A runoff mode intelligent adaptation watershed hydrology prediction method and system

Abstract

The present application relates to the technical field of hydrological forecasting, and more particularly to a method and system for hydrological forecasting of a river basin through intelligent adaptation of runoff generation modes. The method comprises the following steps: collecting hydrological key data by deploying an Internet of Things monitoring array in the river basin, and constructing a three-dimensional layered runoff generation basic data set; calculating the water distribution ratio based on the three-dimensional layered runoff generation basic data set, and performing vertical analysis of runoff generation components to obtain runoff generation component data; performing topological coding based on the runoff generation component data to obtain hydrological response unit topological coding; constructing a multi-way tree confluence network based on the hydrological response unit topological coding, and performing full-basin section analysis to obtain full-basin river section process data. Therefore, the present application solves the problem that traditional single data sources and static analysis models cannot accurately reflect the complex hydrological process of a river basin by constructing a multi-dimensional dynamic data integration and multi-scale hydrological analysis closed loop, and improves the spatiotemporal accuracy of river basin runoff prediction and the scientificity of decision support.

Description

Technical Field

[0001] This invention relates to the field of hydrological forecasting technology, and in particular to a watershed hydrological forecasting method and system with intelligent adaptation to runoff patterns. Background Technology

[0002] Existing watershed hydrological forecasting methods have significant limitations when facing complex changes in runoff generation patterns. These limitations primarily manifest in their inadequate ability to characterize the differences in responses to various runoff generation mechanisms (such as surface runoff, interflow, and groundwater recharge), and the lack of adaptive mechanisms to dynamically adjust forecast model structure and parameters. Most current methods employ fixed runoff generation model structures, making it difficult to accurately identify and adjust applicable runoff generation mechanisms in the face of watershed land use change, spatiotemporal climate variability, or human activity disturbances. This leads to a decline in the model's ability to simulate abrupt events, nonlinear response processes, and local anomalies. Furthermore, existing systems often rely on a single hydrological model or static parameter set, failing to adequately consider the dynamic heterogeneity of different hydrological response units within the watershed, making it difficult to achieve spatial zoning adaptation and intelligent adjustment of process responses for runoff generation mechanisms. Simultaneously, existing methods suffer from low information utilization at the data fusion level, failing to effectively integrate multidimensional observational data such as flow, sediment, pollutants, and soil moisture, resulting in insufficient identification of runoff driving factors and affecting model forecast accuracy. In terms of system architecture, most hydrological forecasting systems lack modularity, multi-scale response and real-time feedback capabilities, and are unable to dynamically reconstruct runoff structures and intelligently switch forecasting strategies, thus limiting the applicability and operational efficiency of the models. Summary of the Invention

[0003] Therefore, it is necessary to provide a watershed hydrological forecasting method and system with intelligent runoff pattern adaptation to solve at least one of the above-mentioned technical problems.

[0004] To achieve the above objectives, a watershed hydrological forecasting method with intelligent runoff generation pattern adaptation is proposed, the method comprising the following steps:

[0005] Step S1: Deploy an IoT monitoring array in the watershed to collect key hydrological data and construct a three-dimensional hierarchical runoff generation dataset;

[0006] Step S2: Calculate the water distribution ratio based on the three-dimensional stratified runoff basic dataset, and perform vertical analysis of watershed runoff components to obtain runoff component data;

[0007] Step S3: Perform topological coding based on runoff composition data to obtain the topological coding of hydrological response units; construct a multi-branch tree confluence network based on the topological coding of hydrological response units, and perform cross-sectional analysis of the entire basin to obtain cross-sectional process data of the entire basin; monitor sediment transport capacity based on the cross-sectional process data of the entire basin, and perform closed-loop modeling to obtain the runoff hydrological prediction model of the river; predict runoff based on the runoff hydrological prediction model of the river to obtain three-dimensional prediction data of flow rate-sediment quantity-pollutant concentration;

[0008] Step S4: Visualize the three-dimensional prediction data of flow rate, sediment load, and pollutant concentration to obtain visualized composite data of river runoff generation; conduct a full-process assessment based on the visualized composite data of river runoff generation to obtain full-process assessment data of river runoff generation; and construct a full-process prediction report of river runoff generation based on the full-process assessment data of river runoff generation.

[0009] The beneficial effects of this invention are as follows: First, it expands hydrological monitoring data from single-point, two-dimensional planar sampling to a three-dimensional vertically layered structure. Through an IoT sensor array, it collects key parameters such as soil moisture, rainfall intensity, groundwater recharge, and surface runoff at high frequency. After mapping with a unified time window and spatial scale, it constructs a three-dimensional layered runoff generation dataset, effectively improving the expressive power of the spatial heterogeneity and vertical structural characteristics of the runoff generation process. Second, the system combines water distribution ratio analysis with vertical analysis of runoff components, refining the model based on the proportion of different runoff sources (such as surface runoff, interflow, and baseflow) within the watershed unit. This allows the flow simulation to not only reflect macroscopic trends but also the evolution of hydrological mechanisms within the watershed. Third, it reconstructs the spatial connection network between hydrological response units based on topological coding, characterizes the hydrodynamic coupling mode between different tributaries and the main channel through a multi-branch tree confluence structure, and introduces cross-sectional hydraulic analysis and sediment transport capacity modeling mechanisms to construct a... The closed-loop updated runoff prediction model possesses dynamic feedback adjustment capabilities in simulating sediment transport and pollutant release. Furthermore, by coupling the output three-dimensional prediction data of flow rate, sediment load, and pollutant concentration, the model introduces a visualization module to structurally encapsulate and graphically represent the multidimensional data, constructing multi-level composite image data represented by runoff composition maps, co-process lines, and pollution risk heatmaps. This provides quantitative support in sub-basin evaluation and full-process report generation, making the analysis results more intuitive, interconnected, and applicable to decision-making. Finally, this process establishes a complete chain from real-time monitoring, spatiotemporal modeling, basin extrapolation, pollution response, graphical encapsulation to analysis report construction, emphasizing the continuity, structural consistency, and semantic transferability of data across multiple scales and process nodes. This significantly improves the data expression depth, prediction accuracy, and comprehensive evaluation capabilities of runoff hydrological analysis, providing a solid data-driven foundation and closed-loop process characteristics for basin-scale hydrological modeling and pollution control decision support. Therefore, by constructing a closed loop of multidimensional dynamic data integration and multi-scale hydrological analysis, this invention solves the problem that traditional single data sources and static analysis models are difficult to accurately reflect the complex hydrological processes of watersheds, and improves the spatiotemporal accuracy of watershed runoff prediction and the scientific nature of decision support.

[0010] Preferably, step S1 includes the following steps:

[0011] Step S11: Install soil moisture sensors in the corresponding watershed hydrological units, with a sampling frequency of 1 / h, and collect data on the moisture content of the upper soil layer and the lower soil layer;

[0012] Step S12: Collect rainfall data at a time resolution of five minutes using a weather station to obtain watershed rainfall data;

[0013] Step S13: Obtain soil organic matter data from the laboratory; use UAV remote sensing to obtain the watershed vegetation cover index, and combine it with soil organic matter data to generate basic parameters for phosphorus adsorption coefficient.

[0014] Step S14: Perform spatiotemporal alignment processing on the moisture content data of the upper and lower soil layers, and use a 1-hour window moving average filter to eliminate noise, to obtain a standardized stratified soil moisture dataset.

[0015] Step S15: Compare the standardized stratified soil moisture dataset, watershed rainfall data, and basic parameters of phosphorus adsorption coefficient, and perform meteorological-soil-topography spatial mapping to obtain a three-dimensional stratified runoff generation basic dataset.

[0016] This invention deploys soil moisture sensors within watershed hydrological response units to acquire moisture content information of both upper and lower soil layers hourly. This not only enhances the temporal dynamics of water transfer along the vertical runoff pathway (surface runoff-interflow-groundwater) but also provides fundamental data for quantifying the contribution of vertical runoff. Simultaneously, the introduction of five-minute resolution rainfall data from meteorological stations significantly improves the response capability to sudden precipitation events, making the input rainfall data highly adaptable to short-term flood simulations. Regarding surface parameter modeling, the invention utilizes UAV remote sensing technology to extract the watershed vegetation cover index and integrates it with laboratory-measured soil organic matter content data to generate basic parameters for phosphorus adsorption coefficients. This data structure enables correlation modeling among vegetation, soil, and pollutants, allowing subsequent pollutant migration simulations to exhibit spatial heterogeneity. In the data processing stage… A mean filtering algorithm based on a sliding window (1-hour pane) was employed to denoise the moisture content data, ensuring that the data retained its fluctuation trend while reducing random disturbances. Further spatiotemporal alignment and interpolation of the moisture content and rainfall data ensured the integrity and consistency of the dataset across the time axis and spatial raster. Finally, three types of data—standardized stratified soil moisture data, meteorological and rainfall data, and basic parameters of phosphorus adsorption—were mapped to a unified spatial reference framework. By combining topographic factors such as elevation data and slope information, a spatial coupling mapping of meteorology, soil, and topography was achieved, constructing a three-dimensional stratified runoff generation dataset with vertical structure, spatial distribution, and temporal evolution characteristics. This significantly enhanced the model's ability to characterize the spatiotemporal heterogeneous response of runoff generation mechanisms, providing a high-density, highly consistent data foundation for subsequent runoff component analysis, topological network construction, and pollutant response prediction. This process effectively improved the input data's structural resolution, spatial registration accuracy, and physical semantic expression, significantly expanding the applicability of traditional two-dimensional static input in dynamic watershed modeling.

[0017] Preferably, step S2, which involves calculating the water allocation ratio based on the three-dimensional stratified runoff dataset, includes:

[0018] The water allocation ratio was calculated based on the three-dimensional hierarchical runoff generation dataset, and the formula for calculating the water allocation ratio is as follows.

[0019]

[0020] in This refers to the proportion of water entering the soil layer, which is also the data on water distribution ratio. This represents the moisture content of the upper soil layer.

