Multi-scale response analysis method for watershed runoff to landscape pattern

Abstract

The invention relates to a multi-scale response analysis method of drainage basin runoff to a landscape pattern, ecological hydrological research is the basis of comprehensive management of a drainage basin, and the response relationship of the runoff to the landscape pattern is an important decision basis for integrated protection and systematic management of landscapes, forests, lakes, grasses and sand in the drainage basin. Taking the Poyang Lake basin as an example, the method comprises the following steps: establishing a basin SWAT model; determining parameter values of the Poyang Lake basin based on an SUFI-2 method and verifying the correctness of the parameter values; inputting parameters, and calculating the runoff volume of the Poyang Lake drainage basin through an SWAT model; constructing a landscape type division system of'landscape, water, forest field, lake, grass and sand 'based on land utilization / cover; selecting landscape indexes through collinearity diagnosis; the influence of landscape composition on the runoff is discussed by utilizing correlation analysis methods such as a scatter diagram, a Pearson's correlation coefficient and a geographically weighted regression model, and the response of the runoff to the landscape pattern is revealed in landscape and type scales. According to the method, the Poyang Lake basin is taken as a research object, runoff simulation and landscape pattern analysis are respectively carried out by utilizing an SWAT model and Fragstats software, and nonlinear characteristics of runoff response to landscape composition and landscape pattern are revealed.

Description

Technical Field

[0001] The present invention belongs to the field of runoff simulation calculation, and in particular relates to a basin runoff calculation simulation and multi-scale analysis technology based on a SWAT model. Background Art

[0002] Water resources are the source of life, the key to production, and the foundation of ecology. As the main way to obtain water resources, runoff is crucial to maintaining life and health, social and economic development, and protecting ecosystems. Accurate runoff simulation can provide a deep understanding of the eco-hydrological processes and mechanisms of the basin, help formulate more accurate water resources management strategies, improve water resources utilization efficiency, and thus ensure the safety of water resources in the basin. Since 1950, the hydrological models used for runoff simulation mainly include lumped hydrological models such as the SWM (Stanford Watershed Model), TANK model, SMAR (Soil Moisture Accounting Routing), and Xin'anjiang model, as well as distributed hydrological models such as the SWAT (Soil and Water Assessment Tool), VIC (Variable Infiltration Capacity), and SHE (System Hydrological European). Among them, the SWAT model, as a typical representative of distributed hydrological models, is widely used in basin runoff simulation at home and abroad due to its mature model and high simulation accuracy.

[0003] Landscape pattern is the spatial arrangement of patches of different sizes, shapes and types, and is the result of the coupling of natural factors and human activities at different scales. Landscape pattern is an important factor affecting basin runoff, which is mainly reflected in the changes in the area, shape, distribution, structure and other characteristics of patches, which affect the water cycle process, including evapotranspiration, interception and infiltration, and then affect the formation and transformation of surface runoff, soil flow and underground runoff. With the rapid development of social economy, large-scale human activities such as urban and rural construction have changed the landscape composition and landscape pattern, interfered with the ecological hydrological process of the basin, and had a significant impact on the formation, transformation and spatial and temporal distribution of runoff, resulting in frequent disasters such as drought, floods, soil erosion and so on. Revealing the response relationship between basin runoff and landscape pattern is particularly important for the integrated protection and systematic management of mountains, rivers, forests, fields, lakes, grasslands and deserts.

[0005] At present, the use of SWAT model to simulate runoff has the following shortcomings: (1) The main research focuses on the spatiotemporal variation and driving factor analysis of runoff in the Poyang Lake Basin. (2) There is little research on the response of runoff to mountains, rivers, forests, farmlands, lakes, grasslands and sand. (3) The research objects are mainly single sub-basins such as Ganjiang River and Fuhe River, and there is little overall analysis of the Poyang Lake Basin. Summary of the invention

[0006] In view of the shortcomings of the existing technology, the present invention establishes a basic database for the target basin and a basin SWAT model. On this basis, the temporal and spatial heterogeneity of runoff in the Poyang Lake basin is analyzed, the spatial correlation between runoff and landscape composition is analyzed, and the nonlinear characteristics of runoff response to landscape pattern are explored.

[0007] In order to solve the above technical problems, the present invention adopts the following technical solution: a method for analyzing the multi-scale response of watershed runoff to landscape pattern, comprising the following steps:

[0008] Step 1, establish the basin SWAT model and basic database;

[0009] Step 2: Determine the model parameter values ​​based on the SUFI-2 method and verify their correctness;

[0010] Step 3: Construct a landscape classification system of "mountains, rivers, forests, farmlands, lakes, grasslands, and sands" based on land use / cover;

[0011] Step 4, select landscape index through collinearity diagnosis;

[0012] Step 5: Analyze the response relationship between runoff and landscape composition and landscape pattern.

