Assessing the Long-Term Hydrological Effects of Rapid Urbanization in Metropolitan Shanghai, China: The Finer the Landscape Classification, the More Accurate the Modeling?

Abstract

Rapid urbanization often leads to increase in surface runoff; its modelling is always the focus in the field of land use effect. One of the methodological issues is how to classify the landscape (land use/land cover) in the model. In this study, the long-term hydrological impact assessment (L-THIA) model was used to simulate the change of annual surface runoff during the rapid urbanization in Shanghai since 1965. Two landscape scenarios, based upon land uses and pervious/impervious surfaces, were compared, and the CN values were adjusted to validate the applicability of the two landscape scenarios. The results showed that there was almost no difference between the results based on the two landscape scenarios, and it was suggested that the simplified landscape scenario based upon pervious/impervious surfaces can be workable and efficient, while the land use scenario may not be necessary for the modelling considering its scale of interpretation of remote sensing data. It was found that there was a clear linear relationship between the percentage of impervious surfaces and surface runoff. For every 1% increase in impervious surface, runoff increased by 0.94%. In addition, the effect of precipitation on the modelling was also discussed, which indicated that with the increase in impervious surface percentage, the response of runoff change in both dry year and dry season was more sensitive.

1. Introduction

In order to maintain the virtuous cycle of urban water, it was essential to evaluate the long-term hydrological effect of urbanization. There were many studies on regional long-term hydrological effects, such as a discussion on the impact of precipitation patterns on water resources and the surrounding environment [1] and a model assessment of precipitation and future prediction [2]. In addition to the impact of rainfall, the increase in surface runoff caused by rapid urbanization is also a matter of concern; its modelling is always the focus in the field of land use effect. While studies on the hydrological effects of urbanization from the perspective of land use/cover change (LUCC) have accumulated a plentiful literature, the difficulty of such research lies in interpreting and classifying land uses using remote sensing data. In contrast, the perspective of pervious/impervious surface does not require such a complex classification. It can simplify the change process of land cover in urban areas, and the hydrological effect can be more intuitively described [3,4,5]. Therefore, the key methodological question is how the two landscape classification systems (land use/land cover) differ in surface runoff modelling.

A variety of models combined with land use data have been used to study the hydrological effects of urbanization, such as SWAT, SCS-CN, MIKE SHE, BASINS-WinHSPF and so on. These models have been applied to the impact of land use change on surface runoff [6,7,8,9,10,11,12]. L-THIA is a SCS-based model developed by Purdue University in the United States to assess runoff change and non-point source pollution caused by long-term land use change. The L-THIA model requires less data than other complex mechanism models, in which the runoff can be simulated only by inputting land use, soil type and long-term rainfall data. With the development of research and applications, the combination of the L-THIA model with GIS [13] could model the spatial distribution of runoff change.

The application of L-THIA has become increasingly mature after long-term development. It is widely used to predict runoff when various external conditions changed [10,14,15,16]. Since L-THIA modelling results were related to land use types of underlying surface, and CN values of land uses were differentiated, the literature therefore classified land use types according to the actual situation; different studies used various classifications of land uses. Common classifications include forests, grasslands, arable land, bare land, water, low-density residential, high-density residential, industrial areas, commercial areas and so on [17,18]. A total of 60 types of land use are provided in L-THIA GIS. Impervious/pervious surface is another landscape classification system, which is simpler than land uses. Directly taking impervious surfaces as a landscape has occasionally appeared in some studies [11,19], but it was not a common practice. It is worth mentioning that very little attention has been given to the impact of complex and simple classification on simulation results.

Since the runoff generation capacity of the same land use type varies, the CN value also varies. In order to ensure the accuracy of L-THIA model simulations, usually, it is also necessary to adjust the CN value according to the comparison of simulated and observed runoff [20,21,22,23]. The basic principle is to obtain the CN value by interpolating measured data of rainfall runoff, and the common way includes the average, median, and arithmetic average [18,24,25].