[0021] This invention constructs a system based on soil moisture content. The nonlinear distribution function is used, and a composite exponent and piecewise weighted parameter structure are introduced to describe the proportion of water entering the deep layers of the soil profile. The formula is first based on soil moisture content data extracted from a three-dimensional stratified runoff generation dataset, and then... Introducing an index control mechanism as an input variable Combined terms with multiple factors This approach constructs a nested structure with nonlinear decay characteristics, thereby more realistically simulating the impact of soil moisture on runoff paths under different wetting levels. At the data level, this representation significantly outperforms traditional linear or threshold-based models, revealing more detailed dynamic changes in water infiltration into lower soil layers. The calculation process logically combines remote sensing estimation, field monitoring, and physical process simulation into a computable data flow structure, not only converting point-based data to spatially distributed data but also constructing a generalizable set of rules for distributed water infiltration simulation. The output of this formula... The data is essentially a runoff distribution factor resulting from the fusion of multiple weights, possessing three major data characteristics: continuity, differentiability, and parameter stability. It is suitable for various downstream model modules, including hydrological response unit division, runoff component coupling analysis, and soil moisture threshold determination. In practical applications, the dataset generated by this formula can not only directly participate in the construction of three-dimensional prediction data for flow-sediment-pollutant systems, but also provide high-resolution input for the decision-making logic of surface runoff and interflow paths. Therefore, from the perspectives of continuous data structure expression, nonlinear modeling capability, and computational compatibility, it exhibits high mathematical plasticity and system adaptability, serving as a crucial computational bridge connecting hydrological observation data and mechanistic modeling parameters.

[0022] Preferably, the vertical analysis of watershed runoff components in step S2 includes:

[0023] The total rainfall intensity per unit area was decomposed from the watershed rainfall data, and the characteristic curves of surface runoff and interflow were constructed to obtain the first characteristic curve and the second characteristic curve.

[0024] The vegetation interception layer-surface runoff layer analysis was performed on the water distribution ratio data using the first characteristic curve to obtain surface water storage data.

[0025] The second characteristic curve was used to perform upper soil layer-lower soil layer analysis on the water distribution ratio data to obtain groundwater storage data;

[0026] Obtain soil temperature data; determine the frozen soil state from the soil temperature data to obtain frozen soil water volume assessment data;

[0027] Based on the data on frozen soil water volume, the surface water volume and groundwater volume data are corrected by exponential decay to obtain the runoff composition data.

[0028] This invention decomposes the total rainfall intensity per unit area into a local hydraulic impact intensity index by performing a decomposition operation on high spatiotemporal resolution watershed rainfall data. Based on this, two types of water response curves, namely the first characteristic curve and the second characteristic curve, reflecting the response characteristics of surface runoff and interflow, are constructed. On this basis, the water distribution ratio data are subjected to hierarchical deconstruction analysis. The former is used to quantify the transformation relationship between vegetation interception and surface runoff, while the latter is used to analyze the accumulation characteristics of soil moisture during its migration from upper to lower layers in a vertical profile. This yields accumulation datasets for surface water and groundwater, significantly improving the granularity and resolution of the runoff composition data structure. Secondly, by introducing high-frequency soil temperature data and identifying frozen soil conditions, dynamic monitoring of the soil freezing process under low-temperature environments is achieved, thereby generating frozen soil water volume assessment data. This dataset not only serves as a basis for judging the state of runoff channel blockage but also is used to perform exponential decay correction on the accumulation data, reflecting the nonlinear influence of freezing conditions on water transport rates and conversion efficiency. Ultimately, surface water and groundwater storage data modulated by freeze-thaw factors are fused into runoff composition data, exhibiting dynamic adaptability and temporal continuity. At the data structure level, this process achieves multi-step fusion from raw rainfall-driven processes to response path selection and environmental factor regulation, expanding static water distribution parameters into dynamically updatable runoff description units with process response capabilities. This enhances the physical consistency and predictive adaptability of subsequent models in areas such as hydrological response, pollutant migration, and flow prediction.

[0029] Preferably, step S3 includes the following steps:

[0030] Step S31: Perform topology coding based on runoff composition data to obtain the topology coding of hydrological response units;

[0031] Step S32: Construct a multi-branch tree confluence network based on the topology coding of the hydrological response unit, and perform cross-sectional analysis of the entire watershed to obtain cross-sectional process data of the entire watershed.

[0032] Step S33: Monitor sediment transport capacity based on cross-sectional process data of the entire watershed, and perform closed-loop modeling to obtain the river runoff hydrological prediction model;

[0033] Step S34: Obtain real-time monitoring watershed runoff data; input the real-time monitoring watershed runoff data into the river runoff hydrological prediction model to predict runoff and obtain three-dimensional prediction data of flow rate, sediment load and pollutant concentration.

[0034] This invention constructs a network-based data structure that supports spatial logic analysis by transforming runoff component data into topological codes for hydrological response units. This transforms the previously unstructured spatial coupling relationships between topography, runoff paths, and response zones into a node-edge multi-branch tree network, giving the data structural characteristics of identifiable upstream and downstream levels and traceable flow direction, thus providing a foundation for spatial modeling of hydrological processes. Subsequently, a full-basin cross-sectional analysis is introduced into this topological framework. Based on confluence paths and hydrodynamic boundary conditions, attribute assignments and dynamic process simulations are performed on each node cross-section, forming a full-basin river cross-sectional process dataset. This dataset not only includes static parameters such as flow velocity, water depth, and cross-sectional morphology but also integrates dynamic process characteristics such as sediment concentration and cross-sectional shear stress, possessing the ability to jointly express multiple variables and multiple time series. The closed-loop construction process of sediment transport capacity monitoring and model modeling based on this dataset enables the river runoff hydrological prediction model to incorporate the co-evolution mechanism of three types of variables: sediment, flow, and pollutants, ultimately forming a three-dimensional prediction result of flow rate, sediment quantity, and pollutant concentration. Furthermore, by embedding real-time monitored watershed runoff data into this coupled model as a dynamic input source, a data input-output pathway supporting time-series updates can be formed, enhancing the system's responsiveness to external disturbances such as sudden rainfall, upstream pollutant input, or changes in underlying surface conditions. From a data perspective, this technical system achieves a closed-loop composite data flow of "structured topological coding + spatial process simulation + multivariate prediction output + real-time driven fusion," which not only improves the consistency of hydrological simulation results in terms of spatial distribution and temporal accuracy but also significantly expands the adaptive boundary of multi-objective collaborative modeling, providing a solid data foundation for constructing a cross-domain integrated, high-dimensional interactive hydrological-sediment-pollutant coupled prediction system.

[0035] Preferably, step S32 includes the following steps:

[0036] Step S321: Based on the parent node being the downstream river channel and the child node being the upstream unit, construct a multi-branch tree confluence network for the topological coding of the hydrological response unit to obtain the watershed multi-branch tree confluence network;

[0037] Step S322: Process the nodes of the watershed multi-branch tree confluence network by channel slope / roughness to obtain water level-discharge mapping data;

[0038] Step S323: When the soil temperature data is less than or equal to 0℃, monitor the soil freezing depth, and use 1 - soil freezing depth / 2 to correct for the frozen soil factor to obtain frozen soil water level mapping data.

[0039] Step S324: Combine the water level-discharge mapping data and the frozen soil water level mapping data with weights to obtain the cross-sectional discharge data of the entire watershed.

[0040] This invention reconstructs the topology of hydrological response units using a multi-branch tree structure encoding method, where parent nodes represent downstream confluence units and child nodes represent upstream runoff generation units. This allows for the encoding of original geographic information and runoff generation mechanisms into a network graph model in data structure form, creating data links with hierarchical relationships and flow direction indications. This enables path dependence and directional constraints in the spatial transmission of subsequent hydraulic and water quality parameters. Next, the Manning formula is used to normalize the channel slope and roughness parameters of each node in the topological network. Based on this normalization result, the numerical mapping relationship between water level and flow rate is derived, thus obtaining a water level-flow rate mapping dataset based on physical mechanism constraints. This data not only reflects the cross-sectional structure and hydraulic conditions but can also be used as an input response function in hydrodynamic prediction models. Furthermore, when the external ambient temperature is below 0℃, soil temperature data is introduced and the freezing depth is dynamically calculated, using an empirical weighting function 1. / 2 (of which The flow estimation results were corrected using the frozen depth as a permafrost influence factor, resulting in permafrost water level mapping data reflecting the impact of permafrost on river hydraulic transport. Finally, the two types of mapping data under normal and permafrost conditions were fused using dynamic weighting to generate a basin-wide river cross-section flow dataset with climate responsiveness. This dataset preserves the consistency of topographic-hydraulic structure and introduces an adaptation mechanism to extreme climate events through a temperature factor, providing a reliable, physically consistent, and iteratively updatable data foundation for downstream modules such as sediment migration modeling, water pollution diffusion prediction, and flow change early warning. In summary, from a data perspective, this process realizes a continuous modeling chain of "topology construction—hydraulic mapping—climate regulation—data fusion," significantly enhancing the overall expressive power of river cross-section flow estimation in terms of dynamism, spatiality, and environmental adaptability.

[0041] Preferably, step S33 includes the following steps:

[0042] Step S331: Perform multi-branch tree confluence analysis on the cross-sectional flow data of the entire basin and calculate the sediment transport capacity to obtain the sediment transport data of the entire basin;

[0043] Step S332: When the sediment transport data of the whole basin is greater than 25 kg / s, the phosphorus uptake module is started to calculate phosphorus release and obtain phosphorus desorption data; when the sediment transport data of the whole basin is less than or equal to 25 kg / s, water quality stabilization treatment is performed to obtain water quality stabilization data.

[0044] Step S333: Couple the phosphorus desorption data and water quality steady-state data with sediment-flow-pollutant output, construct a prediction model, and obtain a river runoff hydrological prediction model.

