Water ecology prediction model generation method and device coupling mechanism model and machine learning model, water ecology prediction method and device and storage medium

Through the coupling mechanism model and machine learning model, combined with the distributed hydrological water quality model in the basin, the problem of insufficient spatial and temporal resolution of water ecological monitoring in the existing technology is solved, and high-temporal resolution prediction of water ecological conditions in the basin is achieved. , serving the optimization decisions of water ecological environment protection projects.

CN119090093APending Publication Date: 2024-12-06BEIJING NORMAL UNIVERSITY

Patent Information

Application Number
CN202411572260.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-11-06
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

The prior art cannot effectively cover the entire basin in water ecological monitoring, the spatial resolution is insufficient, and the time-continuous simulation prediction cannot be achieved, making it difficult to accurately detect the spatial and temporal changes in aquatic biodiversity.

Method used

Using coupled mechanism model and machine learning model, by obtaining water environment quality indicators, hydrodynamic conditions and land use data in water samples, building training sets and verification sets, training initial machine learning models, generating water ecological prediction models, and combining them with the basin Distributed hydrological water quality model simulates hydrodynamic conditions and water environment quality, and then predicts aquatic biological indicators.

Benefits of technology

It realizes continuous simulation and prediction of the temporal and spatial dimensions of the water ecological conditions in the basin, improves the spatial and temporal resolution of the water ecological conditions, eliminates the blind spots of monitoring space, and can predict the water ecological conditions of climate conditions, water resource allocation and water pollution prevention and control measures. The impact of

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119090093A_ABST
    Figure CN119090093A_ABST
Patent Text Reader

Abstract

The invention provides a water ecology prediction model generation method coupling a mechanism model and a machine learning model, a water ecology prediction method and device, and a storage medium, belongs to the technical field of data processing methods for prediction purposes, and solves the problem of insufficient spatial resolution of water ecology prediction. The water ecology prediction model generation method comprises the following steps: acquiring sampling data; calculating aquatic organism indexes of the water sample; training the model to obtain a selected water ecology prediction basic model; hydrological-related current actual survey data serve as input, and a watershed distributed hydrological model is calibrated by means of hydrological station actual measurement data under the hydrodynamic condition; the current water pollution-related actual survey data and the prediction data for the watershed hydrodynamic condition driven water quality model are used as input, and the watershed water quality model is calibrated by using the monitoring data of the water environment quality index; and generating a water ecology prediction model based on the model. The spatial-temporal resolution and precision of water ecology prediction are improved, and the influence of governance measures, water conservancy projects and the like on water ecology can be predicted.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention generally relates to the technical field of data processing methods for prediction purposes, and more specifically, to a method for generating a water ecology prediction model of a coupling mechanism model and a machine learning model, a water ecology prediction method, a device and a storage medium. Background Art

[0002] At present, the focus of water ecological environment protection in the basin has gradually shifted from water pollution prevention and control to the "three water coordination" of water environment, water resources and water ecology. The health of water ecology in the basin has become one of the important guarantee goals of water ecological environment protection. Aquatic biodiversity monitoring is the basic work of water ecological protection. The traditional method of aquatic biological monitoring is morphological identification. With the introduction of gene detection into the field of water environment, a more efficient environmental DNA (eNDA) detection method has emerged. Affected by many factors such as climate, hydrology, water quality, landform and river network structure, aquatic biodiversity shows high temporal and spatial variability in the basin. Regardless of the monitoring method, the early sampling and water sample pretreatment work will consume a lot of manpower, material resources, time and funds. Sampling can only be carried out once or several times at limited local points in the basin. Long-term water ecological data is very lacking, which cannot improve the timeliness and spatial resolution of water ecological monitoring, and it is difficult to accurately detect the temporal and spatial changes of aquatic biodiversity in the basin.

[0003] Fan J, Wang S, Li H et al. proposed a method for constructing a machine learning model in the paper Modeling the ecological status response of rivers to multiple stressors using machine learning: A comparison of environmental DNA metabarcoding and morphological data [J]. Water Research, 2020, 183:116004. DOI:10.1016 / j.watres.2020.116004. This method simulates and predicts the data of 24 monitoring points sampled once in autumn, and evaluates the water ecological status of the sampling points based on ecological indicators; takes ecological indicators as target variables, and takes the water quality index concentration of the monitoring point and the proportion of land use types in the buffer zone delineated with the monitoring point as the center as characteristic variables, trains the machine learning model, and optimizes the parameters; and performs sensitivity analysis of characteristic variables based on the control variable method.

[0004] The above method in the prior art has some shortcomings: first, it does not take into account the seasonal response of water ecological conditions to water physical and chemical indicators and other environmental factors, and the simulation results may be overfitting and insufficient to reflect the actual water ecological conditions; second, this method only simulates and predicts the corresponding points based on the measured data of the monitoring points, which cannot cover the entire river basin and has insufficient spatial resolution; third, this method only uses single sampling detection data as model input, and cannot achieve temporal continuity simulation and prediction. Summary of the invention

[0005] In order to solve the above-mentioned problems in the prior art, in the first aspect, an embodiment of the present invention provides a method for generating a water ecological prediction model by coupling a mechanism model and a machine learning model, the method comprising: obtaining water environment quality indicators and environmental DNA data of water samples collected at sampling points in a specific river basin, as well as hydrodynamic conditions and associated land use data at the sampling points; calculating aquatic biological indicators of the collected water samples based on the environmental DNA data of the collected water samples; constructing a training set and a validation set using the water environment quality indicators of the collected water samples, the hydrodynamic conditions and associated land use data at the sampling points, and the aquatic biological indicators of the collected water samples. The training set and the validation set are used to train the initial machine learning model to obtain a selected basic model for water ecological prediction, wherein the input of the selected basic model for water ecological prediction includes hydrodynamic conditions, water environment quality indicators and related land use data, and the output of the selected basic model for water ecological prediction includes aquatic biological indicators; the current survey data of each of the related land use data, soil cover data, meteorological data, digital elevation information, water diversion data, and water conservancy project scheduling data of the specific river basin are obtained, as well as the hydrological station measured data of the hydrodynamic conditions measured by the hydrological stations deployed in the specific river basin; the specific The current survey data of the associated land use data, soil cover data, meteorological data, digital elevation information, water diversion data, and water conservancy project scheduling data of the basin are used as input data to drive the basin distributed hydrological model of the specific basin to simulate the hydrodynamic conditions, and the measured data of the hydrological station of the hydrodynamic conditions are used as calibration data to calibrate the parameters of the basin distributed hydrological model of the specific basin, so as to obtain the calibrated basin distributed hydrological model of the specific basin, wherein the input of the calibrated basin distributed hydrological model includes the associated land use data, soil cover data, meteorological data, digital elevation information, water diversion data, and water conservancy project scheduling data. Engineering scheduling data, the output of the calibrated basin distributed hydrological model includes hydrodynamic conditions; obtaining the current verified data of point source pollution load and non-point source pollution load of the specific basin, and the monitoring data of water environment quality indicators monitored by water quality monitoring stations arranged in the specific basin; through the calibrated basin distributed hydrological model, based on the current verified data of associated land use data, soil cover data, meteorological data, digital elevation information, water diversion data, and water conservancy project scheduling data of the specific basin, the basin hydrodynamic conditions of the specific basin are predicted, and the prediction data for driving the water quality model of the basin hydrodynamic conditions are obtained;The current actual survey data of the point source pollution load and non-point source pollution load of the specific river basin and the predicted data of the driving water quality model of the water dynamic conditions of the river basin are used as input data to drive the water quality model of the specific river basin to simulate the water environment quality index, and the monitoring data of the water environment quality index is used as calibration data to calibrate the parameters of the water quality model of the specific river basin to obtain the calibrated water quality model of the specific river basin, wherein the input of the calibrated water quality model includes the point source pollution load, non-point source pollution load and hydrodynamic conditions, and the output of the calibrated water quality model includes the water environment quality index; based on the selected water ecological prediction basic model, the calibrated watershed distributed hydrological model and the calibrated flow A water quality model for a specific watershed is used to generate a water ecology prediction model for the specific watershed, wherein the calibrated watershed distributed hydrological model is used to generate current prediction data of hydrodynamic conditions based on currently input associated land use data, soil cover data, meteorological data, digital elevation information, water diversion data, and water conservancy project scheduling data; the calibrated watershed water quality model is used to generate current prediction data of water environment quality indicators based on currently input point source pollution load, non-point source pollution load, and current prediction data of hydrodynamic conditions; the selected water ecology prediction basic model is used to generate current prediction data of aquatic biological indicators based on currently input associated land use data, current prediction data of hydrodynamic conditions, and current prediction data of water environment quality indicators. ;