[0013] Furthermore, a method for analyzing the multi-scale response of watershed runoff to landscape pattern is provided, comprising the following steps:

[0014] Step 1: Establish a watershed SWAT model and a basic database; the data in the basic database include land use data, digital elevation data, meteorological data, soil type data, soil attribute data, and hydrological data; the above data are added to the model after data cleaning, format unification, and projection transformation.

[0015] After the basic data was established, based on DEM data and D8 algorithm, the catchment area threshold (297,838ha) recommended by the SWAT model was used to extract the water system and divide the sub-basin boundaries. The boundaries were corrected and adjusted according to the actual river network in the study area. Finally, the Poyang Lake Basin was divided into 31 sub-basins and 1,152 HRUs. After completing the division of the hydrological response unit Hrus, the meteorological data was loaded into the model and the model was used for hydrological simulation.

[0016] Step 2: Determine the model parameter values ​​based on the SUFI-2 method and verify their correctness.

[0017] The SWAT-CUP software's Sequential Uncertainty Fitting (SUFI-2) algorithm was used to conduct sensitivity analysis on the model parameters. In the order of "upstream first, then downstream", the measured runoff data from Lijiadu Station and Waizhou Station were used for calibration. When the best simulation was achieved, the measured runoff data from Hukou Station were added to complete the runoff simulation of the Poyang Lake Basin. With reference to existing studies, 20 parameters were selected for analysis. The 10 most representative parameters were finally selected through global sensitivity analysis. After 7400 iterative operations, the sensitivity analysis results of the 10 runoff parameters of the SWAT model in the Poyang Lake Basin were obtained. The fitting coefficient R 2 The applicability of the SWAT model was evaluated by using the Nash-Sutcliffe efficiency (NSE) coefficient. 2 The values ​​of R and NSE vary between 0 and 1. The closer they are to 1, the better the model simulation effect. 2 When >0.6 and NSE>0.5, the model simulation results are credible.

[0018] Step 3: Construct a landscape classification system of “mountains, rivers, forests, farmlands, lakes, grasslands, and sands” based on land use / cover.

[0019] According to the land use / cover classification system of the Chinese Academy of Sciences, LULC is divided into six primary types, namely cultivated land, forest land, grassland, water area, construction land and unused land, and 25 secondary types. The Poyang Lake Basin contains all six primary types and 22 secondary types, but lacks three secondary types, namely permanent glaciers and snowfields, Gobi desert and other unused land. In order to analyze the “mountain, water, forest, farmland, lake, grassland and sand” system of the Poyang Lake Basin more specifically, the LULC classification system of the Chinese Academy of Sciences was adjusted (Table 3): “Water” mainly corresponds to the LULC first-level water area (rivers, lakes, reservoirs, ponds, tidal flats and beaches), the second-level lake is separately classified as “lake”, and the second-level swamp is included; the first-level LULC forest land is all classified as “forest (mountain)”; the first-level LULC cultivated land is all classified as “field”; the first-level LULC grassland is all classified as “grass”; “Sand” corresponds to the first-level LULC unused land (sand land, saline-alkali land, swamp land, bare land and bare rock texture), and the second-level swamp is classified as “water”; the construction land remains unchanged.

[0020] Step 4: Select landscape index through collinearity diagnosis.

[0021] The establishment of landscape indices is a key step in landscape pattern analysis, including principles such as comprehensiveness, typicality, and independence. The selected indices should be able to accurately measure and reflect the landscape pattern and have clear meanings. The Variance Inflation Factor (VIF) represents the ratio of the variance of the regression coefficient estimator to the variance when there is no linear correlation between the assumed independent variables. It is an indicator for measuring the severity of multicollinearity. When VIF > 1, the closer VIF is to 1, the smaller the degree of multicollinearity.

[0022] First, refer to existing studies to analyze the landscape pattern from four aspects: area, shape, distribution, and structure. Then, exclude the indices with obvious correlations in the calculation formula, the indices obtained by arithmetic mean, and the indices that do not reflect landscape heterogeneity from the landscape indices of Fragstats 4.2 software (https: / / www.fragstats.org / ). Retain and calculate 25 landscape indices, including 4 in terms of area, 5 in terms of shape, 9 in terms of distribution, and 7 in terms of structure. Finally, conduct a collinearity diagnosis on the above 25 landscape indices and separately select the landscape indices with the smallest VIF in the four aspects.