[0045] This invention employs multi-branch tree confluence analysis on cross-sectional flow data across the entire watershed. This not only identifies the main water conveyance channels between river nodes at various levels but also quantitatively models the distribution of sediment transport within the watershed's spatial structure. By integrating cross-sectional velocity, cross-sectional area, and roughness coefficient data at each node's convergence path, a transport flux model is constructed, generating spatially continuous and temporally consistent watershed-wide sediment transport data. Secondly, a threshold-based response mechanism is established. When the sediment transport capacity exceeds 25 kg / s, the phosphorus adsorption module is triggered, loading experimentally obtained phosphorus adsorption isotherm parameters and a flow velocity shear stress model to dynamically calculate the surface phosphorus release potential of sediment carriers of different particle sizes. This generates phosphorus desorption data, reflecting the active migration risk of sediment as a pollutant carrier. Conversely, when the sediment transport capacity is below or equal to the threshold, a water quality steady-state treatment module is executed to simulate the adsorption-sedimentation-dissociation processes of pollutants under low-disturbance conditions, outputting water quality steady-state data. Finally, the system couples phosphorus desorption data and steady-state water quality data with sediment flow data and river flow data, respectively, to construct a ternary synergistic factor structure of "flow-sediment-pollutant". Based on this structure, a multivariate response mapping relationship is trained to generate a coupled-driven river runoff hydrological prediction model. Compared with the traditional modeling approach that uses flow as the sole driving factor, this type of model integrates pollutant migration dynamics and hydrodynamic processes at the data level. This allows the predictive capability to simultaneously cover multi-field response characteristics under high-velocity disturbances and low-velocity steady-state conditions, thereby improving the simulation accuracy and generalization ability for runoff phenomena involving multiple inputs and multiple mechanisms. Overall, this process, from data acquisition, bifurcation discrimination, coupling output, and modeling closure, completes the numerical encoding and structural integration of complex multivariate processes, providing a solid foundation for building an intelligent hydrological prediction system that emphasizes both data-driven and mechanism-integrated approaches.

[0046] Preferably, step S4 includes the following steps:

[0047] Step S41: Visualize the three-dimensional prediction data of flow rate, sediment load, and pollutant concentration to obtain visualized composite data of river runoff.

[0048] Step S42: Perform sub-basin assessment based on the visualized composite data of river runoff to obtain sub-basin assessment data; perform parallel aggregation of the sub-basin assessment data throughout the entire process to obtain full-process assessment data of river runoff.

[0049] Step S43: Construct a full-process prediction report for river runoff based on the full-process assessment data.

[0050] This invention visualizes three-dimensional predicted data of flow rate, sediment load, and pollutant concentration, achieving the fusion and intuitive expression of multi-source data and enhancing data interpretability and insight. Secondly, it conducts detailed assessments of different sub-basins based on visualized composite data, demonstrating hierarchical and multi-scale data analysis capabilities, which helps reveal the runoff generation characteristics and differences of each sub-basin. Thirdly, it integrates the sub-basin assessment results using a full-process parallel aggregation method, ensuring efficient and complete data processing and avoiding information silos. Finally, by constructing a full-process river flow prediction report, it achieves a systematic summary and application of runoff generation data throughout the river process, providing scientific and comprehensive data support for decision-making. The entire process, from multi-dimensional data collection and fusion to comprehensive assessment and report generation, reflects the refined and intelligent level of data-driven river management and environmental monitoring.

[0051] Preferably, step S41 includes the following steps:

[0052] Step S411: Extract runoff components and spatial topology data based on the three-dimensional prediction data of flow rate, sediment amount and pollutant concentration, and perform Kriging interpolation of surface runoff, interflow and groundwater. Component fusion is performed according to a 1km×1km grid matrix to construct a dynamic runoff component map and obtain a dynamic runoff component distribution map.

[0053] Step S412: Obtain hydrological critical data; extract cross-sectional process-water quality data based on the three-dimensional prediction data of flow-sediment quantity-pollutant concentration, and perform multi-scale visualization of event marking using hydrological critical data to obtain the river flow-sediment co-process line;

[0054] Step S413: Extract pollutant-hydrodynamic data based on the three-dimensional prediction data of flow rate-sediment quantity-pollutant concentration, and perform spatial diffusion simulation to obtain a phosphorus pollution risk warning heat map;

[0055] Step S414: Visualize and encapsulate the dynamic runoff composition distribution map, the river flow-sediment co-process line, and the phosphorus pollution risk early warning heat map to obtain visualized composite data of river runoff.

[0056] This invention extracts runoff components and spatial topological information from three-dimensional prediction data based on flow rate, sediment load, and pollutant concentration. It then employs Kriging interpolation to perform high-precision spatial fusion of surface runoff, interflow, and groundwater components, constructing a dynamic runoff component distribution map in 1km×1km grid units. This achieves spatial refinement and temporal dynamic characterization of the runoff process, providing accurate spatial foundational data for subsequent quantitative analysis of hydrological processes. Secondly, by combining critical hydrological data, event labeling is applied to flow rate, sediment load, and water quality information at cross-sections. Multi-scale visualization methods reveal the coordinated changes in river flow and sediment, deepening the understanding of river dynamic response mechanisms and enhancing the ability to identify key hydrological events. Thirdly, based on the three-dimensional prediction data, pollutant-hydrodynamic coupling information is extracted, and spatial diffusion simulation is conducted to generate a phosphorus pollution risk warning heat map. This effectively reflects the migration and diffusion characteristics of pollutants in the river system and potential risk areas, providing a strong early warning function. Finally, the dynamic runoff composition distribution, flow-sediment synergistic process curve, and pollution risk heatmap—three types of multi-source, multi-dimensional information—are visualized and encapsulated into composite data. This enables integrated expression and interactive display of scene information, enhancing the comprehensiveness and intuitiveness of data representation. Overall, this process, through multi-level, multi-scale, and multi-physical process data extraction, fusion, and dynamic simulation, not only enhances the refined characterization of river runoff processes but also provides scientific data support for water environment risk monitoring and management decisions, demonstrating the advanced level of data-driven dynamic analysis and intelligent early warning of environmental systems.

[0057] This specification provides a watershed hydrological forecasting system with intelligent runoff model adaptation, used to execute the aforementioned watershed hydrological forecasting method with intelligent runoff model adaptation. The watershed hydrological forecasting system with intelligent runoff model adaptation includes:

[0058] The IoT data acquisition and basic data construction module is used to collect key hydrological data from IoT monitoring arrays deployed in the watershed and to construct a three-dimensional hierarchical runoff generation basic dataset.

[0059] The runoff composition analysis and proportion calculation module is used to calculate the water distribution proportion based on the three-dimensional stratified runoff basic dataset, and to perform vertical analysis of the watershed runoff composition to obtain runoff composition data.

[0060] The topology modeling and hydrological response prediction module is used to perform topological coding based on runoff composition data to obtain the topology code of the hydrological response unit; construct a multi-branch tree confluence network based on the topology code of the hydrological response unit, and perform cross-sectional analysis of the entire basin to obtain cross-sectional process data of the entire basin; monitor sediment transport capacity based on the cross-sectional process data of the entire basin, and perform closed-loop modeling to obtain the runoff hydrological prediction model of the river; and predict runoff based on the runoff hydrological prediction model of the river to obtain three-dimensional prediction data of flow rate, sediment load, and pollutant concentration.

[0061] The runoff visualization and assessment report module is used to visualize the three-dimensional prediction data of flow rate, sediment load, and pollutant concentration to obtain composite data of river runoff visualization; to conduct a full-process assessment based on the composite data of river runoff visualization to obtain full-process assessment data of river runoff; and to construct a full-process prediction report of river runoff based on the full-process assessment data of river runoff.

[0062] The beneficial effects of this invention are as follows: First, it expands hydrological monitoring data from single-point, two-dimensional planar sampling to a three-dimensional vertically layered structure. Through an IoT sensor array, it collects key parameters such as soil moisture, rainfall intensity, groundwater recharge, and surface runoff at high frequency. After mapping with a unified time window and spatial scale, it constructs a three-dimensional layered runoff generation dataset, effectively improving the expressive power of the spatial heterogeneity and vertical structural characteristics of the runoff generation process. Second, the system combines water distribution ratio analysis with vertical analysis of runoff components, refining the model based on the proportion of different runoff sources (such as surface runoff, interflow, and baseflow) within the watershed unit. This allows the flow simulation to not only reflect macroscopic trends but also the evolution of hydrological mechanisms within the watershed. Third, it reconstructs the spatial connection network between hydrological response units based on topological coding, characterizes the hydrodynamic coupling mode between different tributaries and the main channel through a multi-branch tree confluence structure, and introduces cross-sectional hydraulic analysis and sediment transport capacity modeling mechanisms to construct a... The closed-loop updated runoff prediction model possesses dynamic feedback adjustment capabilities in simulating sediment transport and pollutant release. Furthermore, by coupling the output three-dimensional prediction data of flow rate, sediment load, and pollutant concentration, the model introduces a visualization module to structurally encapsulate and graphically represent the multidimensional data, constructing multi-level composite image data represented by runoff composition maps, co-process lines, and pollution risk heatmaps. This provides quantitative support in sub-basin evaluation and full-process report generation, making the analysis results more intuitive, interconnected, and applicable to decision-making. Finally, this process establishes a complete chain from real-time monitoring, spatiotemporal modeling, basin extrapolation, pollution response, graphical encapsulation to analysis report construction, emphasizing the continuity, structural consistency, and semantic transferability of data across multiple scales and process nodes. This significantly improves the data expression depth, prediction accuracy, and comprehensive evaluation capabilities of runoff hydrological analysis, providing a solid data-driven foundation and closed-loop process characteristics for basin-scale hydrological modeling and pollution control decision support. Attached Figure Description

[0063] Figure 1 A flowchart illustrating the steps of a watershed hydrological forecasting method that intelligently adapts to runoff patterns;

[0064] Figure 2 for Figure 1 A detailed flowchart illustrating the implementation steps of step S4.

[0065] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0066] The technical method of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of this invention.

