Abstract

To study the impact of land use/cover change (LUCC), the relationship between precipitation and runoff was investigated. Our main objective was to ensure reasonable development, management, and sustainable utilization of water resources at a watershed scale. To investigate the relationship between precipitation and runoff, a SWAT (Soil and Water Assessment Tool) model was developed by analyzing LUCC in Naoli River basin. Then, runoff response was analyzed under different LUCC conditions. The contribution coefficient of different land use types to runoff was calculated. The results of this research study are as follows. From 1986 to 2014, dry land, forest land, paddy fields, and unused land were the major land use types, accounting for more than 93% of the total catchment. On the other hand, grass land, building land, and water bodies accounted for a small proportion. Among the four main land use types, the contribution coefficient of forest land was 3.10 mm·km−2. This indicates that forest land was suitable for runoff generation. The contribution coefficient of dry land, unused land (fluvial wetland in Naoli River basin), and paddy field are −0.11, −0.37, and −0.83 mm·km−2, respectively. This implies that these three land use types were adverse factors for runoff generation.

INTRODUCTION

To understand global climate change and hydrology, it is essential to investigate watershed hydrological processes associated with land use and cover change (LUCC) (Jain et al. 2010). Climate change is one of the main factors that affect watershed hydrologic processes. With population explosion and economic development, LUCC has significantly affected watershed hydrological processes in a short period of time (Li et al. 2010). In recent years, LUCC has dramatically changed due to global climate change (Munroe & Müller 2007). Consequently, surface runoff also changed greatly (Jia et al. 2009). Therefore, it is essential to investigate the impact of LUCC on surface runoff (Nosetto et al. 2005).

Over a long period of time, many researchers have comprehensively investigated LUCC. In a farming-pastoral zone of northern China, Hao & Ren (2009) evaluated eco-environmental responses to LUCC by using the following technologies: remote sensing (RS), global positioning system, geographical information system (GIS), and other statistical methods. Song et al. (2010) analyzed LUCC and landscape pattern variation of wetlands in warm, rainy regions of southern China for over two decades. They used remote sensing data of 1987, 1991, 2006, and 2009. From 1977 to 2007, Yu et al. (2011) simulated LUCC on the conditions of economy, population, and policy in Daqing, China. They evaluated the future of LUCC under three different scenarios of economic growth rate. By considering different patterns of socio-economic development, Schirpke et al. (2012) performed spatial allocation procedures and generated land use/cover scenario maps. Lópezcarr et al. (2012) developed spatial and multi-level statistical models, and Guatemala was chosen as the study case.

The impact of LUCC on surface runoff has been related to land use types. Leitinger et al. (2010) analyzed seasonal dynamics of surface runoff in mountain grassland ecosystems, and they discussed the impact of LUCC on surface runoff of a watershed. Using Soil Conservation Service Curve Number Method, Yao et al. (2015) evaluated potential reductions in surface runoff. These reductions were associated with urban green space in Beijing, China. Jia et al. (2009) analyzed the relationship between forest and runoff with the Mann–Kendall method. They detected a trend of precipitation and temperature. The method of paired catchments was used to calculate the impact of LUCC on surface runoff. Research results indicate that forest can reduce runoff. Samaniego & Bárdossy (2006) built various non-linear integrated models related to several runoff characteristics, which were associated with physiography, land cover, and meteorological factors instead of a standard hydrological model. A sequential Monte Carlo method was used to simulate the impact of land use/cover and climatic change on runoff characteristics in four different scenarios. To simulate the effect of changing land use pattern on environmental flow, Lin et al. (2014) combined the Xinanjiang model with Soil Conservation Service Curve Number.

The aforementioned studies profoundly analyzed LUCC and its impact on surface runoff; however, there are obvious limitations in typical regions where the relationship between precipitation and runoff is complex. This is because watershed hydrology is a complex process, and it is influenced by precipitation. Moreover, it is closely related to LUCC. The relationship between precipitation and runoff is particularly complex, with dramatic changes in land use. In typical areas, it is essential to investigate the impact of land use on the relationship between precipitation and runoff.

The impact of LUCC on the hydrologic process is a controversial issue due to the following reasons. One is that LUCC changes directly with changes in ecological spatial distribution pattern. Owing to changes in the area of different land use types, the hydrologic process changes and leads to runoff change. The other reason is that the hydrologic process in one basin is different from that in another basin; therefore, the impact of LUCC has to be analyzed to understand the relationship between precipitation and runoff in typical areas. To understand the complex relationship between precipitation and runoff, the impact of LUCC on surface runoff was analyzed. Moreover, the contribution of different land use types to surface runoff changes was estimated. The Soil and Water Assessment Tool (SWAT) model was developed by analyzing LUCC. The surface runoff of Naoli River basin was simulated under different land use situations. A multi-objective decision-making approach was used to determine the contribution rate of different land use types and to estimate surface runoff changes.

METHODS

Study area

As illustrated in Figure 1, the Naoli River basin has a catchment area of roughly 20,577 km2, and it is situated in the mid-eastern Sanjiang Plain of China. This region has sub-humid warm temperate continental monsoon climate. The Naoli River basin has a latitude of 32° and 42° N and longitude of 96° and 119° E. The Naoli River basin, which is one of the biggest tributaries, has a length of 596 km. It originates in a southwestern mountainous area, and it flows northeast into Wusuli River. The annual mean precipitation in Naoli River basin is about 580 mm, and the annual mean evaporation rate is 760 mm. The mean annual temperature is in the range 1–3°C. The average wind speed is 4.25 m/s. In winter, the wind blows in a west and northwest direction. In spring, the wind blows in a southwest direction. In summer, the wind blows in an east and southeast direction.

