Impact of urban innovation on urban green development in China’s Yangtze River Economic Belt: perspectives of scale and network

Understanding whether and how urban innovation offers a sound solution to the dilemma of urban green development is a crucial response to mitigate the detrimental effect on natural resources and environment for transitioning to sustainable urban development. To address the critical issue, we propose urban green development evaluation index system, and then examine how the urban innovation affects urban green development from the perspectives of government-scale, enterprise-scale, and spatial correlation network, all of which are originally applied in the 108 cities of Yangtze River Economic Belt of China (YREB) during period 2006–2018. The evaluation results show that urban innovation promotes urban green development, and both government-scale and enterprise-scale contribute to the effects. The constructed spatial correlation network of urban innovation illustrates the network structural form and reveals the network property, and further results tell that increasing network density and centrality would promote green development obviously. More specifically, the network density of urban innovation has been tied to the enhancement of urban green development, which is more significant in middle reaches than in lower and upper reaches of YREB. Similarly, optimizing the network’s degree centrality and closeness centrality can help facilitate urban green development in whole YREB. Thus, the research findings would provide new insights into the essence and driving forces from various scale and hidden network when exploring and seeking urban green development path.


Introduction
China's gradual transition toward an urbanization-oriented development strategy bring social and economic boom as well as detrimental ecological and environmental consequences, which would pose a significant threat to sustainable urban development (Bloom et al. 2008;Yang 2013;Muhammad et al. 2020). As it continues along the path of rapid urbanization and industrialization, the technological innovation is the important power source to promote economic growth in China . In spite of its detrimental effect on natural resources, environment, and social life to some degree, the technological innovation is instrumental in green development gains due to the improvement of green production and the reduction of environmental pressure continuously Cao et al. 2020).
The loss and degradation of natural recourse and ecological environment have raised widespread concern about the green development mode and the consequent impact to the economic activity and human well-being that they support (Tan and Lu 2019;Ouyang et al. 2020). The green development conception has gradually gained wide currency globally. For instance, the United Nations Environment Program (UNEP) launched a "Green New Deal" and a "Green Economy" initiative through the promotion of green development in the short and medium terms. Similarly, as one of China's paramount development conceptions, green Responsible Editor: Eyup Dogan * Zhiyuan Niu ert_new@163.com development emphasizes the coordination of economic growth, environmental protection, and resource conservation (Merino-Saum et al. 2020). In addition, the United Nations Conference on Sustainable Development (UNCSD) considered four components including economic, environmental, social, and institutional for urban sustainability (Panda et al. 2016). The United Nations General Assembly set up 17 interlinked global goals for building a better, planet by 2030 through the achievement of multiple social and environmental goals (United Nations 2015. The goal of sustainable development is more comprehensive than green development. Green development is often seen as a way to achieve sustainable development in China (Lin and Benjamin 2017). Thereby, investigating and assessing urban green development status is an essential prerequisite to improve the urban or regional ecological environment (Fang et al. 2019;Weng et al. 2020;Wei et al. 2020).
Most existing literatures believe that urban innovation is a complex concept (Zhang et al. 2022). In recent years, urban innovation is becoming the driving force in the economic and social development of cities, regions, even countries (Lauer and Liefner 2019). Government takes a smart city project as an important way to promote urban innovation, enhance urban economic competitiveness, reduce urban environmental pollution, and provide residents with a high-quality life (Wang and Deng 2021). The city needs to create a sustainable ecosystem and improve urban green development through some smart technologies (Blasi et al. 2022). As a typical region moving forward on the green development demonstration belt, furthermore, the Yangtze River Economic Belt (YREB) spans three regions of east, middle, and west of China, and it is an important economic and population agglomeration area in China. Meanwhile, it is one of China's three major development strategies and has a prominent position in the overall spatial pattern of the national development strategy. Thereby, it is an important strategic significance for transitioning to the green development model of the Yangtze River Economic Belt.
How to achieve regional multi-scale green development is key to implement the practical needs of China's current ecological civilization practice and policies and to enrich the theoretical needs of global sustainable development. To be specific, understanding whether and how urban innovation should offer a sound solution to the dilemma of urban green development is a crucial response to mitigate the detrimental effect on natural resources and environment for transitioning to sustainable urban development. It has important implications for exploring the current institutional obstacles and policy system of urban innovation, which should be conductive to achieve urban green development effectively. Under this circumstance, this study would join the debate and presents evidence from both perspectives of scales and network when discuss about the Yangtze River Economic Belt of China. These factors interact with the level of innovation and jointly affect urban green development. Moreover, the study used system generalized method of moments (GMM) and threshold methods to tested the relationships. Furthermore, the social network analysis (SNA) is applied to analyze the topological characteristics of spatial innovation correlation network of YERB.
Summarizing the extant literature, this study mainly makes three contributions to the new body of knowledge focusing on the impact of urban innovation on urban green development in China's Yangtze River Economic Belt. First, this study focuses on constructing an index system of urban green development, which helps to find out the main aspects for leading or restricting the urban green development and then puts forward some specific measures relative to the actual situation of the Yangtze River Economic Belt. Second, urban innovation is an important force to promote multidimensional development at various scales. Therefore, this study intends to explore and uncover the impact of urban innovation on urban green development, which can help fill up the deficiency in the field of the interaction between urban innovation and green development. Finally, this study seeks to contribute to the green development from the perspectives of scale and network, which can help provide new insights into the essence and drivers from various scales and hidden network for achieving the sustainable development path.
This paper will be structured as follows: "Literature review" provides a literature review. "Materials and methodology" describes materials and methodology. "Results and analysis" shows results and analysis. "From the spatial innovation correlation network" gives a description for the network correlation of urban innovation and "Discussion and conclusions" contains the main conclusions, policy implications, and limitations.