[0067] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0068] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0069] To achieve the above objectives, please refer to Figures 1 to 2 A watershed hydrological forecasting method with intelligent runoff generation pattern adaptation, the method comprising the following steps:

[0070] Step S1: Deploy an IoT monitoring array in the watershed to collect key hydrological data and construct a three-dimensional hierarchical runoff generation dataset;

[0071] Step S2: Calculate the water distribution ratio based on the three-dimensional stratified runoff basic dataset, and perform vertical analysis of watershed runoff components to obtain runoff component data;

[0072] Step S3: Perform topological coding based on runoff composition data to obtain the topological coding of hydrological response units; construct a multi-branch tree confluence network based on the topological coding of hydrological response units, and perform cross-sectional analysis of the entire basin to obtain cross-sectional process data of the entire basin; monitor sediment transport capacity based on the cross-sectional process data of the entire basin, and perform closed-loop modeling to obtain the runoff hydrological prediction model of the river; predict runoff based on the runoff hydrological prediction model of the river to obtain three-dimensional prediction data of flow rate-sediment quantity-pollutant concentration;

[0073] Step S4: Visualize the three-dimensional prediction data of flow rate, sediment load, and pollutant concentration to obtain visualized composite data of river runoff generation; conduct a full-process assessment based on the visualized composite data of river runoff generation to obtain full-process assessment data of river runoff generation; and construct a full-process prediction report of river runoff generation based on the full-process assessment data of river runoff generation.

[0074] In this embodiment of the invention, reference Figure 1 The diagram shown is a flowchart illustrating the steps of a watershed hydrological forecasting method with intelligent runoff pattern adaptation according to the present invention. In this example, the watershed hydrological forecasting method with intelligent runoff pattern adaptation includes the following steps:

[0075] Step S1: Deploy an IoT monitoring array in the watershed to collect key hydrological data and construct a three-dimensional hierarchical runoff generation dataset;

[0076] In this embodiment of the invention, the watershed is divided into layers based on remote sensing interpretation, DEM topographic analysis, and geomorphological zoning techniques. On this basis, different hydrological units are modeled in three dimensions according to factors such as elevation, soil type, vegetation cover, and slope aspect to determine the role of each layer in the runoff generation process. For data acquisition, an Internet of Things (IoT) sensing system is constructed using a distributed wireless sensor network (WSN) deployed at each layered cross-section and confluence node to collect parameters including rainfall intensity, soil moisture content, infiltration rate, surface runoff, groundwater level, water temperature, conductivity, and dissolved oxygen, forming multi-dimensional dynamic hydrological information covering the surface, root, shallow, and deep layers. The sensor-collected data is preprocessed locally using an embedded edge computing module, and after unifying units and timestamps, it is uploaded in real-time to the central hydrological data platform, constructing a time-space synchronized raw data stream. To ensure data comparability and spatial consistency, a multi-source data fusion algorithm is used to fuse and correct the sensed data with remote sensing images, historical data from hydrological stations, and measured samples. Subsequently, spatial clustering methods supported by machine learning (such as K-means and density peak clustering) were used to aggregate regional features of the collected data, forming a spatial distribution grid. Based on this, the runoff initiation threshold and response curves of different spatial units under different rainfall intensities were derived, thereby generating a three-dimensional hierarchical runoff base dataset with spatial hierarchical attributes and response weights. This dataset is used in subsequent model construction to simulate the response intensity, runoff path, and water contribution rate of hydrological units at each level to precipitation, and is the core data foundation supporting vertical water allocation and topology modeling.

[0077] Step S2: Calculate the water distribution ratio based on the three-dimensional stratified runoff basic dataset, and perform vertical analysis of watershed runoff components to obtain runoff component data;

[0078] In this embodiment of the invention, using the three-dimensional spatial grid data and hierarchical attribute information generated in the previous stage, hydrological response factor datasets are extracted from different vertical profiles, including but not limited to indicators such as surface runoff rate, soil moisture gradient, profile infiltration capacity, capillary water upwelling flux, and interflow and baseflow components. By curve fitting and numerical integration of the temporal evolution trends of these indicators in different layers, the response lag time and response amplitude of each hydrological unit to precipitation events are calculated, thereby deriving the distribution ratio of unit water volume among the surface layer, root layer, shallow layer, and deep layer. Subsequently, the above data are parameterized using a vertical stratified analytical model (such as a multi-layer runoff response factor superposition model) to form an inter-layer flow transfer matrix, and combined with rainfall-driven input, the dynamic migration path of water volume in the vertical profile is simulated. Based on this, by constructing a stratified water flux map, the vertical migration behavior of water volume in the watershed under specific time periods or rainfall patterns is visualized, thereby obtaining the proportion of runoff capacity of each layer under specific topographic units, soil types, and vegetation cover conditions. Furthermore, a clustering algorithm based on multi-objective optimization is introduced to reclassify response units with high runoff contribution in the profile structure, identifying profile clusters with similar hydrological response characteristics, and further refining the runoff component structure. Finally, the above results are transformed into a standardized runoff component database, clarifying the order-of-magnitude relationships and temporal coupling characteristics of surface runoff, interflow, and groundwater baseflow in each runoff path, providing a refined input basis for subsequent topology modeling and hydrological process extrapolation. This process emphasizes the uniformity of data hierarchy, the traceability of spatiotemporal evolution, and the resolvability of response components, and is a crucial prerequisite for hydrological simulation to move towards high-resolution dynamic coupled calculations.

[0079] Step S3: Perform topological coding based on runoff composition data to obtain the topological coding of hydrological response units; construct a multi-branch tree confluence network based on the topological coding of hydrological response units, and perform cross-sectional analysis of the entire basin to obtain cross-sectional process data of the entire basin; monitor sediment transport capacity based on the cross-sectional process data of the entire basin, and perform closed-loop modeling to obtain the runoff hydrological prediction model of the river; predict runoff based on the runoff hydrological prediction model of the river to obtain three-dimensional prediction data of flow rate-sediment quantity-pollutant concentration;

[0080] In this embodiment of the invention, the generated runoff component dataset is spatially partitioned into hydrological units with spatial coordinates, vertical attributes, and runoff weights. Based on topographic elevation data (DEM), slope and aspect information, slope confluence paths, and river network density indices, a unique identifier is assigned to each runoff unit. This identifier is then encoded using hierarchical topological identification rules. The encoding method comprehensively considers the direction of the flow path, hydrological connectivity, vertical runoff layer, and response time window, ensuring the traceability of the relative position and connection order of each unit within the topological structure. After encoding, a multi-branch tree-type confluence network is constructed based on the directed acyclic graph (DAG) construction rules in graph theory, utilizing the encoding results of these hydrological response units. Each node in the network represents a hydrological unit, and each edge represents the hydrological confluence relationship between adjacent units. The transport relationship of water and matter between nodes is defined through a flow direction weight matrix. Time lag functions and storage / discharge function models are introduced into this confluence network to describe the evolution and convergence characteristics of water flow along the transport path. Subsequently, dynamic simulations were performed on the entire network structure to identify the response efficiency and transport stability of the confluence paths under different runoff generation conditions. After the network was constructed, the cross-sectional morphology of the main channel and its tributaries was reconstructed by combining hydraulic geometric profile data and historical channel cross-sectional measurement data through numerical interpolation and three-dimensional spatial fitting algorithms, thereby obtaining full-basin cross-sectional process data, including the water depth-velocity-bottom elevation relationship. Based on this dataset, particulate matter transport equations and moving bed boundary models were further introduced to evaluate the sediment carrying capacity and deposition dynamics of the channel system under different hydrological drives, providing structured input for subsequent model closed-loop training. In this process, hierarchical modeling of the hydrodynamic and material transport processes of the channel system was achieved at the data level through the quantitative expression of runoff topology and the orderly reconstruction of the confluence network.

[0081] Step S4: Visualize the three-dimensional prediction data of flow rate, sediment load, and pollutant concentration to obtain visualized composite data of river runoff generation; conduct a full-process assessment based on the visualized composite data of river runoff generation to obtain full-process assessment data of river runoff generation; and construct a full-process prediction report of river runoff generation based on the full-process assessment data of river runoff generation.

[0082] In this embodiment of the invention, after obtaining the three-dimensional prediction data, high-dimensional feature separation is performed on the information contained in the dataset, such as flow changes, sediment particle concentration distribution, and pollutant diffusion paths. Principal component analysis (PCA) or t-SNE dimensionality reduction methods are used to extract the dominant factors of temporal changes and spatial evolution features. Subsequently, different prediction variables are synchronized in time and aligned in space to construct a unified dynamic data raster structure with "time-space-variable" three axes. Interpolation modeling is used to perform continuous processing on the data between cross sections, filling the time difference and spatial gaps in the prediction data in the hydrological response lag region. On this basis, a GPU-accelerated three-dimensional visualization engine is introduced to construct a dynamic flow field rendering model. The water flow velocity vector, sediment migration trajectory, and pollutant concentration gradient are presented in parallel through streamline diagrams, particle path diagrams, and isosurface diagrams, respectively, to realize a multi-scale, interactive dynamic map of river runoff. The above map is integrated with multi-parameter indicators. The interaction intensity and synergistic effect of each variable at the river cross section are measured by information entropy and structural complexity indicators, thereby constructing a comprehensive indicator system for evaluation. During the assessment phase, Bayesian inference-based process uncertainty analysis was employed to quantify the stability and deviation range of the predicted data under different simulation conditions. This was combined with historical observation data for backtesting of deviations, resulting in a comprehensive assessment that reflects both the overall hydrological evolution trend and the characteristics of local extreme events. Finally, the data was structured and integrated using charts, text descriptions, and interactive layers to automatically generate a comprehensive river runoff prediction report covering start and end times, key sections, hydrodynamic processes, sediment changes, pollutant diffusion paths, model bias analysis, and control recommendations. This achieved a closed-loop processing from raw prediction data to the evaluation report. This process, from data structure unification, variable dimensionality reduction encoding, dynamic map rendering to process assessment modeling, is all based on structured data, ensuring an orderly connection between visual representation, logical evaluation, and result output.

[0083] Preferably, step S1 includes the following steps:

[0084] Step S11: Install soil moisture sensors in the corresponding watershed hydrological units, with a sampling frequency of 1 / h, and collect data on the moisture content of the upper soil layer and the lower soil layer;

[0085] Step S12: Collect rainfall data at a time resolution of five minutes using a weather station to obtain watershed rainfall data;

[0086] Step S13: Obtain soil organic matter data from the laboratory; use UAV remote sensing to obtain the watershed vegetation cover index, and combine it with soil organic matter data to generate basic parameters for phosphorus adsorption coefficient.

