Introduction

The catering industry is an essential part of the service sector in cities. Its growth is a reflection of a city’s culture and vitality, and its prosperity would improve the service level and competitiveness of cities (Osimani and Clementi 2016). As a fundamental geographical unit within the catering industry, restaurants, along with their types and spatial distribution, have a direct impact on a city’s spatial structure, economy, culture, openness, and vibrancy. These attributes of restaurants serve as a critical reference for urban planning and business location (Dong et al. 2019), also for analyzing the development status of a city from the perspectives of economic geography, spatial geography, cuisine culture, and social problem (Schiff 2015; Lee et al. 2016; Lu et al. 2020), etc.

Economic development and urbanization increase the proportion of urban residents, which further boosts the development of catering industry as a basic service sector in cities. According to Maslow’s hierarchical theory of needs, people incline to pursue higher needs after satisfying the most basic physiological need for food provided (Maslow 1987). With an increase in income and consumption capacity, people begin to adjust their dining preferences based on the variety, price, and proximity of restaurants. Eckert and Vojnovic (2017) found that residents with lower incomes may be more influenced by nearby dining locations than residents with higher incomes. Prayag et al. (2012) used GIS techniques to assess the evolution of restaurant locations and found a clustered pattern within a short walking distance from the CBD and spillover effects emerging in the northern part of a city in Hamilton, New Zealand. Dong et al. (2019) found that the geographic concentration and the size of the population led to diversity in the type of restaurants. Mossay et al. (2022) found that an increase in the population density in U.K. city centers was associated with a decrease in the spatial dispersion of both top-and bottom-tier restaurants. The aggregation characteristics of restaurants often arise from the merging of diverse consumer groups (Kim et al. 2020; Zhang et al. 2021b). Han et al. (2024) utilized the spatial co-location pattern mining to capture the spatial correlation of specific restaurant to determine the candidate location selection range. In addition, the distribution characteristics of restaurants with varied tastes (Minner and Shi 2017; Zhang et al. 2021a;), the development of sweet food in specific cities (Chen et al. 2022; Li et al. 2022; Yen et al. 2020; Zhang et al. 2021b; Walter et al. 2022; Wang et al. 2022), and the impact of catering industry on urban form (Shi et al. 2018; Shi et al. 2021) and urban population distribution have been extensively studied. (Xi et al. 2014; Xia et al. 2018; Li et al. 2022). In recent years, due to the globalization of economy and culture, food consumption is not only an economic activity, but also a cultural phenomenon (Zhang et al. 2021c). Economic, social and cultural globalization has had a significant impact on food culture in all regions (Goto et al. 2014; Tricarico and Geissler 2017; Tian et al. 2021; Yigit 2022). What kind of distributional characteristics and relationships exist between local and foreign foods in different regions from the geographical perspective? How is reliable data obtained to explore the integration of foreign and local food cultures in the process of globalization? These challenges are generally encountered by researchers in the fields of geography and GIS.

Improved living standards encourage people to pursue spiritual and cultural needs to realize their societal, respect, or self-realization needs. Tourism can realize people’s third level of social needs and fifth level of self-realization needs. The catering industry, along with the accommodation industry, constitutes one of the three fundamental pillars of tourism. In the context of consumption upgrading, it has played a pivotal role in driving the development of tourism (Zhu et al. 2021; Li et al. 2022; Pilis et al. 2022). People from different regions exhibit diverse preferences on tastes, styles, and ingredients owing to differences in climate, economy, and culture. Then form distinct food culture spaces to a certain extent (Morgan and Sonnino 2010). These unique food culture spaces promote people’s pursuit of gastronomy through tourism. The locational sense and politics of food and the relationship between food and tourism have become popular topics (Cohen and Avieli 2004; Mak et al. 2012; Zeng et al. 2012; Floyd 2013; Boesveldt et al. 2018; Ray 2019; Zhong and Moon 2020).And local gastronomy has established itself as one of the key elements for the enhancement, sustainable and consolidation of tourist destinations (Zhang et al. 2022). For example, Jiménez-Beltrán et al. (2016) found that importance for local gastronomy to tourists through on the relation between the gastronomy, culture and tourism in city of Cordoba, Spain. And get the satisfaction of traveling by tasting the local gastronomy. Even one kind of food in a city can drive tourism for the whole city, like in China’s Chengdu gastronomy, Guangzhou gastronomy and so on.