Literature review
This section discusses the relationship between urban innovation and urban green development and evaluates the moderating role of two factors (government-scale and enterprisescale) in innovation-driven urban green development.
With the continuous emergence of technology innovation and the concept of green development becoming more and more mature, researches on their theory and methods had been deepened (Table 1). Currently, a growing body of literature has studied the interaction between urban innovation and urban green development, and various determinant analysis models have been used in the previous studies. The first strand of research focuses on the notion that urban innovation may exert a positive influence on urban green development. Urban innovations, in particular green technologies, have been regarded as promoting urban green development because they can help mitigate the carbon emission risk (Liu and Zhang 2021;Zhao et al. 2020), air pollution risk (Cui et al. 2011;Zhu et al. 2020), and other environmental risks (Chaudhry et al. 2021). Meanwhile, innovation can also strengthen waste management , which is conducive to waste reuse and improve resource recovery efficiency (Marchi 2012). In addition, innovation improves resource productivity (Pujari 2010;Dai and Sun 2021) to improve urban green development. For instance, the productivity of enterprises in high-technology industries is higher than that in low-technology industries (Khanna and Sharma 2021). Furthermore, Porter hypothesis points out that appropriate environmental regulation can promote enterprises to carry out more innovation activities, which improve the productivity of enterprises, thereby offset the costs brought by environmental protection, then improve the profitability of enterprises in the market and production efficiency (Porter 1991). The second strand of research concentrates on the negative impact of innovation on green development. Traditional neoclassical mainstream economic theory holds that the implementation of green policies can help to improve urban green development. But it will also increase the cost of enterprise pollution discharge and reduce innovation investment (Albrizio et al. 2017). Moreover, technological innovations might cause an increase in excessive energy consumption and carbon emissions (Gunderson and Yun 2017;Acemoglu et al. 2012) and it is much worse for innovation due to deficient talent and capital (Jin et al. 2019). Afterward, there may be an inverted U-shaped relationship between innovation and environmental pressure (Feng and Yuan, 2016).
Existing studies have enlarged the research fields of urban green development, some results of which reveal that both the expansion of government-scale and enterprise-scale can enhance the driving effect from innovation on green development. The inconsistency between private cost and public cost often leads to market failure, according to the externality theory, which can be addressed from the governmentscale (Glemarec and de Oliveira 2012), such as distributing the emission rights (Guo et al. 2021;Tang et al. 2021), highlighting the role of emission tax (Villegas-Palacio and Coria 2010; Li et al. 2021), strengthening the environmental supervision (Cumming 2007;Tang et al. 2021), and improving the green technology innovation level (Martinelli and Midttun 2010). In the meantime, the expansion of enterprise-scale can bring comparative advantage and talent advantage. In light of the talents advantage, for instance, the large-scale enterprises have access to more funds and preferential policies from the government (Beck et al. 2005). Subsequently, the expansion of enterprise-scale can help bring economies of scale and economies of scope. Large-scale enterprises are more likely to implement the eco-friendly measures to improve the quality of green innovation , especially the state-owned enterprises. However, some studies also prove the adverse impact once expanding the scales of government and enterprise. Expanding the scale of government always increases the administrative units and rent-seeking activities, as well as the excessive bureaucracy and low administrative efficiency. Likewise, expanding the scale of enterprise often reduces the innovation efficiency of enterprises (Mei and Shao 2016). For example, the market full of large-scale enterprises or monopoly enterprises is more likely to be short of the competitive power to stimulate technology innovation. It is not only harmful  Albrizio et al. (2017) Community Innovation Survey (CIS) framework Two-part logit model Marchi (2012) An inductive multiple case study methodology Life cycle analysis Pujari (2010) to the sustainable technological innovation of enterprises in the market but also destructive to the long-term health of regional market development.
Most extant studies pointed out that innovation has a positive impact on green development in virtue of technological progress , environmental pollution migitation (Chaudhry et al. 2021;Liu and Zhang 2021), and economic improvement (Batabyal and Beladi 2014;Snieska and Valodkiene 2015). But from another aspect, innovation has been identified as the drivers to improve production capacity when contributing to more serious natural resource consumption as well as environmental pollution (Wang et al. 2014;Zhao et al. 2010). Despite the more and more explorations of interaction between innovation and green development, on the whole, there are scarce theoretical and practical exploration from the coupling perspectives of governmentscale, enterprise-scale, and spatial correlation network.