Figure 1

Location of Naoli River basin.

Figure 1

Location of Naoli River basin.

The Naoli River basin is an important region that produces grain in China. Moreover, it is a national nature reserve for wetland. Water resources are essential for regional economic development and environmental protection. In the last three decades, the runoff in Naoli River basin has changed significantly due to climate change and LUCC. Therefore, it is essential to analyze the impact of LUCC on runoff in typical areas of Sanjing Plain in China. In this study, the Naoli River basin was considered as the case study area.

Data source

LUCC data were gathered with Landsat/thematic mapper (TM) for the following years: 1986, 1996, 2000, 2005, and 2014. LUCC data were received with a resolution of 30 m from China Remote Sensing Satellite Ground Station. To meet application requirements, data were collected from June to September. During this period, the weather was either slightly cloudy or cloudless, so it met application requirements. As shown in Figure 2, LUCC was divided into seven types: paddy field, dry land, forest land, grassland, waters, building land, and unused land. For this classification, geographical features of the study area, image quality, and research objectives were considered.

Figure 2

Land use patterns of the study area in different years.

Figure 2

Land use patterns of the study area in different years.

The soil data (Figure 3) include soil spatial distribution and soil physical properties. The data of soil spatial distribution were collected from Heilongjiang Province in China. In the soil type map, soil spatial distribution ratio was 1:500,000. To gather data of soil physical properties, 22 soil species were considered. Soil texture data were obtained from the Chinese Soil Database (Institute of Soil Science, Chinese Academy of Sciences). International System of Soil Texture Classification was used to analyze soil texture data. Then, it was transformed to the American Soil Texture Classification System, which was appropriate for analysis by SWAT. Cubic spline interpolation of soil texture data was done by MATLAB software (Matrix Laboratory, Math Works Company, USA). The other data were directly checked or calculated by Soil-Plant-Air-Water system, which is a soil water characteristic developed by USDA (United States Department of Agriculture).

Figure 3

Soil type map of the study area.

Figure 3

Soil type map of the study area.

Figure 4 illustrates DEM (digital elevation model) data, which were obtained from geospatial data cloud (http://www.gscloud.cn/search). The resolution is 30 m.

Figure 4

DEM of the study area.

Figure 4

DEM of the study area.

Meteorological data (Figure 5) included precipitation and runoff, which were obtained from 1956 to 2010. These data were obtained from Baoqing and Caizuizi hydrological stations, which are located in the upper and lower sections of Naoli River basin, respectively. At Baoqing hydrological station, the relationship between precipitation and runoff was well established. At Caizuizi hydrological station, the relationship between precipitation and runoff was not well established; therefore, the data from Caizuizi hydrological station were used to analyze the impact of LUCC on runoff.

Figure 5

The relationship between precipitation and runoff at Baoqing and Caizuizi stations.

Figure 5

The relationship between precipitation and runoff at Baoqing and Caizuizi stations.

Research method

CA-Markov model

The CA-Markov (Cellular Automation Markov) model perfectly combines GIS and image processing function, and one of the most distinctive functional modules is land cover change simulation. The CA-Markov model can use multi-criteria evaluation and multi-objective decision support system to define the rules of transfer between types of land use. On the basis of accurate prediction of future land use quantity structure, the CA-Markov model can effectively predict the quantity and spatial distribution of land use structure by the ability of simulating spatial pattern. The working principle of CA-Markov is that the base period of land use is initial state, and it can redistribute land use type on the basis of land use transfer area and the distribution pattern of different land use types of the base period. Practice has proven that the CA-Markov model has a greater precision advantage in land use change simulation than GIS and a number of successful applications have been made.

The advantages of the Markov model and cellular automation (CA) model are presented in the CA-Markov model. The Markov model was used for long-term precipitation prediction, while the CA model was used to simulate changes in complex space–time systems. As a result, the CA-Markov model simulated changing situations in the quantity and structure of LUCC.

The CA model was composed of a cellular space and transfer functions, which were defined in space. The CA model included four parts: cellular, cellular space, neighbor, and regulation. In this paper, CA was defined as grid cell with a scale of 30 m × 30 m, while cellular space was the Naoli River basin. Every cellular space has seven possible states with seven land use types. The cellular state was related to initial conditions and transfer conditions. Transfer conditions were elements that affected all kinds of land use type transformation.

SWAT model

Soil and Water Assessment Tool is a distributed hydrological model which is suitable for large watershed scale and was developed by the Agricultural Research Center of the United States Department of Agriculture (USDA), a task which took more than 30 years. On the base of SWRRB (Simulator for Water Source in Rural Basins) model, the SWAT model combined a few hydrology models and is a physically based distributed hydrological model that can be simulated continuously. The SWAT model is biased towards hydrological simulation and its time step is day. It can simulate the impact of different land use types and various agricultural management measures on water, sand, and chemicals, and can also forecast total runoff volume, sediment loss, and nutrient load for 100 years.