[0006] In some embodiments, calculating the aquatic biological index of the collected water sample according to the environmental DNA data of the collected water sample includes: calculating M aquatic biological indexes of the collected water sample according to the environmental DNA data of the collected water sample, wherein M is an integer greater than 1. Constructing a training set and a validation set using the water environment quality index of the collected water sample, the hydrodynamic conditions and associated land use data at the sampling point, and the aquatic biological index of the collected water sample includes: constructing M groups of training sets and validation sets respectively using the water environment quality index of the collected water sample, the hydrodynamic conditions and associated land use data at the sampling point, and the M aquatic biological indexes of the collected water sample, wherein each of the M groups of training sets and validation sets respectively contains a corresponding one of the M aquatic biological indexes. Using the training set and the validation set to train the initial machine learning model to obtain the selected water ecological prediction basic model includes: using the M groups of training sets and validation sets to train N different initial machine learning models respectively, where N is an integer greater than 1, to obtain M*N trained alternative water ecological prediction basic models and their objective functions, wherein the input of the alternative water ecological prediction basic model includes hydrodynamic conditions, water environment quality indicators and related land use data, and the output of the alternative water ecological prediction basic model includes one aquatic biological indicator among the M aquatic biological indicators corresponding to the training set and validation set used for training the model; selecting the model with the best objective function among the M*N trained alternative water ecological prediction basic models as the selected water ecological prediction basic model.

[0007] In some embodiments, calculating the aquatic biological index of the collected water sample according to the environmental DNA data of the collected water sample includes: calculating M aquatic biological indexes of the collected water sample according to the environmental DNA data of the collected water sample, wherein M is an integer greater than 1. Constructing a training set and a validation set using the water environment quality index of the collected water sample, the hydrodynamic conditions and associated land use data at the sampling point, and the aquatic biological index of the collected water sample includes: constructing M groups of training sets and validation sets respectively using the water environment quality index of the collected water sample, the hydrodynamic conditions and associated land use data at the sampling point, and the M aquatic biological indexes of the collected water sample, wherein each of the M groups of training sets and validation sets respectively contains a corresponding one of the M aquatic biological indexes. Using the training set and the validation set to train the initial machine learning model to obtain the selected water ecological prediction basic model includes: using the M groups of training sets and validation sets to train N different initial machine learning models respectively, where N is an integer greater than 1, to obtain M*N trained alternative water ecological prediction basic models and their objective functions, wherein the input of the alternative water ecological prediction basic model includes hydrodynamic conditions, water environment quality indicators and related land use data, and the output of the alternative water ecological prediction basic model includes an aquatic biological indicator among the M aquatic biological indicators corresponding to the training set and the validation set used for training the model; for the N trained alternative water ecological prediction basic models corresponding to each aquatic biological indicator among the M aquatic biological indicators, selecting the model with the best objective function among the N trained alternative water ecological prediction basic models as the selected water ecological prediction basic model corresponding to the aquatic biological indicator, thereby obtaining M selected water ecological prediction basic models.

[0008] In some embodiments, the aquatic biological indicator is selected from the group including: dominant species, indicator species, sensitive species, richness index, Shannon index, Simpson index, pielou evenness index, Chao1 index, ACE index, fish integrity index F-IBI, plankton integrity index P-IBI, benthic animal integrity index B-IBI; and / or the initial machine learning model is selected from the group including: support vector machine, random forest, gradient boosting machine, artificial neural network, decision tree, elastic network, XGBoost, LightGBM, CatBoost.

[0009] In some embodiments, obtaining water environment quality indicators and environmental DNA data of water samples collected at sampling points in a specific river basin, as well as hydrodynamic conditions and associated land use data at the sampling points, includes: obtaining the water environment quality indicators and environmental DNA data of water samples collected at sampling points in the specific river basin, as well as the specific time point data of each of the hydrodynamic conditions and associated land use data at the sampling points at a plurality of specific time points according to a predefined time interval. Using the water environment quality indicators of the collected water samples, the hydrodynamic conditions and associated land use data at the sampling points, and the aquatic biological indicators of the collected water samples to construct a training set and a validation set includes: using the water environment quality indicators of the collected water samples, the hydrodynamic conditions and associated land use data at the sampling points, and the specific time point data of each of the aquatic biological indicators of the collected water samples, to respectively construct a training set and a validation set corresponding to the specific time points. Using the training set and the validation set to train the initial machine learning model to obtain the selected water ecology prediction basic model includes: using the training set and the validation set corresponding to the specific time point to train the initial machine learning model to obtain the selected water ecology prediction basic model corresponding to the specific time point.

[0010] In the second aspect, an embodiment of the present invention proposes a water ecology prediction method of a coupling mechanism model and a machine learning model, the method comprising: obtaining a water ecology prediction model for a specific river basin generated by the water ecology prediction model generation method described in any of the above embodiments; generating current prediction data of hydrodynamic conditions based on currently input associated land use data, soil cover data, meteorological data, digital elevation information, water diversion data, and water conservancy project scheduling data through a calibrated river basin distributed hydrological model; generating current prediction data of water environment quality indicators based on currently input point source pollution load, non-point source pollution load, and the current prediction data of hydrodynamic conditions through a calibrated river basin water quality model; generating current prediction data of aquatic biological indicators based on currently input associated land use data, the current prediction data of the hydrodynamic conditions, and the current prediction data of the water environment quality indicators through a selected water ecology prediction basic model.

[0011] In some embodiments, the method further includes one or more of the following steps: obtaining future meteorological forecast data for the specific river basin, and inputting the future meteorological forecast data into the water ecological forecast model, and obtaining current forecast data of aquatic biological indicators in response to the future meteorological forecast data via the water ecological forecast model; generating regulated water diversion data according to water resource allocation information of the specific river basin, inputting the regulated water diversion data into the water ecological forecast model, and obtaining current forecast data of aquatic biological indicators in response to the water resource allocation information via the water ecological forecast model; generating point source pollution load and non-point source pollution load control data according to water pollution control information of the specific river basin, inputting the point source pollution load and non-point source pollution load control data into the water ecological forecast model, and obtaining current forecast data of aquatic biological indicators in response to the water pollution control information via the water ecological forecast model; obtaining updated water conservancy project scheduling data, inputting the updated water conservancy project scheduling data into the water ecological forecast model, and obtaining current forecast data of aquatic biological indicators in response to the updated water conservancy project scheduling data via the water ecological forecast model.