Study area and data resources
As a typical region moving forward on the green development demonstration belt, the Yangtze River Economic Belt (YREB) spans three regions in the eastern, middle, and western China, and it is an important economic and population agglomeration area in China (Fig. 1). Meanwhile, it is one of China's three major development strategies and has a prominent position in the overall spatial pattern of the national development strategy. Thereby, it is an important strategic significance for transitioning to the green development model of the Yangtze River Economic Belt. In spite of seeking the common goal of the whole economic belt, there is regional heterogeneity of urban green development tendency due to different resource endowments, geographical positions, infrastructures, policy slants, and so on. In particular, urban innovation is a key element in promoting urban green development. On account of the above, this paper evaluates and explores the interaction between innovation and green development from both the perspectives of scale and network. The required data come from China Industrial Statistical Yearbook (

Index system of green development level
The evaluation index system of urban green development level ought to contain main influencing variables, and some principles, such as comprehensiveness, representativeness, authority, and feasibility, should be taken into consideration when constructing the index system. Based on previous literature (OECD 2017), the core objective of green development level is to balance economic development and environmental protection. Therefore, many scholars comprehensively evaluated urban green development from multiple perspectives (Yuan et al. 2017;Ma and Huang 2017;Chen and Zheng 2017;Liu et al. 2017). In addition, the sustainable development index system and green economy index system of the United Nations Environment Program are all internationally accepted green development evaluation models. Meanwhile, according to China's specific national conditions, many scholars have also put forward China's green development index system, including economic growth, comprehensive carrying capacity of resources and environment, government policy support, and other aspects (Li and Pan 2011;Hu and Ma 2017). This study also refers to the "green development index system" jointly issued by the national development and reform commission and other departments in China and the "green city evaluation index" (Exposure Draft) issued by the China Institute of standardization. Thereby, the index system of urban green development level in this study is characterized and identified by its economic growth (Tan and Lu 2015;Tan et al. 2017), industrial structure (Hu and Ma 2017;Fan et al. 2021;Peng et al. 2021), residents' living standards improvement Liao and Li 2022), and ecological environment management Liao and Li 2022) from an integrated view (Table 2).

