{\left(I-kW\right)}^{-1}$: This term is the inverse of ($\:I-kW$), where $\:kW$ represents the spatially lagged matrix. This transformation accounts for the spatial dependence between cities in the model.

Direct effect

The mean of the elements on the right-hand matrix diagonal represents the direct effect, indicating how changes in the independent variables in the same city (city $\:i$) affect its population distribution.

Indirect effect

The mean of the elements off the diagonal represents the indirect effect, which reflects how the independent variables in neighboring cities ($\:j$) influence the population distribution in city$\:\:i$.

Variable setting and data source

Dependent variable

The focus of this study is the pattern of urban population distribution. To quantify the population distribution accurately, the total urban population ($\:pop$) is the primary indicator. The total urban population can intuitively reflect the differences in population distribution among cities, making it crucial for understanding the patterns of urban population distribution⁵⁸. Notably, the data in this study is sourced from the population census of each city. Precise analysis of these data yields the total population of each city. The meticulous and comprehensive nature of census data enables us to understand the pattern of urban population distribution more accurately, thereby better exploring the relationship between urban blue-green spaces and population distribution.

Core independent variables

In this study, blue-green spaces refer to the combined proportion of blue and green spaces within a city’s boundary. Specifically, blue-green spaces are calculated as the sum of the proportions of blue (e.g., rivers, lakes) and green (e.g., parks, forests) spaces. This definition ensures that each city’s total proportion of blue-green spaces reflects both water bodies and green areas.

This study considers the blue landscape pattern, the green landscape pattern, and the blue-green landscape pattern as the core independent variables. The selection of these core independent variables is grounded in several key considerations, with detailed information on the core independent and control variables constructed in this paper provided in Table 1.

Urban ecological significance

Including blue and green landscape patterns as core independent variables is predicated on their integral role in urban ecosystems⁵⁹. These spaces, encompassing a spectrum from aquatic bodies to verdant areas, are not mere aesthetic elements but are crucial in dictating environmental quality and biodiversity, thus shaping the ecological fabric of urban areas.

Urban attractiveness and habitability

These variables are also selected based on their influence on urban livability. Green spaces, which provide health and recreational benefits, and blue spaces, which contribute to leisure and psychological well-being, collectively enhance urban attractiveness⁶⁰. These spaces serve as pull factors in residential choice-making, influencing population distribution.

Spatial dynamics in urban planning

The spatial configuration and accessibility of these blue-green spaces are of paramount importance in urban planning discourse. Their distribution patterns influence urban population dynamics, including dispersion and agglomeration⁶¹.

Blue landscape pattern

This variable represents the city’s spatial distribution of water bodies. Water spaces, including rivers, lakes, reservoirs, etc., are vital to urban ecosystems and significantly impact population distribution and urban development. This study measures the blue landscape pattern using the proportion of urban water space.

Green landscape pattern

This variable reflects the spatial distribution of vegetation in the city. Urban green spaces, including parks, green spaces, and forests, are critical to urban ecosystems and significantly impact urban residents’ quality of life and health. The green landscape pattern, therefore, plays a vital role in influencing population distribution. This study uses the proportion of urban green space to measure the green landscape pattern.

Blue-green landscape pattern

This variable represents the city’s broad distribution of blue-green spaces. Blue-green spaces are significant components of urban ecosystems and significantly impact urban ecological functions, living environments, and population distribution. Here, the study uses the proportion of blue-green space to measure the blue-green landscape pattern.

The specific calculation formulas are as follows:

$$\:\begin{array}{*{20}c} {blue_{{it}} = \frac{{\sum {\:_{{m = 1}}^{M} } B_{{im}} }}{{All_{i} }}*100\% } \\ \end{array}$$

(4)

$$\:\begin{array}{*{20}c} {green_{{it}} = \frac{{\sum {\:_{{m = 1}}^{M} } G_{{im}} }}{{All_{i} }}*100\% } \\ \end{array}$$

(5)

$$\:\begin{array}{*{20}c} {bg_{{it}} = \frac{{\sum {\:_{{m = 1}}^{M} } BG_{{im}} }}{{All_{i} }}*100\% } \\ \end{array}$$

(6)

In the above equations, $\:{blue}_{it}$, $\:{green}_{it}$, and $\:{bg}_{it}$ represent the blue, green, and blue-green landscape patterns within city $\:i$ at time $\:t$, respectively; $\:{B}_{im}$ represents the area of category $\:m$ blue space within city $\:i$; $\:{G}_{im}$ represents the area of category $\:m$ green space within city $\:i$; $\:{BG}_{im}$ represents the area of category $\:m$ blue-green space within city $\:i$; $\:{All}_{i}$ represents the sum of the areas of all land cover types within city $\:i$; $\:M$ represents the total number of land-use categories. Specifically, to compute the proportion of blue-green spaces, the geographical unit used is the city itself. Using remote sensing data, we extracted the areas of blue spaces and green spaces within the administrative boundaries of each city. These areas were then measured and summed to determine the overall proportion of blue-green spaces for each city.

