Satellite-Based Monitoring of Annual Coastal Reclamation in Shenzhen and Hong Kong since the 21st Century: A Comparative Study

: To solve the shortage of land resources, many coastal cities have implemented coastal reclamation projects over the past few decades. Coastal reclamation can promote the economic development of coastal cities and improve human well-being in coastal zones. However, it can inevitably cause a series of ecological and environmental issues, such as coastal water pollution, ecosystem destruction, habitat loss, and land subsidence. Shenzhen and Hong Kong are two large neighboring Chinese coastal cities in southern China with different systems. As densely populated and economically developed cities, they face similar land shortage issues. However, recent coastal reclamation changes in Shenzhen and Hong Kong are unclear under different social and political systems. To ﬁll this gap, this study aimed to monitor and compare recent annual coastal reclamation in Shenzhen and Hong Kong with totally different systems using free and open satellite products. Then, to compare the results of coastal reclamation between Shenzhen and Hong Kong. Large-scale coastal reclamation in Shenzhen and Hong Kong from 2000 to 2018 can be observed. The total area of coastal reclamation was 4140.7 hm 2 , of which the total coastal reclamation area in Shenzhen was 3409.8 hm 2 and the total coastal reclamation area in Hong Kong was 730.9 hm 2 . Coastal reclamation in Shenzhen showed a temporal characteristic of “increasing slightly and then decreasing sharply”. Before and after 2010, the area of accumulated coastal reclamation were 3202.9 hm 2 and 206.9 hm 2 . Comparatively, coastal reclamation in Hong Kong exhibited a temporal characteristic of “ﬁrst decreasing and then increasing and then decreasing”. In 2002 and 2014, the scale of coastal reclamation was relatively large, with a total area of 501.8 hm 2 , accounting for 69.19% of its total area of coastal reclamation. The comparison between our produced coastal reclamation and the ofﬁcial area on coastal reclamation exhibited a good agreement based on correlation analysis (r = 0.99) and RMSE (Root Mean Square Error) (0.72). Since the 21st century, Shenzhen and Hong Kong have shown different, even opposite, policies on coastal reclamation. It will be necessary to continuously monitor future coastal reclamation driven by policies for better conducting sustainable coastal development in the Guangdong-Hong Kong-Macao Greater Bay Area.