Data standardization and weight determination
This paper uses the dimensionless standardization and weight calculation to evaluate urban green development level. Firstly, the dimensionless standardization method is adopted to standardize the annual index value as Eq. (1), where x ij denotes the raw data and ‾x ij is the corresponding average value; y ij represents the normalization value.
Further, the modified entropy method is used to deal with index weight, and it contributes to assessing the comprehensive urban green development level. Specific steps can be found in many studies (Tan and Lu 2015;Tan et al. 2017), which can help maintain all information of the original samples and can help prevent the elimination of diversity of the raw data. In summary, following the sequence of "bottom to top," each indicator is multiplied by the standardized value of each indicator with its corresponding weight, and subsequently, the values of each dimension are determined to represent the urban green development general trend of each city in YREB during the study period.
(1) The impact of urban innovation on urban green development

All the variables
The independent variable (urban innovation) is measured by the number of patent applications per capita when considering the local urban employees. In general, urban innovation can be quantified either by using patent data to represent innovation capacity or by relying on surveys (Shearmur 2012). Although there is a problem in using patents as innovation indicators that is not all innovation is patented, many existing literatures take patents as a measure of innovation and highlight the quality of innovation (Huang et al. 2012;Ning et al. 2016;Jin et al. 2021Jin et al. , 2022. Meanwhile, in order to control the endogenous problems that may exist in econometric models, this study uses the patent applications per capita as the proxy variable of urban innovation. The share of public expenditure in GDP stands for government-scale (threshold variable) and the log transform of total profits of industrial enterprises above the designated size stands for enterprise-scale (threshold variable).
Urban green development should be affected by urban innovation, as well as by other explanatory variables. Some control variables are selected as follows. Foreign direct investment (FDI) can bring more advanced and cleaner production technology, and further improve the ecological environmental quality of host country, while the foreign enterprises may transfer the high-polluting industries to underdeveloped regions with low environmental standards. Environmental regulation (ER) policy can directly restrain the high pollution emission behavior of enterprises and promote technology innovation. The government should refrain from promoting urban green development (UGD) at the expense of a high unemployment rate (UNE). Generally speaking, with the improvement of the urbanization rate (Urban), the production factors such as talents and capital are gathered in cities for improving the urban innovation level. However, the accelerating urbanization process should be harmful in the green development mode. Technology and education investment (TE) is an important driving factor for the urban green development, providing the funds and talents for urban innovation.
Some data pretreatments should be covered. For instance, all current variables are processed and converted to the real value according to the value of 2006. Besides, taking the logarithm can help eliminate the heteroscedasticity of nonproportional and non-standardized variables, and the linear interpolation techniques will be used once there is some missing data. Consequently, Table 3 show the details of main variables in this study. It should be noted that the variance expansion factor (VIF) test shows that all the selected variables do not have serious multicollinearity (VIF < 10).

Basic regression models
The setting of the basic panel model is used to test and verify the overall impact of urban innovation (UINN) on urban green development (UGD). In addition to the urban innovation, there are some other factors to promote urban green development, such as foreign direct investment (FDI), environmental regulation (ER), unemployment rate (UNE), urbanization rate (Urban), and the technology and education investment (TE). As a result, the benchmark model of urban innovation and urban green development can be established as follows.  Enterprise-scale Logarithmic transformation of total profits of industrial enterprises above designated size Li et al. (2019) where i stands for the individual city, t represents the year, β 0 is the coefficient of the constant term, β 1 ~ β 6 are the coefficients of independent and control variables, δ i denotes the individual effect, and ε it is the error term. The basic panel model needs to address the endogenous problem between urban innovation and urban green development due to the deviation in coefficient estimation. Then, the solution is as follows. The fixed-effects model eliminates heterogeneity between individuals and minimizes the endogenous biases caused by the absence of variables. Here, we use a two-step system GMM for robust estimation. The GMM model of the system is presented as Eq. (3).
UGD it − 1 is the first-order lag term of urban green development level, β 7 is the coefficient of the lag term, and the other variables are the same as above.

Threshold regression models
From the perspective of regional heterogeneity, governmentscale and enterprise-scale are introduced. It shows that urban innovation in different regions has different effects on urban green development by the study of different governmentscale and enterprise-scale. Thus, in order to identify the moderating effect of government-scale and enterprise-scale, we use the threshold models to test how government-scale and enterprise-scale act on the influence process from urban innovation on urban green development.
The models are set as follows.
η 1 is the threshold value to be estimated from governmentscale. This is the expression of single threshold model and so on.
η 2 is the threshold value to be estimated from enterprise-scale.

