Surface water runoff estimation method and system

Abstract

The invention discloses a surface water runoff estimation method and system, and the method comprises the steps: firstly dividing a research region into a plurality of runoff calculation units, and then obtaining the hydrological parameters of each calculation unit through historical data query. A research area is subdivided into runoff calculation units, and then modeling is carried out according to the specific condition of each unit. The method not only can distinguish and reflect the heterogeneity among the regions, but also can capture and simulate the details of the hydrological process on the microscopic level. Through scientific and reasonable model construction and data integration, the surface runoff of each calculation unit is estimated. And finally, summarizing the runoff volume of each calculation unit in a weighted average mode to obtain the total surface water runoff volume of the research area. According to the method, seasonal changes and land utilization differences are considered, the surface water runoff can be estimated more accurately, the accuracy of model parameters is ensured by combining a remote sensing technology and a GIS spatial data analysis means, and the prediction reliability is improved.

Description

Technical Field

[0001] The invention belongs to the field of runoff estimation, and in particular relates to a method and system for estimating surface water runoff. Background Art

[0002] Surface water resources are an important part of the earth's water resources and have important ecological, social and economic significance. Accurate estimation of surface water runoff is the basic work in water resources management and is crucial for the rational allocation, effective utilization and water environment protection of water resources. Estimation of surface water runoff not only helps to make reasonable planning of regional water resources, but also provides a scientific basis for agricultural irrigation, hydropower generation, urban water system management, etc.

[0003] With climate change, land use change and accelerated urbanization, the changes in surface water runoff are becoming more complex, which requires traditional hydrological models to consider more dynamic factors. As the main tool for estimating surface water runoff, hydrological models are based on certain hydrological principles to estimate the water volume in a basin or region by simulating hydrological processes such as precipitation, evaporation, infiltration, and runoff. Traditional hydrological models still face the challenge of fully considering factors such as regional differences, complex terrain, and changes in land use types.

[0004] Land use type has a profound impact on surface hydrological processes. For example, urbanized areas usually have more hardened ground, which makes it difficult for precipitation to penetrate, and most precipitation is converted into surface runoff; while agricultural land (such as paddy fields and dry land) has a strong water retention capacity, and part of the precipitation forms groundwater through soil infiltration or is converted into evaporation; forests and grasslands have strong water absorption and retention capabilities, and the surface runoff is small. Therefore, when estimating surface water runoff, the differences in land use types must be fully considered, especially under complex terrain conditions, the regulatory effect of land use types on hydrological processes cannot be ignored.

[0005] Although the existing hydrological estimation methods have improved the accuracy of surface water runoff estimation to a certain extent, they still have certain limitations when facing factors such as different land use types, complex terrain and climate change. Therefore, how to fully consider the differences in land use types, terrain characteristics and other environmental factors while ensuring the accuracy of the model has become an important topic in current hydrological research. Summary of the invention

[0006] In order to solve the above technical problems, the present invention provides a method and system for estimating surface water runoff based on historical data. By dividing the study area into several runoff calculation units, fully considering the impact of different land use types on the hydrological process and combining remote sensing data, GIS technology and machine learning algorithms, the accuracy of surface water runoff estimation can be effectively improved.

[0007] The technical solution provided by the present invention is specifically as follows:

[0008] A method for estimating surface water runoff, comprising the steps of:

[0009] S1. Divide the study area into several runoff calculation units and obtain the area proportion of each runoff calculation unit;

[0010] S2, respectively obtaining the runoff-related data of each runoff calculation unit in different seasons in the past several years, and respectively constructing a corresponding runoff prediction data set for each runoff calculation unit;

[0011] S3, constructing an annual runoff prediction model for each runoff calculation unit, and using the corresponding runoff prediction data set to train the annual runoff prediction model;

[0012] S4. For the year to be predicted, collect the runoff related data of each runoff calculation unit in that year and input it into the corresponding annual runoff prediction model to obtain the annual runoff prediction value of each runoff calculation unit;

[0013] S5. According to the area proportion of each runoff calculation unit, the annual runoff prediction value of each runoff calculation unit is weighted and summed to obtain the estimated value of surface runoff in the study area in the year to be predicted.

