Evolution Characteristics of Landscape Pa tt erns and the Response of Surface Runo ﬀ in a Rapid Urbanization Area: Focus on the Chang–Zhu–Tan Metropolitan Area of China

: With the acceleration of urbanization, the disturbance to urban landscape pa tt erns causes changes to urban surface runo ﬀ and increases the risk of urban waterlogging. We studied the response relationship between landscape pa tt ern change and surface runo ﬀ in the Chang–Zhu–Tan metropolitan area for the period from 2000 to 2020, analyzing the driving factors that a ﬀ ected surface runo ﬀ . The in ﬂ uence of landscape pa tt ern evolution on surface runo ﬀ was studied using the SCS-CN ﬂ ow generation model, the moving window method, and Pearson’s analysis. The analysis showed that between 2000 and 2020, the forest area decreased, while the impermeable land area increased substantially. At the same time, the landscape spread degree (CONTAG) decreased, whereas the landscape fragmentation degree (DIVISION) increased, and the Shannon diversity index (SHDI) and landscape shape index (LSI) increased. The surface runo ﬀ in the main urban areas of Chang–Zhu–Tan increased substantially. The results showed that surface runo ﬀ is negatively correlated with SHDI, LSI, and DIVISION but displays a positive relationship to CONTAG. Soil texture and precipitation had the greatest impact on surface runo ﬀ . This study emphasizes the importance of landscape pa tt ern evolution to surface runo ﬀ in rapidly developing metropolitan areas in terms of reducing surface runo ﬀ , alleviating urban waterlogging and preventing regional ﬂ oods. Our research additionally seeks to optimize the landscape pa tt ern of the Chang–Zhu–Tan metropolitan area.


Introduction
With the acceleration of urbanization, the interference of human activities in the urban surface landscape is causing changes in the urban climate.Flood disasters caused by extreme rainstorm events occur frequently worldwide, especially in ultra-highintensity urban areas [1].Therefore, these phenomena have been the focus of many scholars worldwide [2].As an important index in efforts to alleviate urban waterlogging, it is of practical significance to explore the influence of landscape pattern change on surface runoff change in the Chang-Zhu-Tan metropolitan area.
The research into relationships in landscape pattern evolution has always been a research hotspot for scholars at home and abroad [3][4][5].This field primarily studies three aspects, namely, city, watershed, and region, and principally explores the driving factors, land use/policy evolution landscape pattern, and the influence of landscape pattern change on the social economy [6], urban climate [7,8], and ecological services [9].At the city level, Wang et al. [10] found that the degree of landscape fragmentation in the study area increased, while connectivity and diversity decreased; at the regional level, Zhang et al. [11] also found that the landscape fragmentation in the study area and the landscape shape index increased; finally, at the watershed level, Xie et al. [12] found that the degree of patch dispersion, spread, and segmentation increased.It has been found that the disturbance of human activities is the main factor causing the change in landscape pattern [13][14][15] and that the landscape pattern has a scale effect [16,17].
Surface runoff refers to residual water flowing on the surface after natural precipitation is absorbed by soil and other surface cover and evaporated into the air as a result of the combined interactions of precipitation and underlying surface factors [18].Runoff has a greater sensitivity to daily precipitation data and to variations in precipitation duration.However, precipitation data often suffer from problems such as missing data, observational uncertainty, and low observational frequency.Therefore, the ability of weather generator models to provide extreme value indications is very important for risk assessment research.Recently, some scholars found that the CFSR and CLIGEN climate generator models are better able to solve this problem [19][20].The methods of simulating surface runoff mainly include SWAT [21,22] and SCS-CN [23,24].Many scholars have carried out research on surface runoff, including the impact of surface runoff changes on soil erosion and the impact of rainfall, land use, and land cover distribution on surface runoff, soil nutrients [25,26], rainfall, land use, and land cover [27][28][29].Rainfall intensity and spatial scale have a certain influence on the SCS-CN runoff model [30], and the SCS-CN runoff model has been studied at the urban and watershed levels.For example, Liu et al. [31] found that the response of urban surface runoff is more obvious when the rainfall is low in a study of impervious surface and surface runoff in Xuzhou City.Fang et al. [32] used waterlogging points to evaluate simulation accuracy and found that the simulated surface runoff was most accurate at a rainfall intensity of 200 mm/d.
When the correlation between landscape index and runoff change is analyzed, the influence of landscape pattern evolution on runoff can be studied [33].ArcGIS, ENVI, and Fragstats software are usually typically used to perform processing.Other studies have shown that surface runoff fluctuations are affected by both rainfall and land use changes [34], and related studies have shown that the impact of landscape pattern evolution on runoff has a greater effect than the impact of rainfall [35].The related research on landscape pattern or land use and surface runoff, whether conducted at home or abroad, is primarily based on watershed and urban scale [36].Previous studies have found that landscape fragmentation is negatively correlated with runoff and that landscape dominance is positively correlated with runoff [37]; the larger and more concentrated the area of artificial landscape patches is, the stronger the surface runoff generation capacity will be [38].When the urban impermeable surface increases and the precipitation falls, the surface runoff response becomes more obvious [31].
From the perspective of research topics, most of the existing research is on the impact of land use/landscape pattern change and surface runoff at the watershed scale or at the urban level.There is presently a lack of research from the perspective of metropolitan areas.In terms of research, most of the objectives of studying urban climate are the temperature of the city or the air quality of the city [39,40].Until now, there have been relatively few studies on hydrology.
The metropolitan area is an important space in which to promote regional highquality development and advance the development of urban agglomerations.In addition to focusing on economic development, attention should be paid to urban ecological security.The Chang-Zhu-Tan metropolitan area belongs to the core part of the Chang-Zhu-Tan urban agglomeration, the fourth largest national metropolitan area in China, and contains world-class ecological green hearts.During the period from 1980 to 2020, regional policies were frequently introduced, the region developed rapidly, and the landscape pattern underwent great changes.As such, it constitutes a typical and representative research object.Therefore, this study explores the influence of landscape pattern change on surface runoff in Chang-Zhu-Tan metropolitan area in order to reduce the risk of urban waterlogging and regional flood.Our research provides theoretical support for maintaining or improving the spatial planning of Chang-Zhu-Tan metropolitan area, furnishing research references for related studies on air quality, temperature and biodiversity.
The main objectives of this study are to study the changes in land use types in the Chang-Zhu-Tan metropolitan area in 2000,2005,2010,2015, and 2020 and to analyze the evolution characteristics of landscape patterns at the class level and landscape level.The surface runoff changes in 2000, 2005, 2010, 2015, and 2020 are simulated, and the evolution characteristics of surface runoff are analyzed.The GIS fishing net method and Pearson correlation analysis method are used to analyze the response relationship between the landscape pattern and surface runoff.Then, the driving factors affecting surface runoff are explored, and optimization suggestions are proposed based on the research results.