Social network analysis
Social network analysis (SNA) is an important analysis method that can depict the form, feature, and structure of a complete network (Scott 2000;Li et al. 2014;Shen et al. 2021), which can be applied in this study when considering each province or sector as a network node. Its advantage (2) especially shows in the expression of correlative relationships, such as revealing the conformity and hierarchy of the overall network, explaining the compactness of network connection, and identifying the dominance of different nodes. In this study, the topological characteristics of spatial innovation correlation networks are presented by the techniques including network density and network centrality. Then, the mathematical expression of network density and network centralities such as degree centrality and closeness centrality are all as follows. Network density (D) reflects the tightness of network connections as Eq. (6), where M denotes the actual number of associations and N denotes the maximum number of associations. Degree centrality (Dc) reflects the position of nodes in the network as Eq. (7), where N denotes the maximum number of directly connected areas and n denotes the number of areas directly associated with an area. Closeness centrality (Cc) reflects the degree to which a node is not controlled by other nodes as Eq. (8), where N denotes the maximum number of directly connected areas and d denotes the shortcut distance between two areas.

Urban green development general trend of the Yangtze River Economic Belt
The green development level showed a steadily increasing tendency for the majority of cities during the period 2006-2018, while it showed a decline for a small minority of cities due to the relatively frequent eco-environmental risk accidents. Figure 2 intuitively reflects the change characteristics of urban green development level in the Yangtze River Economic Belt in 2006, 2012, and 2018. It can be seen that there is seldom any change for most cities in Jiangsu, Shanghai, and Anhui, while there is a noticeable change for other cities in YREB. In particular, some cities had a high level of green development in the early period, and then, they have a gradual decline in the later period. Thereby, it is of great important to explore the green development and its drivers.

Overall impact of urban innovation on urban green development
The fitting degree of regression equation and the accuracy of the model will be improved once increasing the control variables gradually as presented in Table 4. According to the regression results of column (7), the current value of urban green development level and its first-order lag value tell the motivating effect. In other words, urban innovation may accelerate the improvement of urban green development levels to a certain extent. The regression results when covering other control variables verify that the regression coefficients of unemployment rate (UNE) and foreign direct investment (FDI) are significantly negative and that of environmental regulation (ER), urbanization rate (Urban), and technology and education investment (TE) are significantly positive.
In general, urban innovation plays a positive and significant role in urban green development for the three urban agglomerations in the middle, lower, and upper reaches of YREB from Table 5. The urban innovation in the middle reaches gives the most obvious motivating effect to urban green development. To be specific, the upper reaches struggle to bring sufficient technical capital and talent investment for urban innovation, as well as the limited progress of green development level. Meanwhile, the lower reaches have a slight decline in urban innovation, leading to an impenetrable technological threshold. Meanwhile, there is no obvious motivating effect between urban innovation and urban green development for the lower reaches. To sum up, the urban innovation level has been tied to the enhancement of urban green development level, which is more significant in middle reaches than in lower and upper reaches of YREB.

From government-scale
From the perspective of the government-scale, there is a clear threshold effect between urban innovation and urban green development. The results of double-threshold model tell that there are two estimated thresholds when considering the impact of urban innovation on urban green development from the government-scale, and the double threshold values are 0.094 and 0.162 (Table 6). According to the values, 108 cities in YREB were classified as being small-government-scale (GS ≤ 0.094), medium-government-scale (0.094 < GS < 0.162), and large-governmentscale (GS ≥ 0.162).
As shown in Table 7, urban innovation plays a crucial role in promoting the urban green development level despite various government scales. From urban innovation to urban green development, the influence coefficient is significant in the category of large-government-scale. For example, from the angle of increasing government financial investment, cities with large government-scale often provide more financial subsidies to stimulate the innovation power of enterprises and scientific researchers, all of which will help improve the overall innovation level of the city.

From enterprise-scale
From the perspective of enterprise-scale, the results tell that there is also a threshold effect between urban innovation and urban green development. The results of double-threshold model tell that there are two estimated thresholds when considering the impact of urban innovation on urban green development from the enterprises-scale and the double threshold values are 2.642 and 5.362. According to the values, 108    Table 8 shows the regression results after covering the threshold of enterprise-scale. The threshold effect of enterprise-scale is similar to that of government-scale. The coefficients of urban innovation are significant in the cities with various enterprise scales, suggesting that innovation-driven urban green development is effective in the cities with various enterprise scales. In particular, the influence coefficient is more significant in large-enterprise-scale than that in small and medium-enterprise-scale. Thereby, urban innovation should involve urban green development in YREB, which is more obvious in cities with large-enterprise-scale.