[0014] Furthermore, in step S1, the study area is divided into a plurality of grid cells of the same size, each grid cell represents a runoff calculation unit; or, the watershed is divided into a plurality of sub-watersheds using flow direction analysis and watershed extraction algorithms, each sub-watershed being an independent runoff calculation unit; or, the study area is divided into a plurality of runoff calculation units according to existing administrative divisions.

[0015] Furthermore, in step S2, the runoff-related data include the proportion of land area of ​​different utilization types in the runoff calculation unit, the average altitude value of the runoff calculation unit, and the proportion of water area, vegetation coverage area, average precipitation and average evaporation in each season; the runoff prediction data set includes several data samples, in each data sample, the runoff-related data of each year is used as sample data, and the annual surface runoff of the runoff calculation unit is used as corresponding label data.

[0016] Furthermore, the land of different utilization types includes three types: urban land, agricultural land and forest grassland. The urban land includes buildings and roads, the agricultural land includes paddy fields and dry land, and the forest grassland includes forests and grasslands.

[0017] Furthermore, each runoff calculation unit in step S3 adopts an annual runoff prediction model with the same structure; the annual runoff prediction model adopts a neural network model, including an input layer and an output layer, and several hidden layers between the input layer and the output layer.

[0018] Preferably, the output layer includes 1 output neuron for outputting the predicted value of annual runoff; the hidden layer has 4 layers, and the number of neurons contained in each layer is 30, 20, 10, and 5 respectively, and the activation function of the neuron adopts the ReLU function; the loss function of the annual runoff prediction model adopts the mean square error function.

[0019] Furthermore, in step S4, the MNDWI value is used to replace the water area ratio in the runoff related data. The MNDWI is obtained by obtaining the multispectral remote sensing data corresponding to the runoff calculation unit, and extracting the water area coverage MNDWI by calculating the reflectivity difference between the green light and short-wave infrared bands:

[0020] MNDWI=(Green-SWIR) / (Green+SWIR),

[0021] Among them, Green represents the reflectivity of the green light band, and SWIR represents the reflectivity of the short-wave infrared band.

[0022] Furthermore, in step S4, the NDVI value is used to replace the vegetation coverage area ratio in the runoff related data. The NDVI is obtained by obtaining the multispectral remote sensing data corresponding to the runoff calculation unit, and extracting the vegetation coverage NDVI by calculating the reflectivity difference between the red light and near infrared bands:

[0023] NDVI=(NIR-Red) / (NIR+Red),

[0024] Among them, NIR represents the reflectivity of the near-infrared band, and Red represents the reflectivity of the red light band.

[0025] Furthermore, in step S5, it is assumed that the study area is divided into n runoff calculation units, and the area proportion of the i-th runoff calculation unit is S i , the annual surface runoff prediction value of the i-th runoff calculation unit in the predicted year is R i , the annual surface runoff of the entire study area in the year to be predicted is expressed as R, then

[0026] A surface water runoff estimation system based on the above method comprises:

[0027] The runoff calculation unit segmentation module is used to divide the study area into several runoff calculation units and obtain the area proportion of each runoff calculation unit;

[0028] A data acquisition and input module is used to obtain required data from external sources through interfaces or automated scripts and process them into runoff-related data for use in runoff prediction dataset construction and prediction input;

[0029] The runoff prediction data set module is used to store the runoff related data of each runoff calculation unit in previous years and the annual surface runoff value of the corresponding year;

[0030] A runoff prediction model building module is used to build an annual runoff prediction model for each runoff calculation unit, and train the annual runoff prediction model using the runoff prediction data set stored in the runoff prediction data set module;

[0031] The unit runoff prediction module is used to call the prediction model trained in the runoff prediction model construction module, and obtain the annual runoff prediction value of each runoff calculation unit in the year to be predicted based on the input runoff related data of each runoff calculation unit in the year to be predicted;

[0032] The annual runoff estimation module is used to perform weighted summation of the annual runoff prediction values ​​of each runoff calculation unit according to the area proportion of each runoff calculation unit, so as to obtain the estimated value of surface runoff in the study area in the year to be predicted.