Control variables

The research incorporates a meticulously curated set of control variables to discern the specific impact of blue-green spaces on urban population distribution:

Environmental quality indicators (EN_it)

Climatic Factors: This study acknowledges the pivotal role of climate in determining a region’s habitability. It considers climatic conditions a critical environmental quality indicator, influencing residential preferences and population mobility⁶².

Air Quality Parameters: Given the direct correlation between air quality and public health, this variable is essential in assessing its influence on urban population distribution patterns⁶³.

Economic development indicators (EC_it)

Economic Prosperity and Structure: The study incorporates regional GDP per capita and the proportion of secondary and tertiary sectors in the GDP. These indicators collectively reflect a region’s economic health and structural dynamics, which are key determinants in shaping population distribution through job market trends and implications for living standards⁶⁴.

Public service indicators (PU_it)

Total Public Library Collections per Million People: This indicator reflects a region’s cultural and educational environment, which can be a significant draw for populations⁶⁵.

Number of General Primary and Secondary School Teachers per 100 People: The availability of educational resources influences family settlement decisions and demographic patterns⁶⁶.

Number of Practising (Assistant) Physicians per 100 Population: Access to healthcare is a crucial factor in determining the livability of an area, thereby influencing population distribution⁶⁷.

By carefully selecting these variables, the study aims to comprehensively analyze the factors influencing urban population distribution, focusing on the role of urban blue-green spaces.

Consequently, the spatial Durbin model constructed in this paper to study the impact of the blue-green spatial pattern on population distribution is as follows:

$$\:\begin{array}{*{20}c} {pop_{{it}} = \alpha \: + k\sum {\:_{{j = 1}}^{n} } w_{{ij}} pop_{{\begin{array}{*{20}c} j \\ {\:t} \\ \end{array} }} + \beta \:_{1} blue_{{it}} + \beta \:_{2} green_{{it}} + \beta \:_{3} bg_{{it}} + \beta \:_{4} EN_{{it}} + \beta \:_{5} EC_{{it}} + \beta \:_{6} PU_{{it}} + } \\ {\:\theta \:_{1} \sum {\:_{{j = 1}}^{n} } w_{{ij}} blue_{{jt}} + \theta \:_{2} \sum {\:_{{j = 1}}^{n} } w_{{ij}} green_{{jt}} + \theta \:_{3} \sum {\:_{{j = 1}}^{n} } w_{{ij}} bg_{{jt}} + \theta \:_{4} \sum {\:_{{j = 1}}^{n} } w_{{ij}} EN_{{jt}} + \theta \:_{5} \sum {\:_{{j = 1}}^{n} } w_{{ij}} EC_{{jt}} + \theta \:_{6} \sum {\:_{{j = 1}}^{n} } w_{{ij}} PU_{{jt}} + } \\ {\:u_{t} + \lambda \:_{i} + \varepsilon \:_{{it}} } \\ \end{array}$$

(7)

Detailed information on the core independent and control variables constructed in this paper is shown in Table 1.

Table 1 Explanatory variables and corresponding indicators.

Data description

This study uses 352 cities across China, from 2000 to 2020, as the research sample. The data part of this research comprises two types of data: raster data and panel data. Combining these two types of data enables this paper to explore the connection between urban blue-green space landscape patterns and population distribution comprehensively and in-depth, thereby enhancing the reliability of the research results.

The core raster data comes from three primary sources. First, this paper utilises the China Annual Land Cover Dataset, an almanac created based on Landsat satellite data, which provides annual detailed information on China’s land cover changes⁶⁸. Second, this study employs China’s indigenous high-resolution PM2.5 dataset to obtain precise measurements of the air quality index⁶⁹. Third, the national city average temperature record data is obtained from the National Environmental Information Center (NCEI) of the National Oceanic and Atmospheric Administration (NOAA) of the United States, providing critical information for this paper’s climate condition study (https://www.ncei.noaa.gov).