A comprehensive study on the distribution of catering industry and its influencing factors for a certain country can reflect its economy, culture, openness and development vitality. As a multi-ethnic country, China is experiencing rapid urbanization and increasing openness to the outside world. Local and foreign restaurants are widely distributed in Chinese cities, and various catering culture exchanges and integration. The types and distributions of the catering industry differ not only between the east and west but also between the south and north of China. In this context, we are keen to investigate the spatial distribution characteristics of Chinese urban catering culture, as well as the distribution patterns and interrelationships between local and international restaurants in different regions. Additionally, we aim to explore the correlation between these distribution characteristics and regional economic development levels, individual consumption power, and population urbanization, with a focus on identifying the underlying driving factors. More specifically, we seek to uncover the hidden elements influencing the relationship between these factors. We assume that the number of urban restaurants is mainly affected by the level of economic development and the number of urban population. Furthermore, the types of restaurants, in addition to these two factors, are also affected by the degree of openness of a city, living habits of residents and other factors. The spatial distribution of restaurants is comprehensively affected by additional driving factors, such as the type of city (e.g., tourist cities). Now, the impact of these factors on the catering industry is still vague and lacks statistical hypothesis testing. This study will address this gap and provide guidance for the spatial layout of the catering industry, aiding local governments in strategic decision-making for the tertiary industry.

Data and methodology

Data sources and data processing

With the advent of the Geospatial big data era, mounting confidences are built on the research of urban Geospatial entities. The electronic map point of interest (POI) has information on geographical identification, including name, category, latitude and longitude, which are critical for investigating the spatial distribution form and structure of the research object. For this study, POI data for restaurants in March 2020 was obtained from the AMAP database, consisting of over 6.09 million POIs (https://lbs.amap.com/api/webservice). The spatial pattern reflected by this data is not affected by the COVID-19 pandemic, and it can truly mirror the catering consumption situation in China. After thorough data screening and filtering, a total of 4.49 million efficient POI data points were obtained. Then POI data were classified into 4 first-level categories (Chinese restaurants, CRs; Western restaurants, WRs; fast-food restaurants, FFRs; dessert and drink restaurants, DDRs). The CRs represent the traditional cuisine, which are preferred by most Chinese taste and possess regional characteristics. The presence and distribution of WRs imply the openness, internationalization, and diversification of a specific area. FFRs can satisfy the fast-paced life needs of people, and are characterized as convenience and low prices. DDRs’ distribution density is affected by the regional diet culture and life pace. Based on their geographic coordinates, these restaurants are located in 336 cities across China. Due to data unavailability, this study excluded Taiwan, Hong Kong, and Macau regions.

To elucidate the spatial distribution pattern of restaurants and the driving forces, three types of influencing factors were included, i.e., population, economy, and social development. A total of ten socio-economic indicators were collected from the China Statistical Yearbook 2021, including urban population, population density, urbanization rate, GDP per capita, GDP per unit area, the proportion of tertiary industry added value in GDP, the proportion of non-agricultural industry added value in GDP, total retail sales of consumer goods, the per capita disposable income of urban households, and total tourism revenue (Table 1).

Table 1 The influencing factors for three types (population, economic, social development) of city.
Full size table

Methodology

To analyze the spatial distribution characteristics of national catering facilities, we employed the kernel density estimation method, and then adopted the global and local spatial autocorrelation analysis in the exploratory spatial data analysis method (Osullivan and Pawitan 1993; Anselin 1999; Messner et al. 1999) to comprehensively analyze the spatial agglomeration characteristics of national catering facilities of the 4.49 million efficient POI data points. The geographical weighted regression (GWR) model (Brunsdon et al. 1996) was used to explore the spatial distribution of restaurants and their influencing factors.