Introduction
In recent years, with the continuous development of urbanization and industrialization, the demand for land resources has increased over time [1]. To alleviate the shortage of land resources, many coastal countries around the world, such as the Netherlands [2], Italy [3], Japan [4], and Singapore [5], have gained a large amount of land from coastal reclamation activities. China began coastal reclamation activities in the 1950s and 1960s, and by the end of the last century, the area of coastal reclamation in China reached 12,000 km 2 , with an average of 230 to 240 km 2 of reclamation per year, and over 30% of tidal land reclaimed [6]. There are many purposes for coastal reclamation, such as population living, Estuary and Lingding Channel in the west. It connects Hong Kong by the Shenzhen River to the south and borders Dongguan and Huizhou in the north. Although they are two large neighboring Chinese coastal cities, they have totally different systems. Specifically, Shenzhen has a socialist system while Hong Kong has a capitalist system. After the reform and opening up, as a special economic zone, Shenzhen's economy has been developing rapidly since the 1980s. Comparatively, Hong Kong developed as a warehousing and distribution center for U.K. trade with southern China since 1842 after the Treaty of Nanking. Although Shenzhen's social and economic development is far behind that of Hong Kong, Shenzhen exhibits a higher development speed. According to the Shenzhen China website [24], by the end of 2018, Shenzhen had a total land area of 1997 km 2 , a total sea area of 1145 km 2 , a coastline length of 260 km, and a total population of 13 million. Shenzhen is dominated by mountainous areas with few plains. Since the reform and opening up, rapid economic development has led to a significant problem of land scarcity. The limitation of land has become an important factor that restricts the development of Shenzhen [25]. Hong Kong is on the south side of Shenzhen and it is composed of the Hong Kong Island, Kowloon Peninsula, the inland area of the New Territories, and 262 small islands. According to the Hong Kong Lands Department [26] and the Hong Kong Census and Statistics Department [27], by the end of 2018, Hong Kong had a total land area of 1106 km 2 , a total sea area of 1648.7 km 2 , a coastline length of about 360 km, and a total population of 7.5 million. Hong Kong has one of the highest construction density worldwide due to large-scale urban development and limited land resources [28]. Figure 1 shows the map of our study area (i.e., Shenzhen and Hong Kong). J. Mar. Sci. Eng. 2021, 9, x FOR PEER REVIEW 3 of 14 Shenzhen and Hong Kong were selected as our study area. Shenzhen and Hong Kong are located in the Pearl River Delta, belonging to the Guangdong-Hong Kong-Macao Greater Bay Area (GBA). Shenzhen is located on the seashore of Guangdong Province of southern China. It faces Daya Bay and Dapeng Bay in the east and borders Pearl River Estuary and Lingding Channel in the west. It connects Hong Kong by the Shenzhen River to the south and borders Dongguan and Huizhou in the north. Although they are two large neighboring Chinese coastal cities, they have totally different systems. Specifically, Shenzhen has a socialist system while Hong Kong has a capitalist system. After the reform and opening up, as a special economic zone, Shenzhen's economy has been developing rapidly since the 1980s. Comparatively, Hong Kong developed as a warehousing and distribution center for U.K. trade with southern China since 1842 after the Treaty of Nanking. Although Shenzhen's social and economic development is far behind that of Hong Kong, Shenzhen exhibits a higher development speed. According to the Shenzhen China website [24], by the end of 2018, Shenzhen had a total land area of 1997 km 2 , a total sea area of 1145 km 2 , a coastline length of 260 km, and a total population of 13 million. Shenzhen is dominated by mountainous areas with few plains. Since the reform and opening up, rapid economic development has led to a significant problem of land scarcity. The limitation of land has become an important factor that restricts the development of Shenzhen [25]. Hong Kong is on the south side of Shenzhen and it is composed of the Hong Kong Island, Kowloon Peninsula, the inland area of the New Territories, and 262 small islands. According to the Hong Kong Lands Department [26] and the Hong Kong Census and Statistics Department [27], by the end of 2018, Hong Kong had a total land area of 1106 km 2 , a total sea area of 1648.7 km 2 , a coastline length of about 360 km, and a total population of 7.5 million. Hong Kong has one of the highest construction density worldwide due to large-scale urban development and limited land resources [28]. Figure  1 shows the map of our study area (i.e., Shenzhen and Hong Kong).

Data Sources
In this study, three data sources (i.e., Landsat imagery, GSW dataset (Table 1), and the Hong Kong Reclamation Map) were used to monitor annual coastal reclamation in Shenzhen and Hong Kong since the 21st century. Landsat imagery was used for visual interpretation to extract coastlines, the GSW dataset was used for land-ocean classification, and these two data were used to obtain the coastal reclamation results. The

