GPCA-EM Model

The GPCA model is based on the conventional principal component analysis. It overcomes the shortcomings of the diffusion index method in capturing cyclical swings by incorporating cross-sectional data with a temporal dimension [29]. The EM method is an objective weighting approach that determines weights based on the size of indicator variability. The GPCA-EM model combines the GPCA and EM, using the information entropy of the principal components as an assignment criterion to generate the composite index [24]. This approach addresses the shortcomings of the conventional principal component composite score [30]. The GPCA is used to reclassify the regional development index system, and the scores of each primary component are computed first. The final comprehensive values of the green development level are then calculated using the EM.

(1) Constructing a time-ordered stereo data table and normalizing data: X = {X_t∈R_m×_p, t = 1, 2, ..., T} is a set of flat data tables arranged in chronological order, with identical sample points and variables placed in chronological order. Next, as shown below, the data is normalized.

Where r denotes the city (r = 1, 2, ..., n), s denotes the indicator (s = 1, 2, ..., m), and t denotes the year (r = 1, 2, ..., k), X_rst is the original value of the indicator s in the city and r in the year t, X_rst⁺ and X_rst⁺ are the normalized values, minx_st and maxx_st denote the minimum and maximum values of the s indicator.

(2) Determining the normalized data matrix’s correlation coefficient matrix, eigenvalues, and matching eigenvectors.

(3) Computing the matrix of factor loadings. The correlation coefficient matrix B = (b_rs) of X_rt with the principal component F_s needs to be found. The b_rs represents the loading of the variable r in X_r on the principal component F_s. The rotation method with maximum variance yields coefficients that can be used to explain which variables make up the majority of the principal component F_s.

(4) Calculating the regression coefficient for each principal component score.

(5) Calculating the proportion of the indicator r of the evaluation object s in the total by converting the data to be non-negative.

(6) Calculating the entropy value of the indicator r.

Where k>0, k = –1/ln(n), e_r≥0, k.

(7) Calculating the variability of the indicator r. When it comes to indicator r, the more variable the indicator value, the more valuable it is for program evaluation.

Where Ee=∑r=1mer,0≤gr≤1,∑r=1mgr=1.

(8) Calculating weights.

(9) Utilizing the GPCA-EM model to obtain a comprehensive score for the green development level.

Two-Stage Nested Theil Method

Based on the idea of entropy, the Theil index is a particular type of generalized entropy index. When breaking down within-group and between-group differences, it outperforms the Gini coefficient [31]. In the academic world, the Theil index has grown in importance as a measure of income inequality among people or areas. In this paper, we utilize the two-stage nested Theil to analyze the disparities of green development levels within and between provinces, as well as between the middle and lower reaches. We estimate their importance and rates of contribution to the total difference [27, 32].

Assuming that the mid-downstream regions, provinces, and cities are represented by r, s, k. Respectively, the disparities in the overall green development status can be measured by the following Theil index:

where x_rsk is the green development level of city k in province s in region r, X―(=1n∑r=1∑s=1∑k=1xrsk) is the average value of all cities, n indicates the number of all cities.

Accordingly, the following metrics can be used to compare the differences in the overall state of green development:

Then T in Eq. (7) will be broken down into

Where X_r is the average level of green development of region r, TA=nrn∑r(Xr―X―)log(Xr―X―) measures variations in the degree of green development across different regions.

Therefore, the disparity T in green development across different regions is a combination of variations within each region and the difference between them.

Next, if we define disparities within each province T_rs of the green development level for province s in region r as follows to measure:

Then T_i in Eq. (8) can be further decomposed into

Where Tpr=nsnr∑j(Xrz―Xr―)log(Xrz―Xr―) measures disparities of the green development level across different provinces in region r.

By substituting T_i in Eq. (11) into Eq. (9), we obtain T.

Equation (12) is the two-stage Theil index decomposition equation that separates the overall regional disparities in green development into the within-province differences (T_A), the between-province differences (T_B), and the between-region differences (T_C). The within-province differences are represented as a weighted average of within-province differences (T_rs), whereas the between-province differences are a weighted average of between-province differences (T_pi).

Temporal Trend Characteristics

The level of green development in the middle and lower reaches is depicted in Fig. 1 as steadily rising between 2010 and 2020. This suggests that the economy and environmental protection are gradually being coordinated synchronously. Since 2011, the “12^th Five-Year Plan” for National Environmental Protection has been implemented, with the goals of actively constructing an eco-friendly society, reducing major pollutants, and enhancing sustainable development. The YRB’s governments are aggressively balancing the environment and economy, giving higher priority to higher-quality economics and promoting green development. From a numerical standpoint, there are still many areas with low levels of green development. The greatest number is only 0.68, which is less than 0.70. While there has been progress in the green development of the middle and lower reaches of the YRB, there is still much room for improvement that requires immediate attention.

