Impact of Land Use Change Due to Urbanisation on Surface Runoff Using GIS-Based SCS–CN Method: A Case Study of Xiamen City, China

Rapid urban development results in visible changes in land use due to increase in impervious surfaces from human construction and decrease in pervious areas. Urbanisation influences the hydrological cycle of an area, resulting in less infiltration, higher flood peak, and surface runoff. This study analysed the impact of land use change due to urbanisation on surface runoff, using the geographic information system (GIS)-based soil conservation service curve number (SCS–CN) method, during the period of rapid urban development from 1980 to 2015 in Xiamen, located in south-eastern China. Land use change was analysed from the data obtained by classifying Landsat images from 1980, 1990, 2005, and 2015. Results indicated that farmland decreased the most by 14.01%, while built-up areas increased the most by 15.7%, from 1980 to 2015. Surface runoff was simulated using the GIS-based SCS–CN method for the rainfall return periods of 5, 10, 20, and 50 years. The spatial and temporal variation of runoff was obtained for each land use period. Results indicate that the increase in surface runoff was highest in the period of 1990–2005, with an increase of 10.63%. The effect of urbanisation can be realised from the amount of runoff, contributed by built-up land use type in the study area, that increased from 14.2% to 27.9% with the rise of urban expansion from 1980 to 2015. The relationship between land use and surface runoff showed that the rapid increase in constructed land has significantly influenced the surface runoff of the area. Therefore, the introduction of nature-based solutions such as green infrastructure could be a potential solution for runoff mitigation and reducing urban flood risks in the context of increasing urbanization.


Introduction
Urbanisation is a growing concern in the present world. With over 55% of the world's population living in urban areas [1], urban expansion has dramatically influenced the change in urban land use. Land use change due to urbanisation results in more impervious surfaces that have considerable impacts on urban hydrology [2]. Urban expansion leads to large impervious surfaces that reduces rainwater infiltration, generating high surface runoff and peak flow, thus increasing the risks of urban flooding and waterlogging [3]. Urban development is one of the major causes of urban pluvial flooding, aggravated by poor urban drainage systems that become more severe with increasing frequency and use change on surface runoff. The main objective of this research is to identify the characteristics of land use change from 1980 to 2015 and analyse the impact of land use change due to urbanisation on the temporal and spatial distribution of surface runoff under the rainfall return periods of 5, 10, 20, and 50 years.

Study Area
Xiamen City is located on the west coast of the Taiwan Strait, composed of the mainland area along Xiamen Bay, Xiamen Island, Gulangyu, and other islands, including 6 administrative districts: Huli, Siming, Haicang, Jimei, Tong'an, and Xiang'an, as shown in Figure 1. The city has a total land area of 1699 km 2 and a sea area of 324 km 2 . The study area is predominantly flat, low relief, with medium-low mountains, plains, and tidal flats. The slope of the terrain descends from northwest to southeast. The northwest part is mountainous, with the highest elevation of 1175 m above sea level, located on Yunding mountain [31]. Xiamen has a subtropical monsoon climate, with humid and mild climates throughout the year. According to the Xiamen Statistical Yearbook, annual average temperature and rainfall are approximately 21 °C and 1200 mm, respectively, with most rainfall in June, July, and August, accounting for more than half of the annual rainfall. Xiamen is one of the special economic zones in China and has experienced a rapid economic development and urbanisation [31]. According to Xiamen Municipal Bureau of Statistics [32], urban built-up area of the city has expanded from 38.5 km 2 in 1985 to 348.23 km 2 in 2017. The permanent population of the city was 4.29 million in 2019 [32]. As of 2019, the urbanisation rate of the population in Xiamen is 89.2% [33]. Xiamen has experienced several natural disasters, especially flooding and waterlogging induced by sea-level rise and storm surges, which is further intensified by rapid urban growth in past decades. Hence, there is an urgent need for mitigation of the potential hazards induced by urban development and climate change.    The GIS environment was used to build and intersect land use and HGS shapefiles. The land use and soil complex were used to obtain the weighted CN. The CN value of each polygon was estimated using the USDA table. Finally, raster calculator in the GIS was used to calculate runoff depth from CN values for the four rainfall return periods. The methodology was applied to four different years of 1980,1990,2005, and 2015, respectively.
Land 2021, 10, x FOR PEER REVIEW 4 of 19 Figure 2 shows the overall methodology of the study conducted for analysing impact of land use change on surface runoff with the application of GIS and SCS-CN method. The GIS environment was used to build and intersect land use and HGS shapefiles. The land use and soil complex were used to obtain the weighted CN. The CN value of each polygon was estimated using the USDA table. Finally, raster calculator in the GIS was used to calculate runoff depth from CN values for the four rainfall return periods. The methodology was applied to four different years of 1980,1990,2005, and 2015, respectively.