Data Sources
In this study, three data sources (i.e., Landsat imagery, GSW dataset (Table 1), and the Hong Kong Reclamation Map) were used to monitor annual coastal reclamation in Shenzhen and Hong Kong since the 21st century. Landsat imagery was used for visual interpretation to extract coastlines, the GSW dataset was used for land-ocean classification, and these two data were used to obtain the coastal reclamation results. The Hong Kong Reclamation Map was used to produce the accuracy of the results of reclamation. To be specific, the first data source is the Landsat imagery provided by the United States Geological Survey (USGS) (https://earthexplorer.usgs.gov/). Four Landsat scenes (path/row: 121/44, 121/45, 122/44, and 122/45) can cover the entire study area. In this study, 80 Landsat-5 Thematic Mapper (TM) and Landsat-8 Operational Land Imager (OLI) images over 1999-2018 (with 4 images per year) were used. Among them, 56 Landsat-5 TM images were acquired from 1999 to 2012, and 24 Landsat-8 OLI images were acquired from 2013 to 2018. The spatial resolution of the Landsat imagery is 30 m and the temporal resolution is 16 days. To better extract artificial coastlines with coastal reclamation (e.g., dikes, ports, aquaculture ponds, and entertainment), cloud-free Landsat images were collected. The used Landsat images in this study had been pre-processed to L1T level (i.e., Standard Terrain Correction). The geo-registration is consistent and within prescribed tolerances (i.e., georeferencing error < 12 m) (https://lta.cr.usgs.gov/LETMP). The second data source is the GSW dataset, which was available in the Google Earth Engine (GEE) cloud computing platform (https://earthengine.google.com/). Specifically, the annual water classification historical images between 1999 and 2018 across our study area were collected (one image per year, a total of 20 images) from the GSW dataset. These annual water classification maps derived from 30-m Landsat-5, 7, and 8 data have four values: 0 (i.e., no data), 1 (i.e., not water), 2 (i.e., seasonal water), and 3 (i.e., permanent water). Here, permanent water was extracted from the annual images to derive the land-ocean classification map for each year. The third data source is the Hong Kong Reclamation Map with a vector format, which is based on the coastal reclamation data provided by the Hong Kong Lands Department. The map was digitized by using Google Maps Engine for visualization and interactivity online. Thus, the Hong Kong Reclamation Map is reliable for comparison with our coastal reclamation result. This map illustrates the coastal reclamation projects on both sides of the Hong Kong Victoria Harbor and their areas between 1840 and 2018 at 10-year intervals. Here, coastal reclamation data over 2000-2009 and 2010-2018 were used to evaluate our produced coastal reclamation result.

Land-Ocean Classification
In this study, GEE was used to extract the permanent water layers from the GSW dataset for deriving land-water maps. However, owing to many inland water bodies (e.g., coastal wetlands and aquaculture ponds) in coastal zones, it is difficult to accurately distinguish land from the ocean (Figure 2b) using land-water maps derived from the permanent water layer. To solve this problem, an automatic coastline extraction method was adopted [29]. For the land-water maps, functions "imclose" and "imfill" available in MATLAB were applied to generate land-ocean maps. In Figure 2, a Landsat TM image, and a land-ocean map before and after 2003 in the Shekou Peninsula in Shenzhen (113 • 50 -113 • 57 E, 22 • 26 -22 • 33 N) were used to demonstrate the step of generating a land-ocean map. Specifically, the reclaimed lake on the right side of the Shekou Peninsula is an inland water body, which was regarded as a water body in the original land-water map. However, it belongs to the land area after automatic coastline extraction processing. In this study, if a water body near the coastline is not directly connected to the ocean according to satellite imagery, this water body belongs to inland water rather than the ocean. 113°57′ E, 22°26′-22°33′ N) were used to demonstrate the step of generating a land-ocean map. Specifically, the reclaimed lake on the right side of the Shekou Peninsula is an inland water body, which was regarded as a water body in the original land-water map. However, it belongs to the land area after automatic coastline extraction processing. In this study, if a water body near the coastline is not directly connected to the ocean according to satellite imagery, this water body belongs to inland water rather than the ocean.

Post-Processing of Land-Ocean Classification
In this section, the post-processing of land-ocean classification was conducted to remove errors in our produced land-ocean maps induced by the complex landscape in coastal zones (e.g., coastal wetlands, aquaculture ponds, and ports). The specific steps were detailed as follows ( Figure 3). First, coastal land expansions and coastal land fluctuations can be grouped based on their temporal changes based on our produced annual land-ocean maps over 1999-2018. Second, these coastal fluctuations induced by coastal reclamation for lakes and shoals with low distinction were manually checked and refined to obtain land-ocean maps between 1999 and 2018 with the help of the reference land-ocean map in 1999 using Google Earth software. Note that the manual checking can ensure the precision of land-ocean maps with a working resolution (i.e., scale) of about 1:60,000. Finally, for each annual land-ocean map after 1999, we subtracted the land area of last year from the land area of that year to produce the annual coastal land expansion map between 2000 and 2018.