Background
The Chang-Zhu-Tan metropolitan area is the abbreviation for the combination of the districts of Changsha, Zhuzhou and Xiangtan (Figure 1).The Chang-Zhu-Tan metropolitan area is located in Hunan Province, China, and sits in the middle reaches of the Yangtze River [41].The geographical location of the Chang-Zhu-Tan metropolitan area has a continental subtropical monsoon climate.It is rainy in spring and summer, dry in autumn and winter, has a dense river network, and receives concentrated rainfall.Large-scale heavy rainfall causes the water level of rivers to rise, resulting in frequent and very impactful floods in some areas.The research area of this paper includes 20 districts and counties such as the Tianxin District, Furong District and Yuhua District, with a total of 26,917 km 2 .Since the integration of Chang-Zhu-Tan in 1997, it has been codified into the national "Eleventh Five-Year" plan in 2005, and promulgated and implemented policies such as "Regional planning of Chang-Zhu-Tan urban agglomeration", "Overall planning of ecological green heart area of Chang-Zhu-Tan urban agglomeration" and "Development planning of Chang-Zhu-Tan metropolitan area".So far, the Chang-Zhu-Tan urban agglomeration has become the core area of the development of Hunan Province.The developmental background of the study area is shown in Table 1.

Years Events 1950
The idea of building "Mao Zedong City" in Chang-Zhu-Tan is put forward.

1984
The construction of the Chang-Zhu-Tan economic integration program is formally proposed.1997 The integrated development strategy for Chang-Zhu-Tan is implemented.

2005
The Chang-Zhu-Tan urban agglomeration is written into the national "Eleventh Five-Year Plan".The provincial government promulgates and implements the "Chang-Zhu-Tan urban agglomeration regional planning", which is the first urban agglomeration regional planning in mainland China.

2007
The

Data Source and Processing
The main data used in this study are land use data, precipitation data, soil texture data, elevation data, China 's population spatial distribution, China 's GDP spatial distribution, road spatial distribution data, etc.The relevant data and sources are shown in Table 2, and the research technology flow chart is shown in Figure 2.