Shanghai is one of the fastest-growing cities in China and one of the largest cities in the world, with an urbanization level that has developed from 58.7% in 1978 to 89.3% in 2020. The rapid urbanization has resulted in the expansion of impervious water surface, the shrinkage of the natural river system and the rain island effect, which have led to a continuous increase in surface runoff. The flood event risks caused by the rapid urbanization have become a wide concern for both the government and the public [26,27].

This study analyzed the impact of rapid urbanization on runoff in Shanghai in the recent 45 years from the perspective of two landscape scenarios: one with a finer classification system consisting of various land uses and the other simplified to impervious/pervious surfaces. The study also examined the differences between the two scenarios and assessed the applicability of impervious/pervious surfaces. In addition, the relationship in sub-catchments between impervious surfaces and runoffs, CN adjustment and the influence of rainfall conditions on the modelling results were also discussed. The results will be an important practical reference for the hydrological response research in cities under rapid urbanization and will help in making decisions regarding the integral sustainability of cities that must be considered in the present and near future.

2. Data and Methods

2.1. Study Area

Shanghai is one of the largest metropolitan areas in China, covering an area of 6340 km² with a population of over 24 million. In the rapid urbanization, the impervious surfaces were expanding, and the land surface hydrological process was changing. As a typical plain river network area, Shanghai has a reciprocating river pattern and a daily tidal cycle of two fluctuations, making it difficult to define the upstream and downstream boundaries and determine the structure of the river system and sub-catchment boundaries. Therefore, Shanghai has been divided into several sub-catchments for river management through large amount of sluices (Shanghai Water Authority 2001). The study area was covered by 11 sub-catchments as showed in Figure 1A.

2.2. Data Collection

Soil data came from the Key Laboratory of Geographic Information Science, Ministry of Education, East China Normal University. Due to the stability of soil properties, the whole research process adopted the same soil type in all years, and the hydrological soil groups were used as constant values for model calculation.

The land use data include four periods of 1965, 2000, 2006 and 2010. The data in 1965 were aerial data with a spatial resolution of 2 m. The data of 2000, 2006 and 2010 were interpretated from QuickBird satellite remote sensing data. They also came from the Key Laboratory of Geographic Information Science, Ministry of Education, East China Normal University. The scale is 1:10,000, and the spatial resolution is 1.4 m. Based on the existing remote sensing image data, geometric correction, spectral processing, image mosaic, coordinate positioning, water enhancement and other work carried out, it should be noted that there is no land use data of Chongming, Hengsha and Changxing Island in 1965. In order to maintain the consistency of data in whole period, the inland part of Shanghai excluding three islands was selected as the area of this study, with an area of 5199.28 km².

The data of impervious surfaces was based on the permeable performance of various land use types. Land use patches were divided into impervious and pervious surfaces. Specific methods can be found in the author’s studies [28,29]. In fact, impervious surface data can also be interpreted without land use, and it is simpler to use remote sensing algorithms.

Precipitation is also one of the most important meteorological data in L-THIA. There were a total of 11 rain gages in Shanghai, including Longhua, Baoshan, Jiading, Minhang, Songjiang, Jinshan, Nanhui, Fengxian, Chongming, Qingpu and Chuansha. After comparing the total annual precipitation data of various monitoring stations from 1970 to 1999, the gap between the monitoring data of each station over the years is relatively small (Figure S1). Therefore, the precipitation monitoring data of Longhua Station located in the central urban area of Shanghai is selected as the precipitation data for this study. The precipitation monitoring data of Longhua Station are sourced from the Shanghai Central Meteorological Station.

The average annual runoff data of Shanghai were derived from the study of Wang [30]. The methodology of the runoff calculation is the runoff coefficient method. The runoff depth was obtained by multiplying the measured average runoff coefficient in Shanghai by rainfall. The runoff depth was multiplied by the area to obtain the runoff.