[0033] Compared with the prior art, the present invention has at least the following beneficial effects:

[0034] In order to more accurately capture the influence of complex factors on surface water runoff, the present invention subdivides the study area into runoff calculation units, and then models the specific conditions of each unit. It can not only distinguish and reflect the heterogeneity between regions, but also capture and simulate the details of the hydrological process at the micro level. Through scientific and reasonable model construction and data integration, a higher level of hydrological process simulation and runoff prediction accuracy is achieved. At the same time, the present invention combines remote sensing technology and GIS spatial data analysis methods to ensure the real-time and accuracy of model parameters and improve the reliability of prediction. The present invention helps to more deeply understand and respond to complex hydrological environmental changes and provide a solid foundation for water resources management decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0036] Figure 1 It is a schematic flow chart of a method for estimating surface water runoff provided by an embodiment of the present invention;

[0037] Figure 2 It is a structural schematic diagram of an annual runoff prediction model provided by an embodiment of the present invention;

[0038] Figure 3 It is a schematic diagram of the module composition of a surface water runoff estimation system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0039] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, other embodiments obtained by ordinary technicians in this field without making creative work are all within the scope of protection of the present invention.

[0040] Embodiment 1

[0041] This embodiment provides a method for estimating surface water runoff. Figure 1 As shown, the method mainly comprises the steps of:

[0042] S1. Divide the study area into several runoff calculation units and obtain the area proportion of each runoff calculation unit.

[0043] Due to differences in natural factors such as topography, climate, soil, vegetation, and land use types, hydrological processes (such as precipitation, evaporation, infiltration, etc.) usually vary greatly in different regions. Dividing the study area into multiple calculation units can simulate the hydrological characteristics of each region more carefully, avoid using a single homogenization assumption for the entire region, and thus improve the accuracy of runoff estimation.

[0044] In the actual hydrological calculation process, the hydrological model usually processes spatial data in the form of a grid. The study area is divided into multiple grid cells of the same size. Each grid cell represents a geographic location, usually in the form of a square or rectangle. The size of the grid depends on the data resolution and calculation accuracy requirements.

[0045] In some complex watersheds, in addition to conventional rasterization methods, the watershed can also be divided according to the water flow path and confluence process of the watershed, and the local unit can be divided in combination with the river network structure of the watershed. For example, the watershed can be divided into multiple sub-basins using flow direction analysis and watershed extraction algorithms (such as the Watershed algorithm), and each sub-basin is used as an independent calculation unit. This method is particularly suitable for larger watersheds or areas with obvious water flow paths.

[0046] In some cases, the study area can also be divided into calculation units based on existing administrative divisions. This method is usually used for water resource assessment in large areas and can make better use of existing administrative resources and data.

[0047] To facilitate the subsequent description and related calculation expressions, it is assumed that the study area is divided into n runoff calculation units in this embodiment, and the area proportion of the i-th runoff calculation unit is S i (i.e. the area ratio of the calculation unit to the total study area).

[0048] S2. For each runoff calculation unit, through historical data query processing, obtain the water area ratio, vegetation coverage area ratio, average precipitation, average evaporation data of each season (spring, summer, autumn and winter) in the runoff calculation unit in the past few years, as well as the land area ratio of different utilization types and the average altitude value of the runoff calculation unit, and at the same time obtain the annual surface runoff in the runoff calculation unit in the corresponding year.