Data Source and Methods
(a) Land use data Multispectral satellite images from Landsat 3 MSS, Landsat 5 TM, and Landsat 7 ETM+ were obtained to create a land use map of the study area for 1980, 1990, 2005, and 2015. The images were downloaded from Centre for Earth Observation and Digital Earth (CEODE), Chinese Academy of Sciences, and the United States Geological Survey (USGS) [34]. Images were georeferenced to the UTM, Zone 50 North, WGS-84 projection, and Beijing 1954 coordinate systems. The spatial resolution is set to 30 m. The land use was classified according to the National Land Cover Data Sets (NLCD) of China generated by Liu et al. [35] in the construction of the China 20th Century LUCC Spatio-temporal Platform. The method of unsupervised classification with Iterative Self Organizing Data Analysis Technique Algorithm (ISODATA) was used for image classification. Based on the national land use data product by Geographical Information Monitoring Cloud Platform, the land use was initially classified as six first level categories and twenty-five second level categories, which were reclassified into eight classes for the study. High resolution satellite imagery from Google Earth was used for visual interpretation and accuracy assessment. The overall accuracy and kappa coefficient of classified land use types is more than 85%. The high overall accuracy and kappa coefficient suggests a good relationship between classified Hydrological soil group map Soil and Land use complex Determination of maximum soil retention (S) Landsat image (1980,1990,2005,2015) Weighted Curve Number

Data Source and Methods
(a) Land use data Multispectral satellite images from Landsat 3 MSS, Landsat 5 TM, and Landsat 7 ETM+ were obtained to create a land use map of the study area for 1980, 1990, 2005, and 2015. The images were downloaded from Centre for Earth Observation and Digital Earth (CEODE), Chinese Academy of Sciences, and the United States Geological Survey (USGS) [34]. Images were georeferenced to the UTM, Zone 50 North, WGS-84 projection, and Beijing 1954 coordinate systems. The spatial resolution is set to 30 m. The land use was classified according to the National Land Cover Data Sets (NLCD) of China generated by Liu et al. [35] in the construction of the China 20th Century LUCC Spatio-temporal Platform. The method of unsupervised classification with Iterative Self Organizing Data Analysis Technique Algorithm (ISODATA) was used for image classification. Based on the national land use data product by Geographical Information Monitoring Cloud Platform, the land use was initially classified as six first level categories and twenty-five second level categories, which were reclassified into eight classes for the study. High resolution satellite imagery from Google Earth was used for visual interpretation and accuracy assessment. The overall accuracy and kappa coefficient of classified land use types is more than 85%. The high overall accuracy and kappa coefficient suggests a good relationship between classified image and reference image. The detailed descriptions of land use classes are given in Table 1. ArcGIS 10.5 and ENVI 5.3 software were used to generate various layers and land use maps.

Land Use Description
Farmland Areas for growing crops, mainly including paddy fields and arable lands for vegetable farming, with or without regular irrigation facilities. It includes farmland where rice and dry land crops are rotated.

