Landscape metrics for land use/cover change and green space pattern in Can Tho city in 1998-2020 based on remote sensing and GIS

Changes in land cover/green space in Can Tho city in the period 1998-2020 were analyzed based on remote sensing data and the perspective of landscape ecology. The data collected from JAXA on land cover/land use classification entire Can Tho city in the period 1998-2020 was used to extract landscape metrics and to analyze the changes by a conversion matrix. The result shows a complex change in the urban green space in the period 1998 - 2020 in Can Tho. Research results show that Can Tho city had rice paddies and other croplands accounting for the majority of the land use structure. In the period 1998-2020, the city experienced a major shrinkage in croplands. Although other land use/cover types fluctuated differently, the fragmentation process still prevailed with fragments having discrete connections in space. The process of urbanization occurred with the expansion of built-up, but the concentration level was not high, with small fragments appearing sporadically distributed. However, comparing the landscape indices in the period 1998 – 2020 showed the aggregation process was increasingly taking place, and the level of fragmentation had tended to decrease. This trend was evidenced by the sharp increase in the number of patches, effective mesh size, and splitting index simultaneously with the gradual decline of the LPI index. The research results show high potential in expanding green space in Can Tho as well as application in planning urban green space development.


Introduction
Urban land expansion is one of the most visible, irreversible, and rapid types of land cover/land use change in contemporary human history, and is a key driver for many environmental and societal changes across scales [1].Landscape fragmentation is an important indicator of the increase in impervious surfaces related to the process of urbanization and its impact on urban land use/cover structure.Comparing the landscape metrics of objects over time performs the trend of fragmentation and aggregation in spatial.These results clarify how urbanization has an undesirable impact on the fragmentation of the landscape in general and the urban green space in particular.It also allows for forecasting the trend of fragmentation of the urban green space in the future and planning an effective development strategy for land use structure.Remote sensed data and landscape metrics have been commonly used for spatial pattern change in land use/cover and green space.A Future Land-Use Simulation (FLUS) system is built and verified by using remote sensing data to simulate global land use/land cover (LULC) change in relation to humanenvironment interactions [2].The loss in urban green space in relation to LULC change by using a multitemporal land-use transition scenario and to predict the value in the future for planning the strategy monitoring can be achieved by using remote sensing and geographical information system [3][4][5].
Landscape ecology emphasizes the interaction between spatial patterns and ecological process-that is, the causes and consequences of spatial heterogeneity across a range of scales.In landscape ecology, metrics are used as a tool to quantify the spatial pattern.Landscape metrics allow us to determine whether spatial patterns have changed over time, or whether landscapes are different or similar in pattern [6].Landscape metrics are used to measure green area landscape patterns (fragmentation, area-edge, shape, and aggregation) (Jaafari et al., 2020;Nasehi and Namin, 2020); evaluate and monitor of urban greenspace [9], and analyze the cooling effect of green spaces [10,11]; the relationship between urban green space fragmentation and urbanization [12].
In this study, the land use/cover change and evaluate the process of spatial fragmentation of green space in Can Tho from the perspective of landscape ecology based on extracting information from remote sensing data over time.Therefore, this study contributes to assessing land use/cover patterns using landscape metrics and provides basic information for proposing methods to enhance the urban green space structure.

Data and Study area
Can Tho is situated in the mid-downstream and central of the Mekong Delta (alluvial freshwater), stretching over 55 kilometres along the western bank of the Hau River.The city has a total area of 1,401.61km 2 , accounting for 3.49% of the entire region, borders with An Giang province in the north, Dong Thap and Vinh Long province in the east, Kien Giang province in the west, and Hau Giang province in the south.The data collected from JAXA with the research results of [13] on land cover/land use classification of Vietnam's mainland area in the period 1998-2020.This study used multiple image sources from Landsat TM, ETM+, and OLI with 30 m resolution, along with Sentinel MSI Level-2A and SAR GRD image sources with 10 m resolution.Random Forest classification method and post-classification optimization were used.Classification results have a resolution of 30 m with accuracy from 85.7 ± 1.3% to 92.0 ± 1.2%.The study area was divided into nine land cover/use types including Rice paddies, Croplands, Forests, Grasslands, Built-up, Water bodies, Aquaculture, Wetlands, and Barelands.The categories and descriptions of the classification system are presented in Appendix A. This data is used to extract information about the land cover/green space in the Can Tho area in the years 1998, 2008, 2014, and 2020.These data were used to evaluate changes in landscape indices.