In analyzing remote sensing data, we allocated information on land covers and blue-green spaces to cities based on their administrative boundaries. Specifically, we imported city-level administrative boundary data and land-use information into ArcGIS. By utilizing ArcGIS’s partitioning and matching functions, we could assign the corresponding land-use and blue-green space data (e.g., water bodies and green spaces) to each city. This procedure ensured that the final dataset accurately reflected the spatial extent of each city’s land-use categories, resulting in a comprehensive panel dataset for subsequent analysis.

For the panel data portion, this paper mainly uses two sources. On the one hand, it cites the “China City Statistical Yearbook.” This annual publication provides abundant and valuable longitudinal statistical data at the city level, further deepening our analysis. On the other hand, it uses data from China’s 5th, 6th, and 7th population censuses to obtain a more accurate distribution across China.

Moreover, considering the issues of dimensional homogeneity and reducing heteroscedasticity, all variables in the empirical model have been standardised. Table 2 shows the descriptive statistics of the variables in this paper.

Table 2 Variable descriptive statistics.

Evolution characteristics of landscape pattern in blue-green space

Distribution pattern of blue-green space

This study uses the blue-green space data from the first Chinese Annual Land Cover Dataset on the Google Earth Engine. This dataset categorises land use types into nine major categories: Cropland, Forest, Shrub, Grassland, Water, Snow/Ice, Barren, Impervious, and Wetland. Forest, Shrub, and Grassland are classified as green spaces, while Water and Wetland are combined into blue spaces, as shown in Fig. 1.

Fig. 1

Distribution of land cover types in China in 2000, 2010, and 2020. Map generated using ArcGIS Desktop 10.8 (ArcMap 10.8, Release Version 10.8, Product Version 10.7.0.10450), available at https://www.esri.com/en-us/arcgis.

Data analysis reveals that, from 2000 to 2020, green spaces consistently dominated China’s land cover types, slightly increasing from 56.62% in 2000 to 56.68% in 2020. This trend indicates that China’s green spaces remained stable during these 20 years. Meanwhile, the proportion of blue spaces remained unchanged during this period, increasing slightly from 1.51% in 2000 to 1.63% in 2020, as shown in Fig. 2.

Overall, from 2000 to 2020, the proportions of blue and green spaces increased, rising from 58.13% in 2000 to 58.30% in 2020. This finding indicates that China’s blue-green spaces have remained stable overall, reflecting China’s efforts and achievements in ecological conservation and environmental governance.

Fig. 2

Spatial distribution of blue-green landscape in China in 2000, 2010, and 2020. Map generated using ArcGIS Desktop 10.8 (ArcMap 10.8, Release Version 10.8, Product Version 10.7.0.10450), available at https://www.esri.com/en-us/arcgis.

Change patterns of blue-green spaces in cities

A clear trend can be observed by comparing the changes in the area of blue-green spaces between 2000 and 2010 and 2010–2020: the number of areas with a decrease in blue-green space significantly increased. Specifically, the ratio of areas with blue-green space expansion to those with contraction dropped from 1.68 between 2000 and 2010 to 0.46 between 2010 and 2020, as shown in Fig. 3. This shift highlights China’s challenges in environmental management and ecological protection. Despite considerable progress in environmental conservation over the past two decades, the advancing processes of urbanisation and industrialisation still exert pressure on blue-green spaces in many areas.

A closer examination of the spatial distribution reveals that, from 2000 to 2010, blue-green space contraction primarily occurred in South China, Southwest China, and Tibet. However, between 2010 and 2020, contraction mainly occurred in Central China, North China, and Northeast China. This change in the distribution pattern reflects the geographic shift in China’s economic development, specifically the move from coastal areas to inland areas, further putting pressure on blue-green spaces in the latter regions.

Based on the changes in blue-green spaces, all prefecture-level units are divided into four types: continued growth, decline and return to increase, stop growing and shrinking, and continuous contraction. There are 65 prefecture-level units of the continued growth type, mainly distributed in the Sichuan-Chongqing region, Shanxi, northeastern Inner Mongolia, and the Hexi Corridor. These areas have maintained growth in blue-green spaces, likely benefiting from specific land management policies, geographical conditions, and economic development models. Additionally, 45 prefecture-level units of the decline and return to increase type are distributed in Xinjiang, Tibet, and Southwest China. The blue-green space in these areas first contracted and then expanded, a phenomenon related to implementing corresponding environmental protection policies and land management measures. However, 155 prefecture-level units of the stop growing and shrinking type and 55 units of the continuous contraction type are primarily concentrated in Central China, East China, and Northeast China. The blue-green space in these areas has continued to contract, primarily due to the impacts of rapid urbanisation and industrialisation, coupled with the ineffective execution of environmental protection policies and land management measures. This complex pattern of blue-green space changes underscores the necessity of further studying the key factors influencing such changes and the importance of formulating local land management strategies and environmental policies that accommodate the characteristics of each prefecture-level unit.