Extraction of the Coastal Reclamation
In the study area, both natural factors (e.g., sediments from upstream rivers and sealevel rises) and coastal reclamation (e.g., dikes, ports, aquaculture ponds, and

Post-Processing of Land-Ocean Classification
In this section, the post-processing of land-ocean classification was conducted to remove errors in our produced land-ocean maps induced by the complex landscape in coastal zones (e.g., coastal wetlands, aquaculture ponds, and ports). The specific steps were detailed as follows ( Figure 3). First, coastal land expansions and coastal land fluctuations can be grouped based on their temporal changes based on our produced annual land-ocean maps over 1999-2018. Second, these coastal fluctuations induced by coastal reclamation for lakes and shoals with low distinction were manually checked and refined to obtain land-ocean maps between 1999 and 2018 with the help of the reference land-ocean map in 1999 using Google Earth software. Note that the manual checking can ensure the precision of land-ocean maps with a working resolution (i.e., scale) of about 1:60,000. Finally, for each annual land-ocean map after 1999, we subtracted the land area of last year from the land area of that year to produce the annual coastal land expansion map between 2000 and 2018. 113°57′ E, 22°26′-22°33′ N) were used to demonstrate the step of generating a land-ocean map. Specifically, the reclaimed lake on the right side of the Shekou Peninsula is an inland water body, which was regarded as a water body in the original land-water map. However, it belongs to the land area after automatic coastline extraction processing. In this study, if a water body near the coastline is not directly connected to the ocean according to satellite imagery, this water body belongs to inland water rather than the ocean.

Post-Processing of Land-Ocean Classification
In this section, the post-processing of land-ocean classification was conducted to remove errors in our produced land-ocean maps induced by the complex landscape in coastal zones (e.g., coastal wetlands, aquaculture ponds, and ports). The specific steps were detailed as follows ( Figure 3). First, coastal land expansions and coastal land fluctuations can be grouped based on their temporal changes based on our produced annual land-ocean maps over 1999-2018. Second, these coastal fluctuations induced by coastal reclamation for lakes and shoals with low distinction were manually checked and refined to obtain land-ocean maps between 1999 and 2018 with the help of the reference land-ocean map in 1999 using Google Earth software. Note that the manual checking can ensure the precision of land-ocean maps with a working resolution (i.e., scale) of about 1:60,000. Finally, for each annual land-ocean map after 1999, we subtracted the land area of last year from the land area of that year to produce the annual coastal land expansion map between 2000 and 2018.

Extraction of the Coastal Reclamation
In the study area, both natural factors (e.g., sediments from upstream rivers and sealevel rises) and coastal reclamation (e.g., dikes, ports, aquaculture ponds, and

Extraction of the Coastal Reclamation
In the study area, both natural factors (e.g., sediments from upstream rivers and sea-level rises) and coastal reclamation (e.g., dikes, ports, aquaculture ponds, and entertainment) can result in coastal land expansions [30]. Fortunately, it is easy to distinguish coastal reclamation from remote sensing images by their regular shapes. The "Raster to Polygon" function in ArcGIS was applied to obtain original annual coastal land expansion maps with a vector format. Specifically, by comparing the original result with the Landsat image for all years, natural expansion areas were removed to obtain the refined annual coastal reclamation maps in ArcGIS with a working resolution (i.e., scale) of about 1:60,000.

Assessment of the Coastal Reclamation
We assessed the accuracy of our produced coastal reclamation dataset using the coastal reclamation data provided by the Hong Kong Lands Department [31] (i.e., Coastal Reclamation Maps on both sides of Victoria Harbor, Hong Kong, 1840-2018). To be specific, the Root Mean Square Error (RMSE) and correlation coefficient (r) between coastal reclamation areas from our result and official data were calculated by the correlation analysis. RMSE and r were calculated according to Equations (1) and (2). Moreover, the chi-square test (x 2 ) was performed according to Equation (3) to evaluate the agreement of coastal reclamation area results between our result and official data. The Root Mean Square Error (RMSE), correlation coefficient (r), and chi-square test (x 2 ) were calculated as follows.
where X is the coastal reclamation area from our result; Y is the official coastal reclamation area; and n is the number of areas of coastal reclamation projects used for calculation. In this study, four coastal reclamation projects were considered and were detailed in Section 5.2 (i.e., n = 4).