[0012] In the third aspect, an embodiment of the present invention proposes a device for generating a water ecological prediction model of a coupling mechanism model and a machine learning model, the device comprising: a sampling data acquisition module, configured to obtain water environment quality indicators and environmental DNA data of water samples collected at sampling points in a specific river basin, as well as hydrodynamic conditions and associated land use data at the sampling points; a sampling aquatic biological indicator calculation module, configured to calculate the aquatic biological indicators of the collected water samples based on the environmental DNA data of the collected water samples; a training set construction module, configured to construct a training set using the water environment quality indicators of the collected water samples, the hydrodynamic conditions and associated land use data at the sampling points, and the aquatic biological indicators of the collected water samples. and validation set; a basic model training module, configured to use the training set and the validation set to train the initial machine learning model to obtain a selected water ecology prediction basic model, wherein the input of the selected water ecology prediction basic model includes hydrodynamic conditions, water environment quality indicators and related land use data, and the output of the selected water ecology prediction basic model includes aquatic biological indicators; a hydrological related data acquisition module, configured to obtain the current actual inspection data of each of the associated land use data, soil cover data, meteorological data, digital elevation information, water diversion data, and water conservancy project scheduling data of the specific river basin, as well as the hydrological station measured data of the hydrodynamic conditions measured by the hydrological stations deployed in the specific river basin; The model calibration module is configured to use the current actual survey data of the associated land use data, soil cover data, meteorological data, digital elevation information, water diversion data, and water conservancy project scheduling data of the specific river basin as input data to drive the basin distributed hydrological model of the specific river basin to simulate the hydrodynamic conditions, and use the measured data of the hydrological station of the hydrodynamic conditions as calibration data to calibrate the parameters of the basin distributed hydrological model of the specific river basin, so as to obtain the calibrated basin distributed hydrological model of the specific river basin, wherein the input of the calibrated basin distributed hydrological model includes the associated land use data, soil cover data, meteorological data, digital elevation information, water diversion data, and water conservancy project scheduling data. , the output of the calibrated basin distributed hydrological model includes hydrodynamic conditions; a water quality related data acquisition module, configured to obtain the current actual data of the point source pollution load and the non-point source pollution load of the specific basin, and the monitoring data of the water environment quality indicators monitored by the water quality monitoring stations arranged in the specific basin; a basin hydrodynamic condition prediction module, configured to predict the basin hydrodynamic conditions of the specific basin through the calibrated basin distributed hydrological model, based on the current actual data of the associated land use data, soil cover data, meteorological data, digital elevation information, water diversion data, and water conservancy project scheduling data of the specific basin, and obtain the prediction data for driving the water quality model of the basin hydrodynamic conditions;a water quality model calibration module, configured to use the current actual inspection data of the point source pollution load and non-point source pollution load of the specific river basin and the predicted data of the driving water quality model of the watershed hydrodynamic conditions as input data to drive the watershed water quality model of the specific river basin to simulate the water environment quality index, and use the monitoring data of the water environment quality index as calibration data to calibrate the parameters of the watershed water quality model of the specific river basin to obtain the calibrated watershed water quality model of the specific river basin, wherein the input of the calibrated watershed water quality model includes the point source pollution load, non-point source pollution load and hydrodynamic conditions, and the output of the calibrated watershed water quality model includes the water environment quality index; a water ecology prediction model generation module, configured to generate a water ecology prediction model based on the selected water ecology prediction basic model and the calibrated watershed classification system. The distributed hydrological model and the calibrated watershed water quality model are used to generate the water ecology prediction model of the specific watershed, wherein the calibrated watershed distributed hydrological model is used to generate the current prediction data of the hydrodynamic conditions based on the currently input associated land use data, soil cover data, meteorological data, digital elevation information, water diversion data, and water conservancy project scheduling data; the calibrated watershed water quality model is used to generate the current prediction data of the water environment quality index based on the currently input point source pollution load, non-point source pollution load, and the current prediction data of the hydrodynamic conditions; the selected water ecology prediction basic model is used to generate the current prediction data of the aquatic biological index based on the currently input associated land use data, the current prediction data of the hydrodynamic conditions, and the current prediction data of the water environment quality index. ;

[0013] In the fourth aspect, an embodiment of the present invention proposes a water ecology prediction device of a coupling mechanism model and a machine learning model, the device comprising: a water ecology prediction model acquisition module, configured to obtain a water ecology prediction model of a specific watershed generated by a water ecology prediction model generation device described in the above embodiment; a hydrodynamic condition prediction module, configured to generate current prediction data of hydrodynamic conditions based on currently input associated land use data, soil cover data, meteorological data, digital elevation information, water diversion data, and water conservancy project scheduling data via a calibrated watershed distributed hydrological model; a water environment quality index prediction module, configured to generate current prediction data of water environment quality indicators based on currently input point source pollution load, non-point source pollution load, and the current prediction data of hydrodynamic conditions via a calibrated watershed water quality model; an aquatic biological index prediction module, configured to generate current prediction data of aquatic biological indicators based on currently input associated land use data and the current prediction data of the hydrodynamic conditions and the current prediction data of the water environment quality indicators via a selected water ecology prediction basic model.

[0014] In a fifth aspect, an embodiment of the present invention provides a storage medium storing computer-readable instructions, which, when executed by a processor, executes the method described in any of the above embodiments.

[0015] Based on the spatial information data of the watershed, a distributed watershed hydrological and water quality model is constructed to simulate and predict the hydrodynamic conditions and water environment quality at the sub-watershed scale. The land use data at the spatiotemporal scale that matches the simulation results is used as input data, and the watershed aquatic biological conditions are predicted based on the constructed machine learning model. Compared with the existing technology that can only simulate at sampling points, after coupling the distributed watershed hydrological and water quality model, the water ecological conditions that can be achieved at the sub-watershed scale are continuously simulated and predicted in the time and space dimensions, which improves the spatiotemporal resolution of the water ecological condition simulation; at the same time, the water ecological conditions in areas that cannot be reached by human resources in the study area can be simulated and predicted, which eliminates the spatial blind spots of actual water ecological monitoring to a certain extent.

[0016] By constructing a basin hydrology and water quality model, the temporal and spatial distribution of river hydrodynamic conditions and water quality indicators in the basin driven by basin meteorological conditions, water resources allocation and water pollution control measures are simulated. By comparing the temporal and spatial scales of the model simulation, adapting the corresponding associated land use data, and based on the selected water ecology prediction basic model constructed earlier, the water ecology status of the basin can be predicted at a refined scale. Based on the mechanism model, the basin hydrology and water quality indicators are simulated, and then the water ecology status of the basin unit with higher temporal and spatial resolution is simulated and predicted.

[0017] This overall water ecological prediction model can be applied to a variety of scenarios, for example, predicting the water ecological conditions at the sub-basin scale in the entire basin; predicting the water ecological conditions at specific time points throughout the year; and simulating water ecological simulation and prediction with high temporal and spatial resolution at the basin scale driven by climate conditions, water resources allocation and water pollution prevention and control measures.

[0018] The water ecology prediction model generated by the method proposed in the embodiment of the present invention predicts the watershed hydrodynamic conditions and water environment quality driven by climate conditions, water resources allocation and water pollution prevention and control measures through the watershed hydrological and water quality model, and then uses it as input information for the machine learning model to predict the water ecology status of the watershed. Compared with the prior art, the function of the machine learning model has been expanded. By coupling it with the watershed hydrological and water quality model, it can achieve quantitative prediction of the impact of climate conditions, water resources allocation and water pollution prevention and control measures on the water ecology status, and then serve the optimization decision-making of watershed water ecological environment protection projects and non-engineering measures. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The above and other objects, features and advantages of the embodiments of the present invention will become readily understood by reading the following detailed description with reference to the accompanying drawings. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, in which:

[0020] Figure 1 A flow chart showing a method for generating a water ecology prediction model using a coupling mechanism model and a machine learning model according to an embodiment of the present invention;

[0021] Figure 2 A schematic diagram of data set division and cross-validation according to an embodiment of the present invention is shown;

[0022] Figure 3 A flow chart showing a method for predicting water ecology using a coupling mechanism model and a machine learning model according to an embodiment of the present invention;

[0023] Figure 4 A block diagram of a device for generating a water ecology prediction model by coupling a mechanism model and a machine learning model according to an embodiment of the present invention is shown;

[0024] Figure 5 A block diagram of a water ecology prediction device of a coupling mechanism model and a machine learning model according to an embodiment of the present invention is shown.