[0087] Step S14: Perform spatiotemporal alignment processing on the moisture content data of the upper and lower soil layers, and use a 1-hour window moving average filter to eliminate noise, to obtain a standardized stratified soil moisture dataset.

[0088] Step S15: Compare the standardized stratified soil moisture dataset, watershed rainfall data, and basic parameters of phosphorus adsorption coefficient, and perform meteorological-soil-topography spatial mapping to obtain a three-dimensional stratified runoff generation basic dataset.

[0089] In this embodiment of the invention, soil moisture sensors are deployed within the watershed hydrological response unit to collect changes in the moisture content of the upper and lower soil layers at a sampling frequency of 1 hour, forming a dual-layer time-series moisture data stream. In step S12, rainfall data is collected at a time step of five minutes using high-frequency meteorological station equipment to obtain a rainfall-driven sequence corresponding to the soil response process. In step S13, organic matter content data of soil samples at the sampling points are obtained through laboratory analysis. Simultaneously, a UAV equipped with multispectral imaging equipment is used to obtain high spatial resolution vegetation cover index (NDVI or EVI). This is coupled with soil organic matter through a nonlinear regression model (such as support vector regression or radial basis neural network) to generate a spatially continuous phosphorus adsorption coefficient parameter grid. This coefficient can characterize the soil's ability to retain and migrate dissolved phosphorus in water, and has important significance as a parameter of the water quality-water quantity linkage mechanism. In step S14, for moisture content data from different sources, time-series alignment is first performed based on timestamps. A 1-hour fixed-pane moving average model is constructed to smooth high-frequency oscillations and filter out discrete noise caused by sensor sensitivity or environmental transients, forming a standardized stratified soil moisture dataset with temporal stability and spatial consistency. This dataset retains the differences in moisture content between upper and lower layers in its dimensional structure, forming the basis for characterizing the vertical moisture gradient response. Finally, in step S15, combining standardized soil moisture, time-series rainfall-driven data, and the spatial layer of phosphorus adsorption coefficient, a data spatial mapping method based on three-dimensional raster partitioning is used to construct a coupled data structure between meteorological factors, soil response, and topographic attributes. Among them, topographic parameters such as slope, aspect, and runoff accumulation are extracted from the DEM. Using this as a spatial reference frame, multi-source data are registered at the cell level and spatially interpolated to form a three-dimensional stratified, time-series-driven, and attribute-unified runoff generation dataset, providing a spatial-physical integrated input basis for subsequent runoff generation process simulation and pollutant migration coupled analysis. The process employs a multi-source data synchronization mechanism, spatial fusion algorithm, and physical attribute estimation model in synergy to ensure data consistency and resolvability in the vertical, horizontal, and temporal dimensions.

[0090] Preferably, step S2, which involves calculating the water allocation ratio based on the three-dimensional stratified runoff dataset, includes:

[0091] The water allocation ratio was calculated based on the three-dimensional hierarchical runoff generation dataset, and the formula for calculating the water allocation ratio is as follows.

[0092]

[0093] in This refers to the proportion of water entering the soil layer, which is also the data on water distribution ratio. This represents the moisture content of the upper soil layer.

[0094] In this embodiment of the invention, a water distribution ratio estimation model is constructed using mathematical analytical functions. This method uses the soil moisture content parameter... Using functional relationships as the main input variables The infiltration ratio was estimated under different saturation conditions, among which... This represents the proportion of water entering the soil layer, i.e., the percentage of effective water entering the soil after precipitation passes through the surface layer per unit area. This formula essentially simulates the soil's response to precipitation using a nonlinear piecewise function, considering the diminishing marginal returns of soil water absorption under different moisture contents, especially the nonlinear response where the infiltration proportion tends to saturate under high moisture contents. In data processing, soil moisture content data for the corresponding time step and spatial unit must first be extracted from the stratified soil moisture dataset. Typically, this is done pixel-by-pixel reading with an hourly temporal resolution and sub-watershed units or grid units as the spatial resolution. The read data needs to be standardized (e.g., volumetric moisture content converted to a 0-1 range) to ensure consistency of model parameter input. Then, the above formula is substituted into each spatiotemporal unit for batch calculation. The distribution values ​​are used to form a time-series water distribution ratio dataset. This dataset not only contains the distribution characteristics of water between the surface and soil layers after each precipitation event, but also reflects the dynamic changes in soil water storage capacity. It can be used to support subsequent processes such as runoff path tracking, interflow formation identification, and groundwater recharge estimation. In terms of calculation strategy, to avoid numerical instability caused by the denominator approaching zero when the moisture content is close to 3 or 4, boundary conditions are usually set for processing, such as... The values ​​are limited to a reasonable range (e.g., 0.05 to 0.95), or linear interpolation is used to smooth out outliers. Overall, this method realizes the transformation of static soil moisture state parameters into a function mapping of dynamic water migration ratios, which is an important data transformation link in the coupled hydrological response process.

[0095] Preferably, the vertical analysis of watershed runoff components in step S2 includes:

[0096] The total rainfall intensity per unit area was decomposed from the watershed rainfall data, and the characteristic curves of surface runoff and interflow were constructed to obtain the first characteristic curve and the second characteristic curve.

[0097] The vegetation interception layer-surface runoff layer analysis was performed on the water distribution ratio data using the first characteristic curve to obtain surface water storage data.

[0098] The second characteristic curve was used to perform upper soil layer-lower soil layer analysis on the water distribution ratio data to obtain groundwater storage data;

[0099] Obtain soil temperature data; determine the frozen soil state from the soil temperature data to obtain frozen soil water volume assessment data;

[0100] Based on the data on frozen soil water volume, the surface water volume and groundwater volume data are corrected by exponential decay to obtain the runoff composition data.

[0101] In this embodiment of the invention, the watershed rainfall data is normalized per unit area to calculate the total rainfall intensity. An empirical function relating unit rainfall intensity to runoff response is constructed using historical observation data. The main characteristic trends of the watershed response are extracted using regression analysis or kernel density estimation, resulting in dual-characteristic response curves for surface runoff and interflow, denoted as the first characteristic curve and the second characteristic curve, respectively. The first characteristic curve characterizes the flow of rainfall through the vegetation interception layer and the surface runoff layer. It maps the accumulation and confluence of water at the surface layer through a functional relationship between rainfall intensity and surface runoff generation rate. By substituting water distribution ratio data into this curve, surface water accumulation data for each spatial unit under current rainfall conditions is obtained, forming a static distribution map of the contribution to surface runoff. The second characteristic curve characterizes the water transport mechanism between upper and lower soil water layers. A water infiltration capacity model is established based on the relationship function between soil infiltration rate and water content gradient to derive the accumulation capacity of the groundwater system. By combining water distribution ratio data with the second characteristic curve, the infiltration volume and storage capacity of each vertical unit are calculated, thus forming a groundwater accumulation dataset. Based on this, to introduce the impact mechanism of seasonally frozen soil on hydrological processes, soil temperature data is incorporated into the system. The freeze-thaw critical temperature discrimination method (usually setting the critical threshold to 0℃) is used to classify low-temperature periods in the dataset, labeling frozen state spatial units. The degree of freezing is calculated using an empirical model of frozen layer thickness and soil thermal conductivity, ultimately forming frozen soil water volume assessment data. This data is used to determine the blocking effect of frozen soil on infiltration and accumulation processes, and the response coefficients of surface water accumulation data and groundwater accumulation data are corrected using an exponential decay function to obtain the attenuation of runoff paths under different temperature conditions. Finally, the corrected surface and groundwater accumulation data are reintegrated to obtain a runoff component dataset with clear hierarchical classification and dynamic response, providing a structurally consistent and physically meaningful input data foundation for subsequent hydrological simulation and response models.

[0102] Preferably, step S3 includes the following steps:

[0103] Step S31: Perform topology coding based on runoff composition data to obtain the topology coding of hydrological response units;

[0104] Step S32: Construct a multi-branch tree confluence network based on the topology coding of the hydrological response unit, and perform cross-sectional analysis of the entire watershed to obtain cross-sectional process data of the entire watershed.

[0105] Step S33: Monitor sediment transport capacity based on cross-sectional process data of the entire watershed, and perform closed-loop modeling to obtain the river runoff hydrological prediction model;

[0106] Step S34: Obtain real-time monitoring watershed runoff data; input the real-time monitoring watershed runoff data into the river runoff hydrological prediction model to predict runoff and obtain three-dimensional prediction data of flow rate, sediment load and pollutant concentration.

[0107] In this embodiment of the invention, based on runoff composition data, topological coding is performed on each hydrological response unit. This coding process constructs a multi-field topological structure by introducing node attributes (such as surface runoff ratio, interflow ratio, baseflow contribution, etc.) and spatial location indexes. Directed edges and hierarchical indexes from graph theory are used to describe the connection relationships between runoff generating units, ultimately outputting a topological coding result with spatial hierarchy and functional attributes. Step S32 utilizes this topological structure to construct a multi-branch tree-like runoff network through terrain flow direction analysis, runoff accumulation calculation, and river network extraction supported by DEM. This network structurally allows a node to simultaneously receive inflows from multiple upstream nodes, reflecting the true multi-source runoff characteristics of the watershed. Subsequently, combining the cross-sectional geometric information of each runoff node, such as cross-sectional elevation, width-to-depth ratio, water surface morphology, and bed particle size, numerical fitting methods (such as spline interpolation or TIN interpolation) are used to reconstruct the cross-sectional morphology, thereby constructing a full-watershed cross-sectional process dataset. In step S33, based on cross-sectional data and combined with empirical formulas for sediment carrying capacity per unit volume (such as the Meyer-Peter or Engelund-Hansen equations) and hydrodynamic boundary conditions, a sediment transport simulation module is established and embedded into the river runoff model. The model is then calibrated through iterative training and error minimization methods, forming a hydrological prediction model with a self-feedback closed-loop mechanism. This model not only outputs flow information but also embeds response equations for sediment carrying capacity and pollutant concentration, achieving joint prediction of three variables. Finally, in step S34, data streams from the real-time hydrological monitoring system within the basin are introduced to acquire high-frequency data inputs, including river velocity, flow rate, turbidity, conductivity, and rainfall. After timestamp alignment and spatial mapping, these data are imported into the model structure as driving variables. The model runs in real time, generating three-dimensional prediction results of flow rate, sediment load, and pollutant concentration at the corresponding time step, forming a spatiotemporally consistent and index-linked river runoff prediction dataset, providing direct data support for dynamic regulation and early warning. The entire process, from structural topology construction to hydrological element prediction, relies on data-driven approaches, functional expressions, and the coupling of physical mechanisms. It serves as the technical support framework for the evolution of hydrological process modeling towards an integrated flow-sediment-matter model.