Kernel density estimation

The kernel density estimation method is one commonly used nonparametric density estimation technique, and it was used to explore the spatial clustering of the four types of restaurants (CRs, FFRs, WRs, and DDRs) in this study. The formula is as follows:

$$f\left(x\right)=\frac{1}{{nh}}\mathop{\sum }\limits_{i=1}^{n}K\left(\frac{x-{x}_{i}}{h}\right)$$
(1)

where \(f\left(x\right)\) is the distribution density function; \(n\) is the number of restaurants within the bandwidth; \(h(h\, >\, 0)\) is the bandwidth parameter, which reflects the free parameter of the size, \(i\) is each sub-area in the study area, \(K\left(\frac{x-{x}_{i}}{h}\right)\) is the kernel function, and \(\left(x-{x}_{i}\right)\) is the estimated distance from one restaurant to another in the study area. The kernel functions and bandwidth are the parameters that most strongly influence kernel density estimation. However, when the sample size is large enough, the selection of the kernel function has little influence on the estimation results. Bandwidth selection methods include the reference method (Karunamuni and Alberts 2005; Kim and Scott 2012) and least-squares cross-validation (Bowman 1984). This study employs a Gaussian kernel function and the reference method for bandwidth selection (Zhu et al. 2021). According to the size of the study area, the final search radius was set as 2 degrees, and it was implemented in the kernel density module of spatial analysis in ArcGIS.

Spatial data analysis method

The exploratory spatial data analysis was conducted to show the spatial agglomeration characteristics of the restaurants. It includes the global and local spatial autocorrelation analysis methods as formulas (2) and (3):

  1. (1)

    Global spatial autocorrelation analysis: it can quantified the correlation characteristics for restaurants between cities (Cliff and Ord 1981; Wrigley 1982) and it was calculated by Moran’s I index as formulas:

    $$I=\frac{n\mathop{\sum }\nolimits_{i=1}^{n}\mathop{\sum }\nolimits_{j=1}^{n}{W}_{i,j}({x}_{i}-\bar{x})({x}_{j}-\bar{x})}{\mathop{\sum }\nolimits_{i=1}^{n}\mathop{\sum }\nolimits_{j=1}^{n}{W}_{i,j}\mathop{\sum }\nolimits_{i=1}^{n}{\left({x}_{i}-\bar{x}\right)}^{2}}$$
    (2)

    where \(I\) is the index of Moran, and the value range of Moran’s I is [−1, 1]. When the value is (0, 1], it expresses positive correlation, and the feature attribute tends to spatial agglomeration distribution. The value of [−1, 0) means the scattered distribution; 0 showed irrelevant distribution. n is the number of cities; \({x}_{i}\) and \({x}_{j}\) are the sweetness of the i and j cities, respectively; \(\bar{x}\) is the average value of the city’s sweetness; \({W}_{{ij}}\) is the spatial weight matrix, and it shows the spatial dependence between observation objects. It was calculated based on adjacency relationship weight matrix and based on distance relationship weight matrix. In different spatial matrix, the autocorrelation has significant spatial differences.

  2. (2)

    Local spatial autocorrelation analysis was conducted to measure the degree of difference between one city’s restaurants and other surrounding cities’ restaurants. Local Moran’s I index analysis as follows:

$${I}_{i}=\frac{\left({x}_{i}-\bar{x}\right){\sum }_{j=1}^{n}{W}_{i,j}({x}_{i}-\bar{x})}{\frac{\mathop{\sum }\nolimits_{i=1}^{n}{\left({x}_{i}-\bar{x}\right)}^{2}}{n-1}}$$
(3)

where the \({I}_{i}\) is Local Moran’s I index, and value is [−1, 1]. The value in (0, 1] shows the positive spatial relevance between one city’s restaurants with its neighboring cities’; [−1, 0) represents negative spatial relevance, and 0 represents no correlation. The specific spatial relationship contains four types: “High-High Correlation” (H-H), which means high-attribute areas are also surrounded by other high-attribute areas; “Low-Low Correlation” (L-L), which is the low-attribute region still surrounded by the low-attribute region. The \({x}_{i},\) \({x}_{j},\,{W}_{{ij}},\,\bar{x},\) are same as in (2). This analysis was completed in GeoDa.

