Investigating the Patterns and Dynamics of Urban Green Space in China’s 70 Major Cities Using Satellite Remote Sensing

: Urban green space (UGS) plays a pivotal role in improving urban ecosystem services and building a livable environment for urban dwellers. However, remotely sensed investigation of UGS at city scale is facing a challenge due to the pixels’ mosaics of buildings, squares, roads and green spaces in cities. Here we developed a new algorithm to unmix the UGSF derived from Landsat TM / ETM / 8 OLI using a big-data platform. The spatiotemporal patterns and dynamics of UGSs were examined for 70 major cities in China between 2000 and 2018. The results showed that the total area of UGS in these cities grew from 2780.66 km 2 in 2000 to 6764.75 km 2 in 2018, which more than doubled its area. As a result, the UGS area per inhabitant rose from 15.01 m 2 in 2000 to 18.09 m 2 in 2018. However, an uneven layout of UGS occurred among the coastal, western, northeastern and central zones. For example, the UGS percentage in newly expanded urban areas in the coastal zone rose signiﬁcantly in 2000–2018, with an increase of 2.51%, compared to the decline in UGS in cities in the western zone. Therefore, the e ﬀ ective strategies we have developed should be adopted to show disparities and promote green infrastructure capacity building in those cities with less green space, especially in western China.


Introduction
Global warming and rapid urbanization have profound effects on urban settlement environments and human well-being [1][2][3]. Therefore, they have become the core elements of the United Nations Sustainable Development Goals 2030 (SDGs 2030), in order to enhance the resilience of cities and communities, as well as to achieve a livable and sustainable city [4,5]. Fast urbanization results in the transformation from natural vegetation or semi-natural lands to human-dominated lands, i.e., buildings, roads, squares, etc., to accommodate the increasing population accompanied by built-up land expansion [6]. These changes also cause a series of environmental issues, such as an ascending trend in the frequency of urban extreme heat events and flood disasters, increased air pollution, and their related deaths [1,7,8].
Urban green space (UGS) is a patchwork of natural or semi-natural ecosystems in cities with either public or private open spaces covered by vegetation, i.e., forests, lawns, etc. UGS, as an important type of urban green infrastructure, plays a critical role in promoting ecosystem services and maintaining biodiversity in cities, i.e., reducing urban heat islands, removing air pollutants and regulating the urban microclimate [9][10][11][12][13]. So a reasonable and scientific layout of UGS contributes to improving the urban ecosystem services in city habitats, as well as to flood protection, cleaner air and leisure and recreation space [14]. However, gaining knowledge about UGS patterns and dynamics is particularly important in urban planning and management. UGSF in 70 cities in 2000 and 2018 was retrieved from Landsat TM/8. The data for city parks were derived from the Geographical Information Monitoring Cloud Platform (http://www.dsac.cn/). The green coverage area of urban parks was acquired from the annual China City Statistical Yearbook published by the National Bureau of Statistics. In addition, auxiliary data, such as the administrative boundaries within China, were obtained at a scale of 1:250,000 from the National Geomatics Center of China.

Development of UGS Data
The UGS is defined by the green coverage fraction of open space in parks, residential areas and roads located in built-up land. We developed a series of algorithms to unmix the urban impervious surface and UGS components at pixel scale, which delineated the UGSF at each pixel with a spatial resolution of 30 m × 30 m [23,34,36]. The spatially explicit UGS was generated to depict the percentage of open area greening spaces, such as trees, grassland and shrubs, in built-up areas ( Figure 1). The UGSF data were downloaded from the open datasets by Kuang et al. (2020) (https://doi.org/10.5281/zenodo.3778424) [37,38]. First, the monthly maximum normalized difference vegetation index (NDVI) for two years was calculated from Landsat TM in 2000 and Landsat 8 OLI in 2018. The monthly NDVI time-series data in one year were combined to generate an annual NDVI maximum value: where NDVI is the NDVI maximum value of the ith month from Landsat images. We retrieved the annual NDVI maximum value ( ) of the two years from Landsat TM/8 OLI using the GEE platform.
Second, a negative correlation between annual maximum NDVI and urban impervious surface areas fraction was found using prior knowledge at the pixel level based on our previous research [23]. A logistic regression model (LRM) of the relationship between NDVI and the fraction of urban impervious surface areas at provincial level was established to acquire urban impervious surface areas fraction, which can be expressed as First, the monthly maximum normalized difference vegetation index (NDVI) for two years was calculated from Landsat TM in 2000 and Landsat 8 OLI in 2018. The monthly NDVI time-series data in one year were combined to generate an annual NDVI maximum value: where NDVI i is the NDVI maximum value of the ith month from Landsat images. We retrieved the annual NDVI maximum value (NDVI Max ) of the two years from Landsat TM/8 OLI using the GEE platform. Second, a negative correlation between annual maximum NDVI and urban impervious surface areas fraction was found using prior knowledge at the pixel level based on our previous research [23]. A logistic regression model (LRM) of the relationship between NDVI and the fraction of urban impervious surface areas at provincial level was established to acquire urban impervious surface areas fraction, which can be expressed as Remote Sens. 2020, 12,1929 4 of 13 where p, α and β are the regression coefficients of the LRM in some provinces; F UGS (p) is the calculated UGSF in some province in a specific period. A more detailed description of mapping UGSF may be found in previous studies [23,36,37,[39][40][41]. Third, UGSFs with 30 m × 30 m were validated using Google Earth images in 2000 and 2018. The UGSFs in the 70 cities were mapped for this analysis.