Assessment of the Coastal Reclamation
We assessed the accuracy of our produced coastal reclamation dataset using the coastal reclamation data provided by the Hong Kong Lands Department [31] (i.e., Coastal Reclamation Maps on both sides of Victoria Harbor, Hong Kong, 1840-2018). To be specific, the Root Mean Square Error (RMSE) and correlation coefficient (r) between coastal reclamation areas from our result and official data were calculated by the correlation analysis. RMSE and r were calculated according to Equations (1) and (2). Moreover, the chi-square test (x 2 ) was performed according to Equation (3) to evaluate the agreement of coastal reclamation area results between our result and official data. The Root Mean Square Error (RMSE), correlation coefficient (r), and chi-square test (x 2 ) were calculated as follows.
where X is the coastal reclamation area from our result; Y is the official coastal reclamation area; and n is the number of areas of coastal reclamation projects used for calculation. In this study, four coastal reclamation projects were considered and were detailed in Section 5.2 (i.e., n = 4).
where X is the coastal reclamation area from our result; Y is the official coastal reclamation area; and n is the number of areas of coastal reclamation projects used for calculation.
where X is the coastal reclamation area from our result; Y is the official coastal reclamation area; and n is the number of areas of coastal reclamation projects used for calculation. Figure 5 shows the spatial distribution of the coastal reclamation in the study area (i.e., Shenzhen and Hong Kong) from 2000 to 2018 at a 30-m spatial resolution derived from Landsat imagery and GSW dataset. Over the study period, many coastal reclamation activities were conducted in our study area, with a total area of 4140.7 hm 2 . Coastal reclamation in the study area showed a large spatial difference. The amount of coastal reclamation in Shenzhen was much larger than that in Hong Kong. Specifically, the total coastal reclamation area in Shenzhen was 3409.8 hm 2 and the total coastal reclamation area in Hong Kong was 730.9 hm 2 , about 1/5 of that in Shenzhen. Six typical areas of coastal reclamation in the study area were enlarged and displayed in Figure 5 to demonstrate the details of the spatial distribution of coastal reclamation. Note that almost all the coastal reclamation projects were included in the six areas. Specifically, the sum of coastal reclamation in the six areas was 4035.4 km 2 , accounting for 97.5% of the total result of coastal reclamation in the study area. Then, land use information [25] and Google Maps in 2019 across this study area were integrated to interpret the types of coastal reclamation. The online Google maps can provide the name and attribution of each landscape patch, so they can be used to characterize the types of coastal reclamation. Among six areas, three areas (Figure 5a,b,f) are located in Shenzhen. Area a is located on the west coast of Baoan District, which is Shenzhen Baoan International Airport. Area b is located on the southern coast of the Bao'an District and the Shekou Peninsula in the Nanshan District, which consist of harbors and residential land. Area f is located on the east coast of Yantian District, which is the Yantian Port. The other three areas (Figure 5c     were 300.6 hm 2 and 201.2 hm 2 , respectively, totaling 501.8 hm 2 , accounting for 68.7% of the total coastal reclamation area in Hong Kong over the past 19 years. From Figure 6, we can find that the coastal reclamation in Shenzhen exhibited a temporal characteristic of "increasing slightly and then decreasing sharply" while the coastal reclamation in Hong Kong exhibited a temporal characteristic of "first decreasing and then increasing and then decreasing". To be specific, during the first 10 years, the coastal reclamation activities in Shenzhen were relatively frequent, with an average annual coastal reclamation of 320.3 hm 2 , which decreased sharply in the following 9 years, with an average annual coastal reclamation of 23.0 hm 2 . By comparison, the coastal reclamation in Hong Kong only occurred in 2002 (300.6 hm 2 ) and 2014 (201.2 hm 2 ). In other years, coastal reclamation activities were relatively few, with an average of 13.5 hm 2 per year.  Figure 6, we can find that the coastal reclamation in Shenzhen exhibited a temporal characteristic of "increasing slightly and then decreasing sharply" while the coastal reclamation in Hong Kong exhibited a temporal characteristic of "first decreasing and then increasing and then decreasing". To be specific, during the first 10 years, the coastal reclamation activities in Shenzhen were relatively frequent, with an average annual coastal reclamation of 320.3 hm 2 , which decreased sharply in the following 9 years, with an average annual coastal reclamation of 23.0 hm 2 . By comparison, the coastal reclamation in Hong Kong only occurred in 2002 (300.6 hm 2 ) and 2014 (201.2 hm 2 ). In other years, coastal reclamation activities were relatively few, with an average of 13.5 hm 2 per year. To demonstrate more details of temporal changes, Table 3 and Figure 7 were presented to illustrate coastal reclamation results of Shenzhen and Hong Kong over different periods from 2000 to 2018. For Shenzhen and Hong Kong, this Table presents their respective and total reclaimed area over the past 19 years, their respective and total reclaimed area of each year, and their total coastal reclamation area every five years. From 2000 to 2018, the total coastal reclamation area in Shenzhen and Hong Kong was 4140.7 hm 2 , with an average annual coastal reclamation of 217.9 hm 2 . The total coastal reclamation area in Shenzhen was 3409.8 hm 2 , with an annual average coastal reclamation of 179.5 hm 2 , and the total coastal reclamation area in Hong Kong was 730.9 hm 2 , with an annual average coastal reclamation of 38.5 hm 2 , about 1/5 of that of Shenzhen. As shown in Table 2   To demonstrate more details of temporal changes, Table 3 and Figure 7 were presented to illustrate coastal reclamation results of Shenzhen and Hong Kong over different periods from 2000 to 2018. For Shenzhen and Hong Kong, this Table presents their respective and total reclaimed area over the past 19 years, their respective and total reclaimed area of each year, and their total coastal reclamation area every five years. From 2000 to 2018, the total coastal reclamation area in Shenzhen and Hong Kong was 4140.7 hm 2 , with an average annual coastal reclamation of 217.9 hm 2 . The total coastal reclamation area in Shenzhen was 3409.8 hm 2 , with an annual average coastal reclamation of 179.5 hm 2 , and the total coastal reclamation area in Hong Kong was 730.9 hm 2 , with an annual average coastal reclamation of 38.5 hm 2 , about 1/5 of that of Shenzhen. As shown in Table 2  Kong showed a temporal characteristic of "first decreasing and then increasing and then decreasing" and the areas of coastal reclamation during the four periods were 441.6 hm 2 , 59.5 hm 2 , 228.6 hm 2 , and 1.2 hm 2 , respectively.