Woodland
Areas referring to forestry land for growing trees, shrubs, bamboos, and coastal mangroves, including trees and shrubs with canopy density more than 30%.

Grassland
Areas of all kinds of grassland, mainly with herbaceous plants covering more than 5%, including shrub grassland with grassland and canopy density.

Water
Areas of natural land waters and water conservancy facilities, including natural and artificial river canals, lakes, and reservoir ponds.

Coastal wetlands
Areas of tidal flats and beach lands, including lands near water level of the rivers and lakes.

Built-up land
Urban land refers to land in large, medium, and small cities, residential areas, and built-up areas above county towns. It also includes construction sites such as factories and mines, large-scale industrial areas, oil fields, salt fields, and quarries, as well as roads, airports, and special sites.

Rural settlements
Refers to rural settlements independent of cities and towns.

Unused land
Areas of bare land, lands covered with gravel, sand, rocks, and saline-alkali and marsh lands. Generally, vegetation coverage is less than 5%.
(b) Soil data Soil data were obtained from Harmonized World Soil Database (HWSD) and Food and Agriculture Organization (FAO). Soil information for Xiamen city was obtained from 1:1 million soil map of China with a resolution of 1 km (30 × 30 arc seconds) [36]. Hydrological soil groups are classified into A, B, C, and D on the basis of water transmission and infiltration rate of soil when the soil is thoroughly wetted [29]. Soil group A has a high infiltration rate and the lowest runoff potential. These soils consist of sands or gravels that have a high rate of water transmission. Group B and C soils have moderate runoff potential and infiltration rate with a slower rate of water transmission. Soil group D has the lowest infiltration rate and highest runoff potential composed of clay or clayey loam soils with a very slow rate of water transmission [29]. Based on soil texture and soil type, the study area is classified into three HSG types, B, C, and D, as can be seen in Figure 3, which mostly contribute to large surface runoff and less infiltration.
(c) Rainfall data Long-time series rainfall data from 1985 to 2015, collected at the Xiamen meteorological stations, were acquired from the China Meteorological Data Service Centre of the China Meteorological Administration. The rainfall amount for the return periods of 5, 10, 20, and 50 years were obtained from maximum daily rainfall data for the hydrological analysis, as seen in Table 2. The rainfall for corresponding return periods was determined by using Log Pearson Type III distribution, which is the common probability distribution method used in China [37]. The study assumed that the climatic and soil conditions are constant. To investigate spatial heterogeneity, the study required uniformly distributed station data, which are not available. Therefore, for the temporal analysis of rainfall from the stations, the station with the most complete data was selected as a representative of the study area's rainfall. Furthermore, the study area is small, and changes in rainfall variation is considered insignificant.
(d) Digital Elevation Model (DEM) A digital elevation model (DEM) with a 30 m resolution was downloaded from the USGS website. Using hydrology tools in ArcGIS 10.8, 19 catchments were obtained, as seen in Figure 1. The sinks in the DEM data were filled, water flow direction was estimated, and flow accumulation was set with a threshold of 10,000 and 60,000 to obtain the catchments. As the study area has short streams and not many large rivers, areas in the southwest do not have adequate streams to obtain a catchment. Hence, runoff obtained from these areas (c) Rainfall data Long-time series rainfall data from 1985 to 2015, collected at the Xiamen meteorological stations, were acquired from the China Meteorological Data Service Centre of the China Meteorological Administration. The rainfall amount for the return periods of 5, 10, 20, and 50 years were obtained from maximum daily rainfall data for the hydrological analysis, as seen in Table 2. The rainfall for corresponding return periods was determined by using Log Pearson Type III distribution, which is the common probability distribution method used in China [37]. The study assumed that the climatic and soil conditions are constant. To investigate spatial heterogeneity, the study required uniformly distributed station data, which are not available. Therefore, for the temporal analysis of rainfall from the stations, the station with the most complete data was selected as a representative of the study area's rainfall. Furthermore, the study area is small, and changes in rainfall variation is considered insignificant. (d) Digital Elevation Model (DEM) A digital elevation model (DEM) with a 30 m resolution was downloaded from the USGS website. Using hydrology tools in ArcGIS 10.8, 19 catchments were obtained, as seen in Figure 1. The sinks in the DEM data were filled, water flow direction was estimated, and flow accumulation was set with a threshold of 10,000 and 60,000 to obtain the catchments. As the study area has short streams and not many large rivers, areas in the southwest do not have adequate streams to obtain a catchment. Hence, runoff obtained