[0025] In the drawings, the same or corresponding reference numerals represent the same or corresponding parts. DETAILED DESCRIPTION

[0026] The principle and spirit of the present invention will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are provided only to enable those skilled in the art to better understand and implement the present invention, and are not intended to limit the scope of the present invention in any way.

[0027] In one aspect, an embodiment of the present invention provides a method for generating a water ecology prediction model by coupling a mechanism model and a machine learning model. Figure 1 , which shows a flow chart of a method 100 for generating a water ecology prediction model by coupling a mechanism model and a machine learning model according to an embodiment of the present invention. The method 100 includes steps S101-S110.

[0028] In step S101, water environment quality indicators and environmental DNA (eDNA) data of water samples collected at sampling points in a specific river basin, as well as hydrodynamic conditions and associated land use data at the sampling points are obtained. The data obtained in this step is based on field sampling and data processing, including study area division, sampling design and process, data processing and statistics.

[0029] The specific basin can be the entire basin to be studied or a local basin in the entire basin. The sampling points can be arranged according to the natural geographical characteristics of the basin. The sampling points can be sparse sampling points arranged in the entire basin area of ​​the basin to be studied, for example, they can be sampling points that are easy for personnel to reach and convenient for sampling operations. As a specific example, a geographic information system (or geographic information system software, geographic information system application) can be used to divide the study area and arrange sampling points. The arrangement of sampling points should comprehensively consider the accessibility of the sampling points and cover the different types of landforms in the entire basin as much as possible. Then, the latitude and longitude information of the sampling points is counted, and the sampling route is designed in combination with the overview of the study area and the road conditions. On-site sampling is carried out according to water quality indicators and eDNA water sample collection requirements.

[0030] Water environment quality indicators may also be referred to as water quality indicators. As an example only, water environment quality indicators may include water temperature and water quality indicators. Water quality indicators may include dissolved oxygen, biochemical oxygen demand, chemical oxygen demand, ammonia nitrogen, total nitrogen, total phosphorus, etc. Hydrodynamic conditions may include river channel information, river flow direction, slope information, etc. within the basin.

[0031] After collecting water samples, for example, simple indicators (such as water temperature) and hydrodynamic indicators (such as river surface width, water depth, flow rate, and flow) can be measured and recorded on-site by portable water quality sampling instruments. The remaining water chemical indicators and environmental DNA (eDNA) are further tested in the laboratory. For example, the chemical indicators of water samples (biochemical oxygen demand (BOD5), chemical oxygen demand (COD), total nitrogen (TN), total phosphorus (TP), ammonia nitrogen (NH4-N), nitrate (NO3-), nitrite (NO2-), soluble phosphorus (DTP), soluble active phosphorus (PO43-)) are further tested in the laboratory; the water samples are filtered and the filter membranes are stored at -20°C until they are sent to the laboratory for eDNA testing. Based on the eDNA sequencing results obtained by the test, aquatic organisms are classified and counted, such as fish, phytoplankton, zooplankton, benthic animals, etc.

[0032] The associated land use data may be the land use data upstream of the sampling point or the land use data of the buffer zone divided around the sampling point. For example, in a geographic information system, a buffer zone may be divided around the sampling point, and the proportion of land use data (cultivated land, grassland, water body, shrub, woodland, building land) in the buffer zone may be counted. The associated land use data may be obtained, for example, through remote sensing data.

[0033] In step S102, the aquatic biological index of the collected water sample is calculated based on the environmental DNA data of the collected water sample.

[0034] For example, the aquatic biological index may include aquatic biological abundance, diversity index, and integrity index. As an example, the embodiment of the present invention may use one of the above multiple indicators as the aquatic biological index in this step. Optionally, due to differences in regions or environments, different aquatic biological indicators may be selected for different river basins.

[0035] As one embodiment of the present invention, aquatic biological indicators can be selected from the following group: relative abundance indicators of aquatic organisms such as dominant species, indicator species, and sensitive species, aquatic biodiversity indicators such as richness index, Shannon index, Simpson index, Pielou evenness index, Chao1 index, ACE index, etc., aquatic biological integrity indicators such as fish integrity index F-IBI, plankton integrity index P-IBI, benthic animal integrity index B-IBI, etc.

[0036] In step S103, the water environment quality index of the collected water sample, the hydrodynamic conditions at the sampling point and the associated land use data, and the aquatic biological index of the collected water sample are used to construct a training set and a validation set. As an example, the above data can be sorted into a data input format, and the data can be normalized so that the result value is mapped to between [0 - 1]. As an example, the following formula (1) can be used for data normalization.

[0037] Formula (1)

[0038] Among them, X' represents the normalized data, X represents the original data, and X min represents the minimum value of the sample data, X max Represents the maximum value of the sample data.

[0039] As an example, the data can be divided into training set and validation set in a ratio of 4:1.

[0040] In step S104, the initial machine learning model is trained using the training set and the validation set to obtain a selected basic model for water ecological prediction, wherein the input of the selected basic model for water ecological prediction includes hydrodynamic conditions, water environment quality indicators, and related land use data, and the output of the selected basic model for water ecological prediction includes aquatic biological indicators.

[0041] Optionally, due to differences in regions or environments, different initial machine learning models can be selected for different watersheds.

[0042] As one embodiment of the present invention, the initial machine learning model is selected from the group consisting of: support vector machine, random forest, gradient boosting machine, artificial neural network, decision tree, elastic network, XGBoost, LightGBM, CatBoost.

[0043] As an example only, the initial machine learning model can be selected from one of the machine learning models such as support vector machine, random forest, gradient boosting machine, etc.

[0044] During the training process, the aquatic biological indicators obtained by eDNA statistics are used as target variables, and the hydrodynamic conditions, water environment quality indicators and related land use information obtained through field observations are used as feature variables to train the initial machine learning model. The selected water ecological prediction basic model obtained after training can simulate and predict the aquatic biological conditions of the entire basin.

[0045] The machine learning model parameter calibration algorithm (grid search algorithm) is used to divide the data set into training set and validation set based on 5-fold cross validation. Figure 2 , which shows a schematic diagram of data set division and cross-validation according to an embodiment of the present invention. The machine learning model is optimized by adjusting the parameters, and the objective function (R 2 , MAE) to obtain the optimal value and the optimal simulation result.

[0046] Through the above steps S101-S104, a selected water ecological prediction basic model is generated, that is, a response relationship model of aquatic biological conditions to hydrodynamic conditions, water environment quality indicators, and land use is constructed based on the machine learning model.

[0047] In step S105, the current survey data of each of the associated land use data, soil cover data, meteorological data, digital elevation information, water diversion data, and water conservancy project scheduling data of the specific river basin is obtained, as well as the hydrological station measured data of the hydrodynamic conditions measured by the hydrological stations deployed in the specific river basin.

[0048] As an example, based on the digital elevation model (DEM), the GIS software can be used to generate river channel information, river flow direction, slope information, etc. in the basin, and obtain the current verified data of the digital elevation information of a specific basin. The current verified data of the associated land use data, soil cover data, and meteorological data of a specific basin can be obtained based on various related databases, such as the Geospatial Data Cloud, the Nanjing Institute of Soil Science, Chinese Academy of Sciences, the National Meteorological Science Data Center, and so on.

[0049] In step S106, the current survey data of associated land use data, soil cover data, meteorological data, digital elevation information, water diversion data, and water conservancy project scheduling data of the specific river basin are used as input data to drive the basin distributed hydrological model of the specific river basin to simulate the hydrodynamic conditions, and the measured data of the hydrological station of the hydrodynamic conditions are used as calibration data to calibrate the parameters of the basin distributed hydrological model of the specific river basin to obtain the calibrated basin distributed hydrological model of the specific river basin, wherein the input of the calibrated basin distributed hydrological model includes associated land use data, soil cover data, meteorological data, digital elevation information, water diversion data, and water conservancy project scheduling data, and the output of the calibrated basin distributed hydrological model includes hydrodynamic conditions.