Accuracy Assessment
The overall accuracy of urban land or built-up land classification in 2000 and 2018 was more than 93.76%, from an accuracy assessment by CLUD [33][34][35]. The accuracy of UGSF was assessed using Google Earth images. A total of 1000 windows, with 3 × 3 pixels (90 m × 90 m), were randomly selected from the 70 cities. The root mean square error (RMSE) and correlation coefficient (R) were adopted to estimate the accuracy of UGSF in both 2000 and 2018. The accuracies of remotely sensed UGSFs with RMSE and R were 0.14 and 0.91, respectively ( Figure 2). We examined some low estimations of UGSFs in semi-arid and arid cities and found they were due to the inability to see some sparse green coverage in those areas.
Remote Sens. 2020, 11, x FOR PEER REVIEW 4 of 14 where p, α and β are the regression coefficients of the LRM in some provinces; F ( ) is the calculated UGSF in some province in a specific period. A more detailed description of mapping UGSF may be found in previous studies [23,36,37,[39][40][41]. Third, UGSFs with 30 m × 30 m were validated using Google Earth images in 2000 and 2018. The UGSFs in the 70 cities were mapped for this analysis.

Accuracy Assessment
The overall accuracy of urban land or built-up land classification in 2000 and 2018 was more than 93.76%, from an accuracy assessment by CLUD [33][34][35]. The accuracy of UGSF was assessed using Google Earth images. A total of 1000 windows, with 3×3 pixels (90 m × 90 m), were randomly selected from the 70 cities. The root mean square error (RMSE) and correlation coefficient (R) were adopted to estimate the accuracy of UGSF in both 2000 and 2018. The accuracies of remotely sensed UGSFs with RMSE and R were 0.14 and 0.91, respectively ( Figure 2). We examined some low estimations of UGSFs in semi-arid and arid cities and found they were due to the inability to see some sparse green coverage in those areas.

Analysis of UGS Patterns
The 70 cities, distributed in coastal, central, northeastern and western zones, were selected from the representative megacities and large cities with more than one million people across China (Table  1).
UGSF was calculated as the percentage of UGS located in a built-up area. In addition, the UGSFs in the expanded areas were calculated as represents the fraction or percentage of UGS in the i th city, is the UGS area in the i th city, and represents the area of built-up areas in the i th city.

Analysis of UGS Patterns
The 70 cities, distributed in coastal, central, northeastern and western zones, were selected from the representative megacities and large cities with more than one million people across China (Table 1). UGSF was calculated as the percentage of UGS located in a built-up area. In addition, the UGSFs in the expanded areas were calculated as Remote Sens. 2020, 12, 1929 5 of 13 where UGSF i represents the fraction or percentage of UGS in the ith city, UGSA i is the UGS area in the ith city, and UA i represents the area of built-up areas in the ith city.

Analysis of Urban Greening Effectiveness
The effectiveness of urban greening was characterized as the UGSF in newly expanded urban areas in each city. We adopted the National Ecological Garden City standard (2016), published by the Ministry of Housing and Urban-Rural Development. This standard is an assessing index for choosing which cities should be named a National Ecological Garden City, as shown below: where G i represents whether the ith city is a National Ecological Garden City (if it is, the value of G i is 1, if not, the value is 0).