Fig. 3

Growth rate and change pattern of urban blue-green space area in China 2000–2020. Map generated using ArcGIS Desktop 10.8 (ArcMap 10.8, Release Version 10.8, Product Version 10.7.0.10450), available at https://www.esri.com/en-us/arcgis.

Spatial correlation features

Based on each city’s geographical coordinates, a geographic distance spatial weight matrix was constructed, and the global spatial correlation of blue-green spaces was examined using Moran’s I index, as shown in Table 3.

Table 3 Moran’s I value of blue-green spatial landscape pattern in Chinese cities.

Table 3 presents Moran’s I index for the pattern of blue-green spaces in Chinese cities between 2000 and 2020. During this period, Moran’s I indices were all positive, and the index values for all years passed the significance level of 1%, providing evidence of a positive spatial correlation in the pattern of blue-green spaces. This implies that cities with a high level of blue-green space arrangement tend to cluster with other cities geographically, demonstrating a high level of blue-green space arrangement; the distribution of these two factors is not independent in space. Particularly noteworthy is the steady increase in Moran’s I index from 2000 to 2016, indicating a continuous enhancement of spatial correlation. However, in 2017, there was a significant decline in Moran’s I index, which then rapidly rebounded in 2018 and gradually stabilised by 2020. This result underscores that spatial factors cannot be ignored when studying the impact on the pattern of blue-green spaces in Chinese cities. Therefore, a spatial panel econometric model can help avoid bias from overlooking spatial factors.

The impact of urban blue-green space on the population distribution pattern

Benchmark regression and model identification

To avoid spurious regression from non-stationary variables, we first tested for stationarity. The results indicated that all variables were stationary. Based on this, we conducted panel data regression, testing the spatial correlation of the lagged dependent variables and residuals to determine the appropriate spatial econometric model.

Table 4 presents the test results. The model fitting is satisfactory, showing that blue-green space significantly impacts population distribution. The Lagrange Multiplier (LM) and robust LM tests indicate spatial correlation between the lagged dependent variable and residuals. LR and Wald tests confirm that the spatial Durbin model cannot be simplified to a spatial lag or error model. The Hausman test statistic of 86.85, passing the 1% significance level, rejects the random effects model. Therefore, we use the spatial Durbin model with fixed effects for the empirical test. Overall, the blue-green space landscape, environmental quality, economic development, and city public service levels influence population distribution, suggesting a specific correlation between urban blue-green spaces and population patterns.

Table 4 The results of benchmark regression and model identification.

Spatial effect analysis

This study compares the necessity and reliability of spatial Durbin models with different fixed effects by presenting estimation results for spatial fixed effects, time fixed effects, and spatiotemporal double fixed effects, as shown in Table 5. The LR test evaluates the two nested models, comparing the log-likelihood differences to determine which model explains the data more effectively. The results from the LR test show no significant difference between the spatial and spatiotemporal fixed effects models (p-value = 1), while the comparison between the spatiotemporal fixed effects model and the time fixed effects model shows statistical significance (p-value = 0), indicating that the spatiotemporal model is more effective. Thus, the analysis focuses on the SDM with spatiotemporal fixed effects.

The results in Table 5 (columns 4 and 8) show that the coefficient of blue space is negative, while its spatial lag coefficient is positive, both passing significance tests. This suggests that local increases in blue space reduce population distribution, likely due to decreased land availability. However, neighboring areas benefit from the improvement in blue space, attracting more residents and creating a crowding-out effect. The coefficients for green space display similar patterns. The combined effect of blue-green space shows positive coefficients for both local and spatial lag effects, indicating that increases in blue-green space improve local population distribution and attract populations from neighboring areas. This is because blue and green spaces often work together, enhancing the urban ecological environment.

