An Investigation on Shenzhen Urban Green Space Changes and Their Effect on Local Eco-Environment in Recent Decades

Abstract

Rapid urbanization and population growth impact enormous pressures on urban natural, economic and social environments. The quantitative analysis of urban green space (UGS) landscape dynamics and their impact on the urban eco-environment is of great significance for urban planning and eco-environment protection. Taking Shenzhen as an example, the UGS landscape changes and their impact on urban heat islands (UHI), surface wetness, air pollution and carbon storage were comprehensively investigated with Landsat and MODIS images. Results showed a large number of lands transferring from UGS to non-UGS from 1978 to 2018, especially for cropland. Built-up regions have adverse influences on eco-environment factors, and then they suffer high SUHI and AOD and low humidity and carbon storage. The growth of built-up areas not only enlarges the area of SUHI, but also enhances the intensity of heat islands. On the contrary, UGS patches have beneficial influences on all eco-environment factors and then enjoy a better eco-environment, including low SUHII, high surface wetness, high carbon storage and low AOD. It is expected that this study could provide scientific support for UGS plans and for conserving and sustainable urban development for developing cities.

1. Introduction

Rapid urbanization has resulted in a high concentration of people living in cities. In 2020, more than 50% of the world’s population lived in urban areas, and this number is projected to increase to 66% by 2050 [1]. Rapid population growth and urbanization has put enormous pressures on urban natural, economic and social environments [2,3,4], resulting in urban heat islands (UHI), air pollution, water shortage, biodiversity loss, etc. [5,6,7]. For instance, the fragmentation of habitat caused by built-up construction severely influences the living environment of animals and plants [8]. How to mitigate the negative effects of urbanization has attracted substantial attention in recent years [8,9,10].

Urban green space (UGS), including vegetation, water and croplands, is playing an essential role in adjusting the urban climate [11,12] by providing important ecological functions and services for urban environmental protection, public health, biodiversity conservation, etc. How to plan, maintain, utilize and conserve UGS have become important tasks for city authorities, as the effects of UGS in alleviating eco-environmental threats caused by urbanization have been broadly recognized [12,13]. Remote sensing has been widely used in tracking UGS dynamics, due to its high efficiency in large scale, comprehensive and periodical investigations [14,15]. For quantitatively analyzing UGS changes, times-series UGS maps should firstly be interpreted from remote sensing images, then quantitative indicators for evaluating UGS changes would be built and used to find the patterns or characteristics of UGS dynamics.

Besides this topic, many studies have focused on analyzing UGS’s effects on the urban eco-environment, including cooling, humidification, anti-pollution, sterilization, reducing noise, soil and water conservation [16,17]. Some research stated that UGS fragmentation will severely damage the natural habitat of animals and plants [18]. Ref. [19] discovered that increasing UGS coverage can alleviate the concentration of atmospheric particles. Ref. [20] modeled the mechanism of UGS’s effects on local precipitation and temperature by quantitatively analyzing their relationship. Ref. [21] revealed that the area and spatial configuration of impervious surfaces and UGS types have a significant correlation with water organic pollution. With regard to the UGS cooling effects, researchers explored the mechanism and trends of UGS’s impacts on the urban thermal environment by quantitatively analyzing the spatial-temporal dynamics of vegetation coverage, vegetation index and urban thermal environment [22,23,24]. Refs. [15,25] focused on differentiating the cooling effect and impact area of different UGS types. Ref. [26] estimated the forest carbon storage and spatial distribution of China in the past 50 years with forest inventory data.

Although UGS dynamics in terms of LULC change have been widely investigated, more detailed analysis of its landscape changes and related ecological environment responses have been poorly explored for many developing cities [27,28,29,30,31]. The UGS landscape, including distribution, structure, function and development, and its interaction with the ecological environment have great effects on local ecological processes and urban animal and plant communities, climate, hydrology, soil, etc. by influencing local and regional material and energy flow [32,33,34]. Specific and detailed studies on the issue could deepen our understanding of urban environmental problems and help authorities develop more rational and effective polices.