UGS Changes in China's 70 Cities in 2000-2018
Since the beginning of the 21st century, China has undergone rapid urban expansion. As a major type of urban land cover, UGS has also exhibited massive growth, according to the UGSF dataset at 30-m resolution. In the 70 cities, the total area of UGS grew from 2780.66 km 2

Regional Divergence of UGS Growth
The dramatic discrepancies in UGS growth among different zones of China from 2000 to 2018 are shown in Table 2. The UGS in the coastal zone experienced the most rapid growth process since 2000. The areas of UGS in the built-up coastal areas increased from 1675.77 km 2 in 2000 to 4129.02 The top 10 cities, with the highest increments of UGS at the city level from 2000 to 2018, were derived and ranked (Table 1). Among these cities, the highest increase in area of UGS was in Beijing (Figure 3). The increment of UGS was 516.89 km 2 between 2000 and 2018 in Beijing, which was 3.25 times that of the other cities (Table 2). Meanwhile, among the top 10 cities, 8 cities were in the coastal zone, and only Changchun and Hefei were in the northeast and central zones, respectively ( Figure 3, Table 2).

Regional Divergence of UGS Growth
The dramatic discrepancies in UGS growth among different zones of China from 2000 to 2018 are shown in Table 2. The UGS in the coastal zone experienced the most rapid growth process since 2000. The areas of UGS in the built-up coastal areas increased from 1675.77 km 2 in 2000 to 4129.02 km 2 in 2018, with a total increase of 2453.25 km 2 (Table 3). Specifically, the major cities in the coastal zone have paid more attention to designing their forests and parks during the process of urban development. For example, a large number of urban parks, such as Xiangmi, Tanglangshan Country and Nanshan, were built in Shenzhen (Figure 4). The green space area located in Shenzhen's built-up area increased from 156.83 km 2 in 2000 to 301.87 km 2 in 2018 (Table 2). Meanwhile, the proportion of UGS increased from 31.27% to 35.10% in the same period.
Compared to other zones, the northeast zone experienced the lowest increase in UGS. Although the national development strategy of the Northeast Revitalization Plan provided an opportunity for the development of UGS construction, the area of UGS in the northeastern built-up areas increased by only 461.28 km 2 ( Table 3) The proportions of UGS among different zones also showed tremendous differences. The UGS data in 2018 for the built-up areas in the coastal zone showed a higher proportion than that for other zones. The UGS percentage in coastal zone cities was 29.68%, which was higher than the average level of the 70 cities (28.06%). The UGS percentage in western zone cities was the lowest among the four zones. The UGS percentage in those cities was only 24.11%, which was 3.95% lower than the average value in the 70 cities. The UGS percentages of 2018 in the built-up areas of 2000, in the northeast and central zone cities, was 29.46% and 24.83%, respectively. The cities in the western and central zones also have paid more attention to the development of UGS since the beginning of the 21st century. The area of UGS in the built-up areas of the central zone increased from 318.67 km 2 in 2000 to 881.63 km 2 in 2018. For instance, a series of UGSs were built in Wuhan, such as Wuhan East Lake Greenway, Zhengzhou People's Park and Changsha Yang Lake Wetland Park (Figure 4). Meanwhile, the UGS area in the western zone's built-up areas also increased, from 415.55 km 2 in 2000 to 922.15 km 2 in 2018. Many UGSs were built in the western zone, such as Chengdu Tianfu Jincheng ecological park, Chongqing Nanshan Park and Yuanbo Park (Figure 4). For example, the overall goal of urban greening in Chongqing is to build a landscape garden city. The green space area of the main urban area increased from 74.94 km 2 in 2000 to 115.32 km 2 in 2018. The proportions of UGS among different zones also showed tremendous differences. The UGS data in 2018 for the built-up areas in the coastal zone showed a higher proportion than that for other zones. The UGS percentage in coastal zone cities was 29.68%, which was higher than the average level of the 70 cities (28.06%). The UGS percentage in western zone cities was the lowest among the four zones. The UGS percentage in those cities was only 24.11%, which was 3.95% lower than the average value in the 70 cities. The UGS percentages of 2018 in the built-up areas of 2000, in the northeast and central zone cities, was 29.46% and 24.83%, respectively.