2.3. Methods

2.3.1. L-THIA Model

The modelling results of L-THIA can be annual, monthly and daily runoff; the key parameter of L-THIA is SCS-CN, a runoff curve number proposed by the US Soil Conservation Agency. The model uses long-term rainfall, land use and soil data to calculate the average surface runoff of different land use types by daily rainfall. It comprehensively reflects the influence of different soil and land use/land cover conditions on surface runoff. SCS-CN is based on experience rather than on the modelling of the whole physical process of runoff. Its calculation of runoff is derived from the experience of a large number of observations in the United States, which means

$$Q=\frac{{\left(P-I_{\mathrm{a}}\right)}^2}{P-I_{\mathrm{a}}+S} I_{\mathrm{a}}<P$$

$$Q=0 I_{\mathrm{a}}\ge P$$

where Q is the actual surface runoff; P is basin rainfall; I_a is the initial rainfall loss value (the initial rainfall loss includes canopy interception flow, soil surface water storage and evaporation); and S is the maximum possible retention.

The L-THIA model assumes that direct runoff is calculated when rainfall exceeds 20% of recharge capacity. Since the total rainfall is equal to the actual runoff, the sum of the initial loss and the actual retention of the basin, the runoff calculation formula can be transformed into

$$Q=\frac{{\left(P-0.2S\right)}^2}{\left(P+0.8S\right)}$$

where S is a spatial variable, which is related to spatial factors such as soil, land use type and slope. CN (curve number) is a comprehensive dimensionless parameter that can reflect the characteristics of watershed before rainfall events. It has a nonlinear relationship with rainfall runoff. S is derived from CN, which means

$$S=\frac{\mathrm{25,400}}{\mathrm{CN}}-254$$

Total rainfall runoff can be obtained by calculating the sum of runoff in each sub-basin. It can be seen from the above two equations that L-THIA mainly uses the long-term daily rainfall sequence, soil and land use data to calculate the runoff.

2.3.2. Classification of Soil Types

The soil in the study includes paddy soil, coastal saline soil and fluvo-aquic soil, six sub-categories such as submersible paddy soil and de-submersible paddy soil and 22 soil types such as green clay and cyan (Figure 1B). According to the permeability of soils, the US Soil Conservation Agency classified more than 8500 hydrological soil types into A, B, C and D, named the hydrological soil group (HSG), and thus determined their CN values. The permeability of soil types A/B/C/D is in order from strong to weak. In this study, the soil types were classified in Table S1 in the supporting materials. According to the above classification system, the CN types of soil were dominated by B and C, followed by D, and there were few type A in Shanghai. (Figure 1C).

2.3.3. Two Scenarios of Landscape Classification

In the U. S. Department of Agriculture’s National Hydrological Engineering Manual, CN values for different land uses corresponding to different hydrological soil groups in the United States were provided. According to the fact of a metropolitan area, the land use types in Shanghai were divided into 9 first-level types and 32 second-level types (Table S2). Due to the soil properties in different regions, the CN values corresponding to land use types can be varied. Based on the CN values provided in the manual, this study referred to the CN values determined in the relevant studies in Shanghai to determine the CN values [31].

Scenario I: land uses and their CN values

Based on the results of the secondary classification, the land uses in Shanghai were re-divided into 17 categories according to the land use types provided in L-THIA. According to the CN value manual and relevant research in Shanghai [31], the CN value of each land use was determined, and the formula of urban CN value was used to correct the CN value of land use types in urban areas:

$$C{N}_{\mathrm{c}}=C{N}_{\mathrm{p}}+P_{\mathrm{imp}}\left(98-C{N}_{\mathrm{p}}\right)$$

where CN_c is the CN value of comprehensive runoff; CN_p is the CN value of impervious surface percentage of 0; P_imp is the percentage of impervious surface―that is, the impervious surface coefficient corresponding to each land use type. The CN values corresponding to each land use type are shown in Table S3.