As a special economic zone in China, Shenzhen has experienced rapid development and urbanization since 1978. In the urbanization process, the integrity, diversity, and distribution of UGS have been intensively interfered with by various human activities, such as urban extant, road construction, etc. [35]. Taking Shenzhen city as an example, this study aimed to comprehensively explore the spatial-temporal dynamics of UGS and its influences on urban eco-environments in developing cities with times series Landsat images and MODIS images. It is expected that this study could fully reveal the impact of rapid urbanization on the type, number, spatial distribution and configuration of UGS; finely evaluate the ecological effects of UGS changes and provide scientific support for UGS plans and for conserving and sustainable urban development for similar cities.

2. The Study Area and Data

2.1. Study Area

Shenzhen is in the southern part of Guangdong Province, China, on the east coast of the Pearl River Delta adjacent to Hong Kong (Figure 1). The total land area and population of Shenzhen are 1996 km² and 13.02 million, respectively [36]. As the first Special Economic Zone of China, Shenzhen has witnessed remarkable economic development and urbanization since 1978 [37]. By 2019, its gross domestic product (GDP) reached $389.8 billion [38], ranked third in China and 21st in the world.

2.2. Data

A Landsat image is suitable for this study due to its long-time and continuous observation records. Considering that the city has stable weather conditions and evergreen plants in winter, images in this season with little cloud cover were carefully selected and downloaded from the Chinese Geospatial Data Cloud (http://www.gscloud.cn/, accessed on 18 June 2020). As it is difficult to obtain cloud-free Landsat images for the 40 years, rough quinquennial images were selected to depict different stages of Shenzhen’s development.

MODIS products have been successfully used to retrieve aerosol optical depth (AOD) and estimate ground particulate pollution concentrations [34,39,40]. Given cloud-free MODIS images are available for each year after 2000; its yearly products were used to generate high temporal resolution retrieval of environmental factors, so that more about the city’s environmental changes could be exposed. The yearly 1-km MODDIS products were downloaded from the NASA website (https://www.nasa.gov/, accessed on 18 June 2020) for investigating annual AOD dynamics in this city. Image reprojection, Mosaicking and clipping were conducted in the MODIS Reprojection Tool.

High resolution images, including aerial images before 2001, Quickbird images in 2003 and 2008 and Worldview images in 2013 and 2018, were obtained from the Shenzhen Municipal Planning and Natural Resources Bureau.

Meteorology data, vector maps, statistical data and carbon storage data were collected from the Meteorological Bureau of the Shenzhen Municipality, the Resource and environment science and data center of the Chinese Academy of Sciences, the statistical yearbook of the city and the Guangdong Forestry Survey and Planning Institute, respectively. All data information are shown in

Table 1. Data used in this study.

Date	Collection	Spatial Resolution (m)	Temporal Resolution (Day)
2 November 1978	Landsat MSS	80	18
22 November 1984	Landsat TM	30	16
8 November 1988	Landsat TM	30	16
15 November 1993	Landsat TM	30	16
4 November 1998	Landsat TM	30	16
26 October 2003	Landsat TM	30	16
15 November 2008	Landsat TM	30	16
29 November 2013	Landsat OLI&TIRS	30	16
3 October 2018	Landsat OLI&TIRS	30	16
26 October 2003	MODIS MOD02	1000	1
27 October 2004	MODIS MOD02	1000	1
24 October 2005	MODIS MOD02	1000	1
26 October 2006	MODIS MOD02	1000	1
22 October 2007	MODIS MOD02	1000	1
15 November 2008	MODIS MOD02	1000	1
11 November 2009	MODIS MOD02	1000	1
12 November 2010	MODIS MOD02	1000	1
15 November 2011	MODIS MOD02	1000	1
15 November 2012	MODIS MOD02	1000	1
29 November 2013	MODIS MOD02	1000	1
27 November 2014	MODIS MOD02	1000
26 November 2015	MODIS MOD02	1000	1
29 November 2016	MODIS MOD02	1000	1
30 November 2017	MODIS MOD02	1000	1
3 October 2018	MODIS MOD02	1000	1
26 October 2003	Quickbird	0.61	1-6
15 November 2008	Quickbird	0.61	1-6
29 November 2013	WorldView-2	1.85	1
3 October 2018	WorldView-2	1.85	1
29 November 2013	Meteorology data		hourly
2013	carbon storage data		Quinquennial

.