Improvements in Urban Greening in Newly Expanded Areas in Cities
Since the beginning of the 21st century, urban greening has become increasingly important in line with strengthening the cities' ecological construction and urban greening infrastructure. We assessed the UGS changes in the newly urban expanded areas (UEAs) in China, from 2000 to 2018 ( Figure 5). The proportion of UGS in UEAs in the selected 70 cities increased from 2000 to 2018, especially in the eastern zones. The average proportion of UGS in the UEAs of these cities in 2018 was 28.43%, which was higher than the average proportion of UGS (27.54%) in the built-up areas in 2000 (Table 4).
Remote Sens. 2020, 12, 1929 8 of 13 In 2000, the proportion of UGS in the UEAs of 41 of the 70 cities was more than 2% higher than that in the built-up areas, with more than half of them distributed in the coastal and northeastern zones. Among them, the UGS percentage of 11 cities was more than 5% higher in the UEAs than it was in the built-up areas, which were mainly distributed in the coastal zone, such as Beijing, Tianjin, Baoding, Yancheng and Xuzhou.  The proportion of UGS in the UEAs of the coastal zone increased significantly from 2000 to 2018, with an increase of 2.51% percent, from 28.24% to 30.75% (Table 4). In the same period, the proportion of UGS in the central zone increased slightly, from 23.69% in the buit-up areas of 2000, to 25.53% in the newly expanded areas, during 2000-2018. On the contrary, with the built-up areas and newly expanded areas in the northeastern and western zones, the proportions of UGS decreased by 3.71% and 2.04%, respectively.
In 2000, the proportion of UGS in the UEAs of 41 of the 70 cities was more than 2% higher than that in the built-up areas, with more than half of them distributed in the coastal and northeastern zones. Among them, the UGS percentage of 11 cities was more than 5% higher in the UEAs than it was in the built-up areas, which were mainly distributed in the coastal zone, such as Beijing, Tianjin, Baoding, Yancheng and Xuzhou.
The implementation of ecological greening projects in cities since 2000, such as the ecological corridors along the 5th ring road and Olympic Park, has significantly improved the urban greening in Beijing. For example, the road greening with a 100-m width and the large-scale urban parks along the 5th ring road were initiated to form ecological corridors, to prevent urban sprawl and improve the urban environment ( Figure 6). We examined the UGS area's increase continuously, from 46 the 5th ring road were initiated to form ecological corridors, to prevent urban sprawl and improve the urban environment ( Figure 6). We examined the UGS area's increase continuously, from 46.26 km 2 in 2000 to 51.95 km 2 in 2018, along the 800-m buffer zones with the 5th ring road greening. Meanwhile, the area of UGS increased from 304.67 km 2 in 2000 to 821.56 km 2 in 2018 in Beijing. We also investigated the UGS growth in the UEAs of the western city of Lhasa during 2000-2018, with an increased area of 9.25 km 2 , which was a result of improvements in urban greening ( Figure 6). The proportion of UGS in the UEA was 41.32% in 2018-the highest percentage among the 70 citieswhich had increased by 7.29% when compared to the proportion of UGS in the built-up areas (34.03%) in 2000. Figure 6. Landsat Images of typical cities and major parks.

Satellite Remote Sensing Plays a Pivotal Role in Monitoring UGS Change
UGS is an important component of urban land-cover types. It is always mosaicked in different functional areas, such as parks, streets and residential areas. Due to the characteristics of the UGS Figure 6. Landsat Images of typical cities and major parks.

Satellite Remote Sensing Plays a Pivotal Role in Monitoring UGS Change
UGS is an important component of urban land-cover types. It is always mosaicked in different functional areas, such as parks, streets and residential areas. Due to the characteristics of the UGS layout, it is difficult and time-consuming to retrieve details of UGS distribution using the statistical or human investigation method. Satellite remote sensing helps to acquire long-term and large-scale monitoring of UGS patterns and their spatiotemporal changes, which provides a pivotal data source for monitoring and assessing urban surface environments and analyzing their ecological services.
In this research, we mapped the fraction of UGS and other land-cover types. In 2018, the remotely sensed UGS area of 70 major cities was 6764.76 km 2 . In statistical data, the parks are viewed as management units, and impervious land-cover types, such as roads and squares, were also included and added to the area of UGS. However, the differences between UGS and the impervious areas in the parks can be delineated via remotely sensed observation, with more spatially explicit features.