The findings suggest that while increases in blue and green spaces locally may restrict land for construction, their impact on neighboring areas enhances the quality of life and attracts population migration. This highlights the need for urban policies to consider the spatial interaction of environmental resources to improve population distribution and promote socio-economic development. Additionally, the study emphasizes the importance of public service facilities, such as libraries, teachers, and doctors, which positively influence local population distribution and create spillover effects in surrounding areas.

In conclusion, the spatial effect analysis reveals that urban blue-green spaces significantly shape population distribution through local and spatial lag effects. The positive impacts of public service facilities further underline the importance of considering spatial interactions in urban planning to optimize resource allocation, enhance livability, and support socio-economic growth.

Table 5 Regression results of SDM with fixed effects.

Spatial effect decomposition

If the coefficient of the spatial lag term is significantly nonzero, point estimation in the spatial panel Durbin model may bias results. Thus, decomposing the total effects of each variable into direct and indirect effects is necessary. Analysing the regression outcomes from different spatial weight matrices was performed using geographic, economic, and service distances, with effects decomposed via partial differentiation, as shown in Table 6.

Blue and green spaces have significant direct and indirect effects on population distribution, with significant results for blue, green, and combined blue-green across all weight matrices. The service distance matrix shows the most significant effects for most variables. Notably, public service variables like “library,” “teacher,” and “doctor” have higher significance under the service distance matrix, indicating that social public services have a more direct impact on population distribution. The service matrix’s superior interpretability highlights the importance of service resource distribution, whereas geographic and economic distances show less interpretability.

The “blue” variable has a significant negative direct effect, indicating that increased blue space decreases urban population. However, the indirect effect is positive under the geographic matrix, suggesting that blue space influences population distribution indirectly. Similarly, “green” also shows a negative direct effect but has positive indirect effects under the service and economic matrices. For “bg,” the direct effect is positive across all matrices. In contrast, the indirect effect is positive for geographic and service distances but negative for economic distance, indicating the combined effect depends on the spatial weight used.

Control variables such as “library,” “teacher,” and “doctor” show significant direct and spillover effects on population distribution. These resources positively correlate with population distribution and indicate that public services—libraries, education, and healthcare—greatly enhance a city’s appeal. Cities with better service resources attract more residents due to improvements in culture, education, and healthcare.

In summary, the spatial effect decomposition analysis underscores the significance of blue and green spaces in population distribution, with both local and spatial spillover effects. While increased blue and green spaces reduce local population density, they attract populations from neighboring areas. The service distance weight matrix proves most significant, showing that public services like libraries, teachers, and doctors directly influence population distribution more than geographic or economic factors. These findings emphasize the importance of high-quality public services in urban planning and highlight the need to consider environmental and social factors when creating policies for sustainable urban growth.

Table 6 Spatial effect decomposition of SDM with Spatial fixed effect.

Regional heterogeneity and city size test

Grouping by city location

Cities are divided into three groups based on their location: eastern, central, and western. Table 7 examines the influence of blue-green space on population distribution across these regions.

Table 7 Spatial effect decomposition of SDM with Spatial fixed effect by region.

In the eastern region, the direct effects of blue and green spaces on population distribution are negative, reflecting urbanization pressures. However, the indirect effects vary: blue spaces attract populations from other regions, while green spaces cause outflows. Environmental quality indicators like PM2.5 show complex effects, with local increases in PM reducing populations, while increases in neighboring areas attract people. Economic factors reveal that growth in PGDP has negative direct and indirect effects, possibly due to higher living costs, while the secondary industry attracts populations from adjacent areas. Educational resources positively impact population distribution, but healthcare resources show a negative effect, possibly due to uneven distribution or insufficient quality improvements.

In the central region, the direct effects of blue and green spaces are negative, while their indirect effects are positive. This suggests that increasing blue and green spaces locally decreases population, but increases in neighboring regions attract people. Environmental quality indicators show fragile direct effects of PM but positive indirect effects, indicating that air quality deterioration in neighboring areas drives migration. Economic development has positive direct effects but weak indirect effects. Public services, such as educational resources, significantly attract populations, while healthcare resources have a weak effect.

In the western region, blue space has a negative direct effect, while the indirect effect is positive, similar to green space. Both direct and indirect PM effects are positive, suggesting that air quality deterioration attracts populations locally and in adjacent areas. Economic development attracts local populations but decreases neighboring populations. The secondary industry negatively affects both local and neighboring populations, while the tertiary industry has a negative local effect but attracts people from nearby regions. Educational and healthcare resources attract local populations, although healthcare’s indirect effects are weak.