Geographically weighted regression (GWR)

The GWR model is a statistical analysis model for parameter estimation. Compared with the global model, the GWR model can better reflect the heterogeneity of the spatial distribution of factors and has a better fitting effect on the change of factor space. The model structure as follows:

$${y}_{i}={\beta }_{0}({u}_{i,}{v}_{i})+\mathop{\sum }\limits_{k=1}^{p}{\beta }_{k}{({u}_{i,}{v}_{i})x}_{\!i,k}+{\varepsilon }_{i}$$
(4)

where \({y}_{i}\) is a dependent variable via a linear function of a set of \(p\) independent variables, \({x}_{1}\),\(\,{x}_{2}\),\(\ldots \ldots ,{x}_{p}\); \({\beta }_{0}\left({u}_{i},{v}_{i}\right)\) is the regression coefficient at point \(i\), which indicates the degree of influence of the independent variable on the dependent variable; \(\left({u}_{i},{v}_{i}\right)\) is the geographic center coordinate of the \(i\) sample point; \({\beta }_{k}\left({u}_{i},{v}_{i}\right)\) is the sample value of the continuous function, \({\beta }_{k}(u,v)\) in the sample space at point \(i\); \({x}_{{ik}}\) represents the value of the independent variable \({x}_{k}\) at point \(i\); and \(\varepsilon\) is the normal distribution function with constant variance, which represents the random error term.

First, the ordinary least square (OLS) regression method was applied to test the relationship between independent variables and dependent variables (Supplementary Table S1). Based on the OLS results, a geographically weighted evaluation model of restaurant space grid was constructed by selecting the index that passed the model test (Supplementary Table S2). Then, the weights in the GWR model are set as a function of the distance from a certain observation point to other observation points, where the accuracy of the model is largely affected by the bandwidth. Methods for determining the bandwidth include the Akaike information criterion (AIC) and the Cross-Validation (CV) method. As the AIC method takes into account the differences among different models with different degrees of freedom, we adopted the Gaussian function to determine the weights and the AIC method to determine the optimal bandwidth in the calculation. Then we analyzed the driving factors affecting the spatial distribution of the four types of restaurants, and all the data were processed in ArcGIS.

Results

Number and spatial distribution of the restaurants

For the 4.49 million restaurants, CRs, FFRs, DDRs, and WRs account for 67%, 21%, 10%, and 2% of the total, respectively. Spatially, the total number of restaurants showed a gradually reducing trend from southeast to northwest and the pattern is similar to that of the population. The cities with more than 5000 restaurants were predominantly located in the southeast (Fig. 1a). Additionally, a notable cluster of restaurants was observed in the western region, particularly in Chongqing and Chengdu, where an unusually high density of restaurants was observed (Fig. 1a).

Fig. 1: Spatial distribution of the urban restaurants and the urban population in the study area in 2020.
figure 1

a The urban restaurants; b the urban population. *In this paper, all maps are produced according to the Chinese standard map - Audit No GS(2020)4619, with no alterations to the base map.

Full size image

Then 336 cities were divided into eight groups according to their hosting total number of restaurants (Table 2). The cities home to over 100,000 restaurants were Chongqing, Chengdu, and Guangzhou, and these three cities host 7.29% (327,169) of the total restaurants in this study. Chongqing city is the municipality directly under the central government, while Chengdu is the provincial capital. These two cities are the gateway cities of southwest China and they also exhibit strong economic development potential. In addition, these cities are leisure and tourism cities, whose wide variety of cuisines and rich food culture attract a large number of tourists. Guangzhou is the provincial capital of Guangdong Province, which focuses on processing industry and represents rich food culture. The number of cities hosting (50,000–100,000) restaurants was ten, accounting for 15.60% (701,596) of the total restaurants. These cities include Shenzhen, Shanghai, Dongguan, Beijing, Xi’an, Suzhou, Zhengzhou, Wuhan, Hangzhou, and Tianjin. These cities are mainly the municipality directly under the central government, or the provincial capital, which are characterized as strong economic development, a large size of urban area and population. The cities home a total of restaurants (30,000–50,000) were 24, accounting for 19.95% (897,250) of the total restaurants. These cities are relatively underdeveloped provincial capital cities, including Jinan, Changchun, Foshan, and some of them exhibit solid economic development as third-tier cities in China. The cities with (20,000–30,000) restaurants were 19, accounting for 459,495 (10.22%) of the total restaurants, they were distributed in densely populated areas.