The Massive Effectiveness of City Greening Since the Beginning of the 21st Century in China
A series of policies have been aimed at designing ecocities or garden cities that pay great attention to urban greening and ecological construction. In 2016, the Ministry of Housing and Urban -Rural Development published the national ecological garden city assessment criteria. According to the criteria, the urban green coverage percentage in built-up areas needed to be higher than 28% (Table 5). According to our assessment, 41 cities, including Beijing, Tianjin and Nanjing, have fully reached that goal. The construction of urban parks has significantly addressed the lack of green space since 2000. Statistical data showed that the green coverage area of urban parks in 70 cities increased 6. layout, it is difficult and time-consuming to retrieve details of UGS distribution using the statistical or human investigation method. Satellite remote sensing helps to acquire long-term and large-scale monitoring of UGS patterns and their spatiotemporal changes, which provides a pivotal data source for monitoring and assessing urban surface environments and analyzing their ecological services.
In this research, we mapped the fraction of UGS and other land-cover types. In 2018, the remotely sensed UGS area of 70 major cities was 6764.76 km 2 . In statistical data, the parks are viewed as management units, and impervious land-cover types, such as roads and squares, were also included and added to the area of UGS. However, the differences between UGS and the impervious areas in the parks can be delineated via remotely sensed observation, with more spatially explicit features.

The Massive Effectiveness of City Greening Since the Beginning of the 21st Century in China
A series of policies have been aimed at designing ecocities or garden cities that pay great attention to urban greening and ecological construction. In 2016, the Ministry of Housing and Urban -Rural Development published the national ecological garden city assessment criteria. According to the criteria, the urban green coverage percentage in built-up areas needed to be higher than 28% (Table 5). According to our assessment, 41 cities, including Beijing, Tianjin and Nanjing, have fully reached that goal.
The construction of urban parks has significantly addressed the lack of green space since 2000. Statistical data showed that the green coverage area of urban parks in 70 cities increased 6.62-fold between 2000 and 2017, from 504.39 km 2 to 3842.26 km 2 . With the rapid development of cities in the coastal zone, the green coverage area of urban parks has also increased significantly. The UGS in urban parks in 32 coastal zone cities increased from 301.71 km 2

Enhancing Urban Ecosystem Services Induced by Urban Green
The theory of urban ecology has emphasized the understanding and analysis of the complexity of urban ecosystems, including the mechanism of the influence of spatial patterns and structural heterogeneity on urban ecological processes and services, as well as reduction in the negative effects of human activities [42][43][44][45].
As an important part of the urban ecosystem services, UGS plays an important role in improving the urban environment, especially in increasing soil infiltration, reducing air pollution, conserving water and beautifying the city environment [11][12][13]. UGS can also serve as an effective strategy in offsetting CO 2 emissions. The appropriate area and proportion of UGS in cities can regulate the urban microclimate [25] and reduce the urban heat island [46,47]. The experimental observations found that the diurnal air temperature range (DTR) of residential building areas is significantly lower than that of parks (∆DTRa = 2.53 ± 1.93 ºC), with a maximum difference of 3.54 ± 1.96 ºC in autumn. Therefore, UGS shows a cooling effect, especially at nighttime. These ecosystem services of urban greening are beneficial to human health, city environmental improvement, and the achievement of sustainable development goals. To promote the ecosystem services, understanding UGS patterns and dynamics is critical for guiding urban ecological planning and construction. The International Union of Forest Research Organizations established a special project team to support global cooperation and information exchanges, for improving urban ecosystem services with urban greening.

Conclusions
This study delineated UGS patterns and dynamics in 70 Chinese cities in 2000-2018, using the big-data analysis platform. The results show that UGS in major cities has increased significantly from 2000 to 2018 in China. The total area of UGS in these 70 cities increased from 2780.66 km 2 to 6764.76 km 2 during this time period. We also found that the percentage of UGS was strengthened in newly urban expanded areas, especially in the coastal zone.
However, we also found that uneven development appears in urban greening among the four zones. The proportion of UGS in the built-up areas still remained low, especially in the western zone, where the built-up areas still needed to be further optimized in order to enhance the urban greening level in future urban development.