SCS-CN Runoff Model
The runoff curve method is a mathematical model used to estimate surface runoff.It was developed by the United States Department of Agriculture Soil and Water Conservation (USDA SCS) in the 1950s.Among comprehensive evaluation of existing models, the SCS-CN model has a simple calculation process and requires few parameters.The mathematical formula of runoff curve method [42] is as follows: where Q is surface runoff (mm), P is precipitation (mm), and S is the maximum potential stagnant water (mm).The relationship between S and the number of curves is as follows: where CN is the number of curves, which is a dimensionless parameter, indicating the ability to form surface runoff, and the theoretical value range is 0~100.
The runoff coefficient S is determined using the CN value, and the CN value is related to the land use type, soil texture, elevation, and soil moisture.The soil types of the Chang-Zhu-Tan metropolitan area-clay, sandy clay, clay loam, and sandy loam-were divided into four groups using SCS grouping methods (Table 3).In this paper, it is assumed that the soil moisture in the early stage of the study area is in a general state (AMCII [43]).Using the value lookup table provided in the US National Engineering Manual [44], we determined the CN value under different land use types of the Chang-Zhu-Tan to range between 35 and 100 (Table 4).Fragstats software is usually used to calculate the landscape pattern index, which is the most important tool in the quantitative study of landscape pattern change [45].
(1) Landscape pattern index method: Through an analysis of relevant research [46], combined with an assessment of the characteristics of the Chang-Zhu-Tan metropolitan area, a total of 8 types and landscape level indexes were selected and calculated in Fragstats software.Among them, the class level includes the total patch area (CA), patch density (PD), patch number (NP) and the largest patch area ratio (LPI).The landscape level includes landscape fragmentation (DIVISION), Shannon diversity index (SHDI), landscape shape index (LSI) and contagion index (CONTAG).The calculation formula of the relevant landscape pattern index and the significance of landscape ecology are shown in Table 5.The area index can reflect the total area change in each landscape type.The connectivity and fragmentation index reflect the continuity between each landscape type and explore the relationship between the spatial structure of each landscape type and surface runoff.

Scale Name of Indicator Formula Description
Class level Total area of patches (CA) The total area of a certain type of landscape patches.

Patch numbers (NP)
The sum of the number of all patches in the landscape.

PD =
; NP is the total number of certain patch types, and A is the total patch area.
The density of a certain patch in the landscape reflects the overall heterogeneity and fragmentation of the landscape.The fragmentation degree of landscape segmentation reflects the complexity of landscape spatial structure, and to some extent reflects the degree of human disturbance to the landscape.

Shannon diversity (SHDI)
SHDI = -∑  × ; Pi refers to the area ratio of type i to the whole landscape.
Reflecting landscape heterogeneity, the richer the land use, the higher the degree of fragmentation, and the greater the information content of uncertainty.
Landscape shape index (LSI) is the total patch area.
Describes the complexity of the edge of the plaque.The larger the index, the more irregular the shape of the plaque.

Spread index (CONTAG)
Pi refers to the area ratio of type i to the whole landscape, gik refers to the number of patches adjacent to type i and type k, and m refers to the total number of patch types in the landscape.
It reflects the degree of agglomeration or extension trend of different patch types in the landscape.High spread indicates that a certain dominant patch type has formed good continuity and has a high degree of landscape fragmentation.
(2) Moving window: The moving window method is one of the ways to calculate the landscape pattern index in Fragstats software.The overall landscape pattern index can analyze the overall characteristics of the study area, while the moving window method can reflect the detailed characteristics of the landscape pattern and the differences within the space.Based on the characteristics of the Chang-Zhu-Tan metropolitan area, the moving window method was used to study the multi-scale correlation between surface runoff and landscape pattern, and the ArcGIS fishnet tool was used to divide the surface runoff into grids.There are four kinds of moving window scales (Figure 3): 1 × 1 km, 3 × 3 km, 5 × 5 and 10 × 10 km.Two landscape pattern indexes, namely, landscape shape index (LSI) and contagion index (CONTAG), were selected for application in Pearson correlation analysis based on land use data and suggestions by other experts in the subject matter.We removed abnormal data, and the analysis results are shown in Table 6.The comparative study found that the correlation was better at a 5 km scale.

Markov Transition Matrix
The Markov transfer matrix is used in land use transfer analysis.In this study, it is deployed to calculate the land use change and the conversion between various types of land use in the Chang-Zhu-Tan metropolitan area from 2000 to 2020.The Markov transfer matrix formula is as follows [47]: where n is the number of land use types; the formula is the transition probability matrix of i type land use into j type during the study period; the row represents the transition probability of type i from period t to period t +1; and the column indicates the transition probability of the land use type change in the t period to the j type in the t + 1 period.These parameters must satisfy Formula (2).