[0050] As an example, the hydrological and hydrodynamic simulation variables in the distributed hydrological model of the basin are determined by the measured flow rate of the hydrological station, and the simulation effect is evaluated by the Nash efficiency coefficient (NSE):

[0051] Formula (2)

[0052] in, represents the observed flow, represents the simulated flow, Represents the average value of the observed flow. The closer the NSE value is to 1, the better the simulation quality is. The closer it is to 0, the closer the simulation result is to the average level of the observed value, that is, the overall result is credible, but the process simulation error is large. The farther it is less than 0, the model is unreliable.

[0053] In the above example, the basin distributed hydrological model is calibrated based on discharge only, but the output of the basin distributed hydrological model can include all data on hydrodynamic conditions.

[0054] Through the above steps S105-S106, a calibrated watershed distributed hydrological model is generated.

[0055] In step S107, the current actual inspection data of the point source pollution load and the non-point source pollution load of the specific river basin, as well as the monitoring data of the water environment quality indicators monitored by the water quality monitoring stations deployed in the specific river basin are obtained.

[0056] Point source pollution loads come from, for example, sewage outlets and sewage treatment plant outlets, and non-point source pollution loads come from, for example, agricultural planting, livestock and poultry breeding, etc. As an example, the current survey data of point source pollution loads and non-point source pollution loads in a given watershed can be obtained through relevant point and non-point source pollution load statistical yearbooks or environmental statistical yearbooks.

[0057] In step S108, the basin hydrodynamic conditions of the specific basin are predicted through the calibrated basin distributed hydrological model, based on the current actual data of the associated land use data, soil cover data, meteorological data, digital elevation information, water diversion data, and water conservancy project scheduling data of the specific basin, to obtain the prediction data for driving the water quality model of the basin hydrodynamic conditions. It should be noted that the "prediction data for driving the water quality model of the basin hydrodynamic conditions" mentioned here are the data predicted by the calibrated basin distributed hydrological model, and the data will be used as input data to drive the basin water quality model of the specific basin in the subsequent step S109, so it is called "prediction data for driving the water quality model".

[0058] In step S109, the water quality model driven by the water quality model of the specific river basin is driven by the current survey data of the point source pollution load and non-point source pollution load of the specific river basin, and the predicted data of the water quality model driven by the water dynamic conditions of the river basin is used as input data to simulate the water environment quality indicators, and the monitoring data of the water environment quality indicators is used as calibration data to calibrate the parameters of the water quality model of the specific river basin to obtain the calibrated water quality model of the specific river basin, wherein the input of the calibrated water quality model includes the point source pollution load, the non-point source pollution load and the hydrodynamic conditions, and the output of the calibrated water quality model includes the water environment quality indicators.

[0059] As an example, the water quality data monitored by the water quality monitoring station is used to calibrate the water environment quality simulation variables in the basin hydrological and water quality model, and the mean absolute error (MAPE) is used to evaluate the water environment quality simulation results:

[0060] Formula (3)

[0061] in, represents the water quality simulation results, Represents the water quality observation results. The closer the MAE is to 0, the better the simulation result is. The larger the value is, the worse the simulation result is.

[0062] Through the above steps S107-S109, a calibrated watershed water quality model is generated.

[0063] In step S110, a water ecological prediction model for a specific river basin is generated based on the selected water ecological prediction basic model, the calibrated river basin distributed hydrological model and the calibrated river basin water quality model, wherein the calibrated river basin distributed hydrological model is used to generate current prediction data of hydrodynamic conditions based on currently input associated land use data, soil cover data, meteorological data, digital elevation information, water diversion data, and water conservancy project scheduling data; the calibrated river basin water quality model is used to generate current prediction data of water environment quality indicators based on currently input point source pollution load, non-point source pollution load, and current prediction data of hydrodynamic conditions; the selected water ecological prediction basic model is used to generate current prediction data of aquatic biological indicators based on currently input associated land use data, current prediction data of hydrodynamic conditions, and current prediction data of water environment quality indicators.

[0064] Through the above step S110, an overall water ecology prediction model is generated. The hydrological and hydrodynamic conditions, water environment quality obtained by simulating the watershed hydrological and water quality model, and the sub-watershed land use ratio extracted from remote sensing data are used as model inputs, and input into the trained water ecology prediction model, so that the overall model can predict the water ecology status in time and space dimensions. In the process of using the overall model for actual prediction, it gets rid of the dependence on measured water quantity, water quality and other data.

[0065] Based on the spatial information data of the watershed, a distributed watershed hydrological and water quality model is constructed to simulate and predict the hydrodynamic conditions and water environment quality at the sub-watershed scale. The land use data at the spatiotemporal scale that matches the simulation results is used as input data, and the watershed aquatic biological conditions are predicted based on the constructed machine learning model. Compared with the existing technology that can only simulate at sampling points, after coupling the distributed watershed hydrological and water quality model, the water ecological conditions that can be achieved at the sub-watershed scale are continuously simulated and predicted in the time and space dimensions, which improves the spatiotemporal resolution of the water ecological condition simulation; at the same time, the water ecological conditions in areas that cannot be reached by human resources in the study area can be simulated and predicted, which eliminates the spatial blind spots of actual water ecological monitoring to a certain extent.

[0066] By constructing a basin hydrology and water quality model, the temporal and spatial distribution of river hydrodynamic conditions and water quality indicators in the basin driven by basin meteorological conditions, water resources allocation and water pollution control measures are simulated. By comparing the temporal and spatial scales of the model simulation, adapting the corresponding associated land use data, and based on the selected water ecology prediction basic model constructed earlier, the water ecology status of the basin can be predicted at a refined scale. Based on the mechanism model, the basin hydrology and water quality indicators are simulated, and then the water ecology status of the basin unit with higher temporal and spatial resolution is simulated and predicted.

[0067] This overall water ecological prediction model can be applied to a variety of scenarios, for example, predicting the water ecological conditions at the sub-basin scale in the entire basin; predicting the water ecological conditions at specific time points throughout the year; and simulating water ecological simulation and prediction with high temporal and spatial resolution at the basin scale driven by climate conditions, water resources allocation and water pollution prevention and control measures.

[0068] The water ecology prediction model generated by the method proposed in the embodiment of the present invention predicts the watershed hydrodynamic conditions and water environment quality driven by climate conditions, water resources allocation and water pollution prevention and control measures through the watershed hydrological and water quality model, and then uses it as input information for the machine learning model to predict the water ecology status of the watershed. Compared with the prior art, the function of the machine learning model has been expanded. By coupling it with the watershed hydrological and water quality model, it can achieve quantitative prediction of the impact of climate conditions, water resources allocation and water pollution prevention and control measures on the water ecology status, and then serve the optimization decision-making of watershed water ecological environment protection projects and non-engineering measures.

[0069] In the prior art, it is impossible to predict the changes in the water ecological conditions of the basin under the combined influence of the scheduling of water conservancy projects in the basin, the reduction of point source pollution load, and the reduction of non-point source pollution load. Compared with the prior art, the water ecological prediction model finally coupled and formed according to the implementation of the present invention can reflect factors such as water pollution control measures, water resources allocation in the basin, and water conservancy project scheduling. For example, the changes in point and non-point source pollution load after taking water pollution control measures, the impact of the construction of water conservancy project facilities (such as dams) on flow, and the changes in associated land use data caused by changes in the application of fertilizers in the surrounding land, etc., can predict the response of aquatic biodiversity after taking the above measures.

[0070] As an embodiment of the present invention, M aquatic biological indicators and N different initial machine learning models may be set. As an embodiment of the present invention, the M aquatic biological indicators may be selected from aquatic biological relative abundance indicators, aquatic biodiversity indicators, aquatic biological integrity indicators, etc. Specifically, they may be selected from the following groups: relative abundance indicators of aquatic organisms such as dominant species, indicator species, and sensitive species, aquatic biodiversity indicators such as richness index, Shannon index, Simpson index, Pielou uniformity index, Chao1 index, ACE index, etc., aquatic biological integrity indicators such as fish integrity index F-IBI, plankton integrity index P-IBI, benthic animal integrity index B-IBI, etc.