A time-series map of the green growth levels in the five provinces is created to further examine the unique circumstances (Fig. 2). The downstream region generally exhibits a higher level of green development compared to the midstream region, and the upward trend is more significant. Inner Mongolia exhibits slow and erratic features, whereas Shandong Province has experienced the fastest progress, followed by Shanxi, Henan, and Shaanxi provinces. With the exception of Shandong Province, the green development in the other four provinces has exhibited weakness over the last two years, with little increasing tendencies. Moreover, Henan and Shaanxi have seen a decline. The downstream region benefits from a comparatively higher level of social-economic development, which fuels the momentum for green development. Conversely, the economic landscape of the midstream region continues to be dominated by traditional sectors, including heavy industry, coal, and other resource-intensive industries, along with conventional farming and livestock practices that suffer from low productivity. Consequently, the transition and upgrading of these industries towards green development may progress at a slower pace. The COVID-19 pandemic has imposed limitations on domestic economic growth over the past two years [33], affecting the progress of green development initiatives in both midstream and downstream areas. Nevertheless, Shandong Province has consistently witnessed steady improvements in its green development levels. This progress can be largely attributed to its solid economic foundation, favorable geographic position, and continuous refinement of its industrial structure. The government has been fully committed to promoting the transformation between old and new kinetic energy, issuing the “Shandong Provincial New Energy Industry Development Plan (2018-2028)” in 2018. As of 2020, the output value of high-tech industries in Shandong constituted 45.1% of the total industrial output value among enterprises of scale.

Fig. 1.

The general temporal features of the middle and lower reaches of green development.

Fig. 2.

Time-series chart of green development level in the mid-downstream areas of the YRB.

Spatial Distribution Characteristics

The spatial distribution of the mid-downstream cities’ green development level in 2010, 2014, 2017, and 2020 is plotted in this study using ArcGIS10.8 (Fig. 3). We categorize the green development level into four categories: low-level, relatively low-level, relatively high-level, and high-level zone, according to the 2017 categorization standard. Clear regional changes in the green development level can be observed in Fig. 3. As Fig. 3 shows, the number of cities in the high-level zone has increased from 3 to 10. The number in the relatively high-level zone has increased from 6 to 20. It has increased in the relatively low-level zone from 7 to 27. Moreover, there are now 6 cities instead of 48 in the low-level zone. The green development level is trending upward, as seen by the decline in low-level zones and the rise in high-level zones. There are noticeable spatial differences in the degree of green development between the middle and lower reaches of the YRB. With a typical north-to-south gradient, the lower reach has a higher overall level of green development than the middle reach. This is primarily because several midstream cities – whose economies heavily rely on mineral resources – are having difficulty modernizing and changing their industrial structures. These regions’ relative dearth of green resources and renewable energy has, in part, impeded their progress toward green development. Hulunbuir has continuously demonstrated superior green development. This may be because large cities have a trickle-down effect that promotes infrastructure development and economic growth. This has promoted green growth by strengthening the cities’ overall resilience [34].

Fig. 3.

Spatial distribution of green development level in the mid-downstream areas: a) Spatial distribution map of 2010, b) Spatial distribution map of 2014, c) Spatial distribution map of 2017, d) Spatial distribution map of 2020.

Dynamic Evolutionary Trajectories of Spatial Disparities

From the overall distribution pattern of Fig. 4, the kernel density curves are all gradually shifted to the right, indicating that the green development level showed a steadily rising trajectory in the mid-downstream regions. The right tail is obviously lengthened; moreover, the right tail is lengthening year by year, indicating that the difference in green development is gradually increasing. From the distribution pattern of the midstream kernel density plot (Fig. 4(a)), the wave pattern in 2010 has several peaks, and in 2014, 2017, and 2020, double peaks and a steady transfer of the low-level peaks to the right are plainly visible. Over the study period, it can be inferred that more cities in the middle reach are at a low level; however, they are gradually advancing towards a higher stage of progress. The downstream kernel density plot (Fig. 4(b)) shows an overall upward trend and a bimodal distribution, with a relatively constant wave peak width and an uneven right tail. This indicates that the level of green development in the lower reach of the YRB is gradually improving. While there is a polarization phenomenon, its distribution always maintains the club convergence phenomenon.