Topological characteristics of spatial innovation correlation network
The topological characteristics of spatial innovation correlation network are presented in Fig. 3. The network density and the relationship number are increased during the study period in the network of the Yangtze River Economic Belt.
In terms of the network density, the spatial innovation correlation network of YREB is closer and closer. However, it may include some adding redundant lines. For instance, some redundant lines should increase the transaction costs between various cities, leading to a negative impact on urban green development. Ultimately, it is critical to address how to improve the spatial correlation between cities inside YREB.
In terms of the correlation network for YREB, the average value of degree centrality is 62.98. Inside YREB, 64 cities are higher than this average value (Table 9). The degree centrality was roughly in the order of middle, upper, and lower reaches. To be specific, the majority of cities are concentrated in the lower reaches and a minority of cities are distributed in the upper and middle reaches. Each city is connected with another one in the spatial innovation correlation network to a greater extent.
The average value of closeness centrality is 6.603. Inside YREB, 57 cities are higher than the average value (Table 9). Specifically, the cities are concentrated in the provincial capital in the lower, upper, and middle reaches. The higher closeness centrality in the lower reaches suggests that they should associate with other cities through the inherent connection of urban innovation elements at a faster rate. It seems clear that the closeness centrality in the upper reaches is lower than that in the lower and middle reaches, such Table 7 Impact of urban innovation on urban green development under the threshold of government-scale *Significant at 0.1, **significant at 0.5, ***significant at 0.01. (1) (2) (3) (4) (5) as Guiyang and Kunming. In other words, when regarding Guiyang and Kunming as the core of the urban network, the urban innovation level of the upper reaches is weak, and the urban green development level is slow.

On the whole network structure
On the whole network structure, the urban green development and the network density at four scales are chosen as the   explaining and explanatory variables in the OLS regression when in the natural logarithm transformation. The whole spatial innovation correlation network structure gives a significant and positive effect on urban green development, which can be found in Table 10. The improvement of network density can effectively promote the urban green development level in YREB. Notable, the network density of YREB's urban innovation has a positive impact on the urban green development, which is more significant in middle reaches than in lower and upper reaches of YREB. Therefore, it is necessary to continuously strengthen the spatial connection between the lower reaches and the upper reaches by increasing some necessary lines and reducing some redundant lines from associated innovation institutional reforms.

On the individual network structure
On the individual network structure, the network centrality (including the degree centrality and closeness centrality) of 2016 is applied to analyze the effect on the green development from spatial innovation correlation network, because it is an important year when some emerging strategies proposed for seeking the green development (Table 11). This study takes the green development level of 108 cities in YREB as an explanatory variable to analyze the impact of individual innovation network structures on urban green development. The regression coefficient of degree centrality is 0.612, indicating that cities with high degree centrality have a stronger leading role rather than other cities inside the spatial innovation correlation network. Under the circumstance, the government should pay more attention to these cities to promote the whole urban green development level of YREB. However, there are still plenty of cities with low degree centrality and low green development level that need to strengthen the innovation connection with the lower reaches. For example, the talents, technologies, and funds in the lower reaches of YREB should be associated with the upper and middle reaches. What is more, the government should be encouraged to reduce the unemployment rate, increase the urbanization rate, and increase the science and education expenditure. The regression coefficient of closeness centrality is 8.119, presenting that in the spatial innovation correlation network. Cities with high closeness centrality have the comparative advantages with other cities. The cities may control and guide the flow direction of urban innovation elements in more targeted manner, which can strengthen the "spatial spillover" effect on other cities inside the spatial innovation correlation network. On the other hand, cities with low closeness centrality and low urban green development level have to be enhance the potential to obtain innovation profitability inside the network. To sum up, cities with high closeness centrality and high urban green development level should be encouraged to guide and transfer the specific innovation elements to motivate urban green development from the perspective of the network (Table 12).