[0071] As one embodiment of the present invention, N different initial machine learning models are selected from the group consisting of: support vector machine, random forest, gradient boosting machine, artificial neural network, decision tree, elastic network, XGBoost, LightGBM, CatBoost.

[0072] In this embodiment, S102 may include: calculating M types of aquatic biological indicators of the collected water sample according to the environmental DNA data of the collected water sample, where M is an integer greater than 1.

[0073] S103 may include: using the water environment quality indicators of the collected water samples, the hydrodynamic conditions at the sampling points and the associated land use data and the M types of aquatic biological indicators of the collected water samples to respectively construct M groups of training sets and validation sets, wherein each group of the M groups of training sets and validation sets respectively contains a corresponding one indicator of the M types of aquatic biological indicators.

[0074] As an embodiment of the present invention, S104 may include the following steps:

[0075] Using M groups of training sets and validation sets, N different initial machine learning models are trained respectively, where N is an integer greater than 1, to obtain M*N trained candidate water ecological prediction basic models and their objective functions, where the input of the candidate water ecological prediction basic model includes hydrodynamic conditions, water environment quality indicators and related land use data, and the output of the candidate water ecological prediction basic model includes one aquatic biological indicator among the M aquatic biological indicators corresponding to the training set and validation set used for training the model;

[0076] The model with the best objective function among the M*N trained alternative water ecological prediction basic models is selected as the selected water ecological prediction basic model.

[0077] The above implementation method can determine the optimal aquatic biological index and machine learning model combination for simulating different aquatic organisms (fish, phytoplankton, zooplankton, benthic animals, etc.) by comparing the prediction results of different aquatic biological indexes and different machine learning model combinations, as a water ecology prediction model adapted to different aquatic organisms. Based on the machine learning method, a refined water ecology simulation prediction model considering the response of different aquatic organisms to environmental factors is constructed.

[0078] As another embodiment of the present invention, S104 may include the following steps:

[0079] Using M groups of training sets and validation sets, N different initial machine learning models are trained respectively, where N is an integer greater than 1, to obtain M*N trained candidate water ecological prediction basic models and their objective functions, where the input of the candidate water ecological prediction basic model includes hydrodynamic conditions, water environment quality indicators and related land use data, and the output of the candidate water ecological prediction basic model includes one aquatic biological indicator among the M aquatic biological indicators corresponding to the training set and validation set used for training the model;

[0080] For each of the M aquatic biological indicators, the N trained alternative water ecological prediction basic models corresponding to the M aquatic biological indicators are selected as the selected water ecological prediction basic model corresponding to the aquatic biological indicator, thereby obtaining M selected water ecological prediction basic models.

[0081] In this embodiment, an optimal selected water ecology prediction basic model is generated for each of the M aquatic biological indicators, thereby obtaining a total of M selected water ecology prediction basic models. This is conducive to the application of the subsequent overall model, and different selected water ecology prediction basic models can be flexibly selected for different sub-basins or regions, and for different prediction purposes and different application scenarios, or a combination of different selected water ecology prediction basic models can be selected.

[0082] As an embodiment of the present invention, sampling can be performed at predefined time intervals, and the water ecology prediction basic models corresponding to the time can be trained based on the data sampled at different times, and then the water ecology prediction models corresponding to the different times can be generated. The time interval can be, for example, a quarter, a month, or any other time interval. For example, considering that the response of aquatic organisms to seasons varies significantly, in order to reflect this difference in the simulation, the sampling period can be set to once a quarter.

[0083] In this embodiment, step S101 may include: obtaining water environment quality indicators and environmental DNA data of water samples collected at sampling points in a specific river basin at multiple specific time points according to predefined time intervals (sampling periods), as well as specific time point data of each item of hydrodynamic conditions and associated land use data at the sampling points.

[0084] Step S103 may include: using the water environment quality indicators of the collected water samples, the hydrodynamic conditions at the sampling points and the associated land use data, and the specific time point data of each of the aquatic biological indicators of the collected water samples, to respectively construct a training set and a validation set corresponding to the specific time points.

[0085] Step S104 may include: using a training set and a validation set corresponding to a specific time point to train an initial machine learning model to obtain a selected water ecology prediction basic model corresponding to the specific time point.

[0086] Water samples were collected on a seasonal scale in the study area to collect data on water body physical and chemical indicators, water ecological indicators, and land use. The response relationship model between aquatic biological index and environmental factors (water body physical and chemical indicators, buffer zone land use ratio) was trained based on machine learning. Compared with the existing technology, the seasonal differences in the response of water ecology to environmental factors were taken into account, and a water ecology prediction model at different seasonal scales was constructed, which has better explanatory power in terms of mechanism and can improve the accuracy of water ecology prediction.

[0087] On the other hand, the embodiments of the present invention provide a method for predicting water ecology by coupling a mechanism model and a machine learning model. Figure 3 , which shows a flow chart of a water ecology prediction method 300 of a coupling mechanism model and a machine learning model according to an embodiment of the present invention. The method 300 may include steps S301-S304.

[0088] In step S301, a water ecology prediction model for a specific watershed generated by the water ecology prediction model generation method based on the coupling mechanism model and the machine learning model proposed in the above-mentioned embodiment is obtained.

[0089] In step S302, the current prediction data of the hydrodynamic conditions is generated through the calibrated basin distributed hydrological model based on the currently input associated land use data, soil cover data, meteorological data, digital elevation information, water diversion data, and water conservancy project scheduling data.

[0090] In step S303, current prediction data of water environment quality indicators are generated through the calibrated basin water quality model based on the currently input point source pollution load, non-point source pollution load, and current prediction data of hydrodynamic conditions.

[0091] In step S304, by selecting a water ecology prediction basic model, current prediction data of aquatic biological indicators are generated based on currently input associated land use data and current prediction data of hydrodynamic conditions and current prediction data of water environment quality indicators.

[0092] According to the water ecology prediction method of the coupling mechanism model and the machine learning model proposed in the embodiment of the present invention, in the process of using the overall model for actual prediction, it gets rid of the dependence on measured water quantity, water quality and other data, and can predict the watershed hydrodynamic conditions and water environment quality driven by climate conditions, water resources allocation and water pollution prevention and control measures, and then use it as input information of the machine learning model to predict the water ecology status of the watershed. Compared with the prior art, the function of the machine learning model has been expanded. By coupling it with the watershed hydrology and water quality model, it can realize the quantitative prediction of the impact of climate conditions, water resources allocation and water pollution prevention and control measures on the water ecology status, and then serve the optimization decision-making of watershed water ecological environment protection projects and non-engineering measures.

[0093] As an embodiment of the present invention, the water ecology prediction method may also include: obtaining future meteorological forecast data for a specific river basin, and inputting the future meteorological forecast data into a water ecology prediction model, and obtaining current forecast data of aquatic biological indicators in response to the future meteorological forecast data via the water ecology prediction model. Through this embodiment, it is possible to achieve a quantitative prediction of the impact of climate conditions on water ecology, thereby serving the protection of water ecological environment in the river basin. For example, it is only necessary to input the forecast meteorological data for 2050 into the water ecology prediction model to predict the impact of the future weather in 2050 on water ecology.

[0094] As an embodiment of the present invention, the water ecology prediction method may also include: generating regulated water diversion data according to water resource allocation information of a specific river basin (for example, information on water diversion from outside the river basin, i.e., the amount of water from the outside), inputting the regulated water diversion data into a water ecology prediction model, and obtaining current prediction data of aquatic biological indicators in response to the water resource allocation information through the water ecology prediction model. Through this embodiment, quantitative prediction of the impact of water resource allocation on water ecology can be achieved, thereby serving the optimization decision-making of water ecological environment protection projects and non-engineering measures in the river basin.