Fishnet Spatial Analysis Method
(1) Spatial analysis of land use evolution: Analysis of the land use transfer matrix found that the farmland (FL) area changed significantly from 2015 to 2020.In order to study the spatial variation characteristics of FL area during this period, the ecological green center of Chang-Zhu-Tan in the study area was taken as the center, and the sampling points were set at 5 km grid scale.Four sampling lines were selected: west-east, south-north, southeast-northwest and southwest-northeast.The spatial variation characteristics of FL from 2015 to 2020 were studied.Samples were selected as shown in Figure 4. (2) Spatial analysis of surface runoff evolution: In order to study the spatial variation characteristics of surface runoff, the sample point was set at the 5 km fishnet scale and two sample lines (t-c sample line and t-z sample line) were selected along the three cities of Changsha, Zhuzhou, and Xiangtan.The t-c sample line runs from south to north through Xiangtan City and Changsha City, whereas sample 1-26 was marked between the north and south sample line.The t-z sample line runs through Xiangtan City and Zhuzhou City from west to east, as marked by 1-37.Samples were selected, as shown in Figure 4.

Statistical Analysis
This paper mainly uses the Pearson correlation coefficient method.The Pearson correlation coefficient method is widely used to calculate the linear correlation between two variables, the values of which are between −1 and 1.The greater the absolute value is, the stronger the correlation will be.The correlation coefficient formula is as follows [48]: In the formula, ̅ and  are the sample average values of  and  respectively.

Analysis of Land Use Evolution
The spatial distribution of land use in Chang-Zhu-Tan metropolitan area from 2000 to 2020 is shown in Figure 5.The results show that farmland (FL) and forest (FT) in the Chang-Zhu-Tan metropolitan area account for a relatively larger proportion than other land use types.The area ratio is shown in Figure 6b.From 2000 to 2020, FT and impervious land (IL) changed substantially (Table 7).The FT area decreased, whereas the IL area increased substantially.The area of FT and WB decreased by 1479.56 km 2 and 5.26 km 2 , respectively, with a decrease of 9.5% and 0.1.The total area of FL, shrub (SL), grassland (GL), wasteland (WL) and IL increased to 16.26, 345.04, 105.97, 66.23 and 1153.53 km 2 .These data indicate that human activities primary the main cause of land use change during this period.6a.From 2000 to 2005, the area of FL and IL increased by 3.8% and 31% respectively, and the area of FT decreased by 3.6%.The contribution rate of FT to FL was 239%, and the contribution rate of FL to IL was 86%.From 2005 to 2010, FL increased by 6.9% and the contribution rate of FT to FL was 161%.The FT area decreased by 7%, the IL area increased by 40%, and 87% of the contribution came from FT. From 2010 to 2015, the FL area increased by 4%, the IL area increased by 25%, WB area increased by 5%, and the FT area decreased by 4%.In general, during 2000-2015, the areas of FL, WB and IL increased, while the area of FT decreased substantially.From 2015 to 2020, the area of FL decreased by 11.8%, FT increased by 5.1%, and IL increased by 16.3%, respectively.Most of the reduced FL was converted into FT, SL and IL, and a small part was converted into GL and WL.The areas of FL, FT and IL accounted for a large proportion of each land use type, indicating that these three types dominated land use change from 2000 to 2020.
In summary, the FL area changed significantly from 2015 to 2020.The change in FL area in each sampling point is shown in Figure 6c.The analysis showed that the FL decrease was concentrated in the central, western, southern, northern, northwestern, southwestern, and northeastern regions.However, the change in the position near the ecological green heart is the smallest.This shows that the implementation of the policy of "Returning farmland to forest" and "Ecological green heart protection" had a significant effect on land use.