[0023] According to the collinearity diagnosis results, select Edge Density (ED) to characterize the patch area, select Area-Weighted Mean Circularity Index (CIRCLE_AM) to characterize the patch shape, select Interspersion and Juxtaposition Index (IJI) to characterize the patch distribution, and select Patch Richness Density (PRD) to characterize the landscape structure. ED is equal to the total length of all patch edges in the landscape divided by the total landscape area. ED ≥ 0, and the larger ED is, the smaller the average area of a single patch indicates. CIRCLE is equal to 1 minus the patch area divided by the area of the minimum circumscribed circle. 0 < CIRCLE < 1. When CIRCLE is close to 0, the patch is circular; when CIRCLE is close to 1, the patch is linear. IJI is one of the most important indicators for describing the landscape distribution pattern. 0 < IJI ≤ 100. When a patch type is adjacent to only a few patch types, that is, the adjacent distribution between patch types is more uneven, IJI is close to 0. When a patch type is equally adjacent to all other patch types, that is, the adjacency probability of each patch type is equal, IJI = 100. PRD is equal to the number of patch types per unit area (100 hectares) divided by the total landscape area. PRD > 0, and the larger PRD is, the higher the landscape richness.

[0024] Step 5, analysis of the response relationship between runoff and landscape composition and landscape pattern.

[0025] Three correlation analysis methods, namely scatter plot, Pearson correlation coefficient and geographically weighted regression (GWR), were used to explore the response relationship between basin runoff and landscape composition and landscape pattern. Among them, the scatter plot can intuitively reflect the correlation between variables, and the Pearson correlation coefficient can accurately reflect the strength of the correlation between variables. The above two methods are mainly used to analyze the linear relationship between variables, and GWR is used to explore the spatial relationship between runoff and landscape composition and landscape pattern.

[0026] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0027] 1. There is spatiotemporal heterogeneity in the runoff of the Poyang Lake Basin. The runoff depth fluctuated overall from 2011 to 2020, with high-value areas distributed in the Raohe River Basin and the Poyang Lake area, and low-value areas distributed in the middle reaches of the Ganjiang River Basin. The runoff depth has greater spatial differences on a monthly scale than on an annual scale.

[0028] 2. There is a spatial correlation between runoff and landscape composition in the Poyang Lake Basin. The fitting effect of GWR spatial regression is significantly better than that of linear regression. The influence of landscape composition on runoff is ranked from large to small as follows: forest > field > sand > grass, and all are positively correlated.

[0029] 3. The response of runoff in the Poyang Lake Basin to landscape pattern has nonlinear characteristics. At the landscape scale, landscape structure has the greatest impact on runoff, while patch distribution, area and shape have less impact on runoff; the lower the landscape richness, the fewer other types of patches adjacent to the patch, the smaller the area of ​​a single patch, and the closer the patch shape is to a circle, the higher the runoff depth. At the type scale, forest is the dominant landscape type in the Poyang Lake Basin. The patch area of ​​forest has the greatest impact on runoff, while patch distribution and shape have less impact on runoff; the smaller the area of ​​a single forest patch, the fewer other types of patches adjacent to the patch, and the closer the patch shape is to a circle, the higher the runoff depth. The analysis method proposed in the present invention provides new methods and ideas for integrated basin management and integrated protection and restoration of mountains, rivers, forests, farmlands, lakes, grasslands and deserts. BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 This is a diagram showing the temporal and spatial variation of runoff depth in the Poyang Lake Basin. Figure 2 Landscape composition map of the Poyang Lake Basin and its sub-basins Figure 3 This is the geographically weighted regression analysis of runoff depth and landscape index.