[0095] As an embodiment of the present invention, the water ecology prediction method may also include: generating control data of point source pollution load and non-point source pollution load according to water pollution control information of a specific river basin, inputting the control data of point source pollution load and non-point source pollution load into the water ecology prediction model, and obtaining the current prediction data of aquatic biological indicators in response to the water pollution control information through the water ecology prediction model. The control data of point source pollution load comes from, for example, the reduction of pollution load of sewage discharged along the river basin, and the control data of non-point source pollution load comes from, for example, the change of fertilizer usage per unit of agricultural planting area, resource utilization of livestock and poultry breeding, etc. Through this embodiment, it is possible to achieve quantitative prediction of the impact of water pollution control measures on water ecological conditions, thereby serving the optimization decision-making of water ecological environment protection projects and non-engineering measures in the river basin.

[0096] As an embodiment of the present invention, the water ecology prediction method may also include: obtaining updated water conservancy project scheduling data, inputting the updated water conservancy project scheduling data into the water ecology prediction model, and obtaining the current prediction data of aquatic biological indicators in response to the updated water conservancy project scheduling data through the water ecology prediction model. The water conservancy project scheduling data may include, for example, the setting of sluice gates and dams, the amount of water discharged from sluice gates and dams, the data on the amount of water released through other projects, and the like. Through this embodiment, it is possible to quantitatively predict the impact of water conservancy project scheduling on water ecology, that is, to predict the impact of regulation on water ecology, and then serve the optimization decision-making of water ecological environment protection projects and non-engineering measures in the basin.

[0097] In another aspect, an embodiment of the present invention provides a device for generating a water ecology prediction model by coupling a mechanism model and a machine learning model. Figure 4 , which shows a block diagram of a water ecology prediction model generation device for coupling a mechanism model and a machine learning model according to an embodiment of the present invention. The water ecology prediction model generation device may include modules 401-410.

[0098] The sampling data acquisition module 401 may be configured to acquire water environment quality indicators and environmental DNA data of water samples collected at sampling points in a specific river basin, as well as hydrodynamic conditions and associated land use data at the sampling points.

[0099] The sampled aquatic biological indicator calculation module 402 can be configured to calculate the aquatic biological indicator of the collected water sample according to the environmental DNA data of the collected water sample.

[0100] The training set construction module 403 may be configured to construct a training set and a validation set using water environment quality indicators of the collected water samples, hydrodynamic conditions at the sampling points and associated land use data, and aquatic biological indicators of the collected water samples.

[0101] The basic model training module 404 can be configured to use the training set and the validation set to train the initial machine learning model to obtain the selected water ecological prediction basic model, wherein the input of the selected water ecological prediction basic model includes hydrodynamic conditions, water environment quality indicators and related land use data, and the output of the selected water ecological prediction basic model includes aquatic biological indicators.

[0102] The hydrological related data acquisition module 405 can be configured to obtain the current survey data of each of the associated land use data, soil cover data, meteorological data, digital elevation information, water diversion data, and water conservancy project scheduling data of a specific river basin, as well as the hydrological station measured data of the hydrodynamic conditions measured by the hydrological stations deployed in the specific river basin.

[0103] The hydrological model calibration module 406 can be configured to use the current survey data of the associated land use data, soil cover data, meteorological data, digital elevation information, water diversion data, and water conservancy project scheduling data of the specific river basin as input data to drive the basin distributed hydrological model of the specific river basin to simulate the hydrodynamic conditions, and use the measured data of the hydrological station of the hydrodynamic conditions as calibration data to calibrate the parameters of the basin distributed hydrological model of the specific river basin to obtain the calibrated basin distributed hydrological model of the specific river basin, wherein the input of the calibrated basin distributed hydrological model includes the associated land use data, soil cover data, meteorological data, digital elevation information, water diversion data, and water conservancy project scheduling data, and the output of the calibrated basin distributed hydrological model includes the hydrodynamic conditions.

[0104] The water quality related data acquisition module 407 can be configured to obtain the current actual data of point source pollution load and non-point source pollution load in a specific river basin, as well as the monitoring data of water environment quality indicators monitored by water quality monitoring stations deployed in the specific river basin.

[0105] The watershed hydrodynamic condition prediction module 408 can be configured to predict the watershed hydrodynamic conditions of a specific watershed through a calibrated watershed distributed hydrological model, based on the current actual survey data of the specific watershed's associated land use data, soil cover data, meteorological data, digital elevation information, water diversion data, and water conservancy project scheduling data, so as to obtain prediction data for the driving water quality model of the watershed hydrodynamic conditions.

[0106] The water quality model calibration module 409 can be configured to use the current actual inspection data of the point source pollution load and non-point source pollution load of a specific river basin, and the predicted data of the driving water quality model of the river basin hydrodynamic conditions as input data to drive the river basin water quality model of the specific river basin to simulate the water environment quality indicators, and use the monitoring data of the water environment quality indicators as calibration data to calibrate the parameters of the river basin water quality model of the specific river basin to obtain the calibrated river basin water quality model of the specific river basin, wherein the input of the calibrated river basin water quality model includes point source pollution load, non-point source pollution load and hydrodynamic conditions, and the output of the calibrated river basin water quality model includes water environment quality indicators.

[0107] The water ecology prediction model generation module 410 can be configured to generate a water ecology prediction model for a specific river basin based on a selected water ecology prediction basic model, a calibrated river basin distributed hydrological model and a calibrated river basin water quality model, wherein the calibrated river basin distributed hydrological model is used to generate current prediction data of hydrodynamic conditions based on currently input associated land use data, soil cover data, meteorological data, digital elevation information, water diversion data, and water conservancy project scheduling data; the calibrated river basin water quality model is used to generate current prediction data of water environment quality indicators based on currently input point source pollution load, non-point source pollution load, and current prediction data of hydrodynamic conditions; the selected water ecology prediction basic model is used to generate current prediction data of aquatic biological indicators based on currently input associated land use data, current prediction data of hydrodynamic conditions, and current prediction data of water environment quality indicators.

[0108] It should be noted that the functions implemented by each module in the water ecological prediction model generation device of the coupling mechanism model and machine learning model proposed in the embodiment of the present invention correspond one-to-one to the various steps of the water ecological prediction model generation method of the coupling mechanism model and machine learning model described above. For its specific implementation methods, examples and beneficial effects, please refer to the above description of the method.

[0109] In another aspect, an embodiment of the present invention provides a water ecology prediction device that couples a mechanism model and a machine learning model. Figure 5 , which shows a block diagram of a water ecology prediction device of a coupling mechanism model and a machine learning model according to an embodiment of the present invention. The water ecology prediction device may include modules 501-504.

[0110] The water ecology prediction model acquisition module 501 may be configured to acquire a water ecology prediction model of a specific watershed generated by the water ecology prediction model generation device.

[0111] The hydrodynamic condition prediction module 502 can be configured to generate current prediction data of hydrodynamic conditions through a calibrated basin distributed hydrological model based on currently input associated land use data, soil cover data, meteorological data, digital elevation information, water diversion data, and water conservancy project scheduling data.

[0112] The water environment quality index prediction module 503 can be configured to generate current prediction data of water environment quality indicators based on the currently input point source pollution load, non-point source pollution load, and current prediction data of hydrodynamic conditions via a calibrated watershed water quality model.

[0113] The aquatic biological indicator prediction module 504 can be configured to generate current prediction data of aquatic biological indicators based on currently input associated land use data and current prediction data of hydrodynamic conditions and current prediction data of water environment quality indicators via the selected water ecological prediction basic model.

[0114] It should be noted that the functions implemented by each module in the water ecological prediction device of the coupling mechanism model and machine learning model proposed in the embodiment of the present invention correspond one-to-one to the various steps of the water ecological prediction method of the coupling mechanism model and machine learning model described above. For its specific implementation methods, examples and beneficial effects, please refer to the above description of the method.