The nation’s sustained commitment to safeguarding and fostering high-quality development within the YRB is bound to promote green development levels. The spatial distribution characteristics of high-quality development, which are “elevated on both extremities and diminished in the middle” are consistent with the polarization phenomenon of the green development levels in the mid-downstream areas [36]. It is noteworthy that the coal chemical industry and low-end manufacturing sector in the midstream area are facing challenges due to their poor development quality and environmental protection pressures. Because of that, efforts to optimize the industrial structure have progressed more slowly, and the unjust industrial structure has caused ecological space to be occupied. As a result, there has been a clear polarization of the middle region in recent years, which has also contributed to an expansion of the absolute gap.

Fig. 4.

The mid-downstream kernel density plots: a) Midstream kernel density plot, b) Downstream kernel density plot.

Spatial Differences and Their Causes

As is evident from Table 3, it can be seen that the total Theil index increased from 0.281 in 2010 to 0.334 in 2020, with fluctuations between 2015 and 2017. Furthermore, the results have generally increased over time, indicating that the regional disparities in the status of green development are on the rise. The between-region difference is the smallest, yet it is trending upward, increasing from 0.025 in 2010 to 0.066 in 2020. There are slight but consistent swings in the between-province differences between 2010 and 2020, with an upward trend between 2010 and 2015 and a declining trend between 2016 and 2020. The within-province differences are significant, indicating a slight upward trend between 2010 and 2020, accompanied by some minor fluctuations.

First, in terms of contribution rate, the most significant factor contributing to the overall difference is the within-province difference in green development, up to 65.942%. However, the contribution rate exhibits a declining trend, decreasing from 65.942% in 2010 to 59.472% in 2020. Subsequently, differences between provinces have little effect on the overall difference. It does contribute slightly more than the between-region differences, which remain relatively constant at 20% to 26%. Finally, there are between-region difference, which contributes relatively little to the overall difference, but increase annually from 9.054% to 19.834%. We can find that the primary cause of spatial variations in green development in the middle and lower reaches of the YRB may be attributed to varying degrees of green development in each city; however, the contribution to the overall difference has decreased in recent years. In addition, this disparity has progressively grown due to imbalances in regional and provincial development.

Fig. 5 indicates that there is still a considerable distance to cover for the overall coordinated development of the YRB, as the overall difference in green development levels is becoming increasingly evident. The disparities in green development between cities within each province are the main reason for this phenomenon. Due to varying degrees of resource reliance and irregularities in the economic growth process, the imbalances in the degree of green development between cities impact the total green development of the mid-downstream areas [37]. The overall design of a strategy to achieve high-quality development in the YRB has, however, lessened this influence. In recent years, provinces have energetically engaged with the significant strategic decisions pertaining to the YRB, actively coordinating and fostering high-quality development within their borders, thereby diminishing the contribution of within-province disparities. Due to the differences in geographical location and natural resources in the middle and lower reaches, the relationship between overall difference and between-region disparity is becoming more and more obvious. Under the influence of the national strategic adjustment of industrial structures, the advantages of the traditional energy extraction industry are diminishing in the midstream area, while downstream green transformation and pollution control face challenges [38]. This requires that the government take a holistic approach to the problem of high-quality development in the YRB, focusing on the differences between midstream and downstream and placing greater emphasis on regional coordination, collaboration, and integrated development.

Fig. 5.

Results of two-stage nested Theil decomposition.

Driver Analysis Based on the GPCA

Before the GPCA, the normalized data is subjected to the KMO test and Bartlett’s Test of Sphericity to assess its efficacy. The results show that the p-value from Bartlett’s test of Sphericity is 0, and the KMO test value is 0.905, both of which pass the significance tests. Therefore, it follows that the GPCA is suitable for this investigation. The five primary principal components with eigenvalues greater than 1 are extracted using the GPCA, yielding a cumulative contribution of 72%. The factor loading matrix of each principal component is presented in Table 3. These five principal components can be synthesized into the economic development quality index (EDQ), composite social life index (CSL), environmental efficiency index (EEI), sustainable development index (SDI), and ecological balance index (EBI) based on the size of the principal component loading on the variables. On one hand, the five indices can be utilized to calculate a comprehensive index through the EM, providing a comprehensive evaluation of green development. On the other hand, an increase in each index’s weight denotes a greater significance of that index for green development. Hence the weights of the following indexes will be determined year by year in this paper: the EDQ, CSL, EEI, SDI, and EBI. After that, it will carry out a comparative analysis to look at how green development in mid-downstream regions is affected dynamically by these drivers.