3. Methods

The study was conducted according to the following workflow (as shown in Figure 2). At first, the spatial-temporal dynamics of UGS landscapes were analyzed based on nine quinquennial UGS maps, generated from Landsat images; then, urban environment factors (Land surface temperature (LST), Surface wetness, Carbon storage, Air pollution) were retrieved from images with corresponding models; lastly, the influences of UGS dynamics on environment factors were evaluated.

3.1. UGS Mapping and Landscape Pattern Analysis

In order to reveal UGS spatial-temporal dynamics during the 40 years, quinquennial UGS maps were first generated from Landsat images with the random forest classifier. At each mapping year, 100 samples (60% for training and 40% for validation) for each UGS subtype (vegetation, water and croplands) and non-UGS subtype (built-up and barren) were carefully selected from high-resolution images. To guarantee classification accuracy, experts summarized the characteristics of land types through carefully exploring the image, and then they selected representative and widely distributed samples from larger patches. The trial-and-error method was used to obtain better samples, with cross validation and accuracy evaluation.

Then, the spatial-temporal dynamics of UGS were detected by comparing UGS classification results at different mapping years, which is called post-classification change detection [41,42]. Based on the quinquennial UGS maps, the percentage of landscape (PLAND), landscape shape index (LSI), Euclidean area-weighted mean nearest-neighbor distance (ENN_AM) and cohesion index (COHESION) were selected to conduct quantitative investigations on UGS heterogeneity, fragmentation, diversity and evenness. The details of UGS mapping and landscape pattern analysis can be found in Ref. [15].

3.2. Ecological Environment Factors Retrieving

Typical urban environmental factors, including Land surface temperature (LST), surface wetness, carbon storage and aerosol optical depth (AOD), were selected and retrieved from images for evaluating the influences of UGS on the urban environment.

3.2.1. LST and UHI

The mono-windows algorithm [43,44] was used to retrieve the LST in this study. Compared with the radiative transfer equation and single-channel algorithm, it just requires three standard parameters, rather than real-time atmospheric parameters [45,46,47], which are not available for Landsat images before 2000. More details about how to calculate these parameters can be found in [45]. Our experiment proved that the retrieved LST matched meteorological data well, and the difference between them was about 1.0 K.

The Surface Urban heat island (SUHI) of this city was further analyzed. SUHI is considered as the surface LST difference between an urban area and a 10-km buffer region in the surrounding rural area [48]. To illustrate SUHI distribution clearly, pixels were divided into seven categories according to their SUHI values: (1) strong heat island (SHI, SUHII > 5 K); (2) moderate strong heat island (MSHI, 3 K > SUHII > 5 K); (3) weak heat island (WHI, 3 K > SUHII > 1 K); (4) normal area (NA, 1 K > SUHII > −1 K); (5) weak cold island (WCI, −1 K > SUHII > −3 K); (6) moderate strong cold island (MSCI, −3 K > SUHII > −5 K) and (7) strong cold island (SCI, −5 K > SUHII).

3.2.2. Surface Wetness

With the development of remote sensing technology, many humidity indexes have been developed to retrieve surface wetness, such as the Normalized Difference Water Index (NDWI), Modified Normalized Difference Water Index (MNDWI), Tasseled Cap Wetness Index (TC-wetness) and Normalized Difference Moisture Index (NDMI) [49,50,51,52,53].

Our experiment showed that the correlation coefficient between NDMI and meteorological data was 0.94, much higher than others. Thus, NDMI was selected for retrieving surface wetness in this area, the formula of which is shown as [53]

$$\mathrm{NDMI}=\frac{{\rho }_{NIR}-{\rho }_{MIR}}{{\rho }_{NIR}+{\rho }_{MIR}}$$

(1)

where ${\rho }_{NIR}$ and ${\rho }_{MIR}$ represent the near-infrared band and middle infrared band, respectively.