Assessment of changes in land cover/green space
The conversion matrix is used to assess changes in land cover/green space in Can Tho city in the period from 1998 up to now.This is the period that Can Tho began to be rapidly affected by the urbanization process.In the matrix, the columns represent the area reduced due to conversion to another type of coating, and the rows represent the area increased from other types of coating converted during a period.The diagonal cells in the table represent the constant fraction of the cover type over that period.

Extraction of landscape metrics
The set of landscape indices is abundant and can be classified into different groups, for example: size, density, shape, proximity, diversity.Therefore, first of all, it is necessary to select some appropriate indicators for assessing land use/cover changes.Landscape indices are calculated using the FRAGSTATS 4.2 tool for the class level.
Each scene is divided into 5000 x 5000 m segments.To reduce the error caused by scenes containing non-data regions, any scenes containing more than 50% non-data regions will be discarded.Therefore, classification images each year are divided into 56 scenes.The total number of scenes used for analysis in 4 years 1998, 2008, 2014, and 2020 is 224 scenes.
The purpose is to evaluate the change of green space over time and observe the impact of urbanization on green space.Therefore, a number of landscape indices focused on analyzing indicators showing changing trends over time, such as area, shape, and indices showing the process of fragmentation or aggregation were used.Metric indices was used in this study as shown in the table below: DIVISION -This study was used some indices, comprising NP, AREA_MN, PARA-MN, CONTIG_MN, SPLIT, LPI, PLAND, MESH, SHAPE_MN and DIVISION.These indices are often used in urban green space analysis in some studies (Nasehi and Namin, 2020; Zieba-Kulawik and Wezyk, 2022).Through the results of extracting landscape metrics, there are two indices SHAPE_MN and DIVISION that do not show fluctuations across all scenes and are stable over time.Therefore, only the remaining 8 indicators are included in the analysis of landscape changes for Built-up, water bodies, and green space of the study area, Fluctuations of landscape metrics are analyzed through percentage changes with values varying from 0-100% according to the following formula [13]: In which:  is the percentage changes of class;  1 ,  2 are landscape index value at time t1 and t2.
The result of indices analysis was showed in Appendix B. (Figure 2).The green space was classified as: forests, rice paddies, grasslands, croplands and wetlands.

Assessment of land use/cover change and green space pattern
Green space was the main landscape accounting for approximately 90% of the natural area with the majority being rice paddies and other croplands.Other classes occupied a small area, of which built-up accounts for 2.5 to 9% of natural land over the years, concentrated in the southeast area of the city.Table 2 shows the change of land use/cover over different periods from 1998 to 2020.Area changes are the net change which is the combined result of the two simultaneous processes of shrinkage and expansion of a specific layer.Net change does not represent the movement of land use type but demonstrates the general trend of the change process.From the net change data, it showed that built-up, rice paddies, wetlands, and aquaculture had strong fluctuations in area during this time.
Landscape metrics of 224 scenes for green spaces in Can Tho city from 1998 to 2020 were calculated.These metrics were used to evaluate the degree of data dispersion, as well as evaluate the process of spatial fragmentation.Of the 10 calculated indices in landscape metrics, two indices SHAPE_MN and DIVISION had insignificant fluctuations in spatial and temporal, so they were not included in the analysis.The index group describing the degree of fragmentation includes a number of patches (NP), landscape division index (DIVISION), split index (SPLIT), and effective mesh size (MESH).This is a group of metrics closely related to the aggregation metrics class that describes the level of subdivision of the class.The SPLIT represents the number of fragments in the class obtained if the class is divided into effective mesh-sized fragments (MESH).A large SLPIT index together with a small MESH represents high fragmentation of a class.In this study, the huge difference in SPLIT values between subscales described the large dispersion of data, and there was unevenness in the degree of subdivision of the same class in space.Comparing classes, classes with high SPLIT index included aquaculture, barelands, forest, grassland, built-up, and wetlands.These classes were characterized by small fragments (AREA_MN and small LPI), scattered and unfocused distribution, and occupying a small proportion of the percentage of landscape (PLAND).While rice paddies and other croplands occupied a large area and large piece size, they had a lower SPLIT index.
To observe trends and represent the general fluctuations of the landscape class, average indices were calculated for the entire landscape and the percentage variation is calculated.The objects focused on in this study include built-up and classes of main green space including rice paddies and other croplands specifically analyzed in this study.