[0049] Land use type directly affects all aspects of surface hydrological processes, including precipitation infiltration, evaporation, and water storage and loss. Different land use types will have different hydrological responses, so when calculating hydrological processes, the differences in land use must be fully considered. This embodiment mainly considers three types, namely urban land, agricultural land, and forest grassland. Urban land includes buildings, roads, etc. Due to a large amount of hardened ground in urban areas, the infiltration capacity of surface water is poor, and most of the precipitation is converted into runoff. The runoff in these areas is large and the confluence time is short. Agricultural land includes paddy fields, dry land, etc. Agricultural land, especially paddy fields, has a strong water retention capacity. Part of the precipitation is absorbed by plants, and the other part forms groundwater through soil infiltration. Part of the water will also be converted into evaporation. The runoff on the surface of paddy fields is relatively small, and its spatiotemporal distribution is affected by factors such as irrigation and crop growth. Forests and grasslands have strong water absorption and retention capabilities. Precipitation is mainly consumed through evaporation and soil infiltration, and surface runoff is less. In addition, vegetation regulates water cycle through transpiration and increases the evaporation of local precipitation.

[0050] For the same region, the distribution of different land use types and their area ratios are relatively fixed and can be obtained through remote sensing data and GIS technology. If there is no remote sensing data for a specific area, you can query global or national land use datasets. These datasets are usually published by governments, research institutions or international organizations and contain the distribution of various land use types. For example:

[0051] GLC 2000: A global land cover dataset (1 km resolution) suitable for large-scale land use analysis.

[0052] ESACCI Land Cover: A global land cover dataset provided by the European Space Agency (ESA) with high accuracy and long time series.

[0053] FAO Land Cover Database: A global land cover dataset provided by the Food and Agriculture Organization of the United Nations (FAO) that can be used to obtain the distribution of land use types.

[0054] Altitude values ​​are usually derived from digital elevation models (DEMs). By importing DEM data into GIS software, the average altitude of a region can be calculated. The source of DEM data is SRTM data (SRTM is a project for collecting earth surface elevation data jointly initiated by NASA and NSF in the United States, providing digital elevation models on a global scale). Of course, it can also be obtained through online platforms and query tools, such as Google Earth Engine, which provides satellite data and elevation data on a global scale. Users can directly extract DEM data and calculate altitude on the platform.

[0055] Since water areas, vegetation, precipitation, and evaporation have strong seasonal variability, in order to improve the estimation accuracy of surface runoff, this embodiment obtains the water area ratio, vegetation coverage area ratio, average precipitation, and average evaporation data in different seasons (spring, summer, autumn, and winter) in each runoff calculation unit through historical data query processing, and uses it as an important data basis for the subsequent construction of the prediction model. The acquisition of these historical data can be queried and processed through relevant data statistics platforms, such as the Geographic Remote Sensing Ecological Network (www.gisrs.cn), Xihe Energy Meteorological Big Data Platform, local hydrological stations, meteorological stations, etc.

[0056] To facilitate the subsequent description and related calculation expressions, in this embodiment, it is assumed that the average altitude of the i-th runoff calculation unit is A i The proportions of urban land, agricultural land and forest grassland in the runoff calculation unit are B i1 , B i2 , B i3 ; The proportion of water area, the proportion of vegetation coverage area, the average precipitation, and the average evaporation in spring are expressed as C i1 , C i2 , C i3 , C i4 , respectively represented as D in summer i1 , D i2 , D i3 , D i4 , respectively, in autumn, represented by E i1 、E i2 、E i3 、E i4 , respectively, in winter, they are F i1 、F i2 、F i3 、F i4 ; The annual surface runoff in this runoff calculation unit is Qi . Based on the annual i , B i1 , B i2 , B i3 , C i1 , C i2 , C i3 , C i4 , D i1 , D i2 , D i3 , D i4 、E i1 、E i2 、E i3 、E i4 、F i1 、F i2 、F i3 、F i4 As sample data, Q i As the corresponding label data, the sample data and the label data together constitute a data sample, and the historical data of the past few years are obtained accordingly to construct the runoff prediction data set of each runoff calculation unit.

[0057] S3. Build an annual runoff prediction model for each runoff calculation unit, and use the corresponding runoff prediction data set to train the model. The annual runoff prediction model uses a machine learning model, such as SVM, neural network, etc.