SCS-CN Method
The SCS-CN method is the most commonly used empirical hydrological method developed by the NRCS, USDA, and is widely used to simulate runoff [19,29]. The curve number method has been successfully adopted in many ungauged watersheds and has expanded its scope of application in urbanised catchments and forested watersheds [29]. The surface runoff model uses the curve number approach of the US Soil Conservation Service [29], based on combinations of land use, hydrological soil group, and antecedent moisture condition (AMC) for the estimation of runoff. The amount of runoff was estimated using the SCS-CN method in presence of GIS and RS. The curve number is the most important factor in determining runoff via the SCS based method. The runoff of the soil and land use complex is represented by CN, which is a function of soil type, moisture conditions, and land use type [38].
The SCS-CN model is based on the water balance equation as shown in the Equations (1)- (3).
Land 2021, 10, 839 where P is the rainfall depth (mm), Ia is the initial abstraction of the rainfall (mm), F is infiltration, Q is surface runoff depth (mm), S is the potential maximum soil retention, and λ is abstraction coefficient that ranges between 0.0 to 0.2 or 0.05 for urbanised catchments [39]. The value of 0.2 as mentioned by Natural Resources Conservation Service (NRCS) was used in the study [29]. The runoff depth can be obtained for two conditions from the Equations (1) and (2): For P > Ia and If P < Ia, Q = 0 and Q from Equation (4) is expressed as follows: In Equation (5), S was obtained from the dimensionless parameter CN. CN is runoff curve number that ranges from 0 to 100.
In the SCS-CN method, curve number plays an important role in determining the surface runoff of an area, and its value depends on the corresponding soil type and AMC. AMC is antecedent moisture condition present in the soil at the beginning of the rainfall. Areas with a higher curve number represent higher runoff generated from the surface. In our study, we chose B, C, and D as three hydrological soil groups (HSG) present in the study area, and the soil moisture condition (AMC II) was set as moderate according to average runoff condition. A combined map of land use and HSG was generated by combining land use and soil maps in ArcGIS using overlay analysis. Then, the CN values were assigned for each polygon based on the information of land use and soil. The CN values for each land use type under AMC II is obtained from the TR-55 lookup Table 3 [29,40]. Area weighted CN was obtained to simulate the runoff of the whole area using the initial curve numbers from the table as in Equation (8). Combining the CN values of different land use and soil complex polygons, weighted CN was calculated for each catchment. The weighted CN is calculated by taking the sum of each CN value multiplied by its fraction of the total area of each land use type [41]. The Equation is given below: Land 2021, 10, 839 where CNw is the weighted curve number; CNi is the curve number for each land use type; Ai is area of land use with respective curve number; and A is the total area of each land use type. Finally, surface runoff depth was estimated, and runoff coefficient, i.e., the ratio of runoff to rainfall, was calculated.