Table 2 Based on the number of restaurants (%, percentage of the total) to divide the cities types and the number of restaurants in different types cities (%, percentage of the total).
Full size table

Eighty cities (23.81%) had (10,000–20,000) restaurants and the total number of restaurants in these cities was 1,087,972 (24.19%). Most of these cities are located in the central-eastern part with a lower level of economic development and some of them are provincial capitals, such as Lanzhou, Yinchuan and Urumqi in the west. One hundred and three cities (30.65%) had (5000–10,000) restaurants, with their total of 752,935 (16.74%). These cities are primarily prefecture-level cities in the mid- and southwest, where economic development is in relatively low level and population density is also low.

Eighty-two cities (24.40%) had (1000–5000) restaurants, with a total of 263,574 (5.86%) restaurants. These cities are mainly distributed in the northeast, western and southwestern regions in China, and they are less-developed areas with sparse population. Moreover, the per capita income was also low in these regions, with low spending power. Finally, 15 cities (4.4.6%) had less than 1000 restaurants, mainly located in Xinjiang, Tibet, and Hainan, with only 6720 (0.15%) restaurants altogether. These cities have a sparse population, under-developed tourism industry, and insufficient consumption capacity of urban residents. In general, the distribution and the number of restaurants in Chinese cities had obvious hierarchical and spatial heterogeneity, and the pattern was decreasing from the east to the west.

Kernel density estimation of the four types of restaurants

The kernel density estimation of the four types of restaurants revealed that the high-density agglomeration of CRs is concentrated in the Pearl River Economic Delta and Yangtze River Economic Delta (Fig. 2a), including cities such as Guangzhou, Shenzhen, Dongguan, Zhongshan, Shanghai, Hangzhou, and Suzhou. Other dense agglomeration areas were in the Beijing-Tianjin-Hebei region, the Chengdu-Chongqing economic circle, central Henan province, western Shandong province, and other isolated regions, such as Xi’an, Wuhan, and Changsha, etc. (Fig. 2a).

Fig. 2: Kernel density estimation for four types of restaurants.
figure 2

a CRs: Chinese restaurants; b WRs: Western restaurants; c FFRs: fast-food restaurants; d DDRs: dessert and drink restaurants.

Full size image

For the WRs, the high-density agglomeration areas were clustered in the Pearl River Economic Delta and Yangtze River Economic Delta regions, as well as in the Beijing-Tianjin-Hebei region, Chengdu-Chongqing economic circle, central Henan province, Eastern Hubei, Southeastern Fujian, and Dalian in Liaoning province (Fig. 2b). These areas are characterized by dense populations and robust economic development. The three provinces (Heilongjiang, Liaoning and Jilin Provinces) in the northeast of China, and Shandong Peninsula are home to a higher proportion of Japanese and Korean restaurants, which is related to their close distance to, convenient transportation to and more trade opportunities with Japan and South Korea.

FFRs cater to the fast-paced lifestyle needs of the population. Following the rapid economic development in China, this type of restaurant has become widely distributed (Fig. 2c). The high-density agglomeration areas were primarily situated in the Pearl River Economic Delta, Yangtze River Economic Delta, Beijing-Tianjin-Hebei region, central and western Shandong, central Henan, the Liao-dong Peninsula, Xi’an, and Guangzhou cities, the Chengdu-Chongqing economic circle, the middle reaches of the Yangtze River, and Fujian Province (Fig. 2c).

DDRs are more affected by people’s diet habits, climate, urban tourism and leisure characteristics. For example, in the Chengdu-Chongqing economic circle, the eating habits and leisure life style have fostered hot spots for DDRs (Fig. 2d). For the Pearl River Economic Delta and Shanghai area, the climate and economic environment are suitable for growing DDRs. The other relatively high-density agglomeration of DDRs can be found in Beijing, which has a large population, and needs more varied food and beverage services. In addition, medium-density of DDRs was diffused around the high-density agglomeration areas.