Landscape Pattern Evolution Analysis
The change in landscape pattern index in Chang-Zhu-Tan metropolitan area from 2000 to 2020 is shown in Figures 7 and 8.
The change of landscape pattern index in the Chang-Zhu-Tan metropolitan area from 2000 to 2020 is shown in Figures 7 and 8.  (1) Class level: Analysis of Figure 7a-d shows that the total area of FL landscape patches increased from 2000 to 2015 and decreased from 2015 to 2020.The total area of IL patches increased.The total area of water patches decreased slightly, and the total area of FT, SL, and WL decreased from 2000 to 2015 and increased from 2015 to 2020.In general, the most obvious changes in CA were FT and IL.The largest land use type of LPI in 2000-2010 was FT, which was transformed into FL in 2015 and FT in 2020, indicating that FL and FT dominated the landscape pattern of the Chang-Zhu-Tan metropolitan area during 2000-2020.PD represents the density of a certain patch in the landscape, reflecting the degree of fragmentation of such patches.During 2000-2020, the density of FL patches decreased first and then increased, while the density of FT patches simply increased.The density of IL patches increased first and then decreased.Conversely, the density of WB patches continued to decrease, which occurred considerably during 2015-2020.The NP index represents the sum of the number of patches in the landscape, which is basically consistent with the trend in PD time.
The edges of FT landscapes became irregular, while the opposite occurred for GL, WB, FL, SL and IL.The most obvious changes in the landscape pattern index of the four types of levels were for FL, FT, WB, and IL.The change range was the largest during 2015-2020.The analysis of the "Hunan Statistical Yearbook" showed that, from 2000 to 2015, the Chang-Zhu-Tan metropolitan area was dominated by the conversion of agricultural into urban areas, indicating urban expansion.From 2015 to 2020, the urbanization rate increased significantly.Coupled with the implementation of the "Returning farmland to forest" policy, this resulting in a reduction in farmland area, an increase in fragmentation, the acceleration of urbanization, and an increase in impervious land area.
(2) Landscape level: Landscape level reflects the overall structural characteristics under the combination of different types of patches.The analysis of Figure 7e showed that, in terms of time, SHDI and DIVISION in Chang-Zhu-Tan metropolitan area increased slightly from 2000 to 2020, approaching 1.This indicated substantial land use change, as well as rising landscape heterogeneity and fragmentation.To a certain extent, it reflected the increase in human disturbance to the landscape.Additionally, CONTAG gradually decreased, indicating that the dominant patches in the landscape was not connected and the degree of landscape fragmentation was high.The LSI index gradually increased, indicating that the edge shape of the landscape became more and more complex, while the patch shape became more and more irregular.The spatial variation characteristics are shown in Figure 8. From 2000 to 2020, the CONTAG of Chang-Zhu-Tan metropolitan area decreased, while DIVISION, SHDI and LSI increased, which occurred mainly in the central urban areas of the three cities of Chang-Zhu-Tan.The most obvious change was the trend of expansion, and the distance between the three cities grew closer and closer, indicating that the urbanization process of Chang-Zhu-Tan metropolitan area accelerated from 2000 to 2020, which was the main reason for the change in landscape pattern.

Chang-Zhu-Tan Metropolitan Area Surface Runoff Simulation Calculation
The SCS-CN runoff model was used to simulate the spatial and temporal evolution characteristics of surface runoff in the Chang-Zhu-Tan metropolitan area from 2000 to 2020.Using Arcgis10.1 software, precipitation, soil hydrological grouping, soil texture, DEM and land use data were used as basic data.The simulation results are as follows: (1) Precipitation change characteristics: The spatial distribution of annual average precipitation in 2000, 2005, 2010, 2015 and 2020 in Chang-Zhu-Tan metropolitan area is shown in Figure 9.In terms of time change, the average precipitation in the five years was about 1290.8 mm, and the average annual precipitation in 2000, 2005, 2010, 2015 and 2020 was 1240.7,1205.78,1444.8,1273.3 and 1288.9 mm, respectively.The spatial distribution of rainfall in the east was more than that in the west, and the south saw more than the north.Precipitation is mainly related to climate and other factors, and it is also a direct source of surface runoff.The relationship between precipitation and surface runoff is discussed in the following sections.(2) CN value variation characteristics: CN value is a key variable to calculate the maximum potential water retention in soil, which is related to soil texture type, land use, elevation and other factors, indicating the ability to form surface runoff.The higher the value, the stronger the ability to form surface runoff.HEC-GeoHMS10.As shown in Figure 11a, the surface runoff in the southern part of Xiangtan City decreased, but increased as it approached closer towards the urban area of Xiangtan City.The runoff was lowest in the ecological green heart of the Chang-Zhu-Tan urban agglomeration.Similarly, the surface runoff increased gradually from the north of Changsha City but decreased in northernmost edge of Changsha City.The spatial variation of runoff between 2000 and 2020 is shown in Figure 11b.The surface runoff of Chang-Zhu-Tan ecological green heart decreased substantially, indicating that the surface runoff of Xiangtan City and Changsha City increasing from 2000 to 2020.The surface runoff peak appears at the northernmost end, followed by fluctuations between Xiangtan City and Zhuzhou City.From 2015 to 2020, the spatial variation of runoff is as shown in Figure 11b.The surface runoff in the west and east increased but remained varied in other regions.However, the change rule of the t-z line was not more obvious than that of the t-c line.