[0108] Of particular importance, step S31 includes:

[0109] Step S311: Divide the runoff generation units based on the runoff generation component data to obtain the runoff generation component unit division objects;

[0110] Step S312: Perform HRU spatial connectivity analysis on the generated flow component unit division objects to obtain generated flow topology data;

[0111] Step S313: Based on the runoff topology data, perform tree structure coding and hierarchical nested raster coding to obtain the topology coding of the hydrological response unit.

[0112] In this embodiment of the invention, by analyzing the spatial attributes, vertical distribution characteristics, and hydrological factor response relationships of runoff component data, the distribution characteristics of components such as surface runoff, interflow, and groundwater recharge are identified. Based on this, a spatial division operation of runoff-generating units is performed to form runoff component unit division objects with hydrological response characteristics. Subsequently, based on the division results, spatial adjacency analysis between hydrological response units (HRUs) is further introduced. By calculating the spatial connection weight matrix, flow direction logical path, and response time sequence between units, a spatial relationship network between units is obtained. This network not only reflects the physical connection structure in the runoff generation process but also retains the logical association between hydrodynamic and material transport paths. On this basis, a tree structure is constructed using graph theory to encode this spatial relationship network. A hierarchical nested system of trunk-branch-terminal is constructed based on the flow direction trunk, and a regularized rasterization method is used to number different levels to form a multi-dimensional hydrological response unit topology code with spatial location, connection order, and response attributes. This topological coding not only includes spatial location information but also embeds transport paths and runoff levels in hydrological processes, providing a structured data foundation and logical framework for subsequent watershed runoff simulation, cross-sectional process tracking, and flow trend prediction. This process starts with raw runoff data, proceeds through spatial discretization and connectivity modeling to structural coding, forming a rigorous data flow path and topological abstraction system, achieving an efficient mapping from physical space to logical structure.

[0113] Preferably, step S32 includes the following steps:

[0114] Step S321: Based on the parent node being the downstream river channel and the child node being the upstream unit, construct a multi-branch tree confluence network for the topological coding of the hydrological response unit to obtain the watershed multi-branch tree confluence network;

[0115] Step S322: Process the nodes of the watershed multi-branch tree confluence network by channel slope / roughness to obtain water level-discharge mapping data;

[0116] Step S323: When the soil temperature data is less than or equal to 0℃, monitor the soil freezing depth, and use 1 - soil freezing depth / 2 to correct for the frozen soil factor to obtain frozen soil water level mapping data.

[0117] Step S324: Combine the water level-discharge mapping data and the frozen soil water level mapping data with weights to obtain the cross-sectional discharge data of the entire watershed.

[0118] In this embodiment of the invention, based on the topological coding of the hydrological response unit, a hierarchical relationship is established where parent nodes represent downstream main channels and child nodes represent upstream tributaries or sub-units. A directed graph structure is used to construct a watershed multi-branch tree confluence network. This forms a multi-branch tree network model in terms of data structure, composed of node identifiers, node attributes (such as area, runoff ratio, topographic slope, etc.) and work weights (such as flow direction, lag time, path length), ensuring the hierarchy, directionality, and spatial consistency of each confluence path. Subsequently, in step S322, for each level of nodes in the confluence network, the Manning formula is applied... Perform hydraulic calculations, including the slope With roughness The control factors were obtained through digital elevation model (DEM) analysis and land use type inference, respectively, for cross-sectional area. and hydraulic radius The cross-sectional geometry reconstruction module in step S32 provides the data, ultimately establishing a mapping function between water level and flow rate on a node-by-node basis, generating a hydraulic parameter table containing the corresponding flow rates under different water level conditions, i.e., water level-flow rate mapping data. Step S323 introduces a correction mechanism for the influence of temperature on the hydrological process. When the soil temperature data is less than or equal to 0°C, the soil freezing depth is obtained through real-time monitoring or empirical models, and the hydraulic response is adjusted using the mechanism of frozen soil inhibiting infiltration. Specifically, an exponential decay correction coefficient is used. ,in To standardize the soil freezing depth, water level responses under different freezing depths are adjusted, resulting in frozen soil water level mapping data that reflects the mitigation characteristics of watershed flow generation under flow-freezing conditions. Step S324 then performs weighted fusion of the two types of water level-discharge data obtained in steps S322 and S323, using linear combination or Bayesian weight update methods. The weight coefficients are determined based on historical observation accuracy or model fitting error, ultimately outputting a full-basin river cross-section discharge dataset containing three-dimensional coordinates of spatial location, water level, and discharge. This dataset possesses high spatiotemporal resolution, climate adaptability, and structural connectivity, providing a structurally complete hydraulic input for subsequent sediment transport simulation and pollutant load analysis. The overall process consists of spatial network construction, hydraulic function mapping, freeze-thaw effect correction, and data integration, emphasizing the ability to model and meter hydrodynamic elements under complex terrain and dynamic meteorological conditions.

[0119] Preferably, step S33 includes the following steps:

[0120] Step S331: Perform multi-branch tree confluence analysis on the cross-sectional flow data of the entire basin and calculate the sediment transport capacity to obtain the sediment transport data of the entire basin;

[0121] Step S332: When the sediment transport data of the whole basin is greater than 25 kg / s, the phosphorus uptake module is started to calculate phosphorus release and obtain phosphorus desorption data; when the sediment transport data of the whole basin is less than or equal to 25 kg / s, water quality stabilization treatment is performed to obtain water quality stabilization data.

[0122] Step S333: Couple the phosphorus desorption data and water quality steady-state data with sediment-flow-pollutant output, construct a prediction model, and obtain a river runoff hydrological prediction model.

[0123] In this embodiment of the invention, the cross-sectional flow data of the entire river basin obtained in the previous steps is analyzed for confluence path analysis according to the multi-branch tree structure of the confluence network. Based on the flow contribution ratio and time delay function of each child node, the cumulative cross-sectional flow data of each section within a specified time window is calculated. Subsequently, a sediment transport capacity calculation model is introduced. Semi-empirical formulas based on the sediment transport rate per unit width (such as the Engelund-Hansen or Yang model) are often used to fit the sediment transport capacity under different flow velocities, slopes, and water depths in segments, obtaining cross-sectional sediment transport data in kg / s, and integrating them to form a spatially distributed sediment transport capacity dataset covering the entire river basin. After entering step S332, the system uses 25 kg / s as an empirical criterion to perform threshold determination on the sediment transport data. When the value exceeds the threshold, it indicates a significant physical erosion process. The system automatically activates the phosphorus adsorption module and loads a pre-established phosphorus desorption kinetic equation. This equation estimates the concentration of active phosphorus released per unit volume of sediment particles per unit time based on factors such as particle size distribution, phosphorus loading capacity, temperature, and pH, generating a phosphorus desorption dataset. When the sediment transport capacity is less than or equal to 25 kg / s, it indicates that the system is in a relatively steady state of water quality. At this time, a stable distribution function (such as a mimicry normal or gamma distribution) is called to model the fluctuation range of pollutant concentration in the water, forming a steady-state water quality dataset. This data represents the range of concentration changes caused by the system's adsorption, retention, and dilution behavior of pollutants under low-energy perturbations. In step S333, the above two types of data are coupled according to the correspondence between sediment concentration, flow rate change, and pollutant release concentration to construct a data tensor with a ternary output structure, which is used as training input to build a river runoff hydrological prediction model. The model structure can be either data-driven (such as LSTM time series models or Transformer structures) or physically constrained (such as distributed water quality response equations). It fits the dynamic relationships between flow changes, sediment concentration, and pollutant concentration through supervised learning, and embeds a feedback adjustment mechanism to minimize prediction errors. Ultimately, this results in a river runoff hydrological prediction model with real-time extrapolation capabilities, cross-variable collaborative capabilities, and complex process response capabilities, providing structured support for subsequent water quality risk analysis and ecological regulation. The entire process is driven primarily by flow, with sediment as the medium and pollutants as the response variable, emphasizing the integrity of the technical chain and the coherence of the data structure in constructing the coupled model from high-frequency process data.

[0124] As an example of the present invention, reference is made to Figure 2 As shown, step S4 in this example includes:

[0125] Step S41: Visualize the three-dimensional prediction data of flow rate, sediment load, and pollutant concentration to obtain visualized composite data of river runoff.

[0126] Step S42: Perform sub-basin assessment based on the visualized composite data of river runoff to obtain sub-basin assessment data; perform parallel aggregation of the sub-basin assessment data throughout the entire process to obtain full-process assessment data of river runoff.

[0127] Step S43: Construct a full-process prediction report for river runoff based on the full-process assessment data.