3.2.3. Carbon Storage

For vegetation carbon storage estimation, the field measurement needs a lot of labor force, materials and time [54]. Recently, remote sensing technology has been widely used in estimating vegetation biomass and carbon storage, as it can quickly and periodically obtain the vegetation status over a large area [55,56]. The CITYgreen’s carbon module can successfully quantify the role of urban forests through distinguishing atmospheric carbon dioxide and stored carbon dioxide [57,58,59,60,61,62]. In this study, we used a modified model derived from the CITYgreen, based on the ground data of Shenzhen from Ref. [61], to retrieve carbon storage with vegetation indexes, texture and the terrain factor. The model is shown below as

$$\mathrm{Y}=83.859+54.427 \times \mathrm{ARVI} - 85{.816 \times \mathrm{SR}}_{32}+6.920 \times \mathrm{MSR}+4.637 \times \mathrm{G}\_\mathrm{Mean}+0.020 \times \mathrm{DEM}$$

(2)

where

$$\mathrm{ARVI}=\frac{{\rho }_{NIR}-{\rho }_{RB}}{{\rho }_{NIR}+{\rho }_{RB}}$$

$${\rho }_{RB}=2{\rho }_{RED-}{\rho }_{BLUE}$$

$${\mathrm{SR}}_{32}=\frac{{\rho }_{NIR}}{{\rho }_{RED}}$$

$$\mathrm{MSR}=({\rho }_{NIR}/{\rho }_{RED}-1)/(\sqrt{{\rho }_{NIR}/{\rho }_{RED}}+1)$$

Here, $\mathrm{Y}$ indicates the carbon storage; $\mathrm{G}\_\mathrm{Mean}$ and $\mathrm{DEM}$ represent the mean texture and Digital elevation model, respectively.

3.2.4. MODIS AOD Retrieval

In recent years, satellite observation aerosol data has been used to retrieve ground air pollution. Studies have shown that there is a strong positive correlation between aerosol optical depth (AOD) and PM2.5 [63,64,65,66]. We used the Dense Dark Vegetation (DDV) algorithm to retrieval AOD [67]. Compared with the deep blue algorithm, this algorithm has better stability and a higher retrieval accuracy. The details of this algorithm could be found in a paper published by Remer [68].

3.3. Vegetation Fractional Coverage Calculation

The vegetation fractional coverage (VFC), in terms of the percentage of the vertical projected area of vegetation on the ground in the total area [69,70], is widely used to measure the degree of vegetation coverage. In this study, we used VFC to evaluate the growth status of vegetation. A typical formula for VFC is as follows:

$$\mathit{VFC}=\frac{{\mathit{NDVI}-\mathit{NDVI}}_{\mathit{min}}}{{\mathit{NDVI}}_{\mathit{max}}{-\mathit{NDVI}}_{\mathit{min}}}$$

(3)

where NDVI is the normalized difference vegetation index; ${\mathit{NDVI}}_{\mathit{max}}$ and ${\mathit{NDVI}}_{\mathit{min}}$ are considered as the maximum and minimum values in the NDVI image, respectively. In this study, the 5% and 95% levels in the percentile ranks of all the NDVI values were selected as ${\mathit{NDVI}}_{\mathit{min}}$ and ${\mathit{NDVI}}_{\mathit{max}}$, respectively, to avoid the influence of extreme values.

3.4. Spearman’s Correlation Coefficient

In this study, the relationships between the UGS and ecological environment factors were analyzed with the Spearman’s correlation coefficient and 1000 samples evenly selected from UGS patches. The Spearman’s correlation coefficient is an effective way to investigate the relationship between two variables (X and Y) without considering their distribution and sample size. It has been widely used in related studies [15,71,72,73].