Correlation Analysis between Landscape Pattern and Surface Runoff Flow
The runoff is dependent upon land use type and surface runoff of the Chang-Zhu-Tan metropolitan area changed with the change in land use.The landscape metrics: CONTAG, LSI, DIVISION, and SHDI at 5 × 5 km were used at the landscape level to describe the spatial pattern of fragmentation in relation to surface runoff between 2000 and 2020.The results are shown in Figure 12.In 2000, 2005, 2010 and 2020 year, the surface runoff was weakly correlated with CONTAG, DIVISION, LSI and SHDI, and the correlation size from strong to weak was: CONTAG > SHDI > LSI > DIVISION.In 2020, the surface runoff was significantly correlated with CONTAG, LSI and SHDI, and weakly correlated with DIVISION.The correlation size from strong to weak was: CONTAG > LSI > SHDI > DIVISION.From 2000 to 2020, the correlation coefficient between CONTAG and surface runoff increased from 0.38 to 0.7, LSI increased from 0.30 to 0.70, DIVISION increased from 0.22 to 0.40, and SHDI increased from 0.35 to 0.69.The coefficient was positively correlated with CONTAG and negatively correlated with DIVISION, LSI and SHDI.The result also showed that, as the connectivity of impervious surface increased, the surface runoff increased (Figures 8  and 11).The more complex the landscape edge shape is and the more varied the landscape type is, the more conducive conditions are to slowing down surface runoff.The impact of landscape patterns on surface runoff becomes significant with the acceleration of urbanization.
It can be seen that the landscape pattern of Chang-Zhu-Tan metropolitan area has a certain correlation with surface runoff, which is negatively correlated with SHDI, LSI and DIVISION, and positively correlated with CONTAG.Additionally, it is clear that the urbanization process accelerates, the correlation becomes more significant.

Analysis of Driving Factors Affecting Surface Runoff
The factors affecting surface runoff (Q) are diverse and complex.Six driving factors were selected from three aspects: social economy, natural factors, and policy (Figure 13).Among them, social economy factors included population (POP), GDP (GDP) and road (ROAD); natural factors included mean annual precipitation (RAIN), soil texture (SOIL) and elevation (DEM).Among the six factors, population and GDP are the most commonly used social economy drivers [49].Change in road distribution affects the change in landscape pattern, and ultimately has an internal driving force on the change in surface runoff.Natural conditions such as elevation, annual precipitation, and soil texture imposed important constraints on the flow generation, flow direction, and absorption of surface runoff.In addition, due to the frequent promulgation of regional policies in Chang-Zhu-Tan from 1980 to 2020, the implementation of policies has profoundly affected the changes in landscape types and changed the landscape spatial pattern, and thus affecting surface runoff.Taking 2020 as an example, the correlation between each driving factor and surface runoff is shown in Figure 14.
(1) Natural factors: The regression analysis showed that different soil texture types affected surface runoff differently.The higher the soil permeability was, the lower the surface runoff would be.Precipitation is positively correlated with surface runoff; the elevation affects the velocity of surface runoff, and the higher the elevation, the more surface runoff.Correlation comparison of driving factors: SOIL > RAIN > DEM.
(2) Social economy factors: POP and GDP are weakly positively correlated with surface runoff.That is, where there is a large population, buildings are denser, thus affecting rainwater infiltration and easily leading to the formation of surface runoff.However, such events may also be due to the more developed economy, modern infrastructure, and reasonable layout of the drainage network.Even if the rainwater cannot penetrate the ground, it can be discharged in time through the sewer network.The road network distribution is weakly negatively correlated with surface runoff.The correlation comparisons showed that the driving factors were POP > ROAD > GDP.
(3) Related policies: The promulgation and implementation of policies affects the ways in which landscape patterns change, altering surface runoff.Through the analysis of the previous article, we found that since the implementation of the integrated development strategy of Chang-Zhu-Tan three cities in 1984 and 1997, the scale of the three cities has been expanding, the distance between the three cities has gradually narrowed as they have come to occupy a large amount of farmland and forest area.This has resulted in a decrease in the area of ecological space, a substantial increase in the area of impervious land, and an increase in surface runoff.From 2007 to 2020, the Chang-Zhu-Tan urban agglomeration established a "Two-oriented society" reform pilot area and implemented regulations on the protection of ecological green hearts in the Chang-Zhu-Tan urban agglomeration.Through the previous analysis, it was found that the farmland area in Chang-Zhu-Tan metropolitan area decreased sharply from 2015 to 2020 and the surface runoff in the main urban area increased significantly.However, the change was the smallest in the area near the ecological green heart.This shows that the implementation of the overall plan for the ecological green heart area of Chang-Zhu-Tan urban agglomeration (2010-2030) effectively protected the ecological green heart.In summary, the integrated development strategy of Chang-Zhu-Tan led to urban expansion and affected the change in landscape pattern, thus impacting the change in surface runoff.The implementation of the protection policy of "Two-oriented society" and Chang-Zhu-Tan ecological green heart can effectively protect the landscape type of the ecological green heart area and reduce the surface runoff.In order to realize the strategy of a "Strong provincial capital" in Hunan Province and the development goal of Chang-Zhu-Tan metropolitan area by 2035, the Chang-Zhu-Tan metropolitan area should continue to develop rapidly in the future.The promulgation and implementation of relevant policies will have a significant impact on the landscape pattern and surface runoff of Chang-Zhu-Tan metropolitan area.