The SWAT model simulates runoff generation process, which is based on water cycle and water balance. This simulation is divided into two components: land surface component (runoff generation and slope flow) and water surface component (concentration of channel). The former component is known as slope hydrological process. In the main channel of sub-basins, it controls input accounts of water, sand, nutrients, chemicals, etc. Every sub-basin module is divided into eight management components: hydrology, meteorology, sediment, soil temperature, crop growth, nutrients, pesticides, and agriculture. The latter module is known as hydrologic processes of channel. In the river net, it is primarily associated with the transport of water, sand, pesticides, nitrogen, phosphorus, and other nutrients. Channel hydrological processes, such as reservoir, pond, and wetland, include two parts, namely, channel flow calculus and water body flow calculus (Kannan et al. 2007). Equation (1) presents the water balance equation of the SWAT model:  
formula
(1)
Here, SWt is the final water content of soil (SW0 is the initial soil moisture content in day i); i is the time; Rday is the amount of precipitation in day i; Qsurf is the surface runoff in day i; Ea is the surface evaporation in day i; Wdeep is the amount of water entering the aquifer in day i; and Qgw is the back flow in day i. All parameters in Equation (1) are expressed in mm. The SWAT model was built as follows: first, DEM of actual drainage and the basin were divided into sub-basins; second, land use data were reclassified, and soil spatial distribution data were divided into hydrological response units (HRUs). Land use data consisted of interpreted results of 2014. In this experiment, HRUs have the same land use, soil type, and grade level, and they are the basic arithmetic unit in the SWAT model. There are 31 sub-basins and 782 HRUs in Naoli River basin. The calculation of the SWAT model was first based on HRUs. Then, the sub-basin was summarized. Finally, runoff was gathered as the total export of Naoli River basin.

Multi-objective decision function method

Multi-objective decision function method was used to calculate the contribution coefficient of different land use types to runoff.

Rx is the average annual runoff depth; Va, Vb, Vc, and Vd …… are runoff depths affected by dry land (Dx), forest land (Fx), paddy field (Px), and unused land (Ux) in the corresponding unit area, respectively. The unit of measurement is mm·km−2; x are different land use scenarios for the following years: 1986, 1996, 2000, 2005, and 2014. Equation (2) was used to calculate the average annual runoff depth (Rx).  
formula
(2)
A matrix was built by putting runoff depth of different land use scenarios in the year n, m, p, and q. Here, n, m, p, and q are same as x. In this paper, they correspond with the following years: 1986, 1996, 2000, 2005, and 2014. These parameters are substituted into the following formula.  
formula
(3)
Va, Vb, Vc, and Vd …… were obtained by calculating the above matrix, and they were the contribution coefficient of different land use types to runoff.

RESULTS AND DISCUSSION

Analysis of land use change

Dynamic change characteristics of LUCC

The land use classification system was built on the following components: land use status classification, which was published in the year 2007 and the actual situation of land use. The Naoli River basin includes seven land use types: paddy field, dry land, forest land, grassland, water, building land, and unused land. Dry land, forest land, and paddy field have the biggest area, so they are important for runoff. Since 1986, change characteristics of the Naoli River basin have been as follows.

As shown in Table 1, dry land accounts for 37.38% of the total area, and it accounts for the largest proportion of area in 2014. Forest land accounts for 29.54% of total area, and it accounts for the second largest proportion of area in 2014. Finally, paddy field accounts for the third largest proportion of area in 2014. Water bodies, such as lakes, have the smallest percentage of area in 2014. They account for only 0.33% of the total area. The relationship between land use and runoff is analyzed by emphasis on paddy field, dry land, and forest land, especially paddy field. Paddy fields have changed immensely from 1986 to 2014. From 1986 to 1996, the proportion of paddy fields has decreased by 77.24%. From 1996 to 2000, the proportion of paddy field increased by 565.70%. This is a critical period for land use change due to rapid increases in paddy field. Dry land was the biggest proportion in Naoli River basin. From the year 1986 to 2014, there were fluctuating changes in the proportion of dry land. In general, they show a declining trend. Forest land shows a slightly declining trend, which means it changes a little and has no effect on runoff changes from the year 1986 to 2014. In general, paddy fields played the most important role in runoff changes in Naoli River basin because of the following reasons: dry land changes rapidly, whereas forest land changes slowly. The proportions of other types of land use are not significant.

Table 1

The change characteristics of Naoli River basin since 1986

Types of land usePaddy fieldDry landForest landGrasslandWaterBuilding landUnused land
Proportion of 2014 9.05 37.38 29.54 3.50 0.33 1.74 18.46 
Pi 1986–1996 −77.24 30.25 −1.60 −14.73 −23.23 5.37 −18.13 
1996–2000 565.70 −16.87 −7.61 −17.94 43.78 −4.05 −4.83 
2000–2005 −35.47 17.84 0.99 108.33 62.21 −8.24 −37.36 
2005–2014 64.27 −10.91 −0.85 −8.77 10.58 9.12 −0.19 
Types of land usePaddy fieldDry landForest landGrasslandWaterBuilding landUnused land
Proportion of 2014 9.05 37.38 29.54 3.50 0.33 1.74 18.46 
Pi 1986–1996 −77.24 30.25 −1.60 −14.73 −23.23 5.37 −18.13 
1996–2000 565.70 −16.87 −7.61 −17.94 43.78 −4.05 −4.83 
2000–2005 −35.47 17.84 0.99 108.33 62.21 −8.24 −37.36 
2005–2014 64.27 −10.91 −0.85 −8.77 10.58 9.12 −0.19 