Assessing changes in built-up
Figure 3 showed the expansion of built-up in Can Tho city in the period 1998-2020.Since 1998, builtup had focused on development in the southeast of the city.Through the next stages, construction land continued to expand in the west and northwest, originally owning a lot of agricultural land.The percentage of landscape (PLAND) tended to increase and reached the maximum value of 49.7%, representing the expansion of built-up in 2020.Besides, the standard deviation also increased due to the increasing dispersion of data.Built-up was highly concentrated with PLAND accounting for 49.7% on average for all subscene in 2020, but some subscenes were only 0.9%.However, this low average value showed that, despite the expansion in area, built-up was highly concentrated in a few subscene, the remaining subscene contained many scattered and small fragments in space.Similarly, the index group of size such as mean area of patch (AREA_MN) and largest patch index (LPI) all had large differences between the largest and smallest values, while the mean was low.
There was an expansion in area (PLAND) in built-up in 1998-2020, an increase in the number of discrete fragments (NP), however, the appearance of new fragments with small areas made the distribution of built-up very scattered, the low level of concentration.During this period, fluctuations in the indices describing the level of fragmentation (NP, MESH, SLPIT) showed that the aggregation process was increasingly taking place, and the level of fragmentation had tended to decrease.The PARA_MN of built-up tended to decrease in this period, showing a decrease in the complexity of the shape of built-up.The mean patch contiguity (CONTIG_MN) was also a shape index based on spatial connectivity.Large contiguous patches result in larger contiguity index values.This index remained almost unchanged during this period for the construction soil layer.

Assessing changes in green space
Green space in the study area includes forests, rice paddies, grasslands, other croplands, and wetlands.Forests and grasslands tended to fluctuate insignificantly.Rice paddies and other croplands were in the majority of the structure of land use, and also fluctuated the most during this period.Figure 4 showed the spatial variation of other croplands.In particular, there were always two different trends of fluctuations taking place.For other croplands, the increase id shown in green with different levels of intensity, similar to the purple color representing the decrease in other croplands.The increase in other croplands occurred mainly in the central area of Can Tho city, while in the south and southeast regions, there was a decreasing trend during the 22-year period.
Rice paddies and other croplands accounted for the largest proportion of area, creating green areas for Can Tho city.They had different fluctuations in each period, but in general, the area of rice paddies increased slightly, while other croplands tended to decrease sharply.For rice paddies, the NP value decreased, along with an increase in AREA_MN indicating the disappearance of small fragments, and the maintenance of large fragments with a decrease in shape complexity (PARA_MN).The sharp increase in the MESH index from 1998 to 2014 showed that the fragmentation process was occurring strongly, creating many small, discrete fragments in space; while from 2014-2020, MESH values tended to decrease, indicating the predominant aggregation process with the aggregation of small fragments to form larger fragments.Other croplands in the period 1998-2020 had a strong tendency to fragment, especially in the period 2008-2014 with SPLIT change increasing by 96.30%.The PLAND decreased along with the appearance of discrete fragments with a small average fragment size, and the LPI index also decreased deeply during this period.The obvious decreased in effective grid size also indicated high fragmentation of other croplands during the period 2008-2014.
Forests, grasslands, and wetlands all occupied a small area in the landscape, although they tended to fluctuate differently in each period.In general, forests and grasslands tended to reduce the proportion of landscape (PLAND), while wetlands showed signs of increasing.In particular, the forest had tended to decrease deeply since 2008, the number of patches and the mean area of patches decreased due to the disappearance of fragments and the shrinkage of area.Although MESH fluctuated little, PARA_MN increased with a sharp decrease in CONTIG_MN value, showing that the shape of discrete fragments increased in complexity, and decreased in spatial connection.In other words, the forest had a shrinking area, consisting of small scattered fragments, and the largest patch index (LPI) decreased and fluctuated insignificantly.The SPLIT increased, strongest in the period 2014-2020 with an increase rate of 314.18%, representing the ongoing fragmentation of forest cover.
In contrast, wetlands and grasslands tended to aggregate during this period, with an increase in coverage ratio, area expansion, fragment size, the appearance of fragments in space, and a decrease in complexity in shape, with reduced splitting index and better connection to adjacent fragments.

Changes in water bodies
The changes in landscape metrics showed that during this period the water surface had shrunk in area due to the disappearance of small-sized fragments, large sized fragments still remain, so even though the area narrowed the average piece size increased by 15.43% for the period 1998-2020, especially the period 2014-2020 increased the most with 169.62%.The reason for shrinking water surface area was largely due to its use for aquaculture.Therefore, this period saw an expansion in the area of aquaculture with the PLAND increasing by 26.28%.Fluctuations in MESH and SPLIT showed that the aggregation process was taking place for aquaculture, in which the effective mesh size increased significantly, the area expanded with the appearance of new fragments, and the connection between adjacent pieces increased.