The Response of Surface Runoff to Landscape Pattern
The surface runoff is negatively correlated with SHDI and DIVISION, and positively correlated with CONTAG.This is consistent with the results of Wang et al. [47].The characteristics of landscape pattern depend on the scale of analysis scale [16].The positive correlation between LSI and surface runoff differs from that seen in the work of Zhu et al. [42] because of the difference in the scale of analysis.Therefore, it is necessary and important to select the optimal analysis scale for the study.
From 2000 to 2020, the surface runoff in the Chang-Zhu-Tan metropolitan area gradually increased, especially in the main urban area of the city characterized by strong human disturbance intensity.The runoff capacity, from strong to weak, is WB >IL >WL > FL, which is consistent with the results of previous related research [50].However, our conclusion of SL >FT > GL differs.This is because the surface runoff is related to soil texture and soil permeability.The soil texture and permeability of current study are different from those relevant to the study conducted by Tian et al. [50].Therefore, due to the acceleration of urbanization, the increase in construction area and the expansion of impervious land, rainwater cannot be absorbed by soil and the surface runoff increases [51].
In this paper, the driving factors of surface runoff are analyzed.In addition to landscape pattern factors, social economy, nature and policy factors are also considered.This research is far more comprehensive than previous studies [11].

Limitations and Future Research
Land use data optimization: This paper uses 30 m resolution data products, which basically meet the research needs, and whether higher precision has an impact on results still needs to be explored.
Simulation of surface runoff: the SCS-CN model combines soil texture, land use, topography and rainfall to perform the simulation, especially of the monthly average data used in rainfall.The average daily rainfall data may be missing because there are no measured data of surface runoff.As such, the accuracy of the simulation needs to be explored.In addition, this paper does not consider the impact of SCS-CN on correlation at different rainfall or spatial scales.
Selection of surface runoff driving factors: There are still many factors affecting surface runoff.This study only selects six, as well as a number of policy factors.Thus, the selection of driving factors is not comprehensive.
In future, the response relationship between the optimal accuracy of land use and the optimal scale of landscape pattern to surface runoff should be further explored.Using climate models and other methods to obtain high-precision rainfall data, while looking for the measured data of surface runoff, the surface runoff simulation results can be optimized.The selection of driving factors can be further expanded to quantify policy factors.

Suggestions for Alleviating Surface Runoff in the Future
Due to the rapid development of urbanization, human activities have led to changes in the landscape pattern and affected surface runoff.In order to alleviate or prevent urban waterlogging, suggestions for optimizing the landscape pattern of Chang-Zhu-Tan metropolitan area are proposed.
From 2000 to 2020, the area of FT in Chang-Zhu-Tan metropolitan area decreased and the area of IL increased, resulting in an increase in surface runoff, while FT, SL and GL had a strong ability to absorb rainwater.Therefore, in urban development, we should promote compact development, save land, reduce the consumption of ecological space due to urban expansion, and increase the area of FT, SL and GL.
At the scale of 5 × 5 km, the correlation between landscape pattern and surface runoff is the strongest, CONTAG is positively correlated, and DIVISION, LSI and SHDI are negatively correlated.Therefore, under the unit of 25 km 2 area, increasing landscape richness, reducing human disturbance to the landscape, naturalizing the shape of the landscape edge, reducing landscape connectivity, and enhancing the complexity of the internal spatial structure of the landscape can effectively slow down surface runoff.
The introduction of regional policies needs to be considered as a whole.For example, while introducing economic development policies, it is necessary to introduce other related policies, such as controlling urban boundaries and protecting urban ecological space in order to ensure the coordination of urban development and ecological protection.