[0128] In this embodiment of the invention, a three-dimensional tensor prediction dataset is constructed using three variables: flow rate, sediment load, and pollutant concentration. The dataset is then structurally reorganized along three axes: time (t), spatial location (x, y), and index (v). A multi-layered nested graphical rendering strategy is used to visualize the data. Multiple image representation methods are employed, including spatial heat maps, water level profile streamline maps, sediment concentration cloud maps, and pollutant isosurface maps, to encode the hydrodynamic response, sediment distribution characteristics, and pollutant evolution trends during river runoff generation in a multi-scale, multi-variable manner. During visualization, cross-sectional topographic data and GIS geographic elements are combined to construct a composite layer dataset that integrates spatial and attribute data, forming a composite visualized data set for river runoff generation. This dataset possesses a structural interface that can be directly loaded onto visualization terminal platforms (such as WebGIS or C / S structured hydrological analysis platforms). After proceeding to step S42, the system spatially divides the visualized composite data of river runoff into sub-basin units. Taking each sub-basin as the analysis unit, it extracts statistical characteristics such as average, extreme values, and fluctuation coefficients at different time steps within its scope. Combined with the established ecological, hydrological, and water quality evaluation standards, a multi-indicator evaluation matrix is ​​constructed to form a sub-basin evaluation dataset. Subsequently, through a full-process parallel aggregation strategy designed based on DAG (Directed Acyclic Graph) or MapReduce mechanisms, the evaluation results of multiple sub-basins are processed to perform indicator normalization, time step synchronization, and process status classification. The evaluation results of all sub-basins are structurally aggregated while maintaining temporal consistency, ultimately generating full-process evaluation data of river runoff, forming an integrated evaluation result with phased characteristics, spatial integrity, and logical closure. Step S43, based on this assessment data, initiates the automated report generation module. Through template-driven text filling and image embedding mechanisms, it invokes historical process feature libraries, key period extraction algorithms, and anomaly indicator detection logic to summarize the overall runoff generation process of the watershed in stages, regions, and elements. Ultimately, it constructs a full-process river flow prediction report including charts, flowcharts, spatial distribution maps, risk assessment items, and textual conclusions, possessing visualization and structured expression capabilities directly applicable to decision analysis. This technical approach fully embodies the data expression, evaluation, and summarization strategy for hydrological prediction results from data generation to decision support, exhibiting high information compression, spatial fusion, and logical hierarchy traceability.

[0129] Preferably, step S41 includes the following steps:

[0130] Step S411: Extract runoff components and spatial topology data based on the three-dimensional prediction data of flow rate, sediment amount and pollutant concentration, and perform Kriging interpolation of surface runoff, interflow and groundwater. Component fusion is performed according to a 1km×1km grid matrix to construct a dynamic runoff component map and obtain a dynamic runoff component distribution map.

[0131] Step S412: Obtain hydrological critical data; extract cross-sectional process-water quality data based on the three-dimensional prediction data of flow-sediment quantity-pollutant concentration, and perform multi-scale visualization of event marking using hydrological critical data to obtain the river flow-sediment co-process line;

[0132] Step S413: Extract pollutant-hydrodynamic data based on the three-dimensional prediction data of flow rate-sediment quantity-pollutant concentration, and perform spatial diffusion simulation to obtain a phosphorus pollution risk warning heat map;

[0133] Step S414: Visualize and encapsulate the dynamic runoff composition distribution map, the river flow-sediment co-process line, and the phosphorus pollution risk early warning heat map to obtain visualized composite data of river runoff.

[0134] In this embodiment of the invention, the corresponding runoff generation component information is extracted from each spatiotemporal unit in the three-dimensional prediction data of flow-sediment-pollutant concentration, and classified according to their belonging relationship in the topological network (such as belonging to surface runoff path, interflow lag unit or groundwater recharge area); then, the Kriging interpolation method is used to spatially interpolate the simulated values ​​of surface runoff, interflow and groundwater respectively to construct their respective prediction spatial fields, and then the three types of components are fused according to weighted rules to generate a multi-level data structure based on a 1km×1km regular grid, and finally a dynamic runoff generation component distribution map is formed, which reflects the superposition effect of multiple runoff generation mechanisms changing over time on the grid. Step S412 introduces critical hydrological data, including nationally or locally defined warning thresholds (such as flood peaks, sediment surges, and pollutant concentration limits). Based on the river cross-section flow, sediment concentration, and water quality index data in the three-dimensional prediction data, a cross-sectional process line diagram is constructed, and critical event nodes are highlighted or marked with symbols to achieve multi-scale visualization of hydrodynamics and water quality, thereby generating a river flow-sediment co-process line reflecting the changing trends during key periods. Step S413, targeting the pollutant concentration component in the prediction data, combines hydrodynamic data (such as flow velocity, diffusion coefficient, and turbulence intensity) to establish a two-dimensional spatial diffusion equation set (such as based on the ADE model) to simulate the migration and distribution of pollutants in the river network. By calculating the cumulative pollutant concentration in each grid and superimposing the flow velocity vector field, a pollution risk heat map with phosphorus as the indicator is finally generated. This map reflects the distribution pattern and evolution trend of high-risk pollution areas. Finally, in step S414, the three types of visualization outputs are encapsulated using layer overlay and unified coordinate mapping, integrating them into a composite data structure that includes dynamic graphics (such as GIS time series maps), spatial heatmaps, and response curves. This generates interactive, interconnected, and data-traceable composite visualization data of river runoff, providing consistent data input and graphical support for subsequent sub-basin assessments and comprehensive report construction. The entire process, from spatiotemporal data extraction to layer construction, involves multiple data processing techniques, including statistical interpolation, hydrological feature identification, diffusion modeling, and visual fusion, ensuring the dynamic and accurate presentation of various hydrological processes within the same representation space.

[0135] Of particular importance, step S42 includes:

[0136] Step S421: Based on the visualized composite data of river runoff generation, sub-basins are divided and data are correlated to obtain the river runoff generation sub-basin correlation data;

[0137] Step S422: Perform sub-basin assessment calculations on the river runoff sub-basin correlation data to obtain sub-basin assessment data, wherein the sub-basin assessment calculations include preset water volume indicators, sediment indicators, pollution indicators and comprehensive indicators;

[0138] Step S423: Perform parallel aggregation of the sub-basin assessment data throughout the entire process to obtain the full-process assessment data of river runoff generation.

[0139] In this embodiment of the invention, sub-basin boundaries are delineated and data mapped based on visualized composite data of river runoff—which includes three-dimensional predicted information such as flow, sediment load, and pollutant concentration, as well as raster or vector data matched with geographic space. This delineation process typically combines digital elevation models (DEMs) with hydrological response unit coding information, performing watershed line extraction and basin unit attribution determination to establish spatial correlations between the composite data and each sub-basin, generating a river runoff sub-basin association dataset. Next, in step S422, using this dataset as input, a quantitative assessment is performed on each sub-basin. The assessment model consists of multiple preset evaluation dimensions, including water quantity indicators (such as annual average flow and peak flow), sediment indicators (such as sediment output rate per unit area and sediment concentration variation coefficient), pollution indicators (such as average concentrations of pollutants like phosphorus and nitrogen and their temporal variation), and weighted comprehensive indicators reflecting the overall hydrological and ecological state. Each type of indicator undergoes multi-scale time series analysis and statistical calculations based on the original data and trends within the sub-basin, yielding standardized or normalized assessment results. Step S423 employs a full-process parallel aggregation approach to structurally integrate the assessment results of all sub-basins. This process typically involves parallel processing in a multi-threaded or distributed computing environment at the data level. It merges and calculates the assessment data of each sub-basin by constructing an index mapping table and an association matrix, while retaining metadata such as geographical boundaries, indicator weights, and time-series nodes. The result is a unified and logically complete full-process assessment dataset of river runoff generation, providing a data foundation for subsequent regional comprehensive judgment or basin-level response analysis. Throughout the entire process, from sub-basin boundary extraction and indicator calculation to data aggregation, it relies on spatial matching, logical grouping, and multi-indicator statistical processing of high-dimensional data, constructing a rigorous data support mechanism and computational path.

[0140] This specification provides a watershed hydrological forecasting system with intelligent runoff model adaptation, used to execute the aforementioned watershed hydrological forecasting method with intelligent runoff model adaptation. The watershed hydrological forecasting system with intelligent runoff model adaptation includes:

[0141] The IoT data acquisition and basic data construction module is used to collect key hydrological data from IoT monitoring arrays deployed in the watershed and to construct a three-dimensional hierarchical runoff generation basic dataset.

[0142] The runoff composition analysis and proportion calculation module is used to calculate the water distribution proportion based on the three-dimensional stratified runoff basic dataset, and to perform vertical analysis of the watershed runoff composition to obtain runoff composition data.

[0143] The topology modeling and hydrological response prediction module is used to perform topological coding based on runoff composition data to obtain the topology code of the hydrological response unit; construct a multi-branch tree confluence network based on the topology code of the hydrological response unit, and perform cross-sectional analysis of the entire basin to obtain cross-sectional process data of the entire basin; monitor sediment transport capacity based on the cross-sectional process data of the entire basin, and perform closed-loop modeling to obtain the runoff hydrological prediction model of the river; and predict runoff based on the runoff hydrological prediction model of the river to obtain three-dimensional prediction data of flow rate, sediment load, and pollutant concentration.

[0144] The runoff visualization and assessment report module is used to visualize the three-dimensional prediction data of flow rate, sediment load, and pollutant concentration to obtain composite data of river runoff visualization; to conduct a full-process assessment based on the composite data of river runoff visualization to obtain full-process assessment data of river runoff; and to construct a full-process prediction report of river runoff based on the full-process assessment data of river runoff.

[0145] The beneficial effects of this invention are as follows: First, it expands hydrological monitoring data from single-point, two-dimensional planar sampling to a three-dimensional vertically layered structure. Through an IoT sensor array, it collects key parameters such as soil moisture, rainfall intensity, groundwater recharge, and surface runoff at high frequency. After mapping with a unified time window and spatial scale, it constructs a three-dimensional layered runoff generation dataset, effectively improving the expressive power of the spatial heterogeneity and vertical structural characteristics of the runoff generation process. Second, the system combines water distribution ratio analysis with vertical analysis of runoff components, refining the model based on the proportion of different runoff sources (such as surface runoff, interflow, and baseflow) within the watershed unit. This allows the flow simulation to not only reflect macroscopic trends but also the evolution of hydrological mechanisms within the watershed. Third, it reconstructs the spatial connection network between hydrological response units based on topological coding, characterizes the hydrodynamic coupling mode between different tributaries and the main channel through a multi-branch tree confluence structure, and introduces cross-sectional hydraulic analysis and sediment transport capacity modeling mechanisms to construct a... The closed-loop updated runoff prediction model possesses dynamic feedback adjustment capabilities in simulating sediment transport and pollutant release. Furthermore, by coupling the output three-dimensional prediction data of flow rate, sediment load, and pollutant concentration, the model introduces a visualization module to structurally encapsulate and graphically represent the multidimensional data, constructing multi-level composite image data represented by runoff composition maps, co-process lines, and pollution risk heatmaps. This provides quantitative support in sub-basin evaluation and full-process report generation, making the analysis results more intuitive, interconnected, and applicable to decision-making. Finally, this process establishes a complete chain from real-time monitoring, spatiotemporal modeling, basin extrapolation, pollution response, graphical encapsulation to analysis report construction, emphasizing the continuity, structural consistency, and semantic transferability of data across multiple scales and process nodes. This significantly improves the data expression depth, prediction accuracy, and comprehensive evaluation capabilities of runoff hydrological analysis, providing a solid data-driven foundation and closed-loop process characteristics for basin-scale hydrological modeling and pollution control decision support.