Main conclusions
As a typical region moving forward on the green development demonstration belt, the Yangtze River Economic Belt is a valuable subject to explore whether and how urban innovation offers a sound solution to the dilemma of urban green development. Uncovering the influence path can help mitigate the detrimental effect on natural resources and the ecological environment for the green demonstration belt. This study estimated the general trend of urban green development level and the topological characteristic of the spatial innovation correlation network for YREB during the period 2006-2018. It further investigated the specific impact of urban innovation on urban green development through uniting the government-scale, enterprise-scale, and spatial correlation network. The research framework can provide the practical policy implications for the study area, as well as a theory framework for clarifying the interaction between urban innovation and green development.
The research findings indicated that urban innovation in YREB promotes urban green development. However, there is obvious heterogeneity in the interaction of innovation and green development in various cities. For instance, it is more apparent in the middle reaches than in the upper and lower reaches. Moreover, the government-scale and enterprisescale are two important moderating factors between urban innovation and urban green development. The coefficients of urban innovation are all significant in the cities with various government scales and enterprise scales, which means the innovation-driven urban green development is effective in the cities with any scale.
The study further illustrated that the optimization of the spatial correlation network of innovation can also promote the green development of all cities in YREB. Inside the spatial correlation network, the increase in network density can well improve the urban green development level. Moreover, the city's central position in the spatial innovation correlation network has also a significant role in promoting the urban green development inside YREB. In other words, if cities play a crucial role in the spatial innovation correlation network, these cities will have a large spillover effect on the Table 11 Impact of degree centrality on urban green development *Significant at 0.1, **significant at 0.5, ***significant at 0.01. (1) (2) (3) (4) (5) surrounding areas, which is conducive to innovation-driven urban green development.

Policy implications and limitations
The central government of China has put forward the innovation-driven development policy, emphasizing that urban innovation can promote urban development continuously. A more flexible implementation of plenty of policies would promote technology innovation, further facilitate the green development, and enhance the environmental performance, in spite of the negative effects from some specific environmental policies (Albrizio et al. 2017). First, the government often launches some adaptive and appropriate programs to promote the technological innovation of enterprises and reduce environmental problems, such as providing some subsidies for pollution mitigation, setting some targets of environmental supervision, granting the green credits, and collecting the emission taxes. Secondly, regional differences need to be considered in the promulgation of government policies. From the view of regional disparity, it will be critical to avoid a "one-size-fits all" policy approach, both across technologies and across cities. Government should expand the functional scope and strengthen the intervention when focusing on some cities with lower green development level and smaller scale. It is also suggested that government should initiate and launch some preferential policies and projects to drive the cities with small enterprise-scale. For example, some financial subsidies can help reduce the cost and risk of urban innovation (Huang et al. 2019). Finally, it is necessary for deepening the network effect and talent flow effect between cities, promoting the formation of a coordination mechanism of green innovation among cities. For instance, the government needs to incorporate technical and industry expertise into government decision-making, especially for some cities with low network density, centrality, and green development. In terms of spatial correlation network, the network density of urban innovation has been tied to the enhancement of urban green development, which is more significant in middle reaches than in lower and upper reaches of YREB. Due to the motivated effect from network density and network centrality, the cities in middle reaches have reason to act for themselves through attracting adequate talents and strengthen the connections among the three reaches. Some cities and their industries articulate their performance needs, finds supporting funding, and funds academics to develop new or advanced technological technology to achieve the green development goals.
However, because of the restriction of data availability, it is hard to choose indicators that can cover all aspects of the compound urban green development index. We have the limitation that the study puts emphasis only on the past period and does not cover dynamic predictions, yet green development research would be better served by a long-term perspective. The urban green development is brought about by multiple intertwined factors, so we hope to make deeper explorations of the motivated determinants and reveal the interactions among different determinants. In addition, we can add the control index (CI), dependent index (DI), hubs index (HI) and authorities index (AI) in the subsequent research on social network analysis. Moreover, how the four indices (CI, DI, HI, AI) are calculated in details, as well as the related technique methods, can be found in the previous literature (Chen and Chen 2016;Wang et al. 2017). Indeed, inquiring into whether and how urban innovation offers a sound solution to the dilemma of urban green development is just the precondition of achieving a regional sustainability transition, and the journey towards real sustainability has just begun.