Spatial correlation of four types of restaurants

The global spatial autocorrelation analysis revealed that the Moran’s I value for the total restaurants was 0.185, with a standardized Z-statistic of 13.071. This result passed the significant test at the 0.01 level of significance (P = 0.000) (Table 3), indicating a significant positive correlation within China’s catering industry. For four types of restaurants, the Moran’s I were also greater than 0 and p-value was less than 0.01. It indicated that the four types of restaurants in each city was not spatially independent and showed a certain degree of clustering. Therefore, it is necessary and feasible to use the GWR model to analyze the influencing factors on the distribution of restaurants.

Table 3 Spatial global autocorrelation analysis of the restaurants.
Full size table

For local spatial autocorrelation analysis, we utilized the Queen matrix of spatial weights to define the neighborhood and bandwidth of the moving window. For the four types of restaurants, the result showed an agglomeration pattern (Fig. 3). In the 366 cities, the four types of restaurants were primarily observed in the form of High-High (in the eastern coastal cities) and low-low (in the western areas) agglomeration. The Low-High agglomeration was distributed around the High-High agglomeration areas, where were some middle size cities. For the High-Low characteristic of agglomeration, the WRs were the Nanning, Kunming, Lanzhou, Shenyang, and Karamay cities; the CRs were Chengdu, Chongqing, Guiyang, Kunming, Nanning, Changsha, Wuhan, and Xi’an; the FFRs were Shenyang, Harbin, Lanzhou, Chengdu, Kunming, Nanning, Changsha, and Karamay; the DDRs were Shenyang, Changchun, Harbin, Lanzhou and Kunming. Overall, the spatial distribution pattern of the four types of restaurants exhibits both autocorrelation and heterogeneity, with this pattern being more pronounced in the clustering of cities with fewer restaurants.

Fig. 3: The spatial relevance between one type city’s restaurants with its neighboring cities by the Local spatial autocorrelation analysis, four types of restaurants.
figure 3

a CRs; b WRs; c FFRs; and d DDRs.

Full size image

Correlation analysis of the GWR model for restaurants

At first, we utilized the OLS model to analyze the correlations between the 10 influencing factors and distribution of the four types of restaurants. The values of VIF<5 and the tolerance <1 indicate that there was nonlinearity between the dependent variables and the independent variables. For the CRs showed nonlinearity with four variables (Urban population, GDP per unit area, Total tourism revenue, Total retail sales of social consumption). For the WRs, six variables were identified (Urban population, GDP per capita, GDP per unit area, Proportion of tertiary industry added value in GDP, Total retail sales of social consumption, Per capita disposable income of urban residents). For FFRs, five variables were identified (Urban population, GDP per unit area, Urbanization rate, Total retail sales of social consumption, Per capita disposable income of urban residents). For DDRs, three variables were identified (Urban population, GDP per unit area, Total tourism revenue) (Supplementary Table S1). Then based on the correlation analysis by OLS, we compared the model fitting efficiencies between the GWR and the OLS models. The result showed the GWR has advantages in terms of both R2-value and adjusted R2 value (Supplementary Table S2 in Appendix A). The statistical analysis on the regression coefficients in GWR model showed significant difference among the four types of restaurants (Supplementary Table S3).

In the GWR model, for the CRs with the four factors (Urban population, GDP per unit area, Total tourism revenue, Total retail sales of social consumption), showed a positive correlation 98.81% for study and 1.19% a negative correlation with the urban population (Fig. 4a), and the negative correlation was concentrated in Xinjiang and Tibet; with the GDP per unit area showed a positive correlation in 85.12% of the study areas, and focused in Xinjiang, eastern Gansu, and Heilongjiang (Fig. 4b); with the total tourism revenue factor showed a positive correlation in 80.95% of the cities, main distributed in Xinjiang, most of Tibet, eastern Gansu, Ningxia, Shaanxi, and southwest regions (Fig. 4c), and the negative correlations concentrated in the north and southeast; with the total retail sales of social consumption factor were positive and negative in 96.43% and 3.57% of the study area, respectively (Fig. 4d), and the high-value areas were concentrated in Tibet, western Gansu, Sichuan, Yunnan and the three northeastern provinces, where economic development and social goods purchasing power are at low levels.