Claims

1. The multi-scale response analysis method of watershed runoff to landscape pattern includes the following steps: Step 1, establish a watershed SWAT model and a basic database; the data in the basic database includes land use data, digital elevation data, meteorological data, soil type data, soil attribute data, and hydrological data; The above data were added to the model after data cleaning, format unification and projection transformation; After the basic data was established, the catchment area threshold (297,838ha) recommended by the SWAT model was used based on DEM data and the D8 algorithm to extract the water system and divide the sub-basin boundaries. The boundaries were corrected and adjusted according to the actual river network in the study area, and the Poyang Lake Basin was finally divided into 31 sub-basins and 1,152 HRUs. After completing the division of the hydrological response unit Hrus, the meteorological data is loaded into the model and the model is used for hydrological simulation. Step 2: Determine the model parameter values ​​based on the SUFI-2 method and verify their correctness; The SWAT-CUP software's Sequential Uncertainty Fitting (SUFI-2) algorithm was used to conduct sensitivity analysis on the model parameters. In the order of "upstream first, then downstream", the measured runoff data from Lijiadu Station and Waizhou Station were used for calibration. When the best simulation was achieved, the measured runoff data from Hukou Station were added to complete the runoff simulation of the Poyang Lake Basin. With reference to existing studies, 20 parameters were selected for analysis. The most representative 10 parameters were finally selected through global sensitivity analysis. After 7400 iterative operations, the sensitivity analysis results of 10 runoff parameters of the SWAT model in the Poyang Lake Basin were obtained. The fitting coefficient R 2 The applicability of the SWAT model was evaluated by using the Nash-Sutcliffe efficiency (NSE) coefficient. 2 The values ​​of R and NSE vary between 0 and 1. The closer they are to 1, the better the model simulation effect. 2 When >0.6 and NSE>0.5, the model simulation results are credible. Step 3: Construct a landscape classification system of "mountains, rivers, forests, farmlands, lakes, grasslands, and sands" based on land use / cover; According to the land use / cover classification system of the Chinese Academy of Sciences, LULC is divided into six primary types, namely cultivated land, forest land, grassland, water area, construction land and unused land, and 25 secondary types. The Poyang Lake Basin contains all six primary types and 22 secondary types, but lacks three secondary types, namely permanent glaciers and snowfields, Gobi desert and other unused land. In order to analyze the "mountain, water, forest, farmland, lake, grassland and sand" system in the Poyang Lake Basin more specifically, the LULC classification system of the Chinese Academy of Sciences was adjusted (Table 3): "Water" mainly corresponds to the LULC first-level water area (rivers, lakes, reservoirs, ponds, tidal flats and beach), the second-level lake is separately classified as "lake", and the second-level swamp is included; the first-level LULC forest land is all classified as "forest (mountain)"; the first-level LULC cultivated land is all classified as "field"; the first-level LULC grassland is all classified as "grass"; "Sand" corresponds to the first-level LULC unused land (sand land, saline-alkali land, swamp land, bare land and bare rock texture), and the second-level swamp is classified as "water"; the construction land remains unchanged. Step 4, select landscape index through collinearity diagnosis; The establishment of landscape index is a key step in landscape pattern analysis, including the principles of comprehensiveness, typicality and independence. The selected index should be able to accurately measure and reflect the landscape pattern and have a clear meaning. Variance inflation factor (VIF) represents the ratio of the variance of the regression coefficient estimate to the variance when there is no linear correlation between the independent variables. It is an indicator to measure the severity of multicollinearity. VIF>1. The closer VIF is to 1, the smaller the degree of multicollinearity. First, analyze the landscape pattern from four aspects: area, shape, distribution, and structure with reference to existing research. Then, exclude the indices with obvious correlations in the calculation formula, the indices obtained by arithmetic mean, and the indices that do not reflect landscape heterogeneity from the landscape indices of Fragstats 4.2 software (https: / / www.fragstats.org / ), and retain and calculate 25 landscape indices, including 4 in terms of area, 5 in terms of shape, 9 in terms of distribution, and 7 in terms of structure. Finally, conduct a collinearity diagnosis on the above 25 landscape indices and select the landscape indices with the smallest VIF for each of the four aspects respectively. According to the collinearity diagnosis results, select the edge density (ED) to characterize the patch area, the area-weighted mean circularity index (CIRCLE_AM) to characterize the patch shape, the interspersion and juxtaposition index (IJI) to characterize the patch distribution, and the landscape richness density (PRD) to characterize the landscape structure. ED is equal to the total length of all patch edges in the landscape divided by the total landscape area. ED ≥ 0, and the larger ED is, the smaller the average area of a single patch indicates. CIRCLE is equal to 1 minus the patch area divided by the area of the minimum circumscribed circle. 0 < CIRCLE < 1. When CIRCLE is close to 0, the patch is circular; when CIRCLE is close to 1, the patch is linear. IJI is one of the most important indicators to describe the landscape distribution pattern. 0 < IJI ≤ 100. When a patch type is adjacent to only a few patch types, that is, when the adjacent distribution between patch types is more uneven, IJI is close to 0. When a patch type is equally adjacent to all other patch types, that is, when the adjacent probability of each patch type is equal, IJI = 100. PRD is equal to the number of patch types per unit area (100 hectares) divided by the total landscape area. PRD > 0, and the larger PRD is, the higher the landscape richness is. Step 5, analyze the response relationship between runoff and landscape composition and landscape pattern; Use three correlation analysis methods: scatter plot, Pearson correlation coefficient, and Geographically Weighted Regression (GWR) to explore the response relationship between basin runoff and landscape composition and landscape pattern. Among them, the scatter plot can intuitively reflect the correlation between variables, and the Pearson correlation coefficient can accurately reflect the strength of the correlation between variables. The above two methods are mainly used to analyze the linear relationship between variables, while GWR is used to explore the spatial relationship between runoff and landscape composition and landscape pattern.

Images