Discussion and conclusion
In addition, a field survey on urban green spaces in Can Tho was also conducted to supplement data for analysis.The changing characteristics of the green space in Can Tho are as follows: • The landscape size of urban green spaces is relatively reasonable, the scarcity of urban parks and other green spaces (connected green spaces) creates an urgent need among residential areas for better entertainment services.• The form of urban green space landscape is complex, while the spatial dispersion of urban parks and other green areas is relatively high.• Regarding the spatial structure of the urban green space landscape in Can Tho, it is characterized by unbalanced spatial distribution, low landscape heterogeneity and connectivity, and notable ecological functions of regional recreational parks and conservation areas.The above results indicated that the landscape size of urban green spaces is generally appropriate, but the scale of urban parks and affiliated green spaces needs to be expanded.Meanwhile, it is necessary to improve the accessibility and connectivity of the urban green space with the creation of green corridors.In addition, areas with notable ecological functions, such as recreational greenfield areas and protected areas, should be protected.
In conclusion, the relationship between urbanization and the green space landscape, especially the impact of urbanization on the green space in Can Tho, is formed and developed according to the rules that have been mentioned and summarized in many recent scientific publications in the world and in the high urbanization countries in Southeast Asia.This continuous change of land use/cover types involves many factors, which need to be measured both by space-time aspects and by landscape metrics.Therefore, the geographic information system and landscape ecology in land use/cover study are suitable.
The fluctuations assessment using a conversion matrix enables showing the complex fluctuations of the urban green space landscape in the period 2008 -2020 in Can Tho City.Research results showed that Can Tho city has rice paddies and other croplands accounting for the majority of the land use structure.In the period 1998-2020, the city experienced a major change, in which other croplands shrunk in area.The process of urbanization occured with the expansion of built-up, but the concentration level was not high, with small fragments appearing sporadically distributed.Although other land use/cover types fluctuated differently, the fragmentation process still prevailed with fragments having discrete connections in space.
Landscape fragmentation is an important indicator for the increase in economic structures related to the urbanization process in Can Tho City, affecting other land cover types, especially the urban green space.The distribution of built-up was very scattered and a low level of concentration.However, comparing the landscape indices in the period 1998 -2020 showed the aggregation was increasingly taking place, and the level of fragmentation had tended to decrease.This trend was evidenced by the sharp increase in the NP, MESH, and SLPIT simultaneously with the gradual decline of the LPI index.Water bodies were increasingly narrowed and fragmented.The distance between patches tended to increase, and the shape of the patches became more complex.These results allow us to understand how urbanization has an undesirable impact on the fragmentation of the landscape in general and the urban green space in particular.These results allow for forecasting the trend of fragmentation of the urban green space in the future and are meaningful in guiding the planning of urban green space development in Can Tho city.
Results of analyzing the impact of urbanization on the fragmentation of agricultural land, and of the green space show the need to increase the area of urban green space.Can Tho has the ability to expand the urban green space by 31.24% of the current area.
Therefore, the study results can be applied in proposing a method to enhance the urban green space structure in the following aspects: (1) converting the high potential areas into new urban green space, such as urban parks and affiliated green spaces, (2) select green corridors with high potential to expand the urban green space after analyzing the adjacent buffer zones of main roads and rivers and the current urban green space to connect existing green areas, (3) build a systematic urban green space network in the map to optimize the planning with the current situation, the new expanded green space and potential corridors.

Figure 1 .
Figure 1.The study area in Can Tho city The administrative unit of Can Tho city includes 5 urban districts (Ninh Kieu, Cai Rang, Binh Thuy, O Mon, Thot Not) and 4 rural districts (Phong Dien, Co Do, Vinh Thanh, Thoi Lai) with 85

Figure 2 .
Figure 2. Classification of land cover in Can Tho city in 1998 (a), 2008 (b), 2014 (c), and 2020 (d) Land use/cover in Can Tho city was classified into 9 classes for the years 1998, 2008, 2014 and 2020 (Figure2).The green space was classified as: forests, rice paddies, grasslands, croplands and wetlands.Green space was the main landscape accounting for approximately 90% of the natural area with the

Figure 4 .
Figure 4. Changes of croplands in Can Tho in the period 1998-2020

Table 1 .
Landscape metrics used in the study

Table 2 .
Area and change in land use/cover area in Can Tho in the period 1998-2020