For a pixel i, the vegetation fractional coverage (VFC) [74] is assigned as Xi, and a corresponding ecological environment value (e.g., SUHII) is named as Yi. Then, two ranking sets, x and y, were obtained by sorting X and Y in descending or ascending order synchronously. The correlation coefficient between VFC and SUHII can be calculated as follows [75]:

$$r_{xy}=\frac{{\sum }_{i=1}^n\left(x_i-\overline{x}\right)\left(y_i-\overline{y}\right)}{\sqrt{{\sum }_{i=1}^n{\left(x_i-\overline{x}\right)}^2}\sqrt{{\sum }_{i=1}^n{\left(y_i-\overline{y}\right)}^2}}$$

(4)

where the elements x_i and y_i are the rank of X_i and the rank of Y_i, respectively; $\overline{x}$ and $\overline{y}$ indicate the mean value of each variable, respectively. The value range of Spearman’s r is [−1, +1], where 1 indicates the strongest positive correlation, and −1 means the strongest negative correlation.

4. Results

4.1. UGS Dynamics

As shown in Figure 3, nine quinquennial UGS maps were generated from Landsat images with the random forest classifier and samples evenly selected from images by experienced experts. An accuracy assessment using a confusion matrix with a ground truth from high-resolution images showed that the overall accuracy for all UGS classifications was higher than 0.9.

In Figure 3, the tendency of lands transferring from UGS to non-UGS is very clear and remarkable from 1978 to 2018. Specifically, cropland, which occupied 17% of the city in 1978, had totally transferred into built-up and bare lands by 2018. Oppositely, non-UGS (especially the built-up) regions had rapidly expended in the 40 years and accounted for about 60% in 2018. It can be noticed that the western part of expanding built-up area mainly came from land reclamation, which achieved 120 square kilometers by 2020 [76]. The landscape pattern analysis shows that human activities have ever-intensively influenced the area and shape of UGS patches in the developing period. When the awareness of UGS protection has been raised in recent years, UGS landscape patterns remained unchanged successfully.

4.2. Influences of UGS on UHI

The Quinquennial SUHI maps of Shenzhen were derived from Landsat images, shown in Figure 4, for monitoring the thermal environment changes during the study years. It can be noticed that there are no LST or SUHI maps in 1978, as Landsat MSS had no TIR band. Furthermore, the influence of UGS on SUHI has been evaluated and shown in Figure 5.

From 1984 to 1998, there were several cold islands (SCI, MSCI and WCI), while no evident SUHI (WHI, MSHI or SHI) was observed in Shenzhen city. Since 2003, SUHI has dramatically expanded and covered all the city by 2018, except several cold islands located in forest parks. We noticed that the SUHI expanding is consistent with built-up growth.

Figure 5a shows that the vegetation fractional coverage (VFC) of UGS has a significant negative influence on SUHI (r = −0.78), which means that vegetation with a high VFC can effectively alleviate urban heat [77]. Figure 5b shows that all UGS subtypes (water, forest and cropland) had negative SUHII values, while non-UGS subtypes (barren and built-up) had positive SUHII values. This also verifies UGS’s alleviation effects on SUHI. It can be noticed that the SUHII of built-up areas gradually increased during the 40 years, which implies that the growth of the built-up area did not only result in SUHI expanding, but it also enhanced its intensity. Consequently, the SUHII of forest decreased, evidently due to the amplifying temperature difference between vegetation and non-UGS.

4.3. Influences of UGS on NDMI

The Spearman correlation between the relative humidity data observed by meteorology stations and retrieved NDMI was mapped in Figure 6. It shows that the two variables have a strong positive correlation (R² = 0.948), indicating that it is feasible to use NDMI to reflect the surface wetness. Then, nine Quinquennial NDMI maps of Shenzhen were retrieved from Landsat images and are shown in Figure 7. It can be noticed that there is no NDMI map for the year 1978, as Landsat MSS had no TIR band.