Fig. 4: Regression analysis the relation of the CRs and influencing factors in 2020 based on the geographically weighted regression (GWR) model.
figure 4

a Urban population; b GDP per unit area; c Total tourism revenue; d Total retail sales of social consumption.

Full size image

For the WRs with the six factors(Urban population, GDP per capita, GDP per unit area, Proportion of tertiary industry added value in GDP, Total retail sales of social consumption, Per capita disposable income of urban residents) showed strong positive correlation and high spatial heterogeneity. The proportion of tertiary industry in GDP and the Urban population factors all showed weakening trend from east to west, especially for the proportion of tertiary industry in GDP factor in Xinjiang (Fig. 5a, b); On the contrary, the total retail sales of social consumption factor showed a weakening trend from west to east (Fig. 5c); the GDP per unit area showed a weakening trend from northwest to southeast (Fig. 5d). However, the GDP per capita factor showed weakening from southeast to northwest (Fig. 5f); For per capita disposable income of urban residents’ factor, showed a weaken trend from southwest to the border region was displayed (Fig. 5e). Strong correlations were identified in Yunnan, Guangxi, and Guizhou provinces.

Fig. 5: Regression analysis the relation of the WRs and influencing factors based on the GWR model.
figure 5

a The Proportion of tertiary industry added value in GDP; b urban population; c total retail sales of social consumption; d GDP per unit area; e disposable income per capita of urban residents; f GDP per capita.

Full size image

For the FFRs with the five factors (urban population, GDP per unit area, Urbanization rate, total retail sales of social consumption, per capita disposable income of urban residents) were positive in all cities, but they also showed strong spatial heterogeneity (Fig. 6). For example, urbanization rate showed a gradual weakening trend from southeast to northwest, and their strong correlation occurred in the Pearl River Economic Delta (Fig. 6a). Urban population showed a weakening trend from the middle to the borders (Fig. 6b). Strong correlations showed in Heilongjiang, northeastern Inner Mongolia, Shaanxi, Ningxia, Gansu, Sichuan, Chongqing, Yunnan, Guizhou and Guangxi regions. For the total retail sales of social consumption, the correlation was weakened from inland to the border, except for the border of Inner Mongolia (Fig. 6c); The GDP per unit area factor was weakening from northwest to southeast (Fig. 6d). Interestingly, the per capita disposable income of urban residents showed a weakening pattern from southwest and northeast to the center (Fig. 6e) such as Yunnan, and Sichuan in the southwest and Xinjiang total area.

Fig. 6: Regression analysis the relation of the FFRs and influencing factors based on the GWR model.
figure 6

a Urbanization rate; b Urban population; c Total retail sales of social consumption; d GDP per unit area; e disposable income per capita of urban residents.

Full size image

The DDRs showed strong correlations with three factors (urban population, GDP per unit area, total tourism revenue). Spatially, the urban population factor showed a weakening pattern from south to north, except in Xinjiang and Tibet (Fig. 7a), and strong correlations occurred in the Pearl River Economic Delta, Guangxi, Guizhou, and Yunnan. In contrast, the negative correlations were located in the western region; For the GDP per unit area factor, 97.02% of the cities showed a positive correlation and 2.98% showed a negative correlation (Fig. 7b), the strong correlation occurred typically in Tibet (except Nagqu), Sichuan, Yunnan. Meanwhile, the weak correlations were located in the eastern region and Yangtze River Economic Delta; For the total tourism revenue factor showed strong correlation in the west, such as east of Tibet, Qinghai, Gansu, and Ningxia, etc (Fig. 7c), where tourism resources are abundant.

Fig. 7: Regression analysis the relation of the DDRs and influencing factors based on the GWR model.
figure 7

a Urban population; b GDP per unit area; c total tourism revenue.

Full size image

Discussion and conclusion

Food serves as the fundamental material to meet human physiological needs. The quantity and spatial distribution of restaurants reflect human needs and behaviors (Nathan et al. 2012). Increasing urbanization would facilitate development of catering industry. Then how urbanization affects the total number and their spatial distribution pattern of restaurants need mounting attentions (Zhang et al. 2021a; Tu et al. 2020; Zhu et al. 2021; Lan and Tseng 2018; Xia et al. 2018; Wu et al. 2017; Zhou et al. 2015).