Scenario II: impervious surfaces and their CN values

The extraction of impervious surface was based on the permeability of various land uses. In the nine first-level land use types, six types were regarded as impervious surfaces, including industrial, traffic, public building, residential, municipal facilities and other land use types. The other three types, namely green spaces, agricultural land and water, were regarded as pervious surfaces because of the lack of impervious materials on the surface. In this scenario, it was divided into four categories: impervious surface, agricultural land, water area and green spaces.

According to the CN value manual and related research [31], the CN value of each land use type in Shanghai was determined, as shown in Table S3.

3. Results and Analysis

3.1. CN Map under Two Landscape Classification Scenarios

In 1965, the percentage of impervious surface in Shanghai was only 6.20%, which was mostly distributed in the downtown area, while in 2010, it reached 39.45%, expanding rapidly to a larger area surrounding the city center (Figure 2); green space continued to grow by 44.65 times, agricultural land continued to decrease by 43.47%, and water area continued to decrease from 1965 to 2010.

The distribution of CN values under two landscape scenarios was obtained using the hydrological soil and land use data (Figure 2). The deeper the color was, the greater the CN value was, indicating worse permeability of the surface. The CN map of two landscape scenarios both showed that the underlying surface with poor permeability spread from city center to the rural area from 1965 to 2010 and was scattered throughout the study area. Comparing the CN map of two scenarios, it was found that the area with higher CN value under impervious surface scenario is a little smaller than that of the land use one.

3.2. Model Validity and CN Adjustment

In this study, the L-THIA model was used to simulate the average annual runoff under two landscape scenarios: impervious/pervious surfaces and land uses. The average annual runoff of the impervious surface and land use scenario was 1.18 and 1.29 billion m³. The study area does not contain Chongming Island. According to the proportion of the study area to the whole city of Shanghai, the annual average runoff of the two landscape classification systems in Shanghai was estimated to be 1.44 and 1.58 m³, respectively. However, the average annual observed runoff in Shanghai was 2.42 billion m³ [30], and the relative errors were 40.51% and 34.66%. The relative error was so large, indicating that the CN value based on the National Hydrological Engineering Manual and the literature of Shanghai [31] would have resulted in much smaller runoffs.

Therefore, it was necessary to adjust and correct the CN values in L-THIA. In this study, according to the observed runoff data of Shanghai, the appropriated CN value was determined by feedback adjustment under the impervious surface scenario. As shown in Table S4, it showed that the CN values of agricultural land were corrected greatly.

After adjusting the CN values, under the impervious surface scenario, the average annual runoff was 1.93, 2.19, 2.30 and 2.38 billion m³ for the years 1965, 2000, 2006 and 2010, respectively. The total average in 4 periods was 2.20 billion m³. Compared with the observed 2.42 billion m³ in Shanghai, the relative error was 8.90%. Within the allowable range of modelling error, it showed that the L-THIA model was valid for the impact analysis of urbanization effect on hydrological process in Shanghai by adjusting the CN values.

3.3. Finer Landscape Classification May Not Be Necessary for Modelling

3.3.1. Comparison of Annual Average Runoff in Typical Years

The adjusted CN value and daily precipitation data from 1961 to 2012 were input into the L-THIA model as constant values, and only the urbanization situation was changed; that is, the land use data from 1965, 2000, 2006 and 2010 were input. The results of simulated multi-year average runoff based on impervious surface and land use scenarios were shown in Figure 3A. Under the impervious surface scenario, the average annual runoff was 1.01, 1.13, 1.25 and 1.33 billion m³ for the years 1965, 2000, 2006 and 2010, respectively, and the average in four years was 1.18 billion m³, while under the land use scenario, the average runoff was 1.03, 1.27, 1.43 and 1.45 billion m³, and the total average runoff was 1.29 billion m³.