This study reveals a compelling dynamic in urban landscapes: the direct effects of blue and green spaces are predominantly negative, illustrating a crowding-out effect where their expansion limits land availability for residential and commercial use, consequently inhibiting population growth⁷⁰. This finding underscores the inherent land use trade-off in urban development, highlighting the competition between environmental objectives and residential/commercial land needs⁷¹. Intriguingly, the indirect effects of these spaces are positive, demonstrating a spillover effect where enhanced environmental quality in one area can attract populations from neighboring regions, thus beneficially influencing population distribution at a broader spatial scale⁷².

Furthermore, the research identifies the “urban synergy effect of blue-green spaces,” particularly evident in the eastern region. Here, the individual negative effects of blue and green spaces are transformed into a positive impact when combined, underscoring the greater significance of their integrated quality and configuration over isolated spaces^73,74. This synergy effect emerges from the diverse environmental offerings provided by these spaces, effectively catering to a range of needs and enhancing overall attractiveness^75,76.

In summary, blue and green spaces generally negatively impact local population growth, but their indirect effects are positive, attracting populations from neighboring regions. The study highlights the importance of urban synergy in the eastern region, where integrated blue-green spaces have a greater impact than individual spaces. These findings suggest that urban planning should balance environmental conservation with population growth, tailoring strategies to specific regional contexts.

Group by city size

Cities are classified based on their year-end resident populations: large cities (over 3 million), medium cities (1 to 3 million), and small cities (under 1 million). The influence of blue-green space on population distribution is examined for these categories, as shown in Table 8.

Table 8 Spatial effect decomposition of SDM with spatial fixed effect by urban scale.

In examining cities of varying sizes, this study identifies both the “crowding-out effect” and the “synergy effect” of urban blue-green spaces, akin to the phenomena observed across different regions. In large cities, an increase in the individual proportion of blue and green spaces tends to have negative impacts, likely due to high population density and land use pressures⁷⁷. This situation epitomizes the urbanization paradox, where the pursuit of economic growth and urban expansion encounters challenges in environmental protection and ecological balance⁷⁸. However, a holistic increase in blue-green landscapes yields positive effects, indicating that in larger cities, integrated blue-green environments are essential to meet the diverse urban demands, offering ecological services, enhancing aesthetics, and improving residents’ quality of life.

In medium-sized cities, the negative impacts of blue and green spaces suggest that intense natural environments can impede urban development efficiency, reflecting the rapid industrialization and urbanization stage these cities typically undergo. The focus in these cities leans more towards industrial and residential development than green spaces or water bodies. The need for a balanced approach becomes evident, as excessive industrialization and urbanization can lead to environmental stress⁷⁹.

Conversely, in small cities, water bodies significantly improve quality of life, but an overabundance of green spaces may limit urban growth. Blue-green spaces can foster leisure and recreation in these cities, where population and development pressures are lower. However, excessive green space could impede growth in essential urban functions like commerce and transportation.

In large cities, our study finds that temperature (“tmp”) and air quality (“PM”) variables have direct positive and negative effects, respectively, suggesting that warmer January temperatures are beneficial, while poor air quality hinders urban development⁸⁰. Additionally, variables representing cultural, educational, and medical resources (“library,” “teacher,” “doctor”) show direct positive impacts, underscoring the significance of these resources in large urban areas. Medium-sized and small cities exhibit similar patterns regarding these controlled variables.

In conclusion, blue and green spaces affect population distribution across cities of different sizes. Large cities experience “crowding-out” effects due to land constraints, while integrated blue-green spaces have positive impacts. Balancing industrial and environmental development is crucial in medium-sized cities, and in small cities, water bodies play a key role in enhancing the quality of life. These findings underscore the need for tailored urban policies that address the specific environmental, economic, and social dynamics of each city size.

Robustness check with alternative blue-green variables

To verify the consistency and reliability of the baseline regression results, we conduct a robustness analysis by replacing the core explanatory variables with alternative measures of green and blue space. Specifically, we use the NDVI and total water resources as proxies for green and blue areas, respectively. Table 9 reports the SDM estimates using these alternative variables.

Table 9 Spatial effect decomposition of the SDM by region.

Table 10 Spatial effect decomposition of the SDM by city size category.