Analysing Impact of Land Use Change on Surface Runoff
Impact of land use change on surface runoff was analysed by comparing the difference in runoff variables. Runoff depth and runoff coefficient was used as two variables to assess land use change on surface runoff. Surface runoff from the catchments were obtained for the different land use conditions and runoff was calculated under different rainfall return periods. The difference of runoff and runoff coefficient was obtained for the land use period of [1980][1981][1982][1983][1984][1985][1986][1987][1988][1989][1990][1990][1991][1992][1993][1994][1995][1996][1997][1998][1999][2000][2001][2002][2003][2004][2005], and 2005-2015 by using Equations (8)-(10): where Qa and Qb denotes surface runoff depths (mm) of the initial and final land use, respectively, P is the rainfall depth (mm), ∆Q is the change in runoff depth between two periods of land use conditions, ∆α is the absolute change in the runoff coefficient, and ∆β represent relative change in runoff. Land use change leads to an increase in surface runoff if the values of ∆Q and ∆α are positive in the above Equation (10). A relationship between land use and surface runoff was determined by using Pearson's product-moment correlation coefficient. A positive and larger correlation coefficient suggests that the factor is more significant in the change in surface runoff.

Validation of SCS-CN Model
For analysing the performance of the model, the model was validated using observed flow and simulated runoff between 1981 and 2015. Four statistical indices, as shown in Table 4, were used for testing the goodness of fit.

Land Use Change in the Study Area
Land use types for four different years can be seen in Figure 4. The spatial distribution of land use types show that the area has experienced significant change, especially due Volumetric efficiency (VE) represents the fraction of water delivered at the proper time −Inf ≤ VE ≤1 close to 1-efficient

Land Use Change in the Study Area
Land use types for four different years can be seen in Figure 4. The spatial distri tion of land use types show that the area has experienced significant change, especi due to urban expansion. There was a remarkable expansion of built-up areas from 62 to 307.54 km 2 between 1980 and 2015. It was the largest gain, with a net increase of 244 km 2 . The largest net loss from 1980 to 2015 was observed in farmland and coastal wetla with 218.2 km 2 and 46.65 km 2 , respectively. However, constructed land that includes bu up land and rural settlements underwent the largest net increase. The forest area redu substantially with a change of 23.93 km 2 .  Similarly, grassland and farmland were replaced by built-up areas in the south and south-eastern parts of the mainland and major parts of the island. It can be observed that most coastal wetlands decreased from 1980 to 1990, and built-up areas occupied most areas. The land area of Xiamen City expanded outward by reclamation and construction along the coast after the national survey of 1985, which contributed to the expansion of land area and reduced wetlands in 1990 [42]. A considerable increase in water bodies indicate some farmlands being converted to reservoirs, aqua farms, and other constructed wetlands. Table A1 shows the conversion of the land use in the form of a change matrix for the period of 1980 to 2015. There was a major conversion from farmland to constructed land. The trend of land use change from 1980 to 2015 indicates that urban development has dominated the island city, resulting in vast areas of impervious surfaces.
In Table A2, the percent and area change of each land use type is depicted from 1980 to 2015. It can be observed that farmlands have been reduced by 14.01% and built-up area has increased by 15.7%. The transfer process of land use types in between the study years can be seen from to 2015. It can be observed that farmlands have been reduced by 14.01% and built-up area has increased by 15.7%. The transfer process of land use types in between the study years can be seen from Figure 5. It can be noticed that major portion of change in farmlands in 1990 and 2005 have been replaced by built-up land. Overall, constructed land increased from 9.12% in 1980 to 26.1% in 2015. This indicates that urban impervious areas have increased considerably in the last few decades of the study period.

Spatial Distribution of Runoff in Different Years
The spatial distribution of runoff depth for the highest rainfall return period of 50 years is shown in Figure 6. Surface runoff depth ranges from 176.28 mm to 329.16 mm. The area covered by built up land with higher CN depicts a higher amount of runoff in the period from 1980 to 2015. Xiamen Island, particularly dominated by constructed area, shows larger areas of high runoff. The urban areas with high CN are often the areas with high runoff value.