Figure 7 shows that NDMI values of most pixels within the city were between −0.2 and 0.4, which implies that the city does not have humid weather in the winter. During the 40 years, a very slow downward trend was witnessed for the whole city, except for a dramatic decrease in the western part (from 0.6–0.8 to 0.2–0.4), probably caused by reclamation activities. Although a large area of cropland has transformed to built-up or bare land, the change of the NDMI value is not remarkable. This may be because the cropland and bare land have a similar NDMI, which is just a little higher than that of built-ups (as shown in Figure 8b).

Figure 8a shows that the UGS VFC had a significant positive influence on NDMI (r = 0.84), which means that vegetation with a high VFC can effectively enhance local humidity [78]. Figure 8b shows that UGS subtypes, especially water and forest, had a much higher NDMI than others. Given that the area of water and forest remained unchanged while all cropland transferred to non-UGS during the 40 years, the overall NDMI of the city slowly and inevitably decreased, which is consistent with Ref. [79].

4.4. Influences of UGS on Carbon Storage

The Spearman correlation between the ground data and retrieved value was analyzed, as shown in Figure 9. The two had a strong positive correlation (R² = 0.99), which indicates that Equation (2) is suitable for retrieving the carbon storage in city. Then, Quinquennial carbon storage maps of Shenzhen were retrieved from Landsat images and are shown in Figure 10.

As shown in Figure 10, carbon storage was unevenly distributed in the study area. The southeastern part (Dapeng District), which is almost fully covered by arbor forest and economic forest, had the highest carbon storage. Besides this, several forest parks located in the central and northern parts (Longgang and Baoan District) also had high carbon storage [80]. Conversely, developed regions had little carbon storage.

Figure 11a shows that UGS VFC had a high positive influence on carbon storage (r = 0.86). Figure 11b shows that vegetation had the highest carbon storage, followed by croplands, barren, water and built-up. Over the past 40 years, the carbon storage capability of UGS subtypes has decreased a little.

4.5. Influence of UGS on AOD

The Spearman correlation between the PM2.5 observed by meteorology stations and the retrieved AOD shows that they had strong positive correlation (R² = 0.72), indicating that it is feasible to use AOD to represent the surface pollution, as shown in shown in Figure 12. Then, the annual AOD maps of Shenzhen were retrieved from MODIS images according to [67] and are shown in Figure 13. There is no AOD map from 1978 to 1998, as MODIS images were not available at that time.

Figure 13 shows that AOD was unevenly distributed in this city. Coastal regions in the Bao’an District and Nanshan District had higher AOD values. These places are known for airports, wharfs and commercial streets, which are likely to emit large amounts of polluted air [81,82]. What is worse is that the high temperature and low humidity in the regions promoted air pollution accumulation. Conversely, forest parks always have lower AOD values, which proves that UGS can alleviate AOD accumulation to some extent.

Figure 14a shows that UGS VFC had a significant negative influence on AOD (r = −0.83). Thus, UGS patches with a higher VFC can more effectively alleviate local air pollution, as vegetation can block, filter and adsorb atmospheric particulate matter and dust [83,84]. Figure 14b shows that forest and cropland areas had lower AOD values than non-UGS. The AOD of built-up and barren areas has evidently increased in the 15 years, which suggests that the expansion of non-UGS could enhance AOD intensity.

5. Discussion

The city of Shenzhen has undergone one of the most rapid economic, industrial and urban developments in its history, similar to other cities in China, India, Bangladesh, Vietnam and other developing counties [3,85]. Many studies have disclosed that urban expansion and the UGS change of the city are governed by a combination of geographical, environmental and socio-economic factors [3,85,86]. Rapid economic development is the primary cause for rapid urbanization, as economic opportunities in the city fuel rural to urban migration on a massive scale [87]. This process is similar to those that have occurred in Ho Chi Minh [86], Dhaka [3,86], Delhi [87], Hanoi and Anson [88].