Economy and Urbanization Drivers
It is necessary to investigate the relationship between coastal reclamat local economy because coastal reclamation activities can affect the coastal e [7,32]. Specifically, coastal reclamation is strongly associated with soc variables, such as GDP, population, and urbanization level. From the Shenzhe Yearbook 2019 [33] and the Hong Kong Census and Statistics Department [3 2000 and 2018, both Shenzhen and Hong Kong experienced rapid economi development. In 2000, the GDP of Shenzhen was ¥218.7 billion, while the GD Kong was $172.7 billion. In 2018, the GDP of Shenzhen was ¥2422.2 billion, wh of Hong Kong was $362.7 billion. The growth rates are 1007.5% and 110.0% r Besides, in 2000, the population of Shenzhen was 7.0 million, while the po Hong Kong was 6.8 million. In 2018, the population of Shenzhen was 13.0 m the population of Hong Kong was 7.5 million. The growth rates are 85.7% respectively. We can confirm that there is a strong positive relationship betw reclamation, the local economy (i.e., GDP), and the population. Apart from population, urbanization level can also be associated with coastal reclama study, the urbanization rate was used to measure the urbanization level of Sh Hong Kong. From the website of the Guangdong Statistics Bureau [35] and W [36], Hong Kong and Shenzhen are cities with high urbanization rates. The u