In this study, based on actual distribution point data in 2019, the restaurants were categorized into four types (Chinese restaurants (CRs), Western restaurants (WRs), fast-food restaurants (FFRs), and dessert and drink restaurant (DDRs)), and comprehensively analyzed their number, spatial distribution, and influencing factors. The result showed that over 78.86% of the included cities were home to (1000–20,000) restaurants. Spatially, the total number of restaurants decreases from southeast to northwest in China, but Chengdu-Chongqing economic circle in the southwest host a large number of restaurants. One noteworthy point is that the restaurants are mostly concentrated in core region of Chengdu-Chongqing circle, while a low proportion are distributed in the surrounding cities. Along the Yangtze River and Pearl rivers economic belts, restaurants are concentrated not only in major cities such as Wuhan, Nanjing, Hangzhou, Shanghai, Guangzhou and Shenzhen but also evenly distributed in surrounding smaller cities, this indicates that major cities exert a strong influence on their surrounding areas.

The local spatial autocorrelation analysis on the four types of restaurants reveals a High-High agglomeration in the eastern coastal cities, and low-low agglomeration in west regions. The dominant factors in shaping this spatial differentiation model are the diverse regional advantages and economic development levels of coastal and inland areas. WRs are primarily concentrated in eastern coastal urban agglomerations and major cities, showcasing a high level of concentration; As China’s urbanization continues to progress and the fast-paced lifestyle demands increase, the rapidly growing FFR, encompassing honor brand fast food restaurants and casual dining establishments, experienced significant growth across all regions of China. The distribution pattern of DDR was greatly influenced by people’s dietary preferences, climate conditions, as well as urban tourism and leisure characteristics. It main distributed in the Chengdu-Chongqing economic circle, the Pearl River Delta region, Shanghai area with favorable climate and economic environment, along with densely populated Beijing.

The analysis results of the factors influencing the spatial differentiation of these four types of restaurants indicate that both urban population and GDP per unit land area affected the spatial pattern of the four types of restaurants. Additionally, WRs was closely associated with the tertiary industry; FFR exhibited a strong correlation with urbanization rate; DDRS, encompassing dessert shops, cold drinks and coffee shops, were contingent upon the income and purchasing power of urban residents; CRs, represented by restaurants and hot pot restaurants, were the most widely distributed, and its influencing factors were more complex. Based on spatial correlation analysis of the GWR model, the number of restaurants in central and eastern China was mainly regulated by urban population, GDP per capita, urbanization rate, and proportion of tertiary industry. For west and northwest regions, the main factors included GDP per unit area, total tourism revenue, per capita disposable income of urban residents, and total retail sales of social consumption. In northeast, the primary influencing factors were the total retail sales of social consumption, GDP per unit area, urban population, and Proportion of tertiary industry added value in GDP.

The catering industry serves as a barometer for domestic demand in the context of consumption upgrading. China has the number of restaurants about 9 times than the United States, but the total revenue from restaurants is similar between the two countries. China’s per capita food consumption is less than 1/5 of the United States (https://baijiahao.baidu.com/s?id=1688700104778489233&wfr=spider&for=pc). The comparison between the two countries implies that the increasing urban population and total retail sales of social consumption will further boost China’s catering industry.

The rapid advancement of information technology and the continuous enhancement of Internet map data are accompanied by the generation and exploration of big data as well as the new paradigms. This study reveals the space and driving factors of Chinese food culture with big data and catering service POIs with millions of categories. Big data thinking and methods based on data-intensive empirical research improve the accuracy of relevant research results and provide a new perspective for other types of cultural space-related research. However, China is a big country and its land is home to many cultures and histories. In addition to the factors discussed in this article, other factors such as topography, climate, and other natural factors may play a key role in regulating the distribution and type of restaurants. In the future, we will dig deep into the internal correlation of data and show the coupling characteristics between Chinese food, culture, economy and politics from a geographical perspective.