[0058] The annual runoff prediction model in this implementation adopts a neural network model, and the prediction model structure of each runoff calculation unit is the same, such as Figure 2 As shown, it includes an input layer and an output layer, as well as multiple hidden layers located between the input layer and the output layer. Specifically, in this embodiment: the input layer contains 20 input neurons, which are used to input 20 different types of data in the sample data; the output layer contains 1 output neuron, which is used to output the predicted value of annual runoff; there are 4 hidden layers, and the number of neurons contained in each layer is 30, 20, 10, and 5, respectively, which are used to mine hidden features from the input data; the activation function of the hidden layer neurons adopts the ReLU function, and the loss function of the entire model adopts the mean square error function (MSE), which is used to measure the square difference between the predicted value and the true value. MSE can effectively capture the prediction error and is easy to optimize during the training process.

[0059] S4. For the year to be predicted, collect the water area ratio, vegetation coverage area ratio, average precipitation, average evaporation, land area ratio of different utilization types, and average altitude value of the runoff calculation unit in each season in each runoff calculation unit of that year, and input them into the annual runoff prediction model of the corresponding runoff calculation unit to obtain the annual runoff prediction value of each runoff calculation unit.

[0060] Since the year to be predicted is usually the current year or the previous year, the water area percentage and vegetation coverage area percentage in each season may not be able to be obtained in a timely manner through the public statistical resource platform. Therefore, in this embodiment, the MNDWI (Modified Normalized Difference Water Index) value is used to replace the actual water area percentage, and the NDVI (Normalized Difference Vegetation Index) value is used to replace the actual vegetation coverage area percentage. MNDWI and NDVI can be obtained by processing and calculating remote sensing satellite data.

[0061] This embodiment uses Landsat satellite data to calculate MNDWI and NDVI. Landsat satellite data is a set of earth observation satellite data jointly managed by the National Aeronautics and Space Administration (NASA) and the United States Geological Survey (USGS). Since its first launch in 1972, it has collected multispectral remote sensing data for fifty years, which can be used for land use type classification, vegetation index extraction, water body detection, etc.

[0062] Specifically, the multispectral remote sensing data corresponding to the runoff calculation unit is obtained, and the water coverage MNDWI is extracted by calculating the reflectivity difference between the green light and short-wave infrared bands, and the vegetation coverage NDVI is extracted by calculating the reflectivity difference between the red light and near-infrared bands:

[0063] MNDWI=(Green-SWIR) / (Green+SWIR),

[0064] NDVI=(NIR-Red) / (NIR+Red),

[0065] Among them, Green represents the reflectivity of the green light band, SWIR represents the reflectivity of the short-wave infrared band, NIR represents the reflectivity of the near infrared band, and Red represents the reflectivity of the red light band.

[0066] The MNDWI and NDVI values ​​for each season can be obtained by taking the average of multiple calculations within the season. The average precipitation and average evaporation can be obtained through meteorological data published by the meteorological department and hydrological data published by the water conservancy department. Since meteorological data and water conservancy data are highly timely, they can be obtained through inquiries. The proportion of land area with different utilization types can also be obtained through remote sensing satellite data, and the average altitude value for the same runoff calculation unit is basically unchanged, and the data of previous years can be used.

[0067] S5. According to the area proportion of each runoff calculation unit, the annual runoff prediction value of each runoff calculation unit is weighted and summed to obtain the surface runoff prediction value of the study area in that year.

[0068] Assume that the annual surface runoff of the study area in the year to be predicted is expressed as R, and the predicted value of the annual surface runoff of the i-th runoff calculation unit in the year to be predicted is R i , then

[0069] The above is the main content of this embodiment. In this embodiment method, in order to more accurately capture the influence of complex factors on surface water runoff, the study area is subdivided into runoff calculation units, and then modeling is performed for the specific conditions of each unit. The purpose of this is to enable the model to not only distinguish and reflect the heterogeneity between regions, but also to capture and simulate the details of the hydrological process at the micro level. At the same time, the method combines remote sensing technology and GIS spatial data analysis methods to ensure the real-time and accuracy of model parameters and improve the reliability of prediction.

[0070] In summary, this method aims to comprehensively consider the diversified actual influencing factors in its design, and achieve a higher level of hydrological process simulation and runoff prediction accuracy through scientific and reasonable model construction and data integration. This will help to more deeply understand and respond to complex hydrological environmental changes and provide a solid foundation for water resources management decision-making.