The transfer matrix of land use

Table 2 presents a transfer matrix of land use in the Naoli River basin since 1986. From the year 1986 to 1996, 1,699.59 km2 is the area of dry land in paddy field, and unused land is 12.29 km2. During this period, the area of paddy fields decreased rapidly. The area transformed from dry land to paddy field was 244.25 km2, while the area transformed from dry land to forest land was 83.08 km2. Unused land was 82.54 km2. the area transformed from forest land to dry land was 238.15 km2. During this period, it was found that the proportion of area under paddy field decreased rapidly, while the proportion of area under dry land increased significantly. Furthermore, the proportion of area under forest land increased slowly. From the year 1996 to 2000, an area of 32.64 km2 was transformed from paddy field to dry land, while 2,346.18 km2 was transformed from dry land to paddy field and 421.88 km2 was transformed from forest land to dry land. During this period, the proportion of area under paddy field increased sharply while the proportion of area under dry land and forest land decreased slowly. From 2000 to 2005, the areas transformed from paddy field to dry land and forest land were 1,620.77 km2 and 263.35 km2, respectively. During this period, the areas transformed from dry land to paddy field, forest land, and grassland were 821.32 km2, 318.45 km2, and 302.91 km2, respectively. The area transformed from forest land to dry land was 378.19 km2. During this period, it was observed that the proportion of area under paddy field decreased slowly while the proportion of area under dry land and forest land increased slightly. From the year 2005 to 2014, the area transformed from paddy field to dry land was 124.95 km2, while the area transformed from dry land to paddy field was 1,214.61 km2. From the year 2005 to 2014, the area transformed from forest land to dry land was 123.29 km2. Thus, the proportion of area under paddy field increased significantly during this period. At the same time, the proportion of area under dry land and forest land decreased slowly.

Table 2

The transfer matrix of land use since 1986

Paddy fieldDry landForest landGrasslandWatersBuilding landUnused land
1986–1996 Paddy field 135.79 1,699.59 0.27 27.26 1.59 1.39 12.29 
Dry land 244.25 7,278.77 83.08 50.49 0.72 18.48 82.54 
Forest land 0.09 238.15 5,851.91 34.78 0.00 1.75 3.92 
Grassland 1.00 165.27 70.18 453.83 2.26 0.69 32.99 
Waters 0.00 0.25 0.00 1.29 46.61 0.00 21.25 
Building land 0.09 3.14 0.25 0.38 0.00 358.16 0.00 
Unused land 46.30 720.11 26.96 51.20 2.09 0.96 2,983.49 
1996–2000 Paddy field 384.15 32.64 0.09 0.00 0.00 0.09 10.54 
Dry land 2,346.18 7,673.22 4.89 40.89 0.25 3.14 36.71 
Forest land 0.27 421.88 5,562.35 22.02 0.00 0.25 25.88 
Grassland 87.78 81.59 4.93 444.55 0.00 0.38 0.00 
Waters 0.00 2.17 0.00 0.00 50.78 0.00 0.32 
Building land 1.46 15.52 0.75 0.69 0.00 362.13 0.88 
Unused land 26.09 173.72 0.37 0.00 25.56 0.00 2,910.73 
2000–2005 Paddy field 870.09 1,620.77 22.03 51.45 3.99 14.26 263.35 
Dry land 821.32 6,743.06 318.45 302.91 16.91 78.40 119.69 
Forest land 16.40 378.19 5,076.83 45.57 1.94 6.27 48.19 
Grassland 30.52 110.91 150.66 30.83 29.63 1.16 154.45 
Waters 2.60 5.35 1.91 0.64 39.49 0.17 26.44 
Building land 12.24 106.73 6.24 3.03 0.37 235.20 2.16 
Unused land 83.21 934.71 52.71 626.76 31.88 0.36 1,255.44 
2005–2014 Paddy field 1,681.27 124.95 5.23 1.13 3.15 5.71 14.95 
Dry land 1,214.61 8,485.23 96.50 13.88 4.07 46.89 38.55 
Forest land 6.06 123.29 5,472.40 19.29 3.44 1.61 2.73 
Grassland 82.84 38.51 3.39 930.66 0.23 0.54 5.01 
Waters 0.04 0.67 0.32 0.26 121.12 0.00 1.80 
Building land 2.61 21.24 0.99 0.02 0.00 310.75 0.21 
Unused land 29.18 26.25 2.18 2.87 5.39 0.93 1,802.90 
Paddy fieldDry landForest landGrasslandWatersBuilding landUnused land
1986–1996 Paddy field 135.79 1,699.59 0.27 27.26 1.59 1.39 12.29 
Dry land 244.25 7,278.77 83.08 50.49 0.72 18.48 82.54 
Forest land 0.09 238.15 5,851.91 34.78 0.00 1.75 3.92 
Grassland 1.00 165.27 70.18 453.83 2.26 0.69 32.99 
Waters 0.00 0.25 0.00 1.29 46.61 0.00 21.25 
Building land 0.09 3.14 0.25 0.38 0.00 358.16 0.00 
Unused land 46.30 720.11 26.96 51.20 2.09 0.96 2,983.49 
1996–2000 Paddy field 384.15 32.64 0.09 0.00 0.00 0.09 10.54 
Dry land 2,346.18 7,673.22 4.89 40.89 0.25 3.14 36.71 
Forest land 0.27 421.88 5,562.35 22.02 0.00 0.25 25.88 
Grassland 87.78 81.59 4.93 444.55 0.00 0.38 0.00 
Waters 0.00 2.17 0.00 0.00 50.78 0.00 0.32 
Building land 1.46 15.52 0.75 0.69 0.00 362.13 0.88 
Unused land 26.09 173.72 0.37 0.00 25.56 0.00 2,910.73 
2000–2005 Paddy field 870.09 1,620.77 22.03 51.45 3.99 14.26 263.35 
Dry land 821.32 6,743.06 318.45 302.91 16.91 78.40 119.69 
Forest land 16.40 378.19 5,076.83 45.57 1.94 6.27 48.19 
Grassland 30.52 110.91 150.66 30.83 29.63 1.16 154.45 
Waters 2.60 5.35 1.91 0.64 39.49 0.17 26.44 
Building land 12.24 106.73 6.24 3.03 0.37 235.20 2.16 
Unused land 83.21 934.71 52.71 626.76 31.88 0.36 1,255.44 
2005–2014 Paddy field 1,681.27 124.95 5.23 1.13 3.15 5.71 14.95 
Dry land 1,214.61 8,485.23 96.50 13.88 4.07 46.89 38.55 
Forest land 6.06 123.29 5,472.40 19.29 3.44 1.61 2.73 
Grassland 82.84 38.51 3.39 930.66 0.23 0.54 5.01 
Waters 0.04 0.67 0.32 0.26 121.12 0.00 1.80 
Building land 2.61 21.24 0.99 0.02 0.00 310.75 0.21 
Unused land 29.18 26.25 2.18 2.87 5.39 0.93 1,802.90 