Furthermore, Table 10 presents direct and indirect effects by city size. Here the crowding-out vs. spillover pattern holds, but coefficients now vary more distinctly:

The robustness analyses demonstrate that substituting our original green and blue space indicators with NDVI and total water resources leaves the principal findings intact. Specifically, both NDVI and water resources continue to exhibit significant negative direct effects—reflecting local “crowding-out”-while their spillover effects remain positive, drawing population from adjacent municipalities. The NDVI×Water interaction term retains its strong, synergistic impact in large and Eastern cities, yet its magnitude diminishes or even reverses in smaller or Western locales. This heterogeneity highlights the necessity of customizing integrated green–blue infrastructure strategies to regional and urban-scale contexts. By showing that these alternative proxies reproduce the same qualitative spatial patterns, we reinforce the credibility and robustness of our original SDM results.

Discussion

This study unveiled distinctive phenomena in urban blue-green space that impact population distribution, particularly in Chinese cities. Our findings illuminate the “urban crowding out effect,” where an expansion in blue and green spaces paradoxically decreases population distribution due to the consequent reduction in land available for residential and commercial use, a concept resonating with the trade-offs in urban land use discussed by Kronenberg et al., 2021¹⁵​​. However, this study extends beyond by uncovering the positive indirect effects of these spaces, suggesting their role in attracting populations from neighboring areas, an aspect less emphasized in prior research.

Furthermore, the discovery of the “urban synergy effect” in the eastern region of China, where the combined presence of blue and green spaces positively impacts population distribution, marks a significant advancement over earlier works like those by Yang et al., 2020⁹, which highlighted the ecological and aesthetic merits of these spaces without delving into their combined effects on population distribution​​.

The methodological approach of employing various spatial weight matrices to analyze the impact of urban blue-green spaces on population distribution has also proven to be a novel contribution of this study​​. This nuanced analytical framework has yielded insights that surpass the more traditional geographic-focused studies, such as those by Xu et al., 2021³³ and Liu et al., 2021¹³, which predominantly concentrated on natural geographical factors influencing urban development.

Notably, our research has also identified locational and city size heterogeneity in the impact of urban blue-green spaces on population distribution patterns, thus building upon and adding complexity to the findings of researchers like Liu et al., 2020⁴¹ and Chi, G., 2012³⁰ who examined population distribution patterns but did not explore the varying impacts based on urban location and size​​.

This study’s originality lies in its comprehensive temporal and spatial analysis, utilizing panel and remote sensing data across two decades. Incorporating the spatial Durbin model for assessing the impact of blue-green spaces on population distribution presents an innovative methodological advancement over traditional approaches. Furthermore, the empirical examination of urban blue-green space allocation under urbanization and land use pressure offers new insights into urban planning and landscape management​​.

This study explores the relationship between blue-green urban landscapes and population distribution in China from a spatial effects perspective. It complements the theoretical framework of blue-green spaces’ impact on population distribution and landscape planning for high-quality urban development. It empirically examines how the allocation of blue-green spaces influences population distribution under urbanisation and land use pressure. This promotes further research on the theoretical proposition of how blue-green space landscape planning affects urban population distribution and provides empirical support for how cities of different sizes in China can attract and retain populations through optimised blue-green space configurations.

Beyond the Chinese context, our findings hold broader international significance. Blue-green infrastructure is globally recognized as vital for sustainable urban development, aligning with international frameworks like the Sustainable Development Goals. Many cities worldwide are incorporating nature-based solutions into their spatial planning to enhance urban resilience and livability, and our results can inform this global discourse. The “urban crowding out effect” identified in our study underscores a universal challenge: dedicating land to parks and waterways, while environmentally beneficial, may reduce space for housing and commerce—a trade-off likely to emerge in other densely populated cities as well. Conversely, the “urban synergy effect” suggests that strategic integration of blue and green spaces can yield cumulative benefits in attracting and retaining urban populations, highlighting a planning approach that cities around the world could consider to foster sustainable urban growth. Thus, although our analysis is rooted in China, the insights contribute to international discussions on nature-based solutions and urban spatial planning, offering evidence that could guide the optimization of blue-green infrastructure in diverse urban contexts.

However, this study still has some limitations. Firstly, the study has not thoroughly investigated the specific mechanisms through which blue-green landscape patterns influence population distribution. Secondly, the research has not detailed the differential impacts of different types of blue-green spaces in cities of different sizes. Thirdly, the measurement of blue-green urban landscapes mainly relies on the proportion of blue-green landscape facility types within the cities, which is a relatively simple method and overlooks many details of the blue-green landscape patterns, such as connectivity, facility quality, spatial accessibility, and diversity. Additionally, the spatial interaction effects in this study are challenging to fully quantify due to the large number of cities with significant geographic variation in size. While this study used three distinct spatial weight matrices (geographic, economic, and service), defining a clear geographical threshold for the spatial decay effect is difficult, which would be more feasible in studies focused on smaller or more homogeneous areas.