The runoff under the land use scenario was greater than that under the impervious surface scenario. The CN value is the key to affecting the simulation results. The larger the CN value is, the more runoff the land generates. For the same place, the classification of impervious surface scenario and land use scenario is different. This leads to differences in CN values. The difference is the reason for the difference in runoff between the two scenarios.

In both landscape scenarios, with the urbanization, runoff in Shanghai showed a clear increasing trend. In the four typical years, the relative gap of simulated runoff between the two scenarios was about 10%.

3.3.2. Change of Year-by-Year Runoff under Different Urbanization Level

In order to further verify whether the runoff simulation results of the two landscape scenarios can still maintain a small error under the change of precipitation conditions, the annual runoff was simulated by using the landscape conditions of the same four years, respectively. The annual runoff under impervious surface and land use scenario during 1961–2012 was compared in Figure 3B. Under the landscape condition in 1965, the modelling results of the two scenarios were the closest, followed by 2010, 2000 and 2006, which was consistent with the above analysis results. The modelling results of annual runoff of the two scenarios showed the largest gap in 2003, which were about 20% under all landscape conditions, respectively, while the year with the smallest disparity occurred in 1977, which were all less than 7%.

According to the hydrological and meteorological data of Shanghai, 1977 and 2003 were typical wet and dry years. The above analysis showed that the modelling results gap between impervious/pervious surface and land use scenario under larger precipitation was smaller than that under smaller one.

3.3.3. Impervious/Pervious Classification Is Simplified and More Effective

The difference between the results of annual average runoff under two landscape scenarios was only 8.96%. The modelling results of annual runoff were tested by K-S test, which showed that there was a normal distribution. Pearson correlation analysis was performed on the average annual runoff of the two scenarios under the land use conditions in 1965, 2000, 2006 and 2010, and the correlation coefficients were both 0.999, indicating that the simulation results under impervious/pervious surface and land use scenarios were almost the same (Figure 4A). That is, the simpler landscape classification system based on impervious/pervious surfaces can replace the traditional land use system. On this basis, L-THIA model can be used to study the hydrological effect of impervious surface expansion in rapid urbanization areas more easily, because there is no need for complicated and labor-intensive work to interpret enormous land use patches from remote sensing images.

3.4. Quantitative Relationship between Impervious Surface and Runoff

3.4.1. Linear Relationship Displayed between Percent Impervious Surfaces and Runoffs

The relationship between the percentage of impervious surface and average runoff in 1965, 2000, 2006 and 2010 was analyzed, and the results showed that there was a linear relationship between runoff and percentage of impervious surface in Shanghai (Figure 5A). A quantitative analysis of the annual runoff changes for each period, 1965–2000, 2000–2006 and 2006–2010 annual runoff growth rates were 0.37%, 2.13% and 1.89%, which means that the growth rate of annual runoff increased and then decreased. It showed that the average annual growth rate of runoff was consistent with that of impervious surface (Figure 5A). During the period, whenever the impervious surface increased by 1%, the runoff increased by 0.94%.

All the sub-catchments with different urbanization levels were also simulated in the study period. Changes of impervious surface percentage in each sub-catchment are shown in Figure S2A. The results of simulated multi-year average runoff under the land use condition of 1965, 2000, 2006 and 2010 are shown in Figure S2B. Based on the modelling results, the relationship between the percentage of impervious surface and the runoff in each sub-catchment was analyzed. The results showed that there was a strong linear relationship between the growth rate of runoff and that of impervious surface in nine sub-catchments, including Dianbei, Diannan, Jiabaobei, Pudong, Punandong, Punanxi, Qingsonga, Citycenter and Wennan (Figure 5B), indicating that the rapid urbanization process in each sub-catchment leads to increased runoff. The fitting results of each sub-catchment show similar slopes, indicating that the impact of impervious surface on runoff is similar among different sub-catchments.