The Naoli River basin is an integral part of Sanjiang plain, which is a major base of grain production in China. In the next few years agricultural planting structure and land use types will give rise to great changes. According to the state plan, most dry land will be transformed into paddy fields. From the above analysis, we can see dry land, forest, and paddy field are the largest proportion of land use types in proper sequence. In the process of reconstructing the planting structure, it should deserve special mention that forest or other land use types should not be affected in order to avoid the ecological environment of Naoli River basin undergoing great change.

Results of SWAT model

During the operational stage of the SWAT model, many parameters can affect simulation results of runoff. The impact of these parameters is different. For example, parameters with slight impact would increase calibration time. These parameters had slight impact on forecasting results (Li et al. 2010). In order to increase simulation speed of the SWAT model, it was necessary to simplify parameters and to choose parameters sensitive to simulation results. To obtain satisfactory results, it was necessary to have an optimal solution of parameters. This paper used the method from Beven & Binley (1992) to calibrate parameters and to verify SWAT model. The method resolved the problem of fussy parameter calibration and bad convergence rate, which occurred with optimization algorithm SCE (Shuffled Complex Evolution). Thus, a good simulation result was obtained from a series of optimal solution sets. Sensitive parameters and their value ranges, which are simulated by this method in the Naoli River basin, are shown in Table 3.

Table 3

Parameters sensitive to runoff and their value ranges

Name of parameterImplicationMinimum valueMaximum valueCalibrated value
ALPHA_BF Base flow coefficient 0.01 0.07 0.04 
SOL_K(1) Saturation water conductivity in primary soil −0.2 0.8 0.3 
SFTMP Snow temperatures −0.3 0.35 
CN2 The coefficient of CSC runoff curve under soil moisture II −0.07 0.9 0.42 
CH_K2 The effective water conductivity of the channel 5.6 12 8.8 
SOL_AWC(1) Available water capacity in primary soil 0.02 0.4 0.21 
ALPHA_BNK River water return coefficient 0.5 
CH_N2 Manning coefficient of main river 0.08 0.04 
GW_DELAY Groundwater hysteresis coefficient 34 43 38.5 
SOL_BD(1) Wet bulk density in primary soil −0.05 0.06 0.005 
Name of parameterImplicationMinimum valueMaximum valueCalibrated value
ALPHA_BF Base flow coefficient 0.01 0.07 0.04 
SOL_K(1) Saturation water conductivity in primary soil −0.2 0.8 0.3 
SFTMP Snow temperatures −0.3 0.35 
CN2 The coefficient of CSC runoff curve under soil moisture II −0.07 0.9 0.42 
CH_K2 The effective water conductivity of the channel 5.6 12 8.8 
SOL_AWC(1) Available water capacity in primary soil 0.02 0.4 0.21 
ALPHA_BNK River water return coefficient 0.5 
CH_N2 Manning coefficient of main river 0.08 0.04 
GW_DELAY Groundwater hysteresis coefficient 34 43 38.5 
SOL_BD(1) Wet bulk density in primary soil −0.05 0.06 0.005 

Note: The number of CN under the condition that early soil moisture is dry is called CNI, the number of CN under normally condition is CNII, and the number of CN under wet condition is CNIII. In this table CNII is needed.

The evaluation of SWAT model simulation results are as follows: Nash–Sutcliffe efficiency coefficient (NS) and correlation coefficient (R2):  
formula
(4)
 
formula
(5)

Here, is the measured value, and are simulated values; is the average value of the measured value. If NS > 0. 5 and R2 > 0. 6, the simulated result was considered to be acceptable.

Kannan et al. (2007) suggested that the period during which runoff is steady should be chosen to calibrate and verify the SWAT model. During the period 1996–2009, the runoff in Naoli River basin was steady, based on the analyses of the measured runoff data (1956–2010) in the station, which was located in the watershed outlet. Therefore, the data of 1996–2000 were used for model calibration, and the data of 2001–2009 were used for model verification. First, the measured runoff data of 1996–2000 were entered into the SWAT model, and the value range of parameters adjusted according to sensitivity in order to achieve good simulation results. Then, the measured data of 2001–2009 were used to verify the model, which was calibrated by the data of 1996–2000. Figure 6 and Table 4 present the simulation results of calibration period and verification period, respectively.