Based on a 40-years investigation, the study revealed that in Shenzhen, the majority of built-up and bare land has been converted from previously being agricultural land, forest and water bodies. This suggests the existence of increased pressure on UGS in the city to meet the increasing demand for urban land. For instance, croplands were completely transformed into built-up and bare lands by 2018. Another noticeable fact is that the west coastal commercial regions were transferred from coastal ponds and crop lands. This ultimately resulted in considerable impacts on the UGS landscapes in that region. Although the forest area has barely changed, the number and density of its patches increased. This shows that the relatively natural state was intensively interfered with by human activities and then changed into an artificial form, as many developing cities did [3,85,86,87,88]. The fluctuating tendency of the UGS patch shapes was evidently influenced by the changing awareness of local government regarding the priority of balance between economic development and environmental protection. This resembles the findings in previous studies that the rapid human activity in Dhaka, Bangladesh has leaded to numerous unconnected small patches of vegetation and cultivated land and greater isolation, as well as a higher percentage of edge areas in patches [86].

Numerous built-up and bare lands emerging along with urbanization processes have cast irreversible impacts on the urban environment, such as air pollution and soil loss [16,17,18,19,20]. This brings serious threats to citizens’ health and economic activities and the urban environment [21,22,23,24,25,26,27]. This study revealed that when the urban area expands, it becomes difficult to alleviate urban heat by exchanging energy between built-up and suburb regions; thus, the SUHI intensity will rise. Besides this, built-up regions have adverse influences on other eco-environment factors, and they also suffer terrible environments, such as low humidity, low carbon storage and high AOD. On the contrary, UGS is likely to promote beneficial effects for all eco-environment factors. Specifically, urban forest parks, such as Yangtaishan Forest Park and Wutongshan Forest Park, have lower SUHI and can significantly alleviate urban SUHI. Besides this, UGS patches enjoy a higher surface wetness and carbon storage and a lower AOD. This study showed that UGS has a significantly mitigating impact on urban LST, while non-UGS strongly enhances SUHI.

It is noticeable that rapid urbanization, economic development and local climatic conditions resulted in a high concentration of O₃ and PM2.5 in this city [12]. The serious photochemical pollution has brought unignorable threats to the eco-environment and public health. This study revealed that high AOD is mainly distributed in transportation sites, such as terminals, airports and ports, which play important roles in urban economic development. Today, Shenzhen is one of the greatest economic and transportation centers in China. For rapidly developing cities, strategies for alleviating air pollution from transportation and industry should be carefully designed.

Due to limited data and errors in the retrieval methods, some uncertainties should be carefully estimated. First, due to a lack of grided ground observations, MODIS products were used to retrieve environmental factors; their spatial resolution may have affected the accuracy of factors’ maps. Secondly, the retrieved value cannot be the same as ground observations, no matter which algorithm is used [89]. The differences between them may have influenced the quantitative analysis. However, these uncertainties cannot reverse our general findings, given that there was a very strong positive correlation between them. Further work will involve integrating more environment sensors and developing novel retrieval models to conduct long time, fine-scale analyses, such as the Ecosystem Spaceborne Thermal Radiometer Experiment on Space Station (ECOSTRESS) thermal infrared sensor and heat vulnerability index (HVI) models [90].

6. Conclusions

Shenzhen has experienced rapid economic development and urban expansion from 1978 to 2018. In the 40 years, UGS areas and landscapes have evidently been disturbed by human activities. In this study, UGS landscape dynamics and their impacts on the urban eco-environment were comprehensively investigated and analyzed. Results indicated that the transformation from UGS to non-UGS is clear and remarkable, especially for cropland, which completely disappeared by 2018. Secondly, non-UGS patches have adverse influences on urban eco-environment factors, and they suffer terrible environments, such as high SUHII, low humidity, low carbon storage and high AOD. What is worse is that the expanding and connecting of non-UGS patches enhanced the intensity of urban heat islands. On the contrary, UGS patches have beneficial influences on all eco-environment factors, and they then enjoy a better eco-environment, including low SUHII, high surface wetness, high carbon storage and low AOD. This study will be of great help for UGS protection and urban sustainable development for rapid developing cities, as the detailed change of UGS patches and their influences on urban environment have been comprehensively discovered.