Spatial Distribution of Runoff in Different Years
The spatial distribution of runoff depth for the highest rainfall return period of 50 years is shown in Figure 6. Surface runoff depth ranges from 176.28 mm to 329.16 mm. The area covered by built up land with higher CN depicts a higher amount of runoff in the period from 1980 to 2015. Xiamen Island, particularly dominated by constructed area, shows larger areas of high runoff. The urban areas with high CN are often the areas with high runoff value.
Under the land use conditions of 1980,1990,2005, and 2015, the average surface runoff depth and runoff coefficient show an increasing trend as shown in Figure 7. The average surface depth of the area under four different periods differ from 117.2 to 271.6 mm, and runoff coefficient fluctuated from 0.6 to 0.8. The calculated value of average surface runoff and runoff coefficient increases as the rainfall return period increase from 5 years to 50 years. It can be observed that surface runoff and runoff coefficient significantly increase from the year 1990 to 2015. The amount of runoff percent for different land use types can be noticed in Figure 8. A major portion of runoff is contributed by built-up land and rural settlements which are the major constructed land in the study area with the rise from 38.2% to 48.4%. Runoff contributed by built-up land alone in the study area increased from 14.2% to 27.9% from 1980 to 2015. Similarly, farmlands and coastal wetlands also contribute to the significant portion of surface runoff in the study area.

Change in Surface Runoff in Different Land Use Conditions
Increase in urbanised area gives rise to impervious surfaces, resulting in increased surface runoff. Table 5

Change in Surface Runoff in Different Land Use Conditions
Increase in urbanised area gives rise to impervious surfaces, resulting in increased surface runoff. Table 5

Relationship between Surface Runoff and Land Use
Land use exhibits a high relationship with surface runoff with a statistically significant correlation (p > 0.05). As shown in Table 6, farmlands (−0.97), forestland (−0.96), and grassland (−0.97) contribute negatively to the surface runoff, which indicates that increase in these land use types contribute to a decrease in surface runoff. Increase in urban built-up land (0.98) and rural settlements (0.99) corresponded to an increase in average surface runoff, while the decrease in farmland, forestland, and grassland contribute to higher runoff. Furthermore, increase in coastal wetlands contribute to a decrease in runoff, and unused land contributed positively to average runoff. The results are consistent with the common knowledge that increase in urban constructed land causes an increase in surface runoff; however, increases in farmland, grassland, forestland, and wetlands lead to a reduction in surface runoff. Water bodies are considered to have less effect on runoff depth, hence the relationship is less significant. Table 6. Relationship between land use and surface runoff under rainfall return period of 50 years. t-t-statistic, p-p-value, and R 2 -Pearson's correlation coefficient. Areas in square kilometres.

Farmland Forestland Grassland
Water Body

Rainfall-Runoff Correlation Analysis
The correlation analysis shown in Figure 9 indicates a strong linear relationship between the SCS-CN runoff and maximum daily rainfall, with a correlation coefficient of 0.99. The study results are comparable with the findings of Rawat and Singh [43] who found a good coefficient of determination (0.91) in a small study area using the SCS-CN model. The slope of the line determines the runoff coefficient, i.e., 0.84. The resulting findings are similar to Al. Ghobari et al. [44], who came to the conclusion that the SCS-CN model has a better simulation effect on study areas with a coefficient of runoff greater than 0.5 than those with a coefficient of runoff less than 0.5. This coefficient may provide valuable information on the extent of the basin response to runoff generation.
Land 2021, 10, x FOR PEER REVIEW 14 of 19 than 0.5 than those with a coefficient of runoff less than 0.5. This coefficient may provide valuable information on the extent of the basin response to runoff generation.