Table 4

Calibration and verification results of SWAT model

NSGradeR2Grade
Calibration period 0.72 0.76 
Validation period 0.71 0.75 
NSGradeR2Grade
Calibration period 0.72 0.76 
Validation period 0.71 0.75 
Figure 6

Comparison of simulated and observed yearly runoff data during calibration (1996–2000).

Figure 6

Comparison of simulated and observed yearly runoff data during calibration (1996–2000).

As shown in Figure 6, the simulation result of the calibration period was identical with the observed data. This indicates that the simulation result of the model was qualified. Figure 7 shows that the simulation result during the verification period was superior to the result of the calibration period because the annual runoff during verification period was greater than that observed during calibration period.

Figure 7

Comparison of simulated and observed monthly data during calibration and validation.

Figure 7

Comparison of simulated and observed monthly data during calibration and validation.

The simulation results were divided into four grades based on NS and R2: grade A was greater than 0.9; grade B was in the range 0.7–0.9; grade C was in the range 0.5–0.69, and grade D was less than 0.5. Table 4 shows simulation results of the calibration period and verification period. Both results achieved grade B, and NS and R2 were greater than 0.7. This indicates that simulation precision of the SWAT model in Naoli River basin was good, and the model was applicable to the Naoli River basin.

Impact of different land use types on runoff

Runoff simulation of different land use scenarios

Land use data of 1986, 1996, 2000, 2005, and 2014 were entered into the SWAT model, which have been calibrated and simulation results analyzed. Figure 8 shows simulation results of annual runoff depth under different land use scenarios (1986–2010).

Figure 8

Annual runoff depth under different land use scenarios.

Figure 8

Annual runoff depth under different land use scenarios.

Under different land use scenarios, annual runoff depth assumed disciplinary changes with respect to changes in land use pattern. From 1986 to 2010, annual runoff depth under different land use scenarios was 67.63, 73.91, 52.93, 61.48, and 48.25 mm. The change of runoff was in line with the area of paddy fields. During the period 1986–1996, the area of paddy fields decreases from 8.24% to 2.06%. The annual runoff depth increases from 67.63 mm to 73.91 mm, eliciting a change of 9.29%. From 1996 to 2000, the area of paddy fields increased from 8.24% to 12.29%. Moreover, the annual runoff depth decreased from 73.91 mm to 52.93 mm, indicating a change of 39.64%. The same change was observed for the period ranging from 1986 to 1996. During the period ranging from 2000 to 2005, the area of paddy fields reduced from 12.29% to 10.22%. The annual runoff depth increased from 52.93 mm to 61.48 mm, indicating a change of 16.15%. The period 2005–2014 was in line with the period 1996–2000. The area of paddy fields increased from 10.22% to 16.19%, and the annual runoff depth decreased from 61.48 mm to 48.25 mm, indicating a change of 27.42%. It can be seen from the above analysis that the biggest annual runoff depth was in 1996. In fact, the annual runoff depth was bigger than that in the following years: 1986, 2000, 2005, and 2014. It follows that paddy field affects runoff most among all the land use types and it is identical with annual runoff depth.

Analysis of contribution coefficient of different land use types to runoff

Multi-objective decision method was applied to analyze contribution coefficient of different land use types to runoff in Naoli River basin. As pointed out earlier, the four main land use types in Naoli River basin are as follows: dry land, forest land, paddy field, and unused land. The four main land use types accounted for more than 94% of the total area of Naoli River basin. These land use types also underwent apparent changes annually. Grassland, building land, and water body are small, and they constitute a small proportion of Naoli River basin. These components have slight impact on runoff. In this study, we analyzed the contribution coefficients of dry land, forest land, paddy field, and unused land.

Multi-objective decision method was used to calculate the contribution coefficient of different land use types to runoff. The contribution coefficient of dry land, forest land, paddy field, and unused land were −0.11, 3.10, −0.83, and −0.37 mm·km−2, respectively (Table 5).

Table 5

Coefficients of different land use types on runoff depth

Land use typesForest landUnused landPaddy fieldDry land
Contribution coefficient V 3.1 −0.37 −0.83 −0.11 
Land use typesForest landUnused landPaddy fieldDry land
Contribution coefficient V 3.1 −0.37 −0.83 −0.11 

As can be seen from Table 5, the contribution coefficients of four land use types in descending order are as follows: forest land, paddy field, unused land, and dry land. This indicates that in the Naoli River basin, the contribution coefficient of forest land was highest, followed by paddy field, unused land, and dry land.

The contribution coefficient of forest land was 3.10 mm·km−2, which demonstrates that forest land was beneficial to runoff generation in Naoli River basin. Forest land was located in hilly and mountainous areas. The greater the degree of slope, the higher would be the runoff generated under such circumstances. Compared with other land use types, forest land has little reclining edge due to the following reasons: thinner soil layer, low vegetation coverage fraction, small forest density, lower vegetation leaf area index, shallow ground litter, and humus layer. The Naoli River basin has a semi-humid temperate continental monsoon climate. Precipitation normally occurs in summer in this region. In general, the flow increases with greater rainfall. Hence, forest land was beneficial to direct runoff and has small infiltration. From the year 1986 to 2010, there was a steady decline in forest land. In fact, a great deal of forest land was converted into paddy fields and dry land during this period. Consequently, runoff depth declined during this period. By 2014, runoff depth became smaller due to the decrease in forest land.