Conclusions
By exploring the evolution characteristics of landscape pattern and the response of surface runoff in Chang-Zhu-Tan metropolitan area from 2000 to 2020, the following conclusions were drawn: From 2000 to 2020, the area of FT in Chang-Zhu-Tan metropolitan area decreased by 1479.56 km 2 , while the area of IL increased by 931.16 km 2 .The landscape connectivity decreased, the landscape fragmentation and richness increased, and the landscape edge shape tended to be complex.
The surface runoff of the Chang-Zhu-Tan metropolitan area increased year by year from 2000 to 2020, especially in the vicinity of the main urban area, indicating that the impact of impervious surfaces on surface runoff is significant.
Surface runoff is positively correlated with CONTAG and negatively correlated with DIVISION, LSI, and SHDI.In particular, the higher the landscape connectivity of the impervious surface is, the easier it is to produce surface runoff.
The surface runoff is positively correlated with RAIN, DEM, GDP, and POP and negatively correlated with SOIL and ROAD.Among these factors, it has the greatest correlation with soil texture and rainfall and is also affected by regional policies.

Figure 1 .
Figure 1.Description of study area.
= ( ) × 100; () refers to the largest patch area in a patch type.The proportion of the largest patch in a landscape type to the entire landscape area is used to determine the dominant patch type and indirectly reflects the direction and size of human activity interference.landscape level Landscape fragmentation (DIVISION) DIVISION = ; Ni is the number of patches of type i, and Ai is the total area of type i.

Figure 5 .
Figure 5. Spatial distribution of land use in Chang-Zhu-Tan metropolitan area.

Figure 6 .
Figure 6.Landuse in the Chang-Zhu-Tan metropolitan area.(a)-land use transfer matrix; (b)-the proportion of each land use type area; (c)-farmland spatial change from 2000 to 2020.

Figure 7 .
Figure 7. Characteristics of landscape pattern change from 2000 to 2020.

Figure 8 .
Figure 8. Spatial distribution characteristics of landscape pattern change from 2000 to 2020.
1 and Arc-Hydro Tools10.1 were used to calculate the CN value of the Chang-Zhu-Tan metropolitan area from 2000 to 2020.The change in CN value between 2000 and 2020 is shown in Figure 10.The CN value increased from 2000 to 2020 in various spatial locations.The area with the highest CN value is concentrated in the center of the Chang-Zhu-Tan metropolitan area, which is the main urban center of the three cities of Chang-Zhu-Tan.

Figure 10 .
Figure 10.Spatial distribution of CN from 2000 to 2020.(3) Characteristics of surface runoff variation: The average annual precipitation in 2000, 2005, 2010, 2015 and 2020 was input into the Arcgis10.1 grid calculator.The average annual runoff in five years was 1138.7,1105.5, 1345.18,1175.38 and 1185.95mm, respectively.The runoff capacity, from strong to weak, was WB > IL > WL > FL > SL > FT > GL.The surface runoff of Changsha, Zhuzhou and Xiangtan increased substantially.As shown in Figure11a, the surface runoff in the southern part of Xiangtan City decreased, but increased as it approached closer towards the urban area of Xiangtan City.The runoff was lowest in the ecological green heart of the Chang-Zhu-Tan urban agglomeration.Similarly, the surface runoff increased gradually from the north of Changsha City but decreased in northernmost edge of Changsha City.

Figure 11 .
Figure 11.Spatial distribution characteristics of surface runoff from 2000 to 2020.a-spatial distribution of surface runoff; b-two sample lines of runoff change from 2000 to 2020.

Figure 12 .
Figure 12.Correlation analysis between landscape pattern and surface runoff from 2000 to 2020.

Figure 13 .
Figure 13.Spatial distribution of driving factors.

Figure 14 .
Figure 14.Correlation analysis between driving factors and surface runoff.

Table 1 .
Development Background of Chang-Zhu-Tan Urban metropolitan.

Table 2 .
Research data sources.

Table 4 .
CN value index of the Chang-Zhu-Tan metropolitan area under AMCII state.

Table 5 .
Selection of landscape pattern index.

Table 6 .
Correlation comparison at different scales.