Validation of SCS-CN
The SCS-CN model was validated using historical observations and simulated flow from 1980 to 2015, as shown in Figure 10. Some statistical efficiency criteria are used to perform evaluation of the validation results between simulated output and observed data which are percent bias (PBIAS), correlation coefficient (r), Nash-Sutcliffe efficiency (NSE) and volumetric efficiency (VE). These statistical indices indicate the goodness of fit between simulated and observed data. The model successfully predicts the annual flow with the high accuracy as depicted by the indices. The PBIAS, r, NSE, and VE were −5.7, 0.82, 0.64, and 0.86, respectively. Although there is an underestimation of streamflow due to

Validation of SCS-CN
The SCS-CN model was validated using historical observations and simulated flow from 1980 to 2015, as shown in Figure 10. Some statistical efficiency criteria are used to perform evaluation of the validation results between simulated output and observed data which are percent bias (PBIAS), correlation coefficient (r), Nash-Sutcliffe efficiency (NSE) and volumetric efficiency (VE). These statistical indices indicate the goodness of fit between simulated and observed data. The model successfully predicts the annual flow with the high accuracy as depicted by the indices. The PBIAS, r, NSE, and VE were −5.7, 0.82, 0.64, and 0.86, respectively. Although there is an underestimation of streamflow due to static land use, annual flow statistics indicate that there is a good relationship between observed and simulated streamflow. Hence, the model performance was satisfactory and responded well in simulation of runoff.

Validation of SCS-CN
The SCS-CN model was validated using historical observations and simulated flow from 1980 to 2015, as shown in Figure 10. Some statistical efficiency criteria are used to perform evaluation of the validation results between simulated output and observed data which are percent bias (PBIAS), correlation coefficient (r), Nash-Sutcliffe efficiency (NSE) and volumetric efficiency (VE). These statistical indices indicate the goodness of fit between simulated and observed data. The model successfully predicts the annual flow with the high accuracy as depicted by the indices. The PBIAS, r, NSE, and VE were −5.7, 0.82, 0.64, and 0.86, respectively. Although there is an underestimation of streamflow due to static land use, annual flow statistics indicate that there is a good relationship between observed and simulated streamflow. Hence, the model performance was satisfactory and responded well in simulation of runoff.

Discussion
The study area has undergone significant land use change from 1980 to 2015. A significant loss of farmland was observed in the period between 1980 and 2015. During this period, most of the farmlands are replaced by built-up lands in the south-eastern part of the study area. After the national survey was carried out in 1985, the land area of Xiamen City expanded outward through reclamation [42]. The major land use change in Xiamen is attributable to land reclamation and urban development in the past years. The process of urban construction in reclaimed land and building new residential areas started to rise after 1985, which is reflected in the runoff increase after 1990 [45,46]. In 1980, after the city was declared as specific economic zone, most of the farmlands and forestlands were converted into urban areas between 1985 and 2005. After 2005, there was major sea reclamation which increased the urban area. New initiatives for industrial and economic development occurred in the period of 1990 to 2005, and an urban renewal program took place between 2003 and 2012 [31]. The change in land use significantly affected the runoff hydrology of the city as a result of urban development. The high values of runoff gradually expanded outside Xiamen Island and were mainly distributed in the areas with increased urban construction [40]. Results show that particular increase in constructed land with higher CN value contributes to higher runoff, whereas farmland, forestland, and grassland with lower CN contribute to a lower percent of runoff [30]. Therefore, change in land use, particularly an increase in urban areas, corresponds to an increase in runoff. As a result of rapid urbanisation, significant increase in surface runoff is observed in the island and outer areas of mainland, which poses higher risks for urban as well as coastal floods [47]. Similar results depicting the strong impact of urbanisation on surface runoff has been obtained by previous studies [2,7,17,48]. Relevant studies have shown that the extent to which urbanisation affects hydrological response depends on spatial and temporal scale, physical geography, landscape composition, and physical and climatic characteristics [49]. The study area, being a lowland coastal area, is relatively flat, and a major part of the area has a slope of less than 15%. Therefore, the slope of the area had a less significant effect on the surface runoff. Additionally, the main objective of the study is to analyse the impact of land use change on the surface runoff, especially due to rapid urbanisation. The impact of slope on the runoff characteristics (peak flow and runoff velocity) are not studied in detail.
However, future studies could help better understand the effect of slope on peak flow and runoff velocity.
This study demonstrated the use of a GIS-based SCS-CN method to assess the effects due to land use change on surface runoff by integrating spatial data and hydrological parameters. A GIS-based approach proved to be a reliable tool for quantifying the impact of land use change on runoff with respect to change in CN, which is a function of soil, land use, and moisture conditions [50]. The input data of the soil map and rainfall were based on actual field data, and the CN values of each land use type were obtained from the USDA standard table. The simulation of runoff was validated by comparing the observed and simulated annual flow in the study area, which shows that the design runoff simulated by the model is well accepted. Therefore, the model produced accurate and reliable results of runoff incorporating different spatial aspects. The results were consistent with the results obtained by previous studies of runoff simulations in Xiamen [40,42,51].
To fully comprehend the impacts of spatiotemporal land use change on abrupt or gradual flood peaks and runoff flow, various geographical parameters that influence the runoff should be considered in future studies. In the study, other human and environmental factors affecting runoff, such as the construction of dams, reservoirs and underground extraction, drainage systems, temperature, canopy cover, and soil loss, etc. are given little attention. Therefore, it is necessary to conduct extensive studies including these factors in the future.