The correlation test p value in Shangta, Taibei and Tainan sub-catchments was higher than 0.05, and there was no linear relationship between impervious surface and runoff in the area. These three sub-catchments were all smaller sub-catchments. Since the L-THIA model is usually applied to large-scale urban areas, the accuracy of its application in smaller areas may need to be further studied.

3.4.2. Rainfall Conditions Influence the Relationship between Impervious Surface and Runoff

According to the previous research [32], the hydrological effects with same land use change may also be varied under different precipitations. The average annual precipitation in Shanghai was 1096.4 mm. In this study, 1978, 2008 and 1977 were selected as three typical years of dry, normal and wet conditions, with an annual precipitation of 667.1 mm, 1086.5 mm and 1728.7 mm, respectively. Under different rainfall conditions in these typical years, the impact of urbanization under different rainfall conditions was discussed (Figure 6A).

Based on the runoff in 1965, the growth rates of runoff in 2000, 2006 and 2010 were calculated. The results showed that the heavier the precipitation was (1977), the smaller the change in runoff with the increase in impervious percentage was; the smaller the precipitation was (1978), the greater the runoff changes with the increase in impervious surface were. In the dry year, the runoff changed most with the increase in impervious surface, followed by the normal year, but the wet year showed the smallest change.

Based on the analysis of monthly rainfall in Shanghai (Figure S3), May to September was selected as wet season, and January to April and October to December were selected as dry seasons. The runoff changes in dry season and wet season in different years were shown in Figure 6B. With the expansion of impervious surface, the runoff in dry season and wet season increased, but the growth rate of runoff in wet season was less than that in dry season. It also showed that, under the same urbanization conditions in Shanghai, the runoff in dry season was more affected by impervious surface than that in wet season.

4. Discussions

4.1. Comparative Analysis of Similar Studies

This study concluded that runoff increased with the rapid expansion of impervious surface, which is consistent with some similar studies based on various modelling methods. A study in the Beijing–Tianjin–Hebei region showed that surface runoff increased by 11.83% as impervious surface increased by 13.7% from 1980 to 2015 [16], and it was very close to the results of this study. Zare [14] also showed that the growth of residential areas, the only impervious surface in their classification, contributed the most runoff on the Kasilian watershed in the north of Iran. They set a variety of future scenarios for simulation, and the results showed that the increased residential areas scenario increase the runoff by 36.6%, while the afforestation scenario decrease the runoff by 7.8%.

In different studies, although the relationship between runoff and impervious surface was linear, the coefficient of increase in runoff with impervious surface was not a constant. In this study, for every 1% increase in impervious surface, rainfall runoff increased by 0.94%. In some other studies, the value was 0.85, 0.92 [16,33]. It should be stressed that these studies were not limited to only using the L-THIA model, but also the SCS-CN model, RHESys and so on [33,34]. It all indicates that there is a linear relationship, but the correlation coefficients varied.

On the impact of rainfall on impervious surface and runoff, the results of this study were also consistent with the findings of Guzha [35] in the East African region, the impact of impervious surface on runoff in the wet season was smaller. We deduced that the soil water content in the dry/wet season is different, resulting in different permeability of the pervious surface to rainwater in different seasons. In the dry season, the pervious surface can infiltrate more rainwater, while the soil permeability is poor due to the more saturated soil moisture in the rainy season. As a result, the impact of impervious surfaces is expanded in the dry season.

4.2. Modelling Errors

After analysis, the possible reasons for the relatively lower runoffs were as follows:

The initial loss value I_a was too large. At the beginning of a precipitation event, the rain falling to the surface does not turn to runoff immediately but is intercepted by plants and infiltrated into soils. This part of rainfall was the initial loss I_a, which was all the rainfall loss before surface runoff began. The initial loss was affected by many factors, such as soil infiltration capacity, surface land cover, land use and soil water holding capacity. The soil water content, soil flow and underground runoff were the main losses. The initial loss I_a = 0.2S in the L-THIA model was based on the situation in United States but not in China or any other countries. It is argued that the initial loss value I_a = 0.2S may not be applicable in other regions [36,37].