The contribution coefficient of unused land is −0.37 mm·km−2, which shows that unused land would make runoff small. In Naoli River basin, unused land contained mostly natural reserves. In fact, unused land was fluvial wetland. Compared to other land use types, flood water was stored in unused land. During the rainy season, flood water is retained in unused land. During summer, this water would be released from unused land to supplement surface runoff. From 1986 to 2014, the area of unused land declined drastically. Consequently, the unused land had a miniscule impact on runoff.

The contribution coefficient of paddy field is −0.37 mm·km−2. It plays an important role in changing the process of surface and underground runoff. Irrigation facilities must be provided to ensure steady growth of paddy in fields. Paddy fields consume a great deal of water and increase the infiltration of surface water. Due to ridge intercept, it is not easy to form surface runoff. This leads to a reduction in runoff depth. During the study period, paddy field changes obviously and has a rising trend. Runoff depth is reduced as a consequence.

The contribution coefficient of dry land is −0.11 mm·km−2. Obviously, the changes in dry land are opposite to runoff depth. Runoff depth would decrease with increase in dry land. Moreover, dry land would decrease with increasing runoff. This is because the flat ground of dry land is suitable for infiltration. Infiltration process was also accelerated with surface tillage. All the above factors restricted the transformation from precipitation to runoff. During the study period, dry land increased constantly and led to a decrease in runoff. Consequently, the impact of dry land on runoff increased sharply.

Among the four main land use types, the contribution coefficient of forest land was 3.10 mm·km−2. This indicates that forest land was suitable for runoff generation. The contribution coefficient of dry land, unused land (fluvial wetland in Naoli River basin), and paddy field are −0.11, −0.37, and −0.83 mm·km−2, respectively. This implies that these three land use types were adverse factors for runoff generation. According to the state plan, most of the dry land would be transformed into paddy field. However, the most important thing is to maintain channel runoff and avoid the ecological environment of Naoli River basin undergoing great change during the planting structure and land use type change.

CONCLUSIONS

Naoli River basin, which is located in Sanjiang plain, was considered as the study area. This is because there was a good relationship between precipitation and runoff in its upstream; however, the relationship between precipitation and runoff was not good in its downstream due to LUCC (Figure 5). Hence, the Naoli River basin exhibited good representation. It was used as a typical area to study the impact of LUCC on runoff in a complex environment.

Since the 1980s, Naoli River basin has the following land use types: dry land, forest land, paddy field, and unused land. The SWAT model was used to analyze the impact of different land use types on runoff depth. The results show that forest land can promote runoff generation, and its contribution coefficient is 3.10 mm·km−2. On the other hand, paddy field, unused land, and dry land have an inhibiting effect on runoff. The contribution coefficients of paddy field, unused land, and dry land were −0.83 mm·km−2, −0.37 mm·km−2, and −0.11 mm·km−2, respectively.

Many researchers have investigated the complexities of the relationship between forest land and runoff; however, a unified conclusion has not been reached to date. Li et al. (2001) proposed that a complicated relationship exists between forest land and runoff. The impact of forest land on runoff was different in different regions; forest type and forest management were the main influencing factors. Consequently, runoff depth would also increase in Northern China. Jin (1989) and Hao et al. (2004) drew the conclusions that higher forest cover rate was beneficial for increasing runoff depth. In Naoli River basin, forest land is located on the border of the watershed where the elevation is high and terrain mountainous. It has a steep slope and is good for precipitation. In addition, Naoli River basin is located at high latitudes. Its coniferous forest has a small capability of canopy interception. Compared with Northern China, the storage capacity of ground dead leaves is lower in Naoli River basin. Therefore, forest land promotes runoff generation due to its small canopy interception, low-water capacity, and steep slope. The conclusion that forest land is good for runoff is in complete agreement with the results of the above-mentioned studies.

Researchers have different opinions about the complexities of the relationship between fluvial wetland and runoff. Li & Shi (2015) believe that there is a complicated relationship between fluvial wetland and runoff. With the shrinking in wetland, runoff coefficient would become smaller or bigger depending on the change in circumstances. According to a study conducted by Mosquera et al. (2015), there is a close relationship between runoff coefficient and wetland. With the enlargement of wetland, runoff coefficient would become larger. Meanwhile, wetland has higher evapotranspiration than other land use types. The discharge of wetland is bound up with wetland slope. The results of this paper completely agree with the conclusions mentioned above.

In the upstream regions of Naoli River, the impact of human activity is small, but farmland cover most of the downstream basin and the unused land (fluvial wetland in Naoli River basin) plays an important role in runoff of the downstream. Dry land, paddy field, and fluvial wetland were all adverse factors for runoff generation. When dry land transforms into paddy field runoff is significantly affected and fluvial wetland changes as well. The downstream basin is a wetland nature reserve, moreover, it is an integral part of Sanjiang plain which is a major base of grain production in China. As a result, runoff in the downstream of Naoli River is very complicated. To analyze the impact of different land use types on runoff is very necessary.