Driver Analysis Based on a Dynamic Perspective

The EDQ holds relatively significant weight as an indicator of scientific and technological innovation across midstream and downstream regions, depicted in Fig. 6. In the midstream region, the overall weight of the index shows an increasing trend, while the downstream region shows a trend of increasing, decreasing, and then increasing again. It is not difficult to find that the most crucial elements of green development have always been the EDQ in the midstream region. However, the area exhibits a relatively low level and constrained capability in terms of such indicators. Consequently, the progress of the economy, science, and technology is more and more important for green development in the middle reaches. Although the downstream region has a higher level of EDQ, the promotion of green development through that is less evident than in the middle reach. This is due to the large population, more complex resources, and environmental problems necessitating continual improvement of their extensive economic growth paradigm [39].

Representing life and well-being in the midstream and downstream regions, the CSL is a crucial indicator for gauging urban environmental planning in green development as well as economic and social vitality. As can be seen in Fig. 6, the CSL’s score has a relatively low weight. The midstream and downstream regions exhibit a decreasing trend and a noticeable amount of volatility, respectively, suggesting that the CSL’s driving force behind green development is unstable. In the midstream region, the ecological environment is delicate. While environmental protection measures are being built and urban economic activities are taking place, these efforts are not yet sufficient to yield positive environmental effects. As a result, the incentive for green development is lessened. Although the economic growth in the downstream area is high, the drawbacks of the extensive “high pollution + high consumption + high input” development model have started to become apparent due to the disregard for livelihoods and environmental protection. Because of this, the advantages to the economy now exceed the advantages to the environment, underscoring the need for social welfare to be improved. Improving the population’s well-being ought to be the top priority for governments in their future endeavors [40].

Fig. 6.

Line chart of inner motivators of green development level changes in the mid-downstream.

The synergistic economic and environmental benefits of pollution emissions are measured by the EEI. As can be seen in Fig. 6, the driver of green development in the middle and lower regions is given comparatively less weight. It exhibits a slow rising tendency in the downstream region and a little fluctuation in the midstream region. We can learn that enhancing the EEI under carbon emission constraints significantly propels green development. Preserving the environment is a necessary condition for attaining sustainable green development. Three key components of this development strategy include improving the CUISW, cutting back on industrial pollutants, and lowering CO₂ emissions. On the one hand, countries’ development strategies are now oriented toward carbon neutrality as the enhanced greenhouse effect has grown more severe in recent years, aggravating environmental degradation [41]. By including carbon peaking and carbon neutrality in its growth plans, China hopes to achieve a thorough green transformation of its economy and society, promoting synergistic effects in pollution reduction and carbon emissions. The middle and lower reaches are the focus of efforts to balance environmental preservation with economic expansion [42]. On the other hand, technical developments and a growing awareness of environmental protection in cities are forcing them to improve their pollution reduction tactics as urban economic development moves toward high-quality intensification [34].

In the mid-downstream areas of the YRB, the SDI primarily represents the total social water consumption and the level of investment made in environmental preservation. Fig. 6 illustrates that the weight of this index shows a gradually decreasing trend in the middle region. Whereas in the downstream region, it experiences noticeable fluctuations but ultimately tends towards stability overall. In the midstream region, economic expansion leads to significant resource consumption and increased environmental pressure. Coupled with the impacts of urbanization – characterized by infrastructure construction and industrial expansion – this dynamic has positioned sustainable development as an urgent matter that necessitates immediate and sustained attention. The downstream region must increase spending on environmental preservation and improve the efficiency of water resource utilization in order to achieve more sustainable and harmonious green development [13].

The distribution of water resources and population density are the main factors that the EBI represents. From Fig. 6, the weight of this indicator generally increases in the midstream and is comparatively high in the downstream region. However, the volatility is clear, particularly over the last two years. In the midstream region, harmony between humans and the environment consistently promotes green development. The downstream region has a precarious balance between population density and water demand, and the promotion of green development is waning. This can be attributed to the fact that, despite water scarcity in the middle reach, there has been a concerted effort in recent years to protect groundwater resources. The government is committed to the utilization of urban spaces, thereby actively fostering ecological equilibrium. The ecological environment in the downstream area is relatively stable, but in contrast to the middle region, there has been a rapid increase in urbanization in recent years [43]. This rapid urban expansion has disrupted the ecological balance and has gradually diminished the positive impact of green development when assessed.