The annual precipitation in the United States was averagely allocated in each month, and more than 70% of precipitation infiltrated into the soil. However, the rainfall in Shanghai has large seasonal and interannual variations, and storms usually happened in wet season, with an infiltration percentage of about 40% [38]. It can be seen that, for the model I_a = 0.2S, the estimated initial loss was too large, resulting in a lower run-off than the actual situation.

On the other hand, the CN values of agricultural land, which covered the majority of Shanghai, were quite difficult to determine. The percentages of agricultural land in Shanghai in 1965, 2000, 2006 and 2010 were 74.67%, 60.06%, 48.23% and 40.54%, respectively. Therefore, the adjustment of CN values for agricultural land has a great influence on the modelling results. It can be seen from Table S4 that the adjustment of CN value of agricultural land was the largest. The adjusted CN value was larger than that in other regions in China [18]. The main reason was that some agricultural land in the suburban area was the farming mode of water and drought rotation, and the CN value of paddy and dry farmland was quite different [39]. Taking the most important crop rice as an example, water is needed at the tillering stage and the heading stage; at this time, the CN value is large. At the stage of yellow maturity, the field needs to be kept dry; then, the CN value turns smaller.

In this study, the paddy and dry farmland were classified as agricultural land in the land use of Shanghai. Therefore, it was necessary but difficult to adjust the CN value of agricultural land, and it was often difficult to obtain reasonable modelling results. It can be seen that the lower CN values for agricultural lands will lead to lower runoff than the actual situation.

4.3. Adjustment of CN Value Does Not Mean Too Much

After adjusting the CN value of impervious surfaces, the K-S test showed that the annual runoff modelling results were normally distributed. Pearson correlation analysis was carried out on the simulated results under impervious/pervious scenario using precipitation data from 1961 to 2012 before and after the adjustment of CN value using land use conditions in 1965, 2000, 2006 and 2010. The correlation coefficients in different years were all greater than 0.9, which showed that there was a strong correlation between the two scenarios (Figure 4B).

The CN values of different land uses determined in the U. S. Department of Agriculture Hydrology National Engineering Manual and studies in Shanghai will bring a lower result. However, if the results were only used to analyze the growth rate of surface runoff with the expansion of impervious surfaces, it can be held that the CN value was not necessary to be adjusted in L-THIA modelling.

5. Conclusions

The results of the L-THIA model based on impervious/pervious surface and land use type showed a good consistency; the difference of total annual runoff of the two scenarios was only 8.96% in 4 typical years, and there was also a very significant linear correlation of the modelling annual runoff between the two scenarios. It can be concluded that the impervious/pervious surface, a simpler landscape scenario, can significantly improve the modelling efficiency considering the scale of interpretation of land use data from remote sensing images.

The L-THIA model was modified by adjusting the CN value under the impervious/pervious surface scenario. The relative error of the modelling results after adjustment was only 8.90% compared with the observed runoff. There was also a quite significant linear correlation between the annual runoff before and after CN value adjustment. It was suggested that the CN value adjustment may not be necessary, when the L-THIA model was just applied to analyze the growth rate of surface runoff during rapid urbanization.

The impervious surface in rapid urbanization in Shanghai increased from 6.20% to 39.45%, and the runoff increased by 31.33%. The linear relationship between impervious surface and runoff showed that the rapid urbanization was an important reason for the continuous increase in surface runoff. At the same time, it was found that under the same urbanization conditions, the response of different hydrological years or periods was not the same. The runoff in dry year changed more evidently with the increase in percentage of impervious surface, followed by normal and wet year. Similarly, runoff in dry season could be more affected by the change in impervious surface than that in wet season.