[0146] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0147] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A watershed hydrological forecasting method with intelligent adaptation to runoff generation patterns, characterized in that, Includes the following steps: Step S1: Deploy an IoT monitoring array in the watershed to collect key hydrological data and construct a three-dimensional hierarchical runoff generation dataset; Step S2: Calculate the water distribution ratio based on the three-dimensional stratified runoff generation dataset, and perform vertical analysis of the runoff generation components in the watershed to obtain runoff generation component data. The vertical analysis of the runoff generation components in the watershed includes: The total rainfall intensity per unit area was decomposed from the watershed rainfall data, and the characteristic curves of surface runoff and interflow were constructed to obtain the first characteristic curve and the second characteristic curve. The vegetation interception layer-surface runoff layer analysis was performed on the water distribution ratio data using the first characteristic curve to obtain surface water storage data. The second characteristic curve was used to perform upper soil layer-lower soil layer analysis on the water distribution ratio data to obtain groundwater storage data; Obtain soil temperature data; determine the frozen soil state from the soil temperature data to obtain frozen soil water volume assessment data; Based on the data on frozen soil water volume, the surface water volume and groundwater volume data are corrected by exponential decay to obtain the runoff composition data; Step S3: Based on the runoff composition data, perform topological coding to obtain the topological coding of the hydrological response unit; construct a multi-branch tree confluence network based on the topological coding of the hydrological response unit, and perform cross-sectional analysis of the entire basin to obtain cross-sectional process data of the entire basin; monitor sediment transport capacity based on the cross-sectional process data of the entire basin, and perform closed-loop modeling to obtain the river runoff hydrological prediction model; predict runoff based on the river runoff hydrological prediction model to obtain three-dimensional prediction data of flow rate-sediment quantity-pollutant concentration; wherein, step S3 includes the following steps: Step S31: Perform topology coding based on runoff composition data to obtain the topology coding of hydrological response units; Step S32: Construct a multi-branch tree confluence network based on the topology coding of the hydrological response unit, and perform cross-sectional analysis of the entire watershed to obtain cross-sectional process data of the entire watershed. Step S33: Monitor sediment transport capacity based on cross-sectional process data of the entire watershed, and perform closed-loop modeling to obtain the river runoff hydrological prediction model; Step S34: Obtain real-time monitoring watershed runoff data; input the real-time monitoring watershed runoff data into the river runoff hydrological prediction model to predict runoff and obtain three-dimensional prediction data of flow rate, sediment load and pollutant concentration. Step S4: Visualize the three-dimensional prediction data of flow rate, sediment load, and pollutant concentration to obtain visualized composite data of river runoff generation; conduct a full-process assessment based on the visualized composite data of river runoff generation to obtain full-process assessment data of river runoff generation; and construct a full-process prediction report of river runoff generation based on the full-process assessment data of river runoff generation.

2. The watershed hydrological forecasting method with intelligent runoff pattern adaptation according to claim 1, characterized in that, Step S1 includes the following steps: Step S11: Install soil moisture sensors in the corresponding watershed hydrological units, with a sampling frequency of 1 / h, and collect data on the moisture content of the upper soil layer and the lower soil layer; Step S12: Collect rainfall data at a time resolution of five minutes using a weather station to obtain watershed rainfall data; Step S13: Obtain soil organic matter data from the laboratory; use UAV remote sensing to obtain the watershed vegetation cover index, and combine it with soil organic matter data to generate basic parameters for phosphorus adsorption coefficient. Step S14: Perform spatiotemporal alignment processing on the moisture content data of the upper and lower soil layers, and use a 1-hour window moving average filter to eliminate noise, to obtain a standardized stratified soil moisture dataset. Step S15: Compare the standardized stratified soil moisture dataset, watershed rainfall data, and basic parameters of phosphorus adsorption coefficient, and perform meteorological-soil-topography spatial mapping to obtain a three-dimensional stratified runoff generation basic dataset.

3. The watershed hydrological forecasting method with intelligent runoff pattern adaptation according to claim 1, characterized in that, Step S2, which calculates the water allocation ratio based on the three-dimensional stratified runoff dataset, includes: The water allocation ratio was calculated based on the three-dimensional hierarchical runoff generation dataset, and the formula for calculating the water allocation ratio is as follows. in This refers to the proportion of water entering the soil layer, which is also the data on water distribution ratio. This represents the moisture content of the upper soil layer.

4. The watershed hydrological forecasting method with intelligent runoff pattern adaptation according to claim 1, characterized in that, Step S32 includes the following steps: Step S321: Based on the parent node being the downstream river channel and the child node being the upstream unit, construct a multi-branch tree confluence network for the topological coding of the hydrological response unit to obtain the watershed multi-branch tree confluence network; Step S322: Process the nodes of the watershed multi-branch tree confluence network by channel slope / roughness to obtain water level-discharge mapping data; Step S323: When the soil temperature data is less than or equal to 0℃, monitor the soil freezing depth, and use 1 minus half of the soil freezing depth to correct for the frozen soil factor, and obtain the frozen soil water level mapping data. Step S324: Combine the water level-discharge mapping data and the frozen soil water level mapping data with weights to obtain the cross-sectional process data of the entire watershed.

5. The watershed hydrological forecasting method with intelligent runoff pattern adaptation according to claim 1, characterized in that, Step S33 includes the following steps: Step S331: Perform multi-branch tree confluence analysis on the cross-sectional process data of the entire watershed and calculate the sediment transport capacity to obtain sediment transport data for the entire watershed; Step S332: When the sediment transport data of the whole basin is greater than 25 kg / s, the phosphorus uptake module is started to calculate phosphorus release and obtain phosphorus desorption data; when the sediment transport data of the whole basin is less than or equal to 25 kg / s, water quality stabilization treatment is performed to obtain water quality stabilization data. Step S333: Couple the phosphorus desorption data and water quality steady-state data into sediment-flow-pollutant output, and construct a prediction model to obtain a river runoff hydrological prediction model.

6. The watershed hydrological forecasting method with intelligent runoff pattern adaptation according to claim 1, characterized in that, Step S4 includes the following steps: Step S41: Visualize the three-dimensional prediction data of flow rate, sediment load, and pollutant concentration to obtain visualized composite data of river runoff. Step S42: Perform sub-basin assessment based on the visualized composite data of river runoff to obtain sub-basin assessment data; perform parallel aggregation of the sub-basin assessment data throughout the entire process to obtain full-process assessment data of river runoff. Step S43: Construct a full-process prediction report for river runoff based on the full-process assessment data.

7. The watershed hydrological forecasting method with intelligent runoff pattern adaptation according to claim 6, characterized in that, Step S41 includes the following steps: Step S411: Extract runoff components and spatial topology data based on the three-dimensional prediction data of flow rate, sediment amount and pollutant concentration, and perform Kriging interpolation of surface runoff, interflow and groundwater. Component fusion is performed according to a 1km×1km grid matrix to construct a dynamic runoff component map and obtain a dynamic runoff component distribution map. Step S412: Obtain hydrological critical data; extract cross-sectional process-water quality data based on the three-dimensional prediction data of flow-sediment quantity-pollutant concentration, and perform multi-scale visualization of event marking using hydrological critical data to obtain the river flow-sediment co-process line; Step S413: Extract pollutant-hydrodynamic data based on the three-dimensional prediction data of flow rate-sediment quantity-pollutant concentration, and perform spatial diffusion simulation to obtain a phosphorus pollution risk warning heat map; Step S414: Visualize and encapsulate the dynamic runoff composition distribution map, the river flow-sediment co-process line, and the phosphorus pollution risk early warning heat map to obtain visualized composite data of river runoff.

8. A watershed hydrological forecasting system with intelligent runoff pattern adaptation, characterized in that, For executing the watershed hydrological forecasting method with intelligent runoff pattern adaptation as described in claim 1, the watershed hydrological forecasting system with intelligent runoff pattern adaptation includes: The IoT data acquisition and basic data construction module is used to collect key hydrological data from IoT monitoring arrays deployed in the watershed and to construct a three-dimensional hierarchical runoff generation basic dataset. The runoff composition analysis and proportion calculation module is used to calculate the water distribution proportion based on the three-dimensional stratified runoff basic dataset, and to perform vertical analysis of the watershed runoff composition to obtain runoff composition data. The topology modeling and hydrological response prediction module is used to perform topological coding based on runoff composition data to obtain the topology code of the hydrological response unit; construct a multi-branch tree confluence network based on the topology code of the hydrological response unit, and perform cross-sectional analysis of the entire basin to obtain cross-sectional process data of the entire basin; monitor sediment transport capacity based on the cross-sectional process data of the entire basin, and perform closed-loop modeling to obtain the runoff hydrological prediction model of the river; and predict runoff based on the runoff hydrological prediction model of the river to obtain three-dimensional prediction data of flow rate, sediment load, and pollutant concentration. The runoff visualization and assessment report module is used to visualize the three-dimensional prediction data of flow rate, sediment load, and pollutant concentration to obtain composite data of river runoff visualization; to conduct a full-process assessment based on the composite data of river runoff visualization to obtain full-process assessment data of river runoff; and to construct a full-process prediction report of river runoff based on the full-process assessment data of river runoff.

Images