Conclusions
With the gradual increase in urban areas from 1980 to 2015, a major fraction of land was converted to areas with poor infiltration and low potential storage, which significantly influenced surface runoff. In order to analyse the effect of land use change due to urbanisation on surface runoff, we used GIS-based SCS-CN model. The change in average runoff during the study period was determined, and relation between land use and runoff was obtained. A relation between rainfall and SCS runoff showed that the performance of SCS-CN model is suitable for runoff estimation. The GIS-based approach seemed to be an efficient tool for assessing the land use change and surface runoff through spatial analysis. Hence, the conclusions derived are as follows: 1.
The major changes in land use were observed at the expense of conversion of farmland to built-up land. Farmland decreased by 14.02%, and built-up land increased by 15.7%, from 1980 to 2015. Another significant change can be observed in the reduction in coastal wetlands by 2.99% which is attributed to land reclamation and conversion of reclaimed land to constructed land. Overall, the constructed land in the study area increased from 9.12% in 1980 to 26.1% in 2015; 2.
Spatial change in surface runoff was noticed from 1990 to 2015 in the south-eastern part of the study area, in which there are areas with higher urban built-up land. Therefore, the increase in runoff in the study area indicates the positive impact of urbanisation. The amount of runoff contributed by land use type shows that, with the increase in total constructed land, the amount of runoff significantly increased from 38.2 to 48.4%. The amount of surface runoff is noticed to be increased from 1990, which is consistent with the rise in urban development that occurred since 1990; 3.
The average surface runoff was positively correlated with the built-up and rural settlements, but negatively correlated with the areas of farmland, forestland, grassland, and coastal wetlands. The urbanised land use was determined as a dominant factor for surface runoff increase during the period from 1980 to 2015.
The areas with higher runoff, especially in Xiamen Island, dominated by urban built up land, should be given more attention during land use planning. Forestland, grassland, and farmland in the area have a higher significance for storing runoff. Therefore, these natural green infrastructures should be considered as potential areas for runoff storage. Further research can be focused in exploring the effectiveness of natural infrastructure and nature-based solutions for runoff mitigation and reducing urban flood risks.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.
Acknowledgments: Authors would like to extend their sincere gratitude to the USGS and Geographical Information Monitoring Cloud Platform is acknowledged for making the Land use data product available for the study. We also immensely thank the UCAS International Scholarship for the international Masters Student program. We thank the anonymous reviewers, editor for their constructive comments on the manuscript.

Conflicts of Interest:
The authors declare no conflict of interest.