[0071] Embodiment 2

[0072] Based on the above method, this embodiment provides a surface water runoff estimation system based on historical data, such as Figure 3 As shown, the system mainly includes the following modules:

[0073] (1) Runoff calculation unit segmentation module

[0074] This module is used to divide the study area into several runoff calculation units and obtain the area proportion of each runoff calculation unit. It provides users with input or uploads of regional division information, allowing users to customize the study area.

[0075] (2) Data acquisition and input module

[0076] This module is used to obtain required data from external sources through interfaces or automated scripts, including geospatial data, remote sensing images, digital elevation models (DEMs), meteorological data, etc., and process them into runoff-related data for runoff prediction dataset construction and prediction input.

[0077] (3) Runoff prediction dataset module

[0078] This module is used to store the runoff-related data of each runoff calculation unit in previous years and the annual surface runoff value of the corresponding year. It can be divided into multiple independent storage areas to store the data of each runoff calculation unit respectively.

[0079] (4) Runoff prediction model construction module

[0080] This module is used to build an annual runoff prediction model for each runoff calculation unit, and train the annual runoff prediction model using the runoff prediction dataset stored in the runoff prediction dataset module.

[0081] (5) Unit runoff prediction module

[0082] This module is used to call the prediction model trained in the runoff prediction model construction module, and obtain the annual runoff prediction value of each runoff calculation unit in the year to be predicted based on the input runoff related data of each runoff calculation unit in the year to be predicted.

[0083] (6) Annual runoff estimation module

[0084] This module is used to perform weighted summation of the annual runoff forecast values ​​of each runoff calculation unit according to the area proportion of each runoff calculation unit, and obtain the estimated value of surface runoff in the study area in the year to be predicted.

[0085] (7) Result output and visualization module

[0086] This module is responsible for visualizing the calculation results and generating charts, reports and maps. Specifically, based on GIS and data visualization tools, maps can be generated to show the distribution of surface water runoff in each calculation unit; statistical charts, such as bar charts and line charts, can be generated to show the hydrological response of each region and the relationship between precipitation and runoff; report output function can be provided to generate detailed reports including surface water resource assessment, runoff estimation, land use analysis, etc.

[0087] The system is modularly designed and combines GIS, remote sensing, historical meteorological data and hydrological models to effectively estimate surface water runoff. Each module has clear and independent functions, but can work closely together to provide accurate water resource assessment results and support applications such as scientific research, decision-making and water resource management.

[0088] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them. Under the concept of the present invention, the technical features in the above embodiments or different embodiments may also be combined, the steps may be implemented in any order, and there are many other changes in different aspects of the present invention as described above, which are not provided in detail for the sake of simplicity. Although the present invention has been described in detail with reference to the aforementioned embodiments, a person of ordinary skill in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for estimating surface water runoff, characterized in that: Includes steps: S1. Divide the study area into several runoff calculation units and obtain the area proportion of each runoff calculation unit; S2, respectively obtain the runoff related data of each runoff calculation unit in different seasons in the past several years, and respectively construct a corresponding runoff prediction data set for each runoff calculation unit; S3, constructing an annual runoff prediction model for each runoff calculation unit, and using the corresponding runoff prediction data set to train the annual runoff prediction model; S4. For the year to be predicted, collect the runoff related data of each runoff calculation unit in that year and input it into the corresponding annual runoff prediction model to obtain the annual runoff prediction value of each runoff calculation unit; S5. According to the area proportion of each runoff calculation unit, the annual runoff prediction value of each runoff calculation unit is weighted and summed to obtain the estimated value of surface runoff in the study area in the year to be predicted.

2. A surface water runoff estimation method according to claim 1, characterized in that: In step S1, the study area is divided into multiple grid cells of the same size, each grid cell represents a runoff calculation unit; or, the watershed is divided into multiple sub-basins using flow direction analysis and watershed extraction algorithms, each sub-basin serves as an independent runoff calculation unit; or, the study area is divided into several runoff calculation units according to existing administrative divisions.