Moreover, this study has not sufficiently examined localized population density changes near blue-green spaces. While the study has explored the overall spatial effects of blue-green spaces, it does not explicitly address how population density increases or decreases at varying distances from these spaces. This could be particularly important in understanding the localized effects where population densities might increase around BG spaces but sharply decrease further away. Given the large scale of the study, which covers 352 cities, it is challenging to define specific proximity thresholds for such effects, especially considering the varied sizes and urban structures of the cities involved.

Therefore, future research should delve deeper into the underlying mechanisms and spatial dynamics of blue-green spaces, including examining how localized population densities respond to the proximity of blue-green spaces. Additionally, analyzing the impacts of different types of blue-green spaces on urban population distribution and exploring how these effects differ across cities of varying sizes would further enhance the understanding of these dynamics. Future studies could also adopt more sophisticated and comprehensive methods for measuring blue-green landscape patterns, such as incorporating connectivity, spatial accessibility, and facility quality. Moreover, focusing on smaller, more homogeneous regions would allow for a more accurate examination of spatial decay effects, providing clearer insights into how spatial interactions diminish with distance and enabling more explicit analyses of the effects of blue-green spaces on population distribution at finer spatial scales. This approach could significantly refine the current findings and offer more actionable insights for urban planning and policy development.

Conclusion

This study is based on panel data and remote sensing data from 352 cities in China from 2000 to 2020. Firstly, ArcGIS 10.8 was used to identify and extract the blue-green spaces in the cities. Then, various spatial weights were constructed, and the spatial Durbin model was employed to empirically analyse the impact of urban blue-green spaces on population distribution patterns, considering the heterogeneity of urban location and size. The study found the following:

• (1) Considering different spatial weights, urban blue-green spaces significantly impact population distribution patterns, and this impact becomes more pronounced when considering the distribution disparities of public service resources among cities. Specifically, regardless of the distance standard used, the estimated coefficients for blue, green, and the combination of blue-green passed different degrees of significance tests. Among the three weight matrices, the service distance weight matrix demonstrated more significant direct and indirect effects in most variables. This is because the service distance weight matrix can reflect the distribution disparities of service resources among cities, which directly correlate with population distribution. Although geographic and economic distances have specific effects, they are slightly inferior in explanatory power.

• (2) Blue and green spaces, separately and in integrated configurations, have different paths of impact on population distribution patterns. In all the spatial weight matrices, the direct effects of blue and green spaces are significant and negative, indicating that increasing blue and green spaces in cities leads to decreased population. However, the indirect effects vary in weight matrices, suggesting that these spaces have different indirect paths of influence on population distribution. In contrast, the direct effects of blue-green spaces are significant and positive, revealing that an overall increase in blue-green spaces triggers population growth in cities. However, the indirect effects are positive under the geographic and service distance weight matrices and negative under the economic distance weight matrix, indicating that the comprehensive effects of blue-green spaces on population distribution are constrained by economic distance.

• (3) The impact of urban blue-green spaces on population distribution patterns exhibits locational heterogeneity. This study identifies two observed phenomena: the “crowding-out effect of urban blue-green spaces” and the “synergy effect of urban blue-green spaces.” Increasing individual blue or green spaces in the eastern region leads to a decrease in population. However, an increase in the overall proportion of blue-green spaces attracts more population, highlighting the importance of the synergy between green and blue spaces. Additionally, environmental quality improvements and public services contribute to population attraction. In the central and western regions, an increase in blue-green spaces leads to a decrease in population locally. However, it attracts population in neighbouring areas, indicating the crowding-out effect of urban blue-green spaces.

• (4) The impact of urban blue-green spaces on population distribution patterns exhibits heterogeneity based on city size. In large cities, increasing the proportion of individual blue and green spaces has negative effects, likely due to high population density and land use pressure. This reflects the contradiction between economic development and environmental protection in urbanisation. However, an overall increase in the proportion of blue-green spaces has positive effects, indicating that comprehensive blue-green spaces play a positive role in providing ecological services, enhancing urban aesthetics, and improving quality of life. However, the indirect effects are positive under the geographic and service distance weight matrices and negative under the economic distance weight matrix, indicating that the comprehensive effects of blue-green spaces on population distribution are constrained by economic distance.