Economy and Urbanization Drivers
It is necessary to investigate the relationship between coastal reclamation and the local economy because coastal reclamation activities can affect the coastal environment [7,32]. Specifically, coastal reclamation is strongly associated with socioeconomic variables, such as GDP, population, and urbanization level. In 2018, the GDP of Shenzhen was ¥2422.2 billion, while the GDP of Hong Kong was $362.7 billion. The growth rates are 1007.5% and 110.0% respectively. Besides, in 2000, the population of Shenzhen was 7.0 million, while the population of Hong Kong was 6.8 million. In 2018, the population of Shenzhen was 13.0 million, while the population of Hong Kong was 7.5 million. The growth rates are 85.7% and 10.3% respectively. We can confirm that there is a strong positive relationship between coastal reclamation, the local economy (i.e., GDP), and the population. Apart from GDP and population, urbanization level can also be associated with coastal reclamation. In this study, the urbanization rate was used to measure the urbanization level of Shenzhen and Hong Kong. From the website of the Guangdong Statistics Bureau [35] and World Bank [36], Hong Kong and Shenzhen are cities with high urbanization rates. The urbanization rate of Shenzhen was 92.46% in 2000 and reached 100% in 2010. For Hong Kong, as early as 1993, its urbanization rate has reached 100%. Owing to the high urbanization rates, rapidly growing economy, and population, the Shenzhen and Hong Kong governments obtained land from the ocean for increasing industrial lands, commercial lands, construction lands, and residential lands, which resulted in an increasing demand for land area through coastal reclamation.
In summary, GDP, population, and urbanization level are associated with coastal reclamation. From Table 3, compared with Hong Kong, Shenzhen exhibited a greater increase in both economy, population, and urbanization rate from 2000 to 2018. In the future, the potential correlation between coastal reclamation events and economic-social development in coastal zones can be further examined. Note: In Table 3, B represents billion and M represents million. In this study, the urbanization rate was defined as the proportion of the population of a country (region) resident in urban areas to the total population of the country (region) [37]. Figure 8 shows a comparison between the coastal reclamation data obtained in this study and the Hong Kong reclamation map in the areas on both sides of Victoria Harbor from 2000 to 2018. According to our calculations, coastal reclamation areas from our study and official data are 35.4 hm 2 and 37.1 hm 2 , which differ by 4.6%. From Figure 8, areas of four coastal reclamation projects from our calculations are 8.3 hm 2 , 19.8 hm 2 , 4.3 hm 2 , and 3.0 hm 2 , and the area of four coastal reclamation projects from official data are 8.6 hm 2 , 21.1 hm 2 , 4.5 hm 2 , and 2.9 hm 2 . We can find a good agreement (RMSE = 0.72, r = 0.99) between coastal reclamation areas from our result and official data by correlation analysis. According to the chi-square test, our coastal reclamation data and official data exhibit a good agreement at the 0.05 significance level. It can also be seen that the spatial distribution of coastal reclamation from our study is consistent with the official data, indicating that our coastal reclamation data have high accuracy. 21.1 hm 2 , 4.5 hm 2 , and 2.9 hm 2 . We can find a good agreement (RMSE = 0.72, r = 0.99 between coastal reclamation areas from our result and official data by correlation analysis According to the chi-square test, our coastal reclamation data and official data exhibit a good agreement at the 0.05 significance level. It can also be seen that the spatia distribution of coastal reclamation from our study is consistent with the official data indicating that our coastal reclamation data have high accuracy.