3. A surface water runoff estimation method as claimed in claim 1, characterized in that: In step S2, the runoff related data include the proportion of land area of ​​different utilization types in the runoff calculation unit, the average altitude value of the runoff calculation unit, and the proportion of water area, vegetation coverage area, average precipitation and average evaporation in each season; the runoff prediction data set includes a number of data samples, in each data sample, the runoff related data of each year is used as sample data, and the annual surface runoff of the runoff calculation unit is used as the corresponding label data.

4. A surface water runoff estimation method as claimed in claim 3, characterized in that: The land of different utilization types includes three types: urban land, agricultural land and forest grassland. The urban land includes buildings and roads, the agricultural land includes paddy fields and dry land, and the forest grassland includes forests and grasslands.

5. A surface water runoff estimation method according to claim 1, characterized in that: In step S3, each runoff calculation unit adopts an annual runoff prediction model with the same structure; the annual runoff prediction model adopts a neural network model, including an input layer and an output layer, and several hidden layers between the input layer and the output layer.

6. A surface water runoff estimation method as claimed in claim 5, characterized in that: The output layer includes 1 output neuron for outputting the predicted value of annual runoff; the hidden layer has 4 layers, and the number of neurons contained in each layer is 30, 20, 10, and 5 respectively, and the activation function of the neuron adopts the ReLU function; the loss function of the annual runoff prediction model adopts the mean square error function.

7. A surface water runoff estimation method as claimed in claim 3, characterized in that: In step S4, the MNDWI value is used to replace the water area ratio in the runoff related data. The MNDWI is obtained by obtaining the multispectral remote sensing data corresponding to the runoff calculation unit, and extracting the water area coverage MNDWI by calculating the reflectivity difference between the green light and short-wave infrared bands: MNDWI=(Green-SWIR) / (Green+SWIR), Among them, Green represents the reflectivity of the green light band, and SWIR represents the reflectivity of the short-wave infrared band.

8. A surface water runoff estimation method as claimed in claim 3, characterized in that: In step S4, the NDVI value is used to replace the vegetation coverage area ratio in the runoff related data. The NDVI is obtained by obtaining the multispectral remote sensing data corresponding to the runoff calculation unit, and extracting the vegetation coverage NDVI by calculating the reflectivity difference between the red light and near infrared bands: NDVI=(NIR-Red) / (NIR+Red), Among them, NIR represents the reflectivity of the near-infrared band, and Red represents the reflectivity of the red light band.

9. A surface water runoff estimation method according to claim 1, characterized in that: In step S5, it is assumed that the study area is divided into n runoff calculation units, and the area proportion of the i-th runoff calculation unit is S i , the annual surface runoff prediction value of the i-th runoff calculation unit in the predicted year is R i , the annual surface runoff of the entire study area in the year to be predicted is expressed as R, then 10. A surface water runoff estimation system based on the method according to any one of claims 1 to 9, characterized in that: include: The runoff calculation unit segmentation module is used to divide the study area into several runoff calculation units and obtain the area proportion of each runoff calculation unit; A data acquisition and input module is used to obtain required data from external sources through interfaces or automated scripts and process them into runoff-related data for use in runoff prediction dataset construction and prediction input; The runoff prediction data set module is used to store the runoff related data of each runoff calculation unit in previous years and the annual surface runoff value of the corresponding year; A runoff prediction model building module is used to build an annual runoff prediction model for each runoff calculation unit, and train the annual runoff prediction model using the runoff prediction data set stored in the runoff prediction data set module; The unit runoff prediction module is used to call the prediction model trained in the runoff prediction model construction module, and obtain the annual runoff prediction value of each runoff calculation unit in the year to be predicted based on the input runoff related data of each runoff calculation unit in the year to be predicted; The annual runoff estimation module is used to perform weighted summation of the annual runoff prediction values ​​of each runoff calculation unit according to the area proportion of each runoff calculation unit, and obtain the estimated value of surface runoff in the study area in the year to be predicted; The result output and visualization module is used to visualize the estimation results and generate charts or reports.

Images