This paper analyzed the impact of different land use types on runoff depth. The relationship between precipitation and runoff was also explored. There was significant impact of LUCC on the hydrologic process. This study has certain limitations: soil and meteorological data were obtained with limited precision. Defaults were observed in the relationship between LUCC and runoff depth by the method of multi-objective decision analysis. Further studies must be conducted on data accuracy and research method.

ACKNOWLEDGEMENTS

The authors express sincere thanks to the Water Conservancy Science and Technology Project of Heilongjiang Province (201305), National Natural Science Foundation Grants Program (51509264), National Natural Science Project of China (51679252), and the National Key Research and Development Program of the 13th five-year plan (2016YFA0601500) for supporting this study. Baoqi Li is the lead author of this paper. Weihua Xiao and Yicheng Wang have given advice and modified the paper. Mingzhi Yang and Ya Huang provided support for calculation. The authors declare that they have no conflict of interest.

REFERENCES

REFERENCES
Hao
,
H. M.
&
Ren
,
Z. Y.
2009
Land use/land cover change (LUCC) and eco-environment response to LUCC in farming-pastoral zone, China
.
Journal of Integrative Agriculture
8
(
1
),
91
97
.
Hao
,
F. H.
,
Chen
,
L. Q.
,
Liu
,
C. M.
&
Dai
,
D.
2004
Impact of land use change on runoff and sediment yield
.
Journal of Soil and Water Conservation
18
(
3
),
5
8
.
Jain
,
S. K.
,
Tyagi
,
J.
&
Singh
,
V.
2010
Simulation of runoff and sediment yield for a Himalayan watershed using SWAT model
.
Journal of Water Resource & Protection
2
(
3
),
267
281
.
Jia
,
Y. W.
,
Zhao
,
H. L.
,
Niu
,
C. W.
,
Jiang
,
Y. Z.
,
Gan
,
H.
,
Xing
,
Z.
,
Zhao
,
X. L.
&
Zhao
,
Z. X.
2009
A WebGIS-based system for rainfall-runoff prediction and real-time water resources assessment for Beijing
.
Computers & Geosciences
35
(
7
),
1517
1528
.
Jin
,
D. L.
1989
The impact of forest on hydrologic factors
.
Yangtze River
1
,
30
37
.
Kannan
,
N.
,
White
,
S. M.
,
Worrall
,
F.
&
Whelan
,
M. J.
2007
Sensitivity analysis and identification of the best evapotranspiration and runoff options for hydrological modelling in SWAT-2000
.
Journal of Hydrology
332
(
3–4
),
456
466
.
Leitinger
,
G.
,
Tasser
,
E.
,
Newesely
,
C.
,
Obojes
,
N.
&
Tappeineret
,
U.
2010
Seasonal dynamics of surface runoff in mountain grassland ecosystems differing in land use
.
Journal of Hydrology
385
(
1–4
),
95
104
.
Li
,
W. H.
,
He
,
Y. T.
&
Yang
,
L. T.
2001
A summary and perspective of forest vegetation impacts on water yield
.
Journal of Natural Resources
16
(
5
),
398
406
.
Li
,
C. B.
,
Qi
,
J. G.
,
Feng
,
Z. D.
,
Yin
,
R. S.
,
Zou
,
S. B.
&
Zhang
,
F.
2010
Parameters optimization based on the combination of localization and auto-calibration of SWAT model in a small watershed in Chinese Loess Plateau
.
Advances in Earth Science
4
(
3
),
296
310
.
Lin
,
K.
,
Lv
,
F.
,
Chen
,
L.
,
Singh
,
V. P.
,
Zhang
,
Q.
&
Chen
,
X. H.
2014
Xinanjiang model combined with Curve Number to simulate the effect of land use change on environmental flow
.
Journal of Hydrology
519
,
3142
3152
.
Lópezcarr
,
D.
,
Davis
,
J.
,
Jankowska
,
M.
,
Grant
,
L.
,
Lópezcarr
,
A.
&
Clark
,
M.
2012
Space versus place in complex human-natural systems: spatial and multi-level models of tropical land use and cover change (LUCC) in Guatemala
.
Ecological Modelling
229
(
4
),
64
75
.
Mosquera
,
G. M.
,
Lazo
,
P. X.
,
Célleri
,
R.
,
Wilcox
,
B. P.
&
Crespo
,
P.
2015
Runoff from tropical alpine grasslands increases with areal extent of wetlands
.
Catena
125
,
120
128
.
Schirpke
,
U.
,
Leitinger
,
G.
,
Tappeiner
,
U.
&
Tasser
,
E.
2012
SPA-LUCC: Developing land-use/cover scenarios in mountain landscapes
.
Ecological Informatics
12
(
11
),
68
76
.
Song
,
K.
,
Zhao
,
J.
,
Ouyang
,
W.
,
Zhang
,
X.
&
Hao
,
F.
2010
LUCC and landscape pattern variation of wetlands in warm-rainy Southern China over two decades
.
Procedia Environmental Sciences
2
(
6
),
1296
1306
.
Yao
,
L.
,
Chen
,
L.
,
Wei
,
W.
&
Sun
,
R.
2015
Potential reduction in urban runoff by green spaces in Beijing: a scenario analysis
.
Urban Forestry & Urban Greening
14
(
2
),
300
308
.
Yu
,
W.
,
Zang
,
S.
,
Wu
,
C.
,
Liu
,
W.
&
Na
,
X.
2011
Analyzing and modeling land use land cover change (LUCC) in the Daqing City, China
.
Applied Geography
31
(
2
),
600
608
.