Future Reclamation Planning in the Study Area
In general, we can find that national policy plays an essential role in promoting coastal reclamation activities in Shenzhen and Hong Kong. According to the Shenzhen Coastal Zone Comprehensive Protection and Utilization Plan (2018-2035) [38], from 2018 to 2035, the coastal reclamation area in Shenzhen will be concentrated on the west coast of Dakonggang New City, Bao'an District, with a total reclamation area of 1724.9 hm 2 , mainly used for port constructions. Compared with 2000-2018, the scale of the coastal reclamation activities in Shenzhen will greatly decline, with more emphasis on coastline restoration and coastal zone development. By contrast, from 2018, Hong Kong has planned many coastal reclamation projects, such as airport city construction in the north of Hong Kong International Airport (650 hm 2 ), Tung Chung New Town Expansion (130 hm 2 ), Lantau Island Xinao Reclamation Project (80 hm 2 ), New Territories Longgutan Reclamation Project (220 hm 2 ), and Lantau Tomorrow Vision project (1700 hm 2 ). According to the Hong Kong Sustainable Lantau Office [39], compared with 2000-2018, the scale of the coastal reclamation activities in Hong Kong will increase significantly, mainly concentrated on the less developed Lantau and surrounding islands.
At the end of the 20th century, coastal reclamation projects in Hong Kong caused dissatisfaction among environmentalists, who believed that coastal reclamation projects were unnecessary and can destroy the coastal environment. To protect the feature of Victoria Harbor as a natural and cultural heritage and the marine ecological environment, some councilors started a series of campaigns [40]. In 1997, the "Protection of the Harbor Act Ordinance" was performed to restrict the coastal reclamation projects dominated by the Hong Kong government. In 2003, some people overturned the "Wan Chai Reclamation Plan" in the Hong Kong Court of Final Appeal. Since then, these campaigns have directly affected the future planning of coastal reclamation in Hong Kong. As a result, more than 600 hm 2 of planned coastal reclamation in Hong Kong has been "deleted" [40], resulting in fewer coastal reclamation projects than Shenzhen in the 21st century. After 2018, to promote economic development and solve the housing tension problem, the Hong Kong government has put forward a series of coastal reclamation plans under pressure from protests. These plans will greatly increase the amount of coastal reclamation in Hong Kong in the future. The policy of the Shenzhen government is different from Hong Kong. According to the China Academy of Urban Planning and Design Shenzhen [41], Shenzhen released the "Shenzhen City Master Plan (2010-2020)" to protect the coastal environment in Shenzhen in 2010, and there were few planned coastal reclamation projects. The shift of planning agreed well with the decline of coastal reclamation in Shenzhen before and after 2010 (Table 3 and Figure 7).
In addition, it is important to understand the future reclamation development policies to better monitor the future reclamation development of Shenzhen and Hong Kong using satellite data. According to the Shenzhen Coastal Zone Comprehensive Protection and Utilization Plan (2018-2035), Shenzhen will pay more attention to strengthening the protection and restoration of habitats and improving the ecological service function after 2018. As a consequence, the coastal reclamation area will be further reduced in the future. Overall, under the current and future coastal reclamation planning in Shenzhen and Hong Kong, the area of coastal reclamation in Hong Kong will exceed that in Shenzhen. Therefore, in the coming decades, it will be necessary to continuously monitor the coastal reclamation in Shenzhen and Hong Kong to provide an accurate and updated dataset for better understanding and evaluating the role of coastal reclamation related policies in coastal environmental changes.

Conclusions
In this study, we monitored annual coastal reclamation in Shenzhen and Hong Kong since the 21st century and obtained several conclusions. First, the method based on Landsat imagery and GSW dataset in this study could be applied to monitor coastal reclamation at the annual scale. Second, according to our results derived from our method, a large-scale coastal reclamation in Shenzhen and Hong Kong with a total coastal reclamation area of 4140.7 hm 2 over 2000-2018 could be observed. Shenzhen and Hong Kong have exhibited different temporal characteristics in coastal reclamation since the 21st century. The area of coastal reclamation in Shenzhen (i.e., 3409.8 hm 2 ) was much larger than that in Hong Kong (i.e., 730.9 hm 2 ). In addition, coastal reclamation in Shenzhen exhibited a significant declining trend. Third, as coastal reclamation is dominated by policies, it is necessary to produce an accurate and updated dataset on coastal reclamation from satellite data to further investigate coastal reclamation related environmental issues and feedback to policy-making in the future. For example, coastal reclamation may result in coastal land subsidence, which can increase the vulnerability of coastal zones especially in the context of rising sea levels. An accurate and updated dataset on coastal reclamation could be used to guide the coastal land use planning to avoid the coastal area with a higher vulnerability. Moreover, by benefiting from the global coverage and free access to Landsat and GSW datasets, this method can be extended to the globe in the future. Furthermore, our method could be extended to an earlier year to provide a longer history of coastal reclamation.
Author Contributions: T.F. contributed to the data collecting, analyzing and writing of this study. N.X. contributed to the overall idea and editing of this study. All authors have read and agreed to the published version of the manuscript. Institutional Review Board Statement: "Not applicable" for studies not involving humans or animals.
Informed Consent Statement: "Not applicable" for studies not involving humans.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.