[0115] On the other hand, an embodiment of the present invention proposes a device, which includes a memory and a processor, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the method described in any of the above embodiments is implemented, including a method for generating a water ecological prediction model by coupling a mechanism model and a machine learning model, and a method for predicting water ecology by coupling a mechanism model and a machine learning model.

[0116] On the other hand, an embodiment of the present invention proposes a storage medium storing computer-readable instructions, which, when executed by a processor, executes the method described in any of the above embodiments, including a method for generating a water ecological prediction model by coupling a mechanism model and a machine learning model and a method for predicting water ecology by coupling a mechanism model and a machine learning model.

[0117] To make decisions on the planning of water ecological environment protection in a river basin, it is necessary to evaluate the implementation effects of various water resources allocation, water pollution control and water ecological restoration projects and non-engineering measures, evaluate the degree of achievement of water ecological environment protection goals under different combinations of measures, and then optimize various engineering and non-engineering measures. Biodiversity indicators are important indicators to characterize the health of water ecology in a river basin. Constructing a mathematical model that characterizes the response of aquatic biodiversity indicators to various water ecological environment protection projects and non-engineering measures in a river basin has important scientific value and practical significance for the scientific formulation of water ecological environment protection plans in a river basin.

[0118] Mathematical models can simulate the variables of interest continuously in time and space. Compared with one or more local sampling in the basin, model simulation has obvious advantages in improving the temporal and spatial resolution. The relevant variables affecting aquatic biodiversity are used as input variables to simulate and predict the spatial heterogeneity and temporal variability of the distribution of aquatic biodiversity in the river network in the basin, which can reduce the direct and indirect economic costs of directly monitoring the water ecological status. At the same time, the model can also scientifically evaluate the impact of various watershed water ecological environment protection projects and non-engineering measures on the aquatic biodiversity of the basin after implementation. There are many factors that affect the aquatic biodiversity of the basin, and the mechanism of the evolution of the water ecological status of the basin is unclear, so it is difficult to establish a mathematical model based on the impact mechanism. Machine learning models that are completely data-driven have been widely used to predict water ecological conditions. However, the input variables in the machine learning models in the prior art, especially the hydrodynamic conditions and water environment quality, are still obtained through field observations, spatially dependent on monitoring points, and temporally dependent on monitoring cycles, without improving the temporal and spatial resolution of water ecological prediction. At the same time, the changes in hydrodynamic conditions and water environment quality are driven by the relationship between rainfall and runoff, water resources allocation and water pollution control measures in the basin. The current machine learning model cannot reflect these watershed water ecological environment protection projects and non-engineering measures, which affect the hydrodynamic conditions and water environment quality, and then affect the process of aquatic biodiversity. The basin hydrological model with a physical mechanism coupled with a water quality model can realize the continuous estimation of river hydrodynamic conditions and water environment quality in time and space in the basin, provide more abundant data input for water ecological prediction, and improve the time and space continuity of water ecological prediction. To this end, the embodiment of the present invention innovatively couples the mechanism model with the machine learning model, and proposes a water ecology prediction method that couples the mechanism model and the machine learning model. First, the water ecology simulation prediction model is trained based on field sampling data, and then the hydrological and hydrodynamic conditions and water environment variables simulated based on the physical mechanism model are used as input to predict the water ecology status with higher temporal and spatial resolution in the basin, and the continuous estimation of the water ecology status in time and space is realized, which can predict the impact of water resources allocation and water pollution control measures on the water ecology status in the basin.

[0119] In the prior art, it is impossible to predict the changes in the water ecological conditions of the basin under the combined influence of the scheduling of water conservancy projects in the basin, the reduction of point source pollution load, and the reduction of non-point source pollution load. Compared with the prior art, the water ecological prediction model finally coupled and formed according to the implementation of the present invention can reflect factors such as water pollution control measures, water resources allocation in the basin, and water conservancy project scheduling. For example, the changes in point and non-point source pollution load after taking water pollution control measures, the impact of the construction of water conservancy project facilities (such as dams) on flow, and the changes in associated land use data caused by changes in the application of fertilizers in the surrounding land, etc., can predict the response of aquatic biodiversity after taking the above measures.

[0120] For the purpose of illustration, the foregoing description of the embodiments of the present invention has been given, which is not exhaustive nor intended to limit the present invention to the disclosed exact form. It will be appreciated by those skilled in the art that various changes can be made without departing from the scope of the present invention, and that the elements therein can be replaced with equivalents. In addition, without departing from the basic scope of the present invention, many modifications can be made so that specific situations or materials are adapted to the teachings of the present invention. Therefore, the present invention is not intended to be limited to the specific embodiments disclosed as the best mode for realizing the present invention, and the present invention will include all embodiments falling within the scope of the appended claims.

Claims

1. A method for generating a water ecological prediction model by coupling a mechanism model and a machine learning model, characterized in that: The method comprises: Obtain water environment quality indicators and environmental DNA data of water samples collected at sampling points in a specific watershed, as well as hydrodynamic conditions and associated land use data at the sampling points; Calculating aquatic biological indicators of the collected water samples according to the environmental DNA data of the collected water samples; Constructing a training set and a validation set using the water environment quality indicators of the collected water samples, the hydrodynamic conditions and associated land use data at the sampling points, and the aquatic biological indicators of the collected water samples; Using the training set and the validation set, the initial machine learning model is trained to obtain a selected water ecology prediction basic model, wherein the input of the selected water ecology prediction basic model includes hydrodynamic conditions, water environment quality indicators and associated land use data, and the output of the selected water ecology prediction basic model includes aquatic biological indicators; Obtaining current survey data of each of the associated land use data, soil cover data, meteorological data, digital elevation information, water diversion data, and water conservancy project dispatching data of the specific river basin, as well as hydrological station measured data of the hydrodynamic conditions measured by the hydrological stations deployed in the specific river basin; Using the current survey data of the associated land use data, soil cover data, meteorological data, digital elevation information, water diversion data, and water conservancy project scheduling data of the specific river basin as input data to drive the river basin distributed hydrological model of the specific river basin to simulate the hydrodynamic conditions, and using the measured data of the hydrological station of the hydrodynamic conditions as calibration data to calibrate the parameters of the river basin distributed hydrological model of the specific river basin, to obtain the calibrated river basin distributed hydrological model of the specific river basin, wherein the input of the calibrated river basin distributed hydrological model includes the associated land use data, soil cover data, meteorological data, digital elevation information, water diversion data, and water conservancy project scheduling data, and the output of the calibrated river basin distributed hydrological model includes the hydrodynamic conditions; Obtaining current survey data on point source pollution load and non-point source pollution load in the specific river basin, as well as monitoring data on water environment quality indicators monitored by water quality monitoring stations deployed in the specific river basin; Through the calibrated basin distributed hydrological model, based on the current survey data of the associated land use data, soil cover data, meteorological data, digital elevation information, water diversion data, and water conservancy project scheduling data of the specific basin, the basin hydrodynamic conditions of the specific basin are predicted to obtain prediction data for the driving water quality model of the basin hydrodynamic conditions; Using the current actual inspection data of the point source pollution load and non-point source pollution load of the specific river basin and the prediction data of the driving water quality model of the water dynamic conditions of the river basin as input data, the river basin water quality model of the specific river basin is driven to simulate the water environment quality index, and the monitoring data of the water environment quality index is used as calibration data to calibrate the parameters of the river basin water quality model of the specific river basin, so as to obtain the calibrated river basin water quality model of the specific river basin, wherein the input of the calibrated river basin water quality model includes the point source pollution load, the non-point source pollution load and the hydrodynamic conditions, and the output of the calibrated river basin water quality model includes the water environment quality index; Based on the selected water ecological prediction basic model, the calibrated watershed distributed hydrological model and the calibrated watershed water quality model, the water ecological prediction model of the specific watershed is generated, wherein the calibrated watershed distributed hydrological model is used to generate current prediction data of hydrodynamic conditions based on currently input associated land use data, soil cover data, meteorological data, digital elevation information, water diversion data, and water conservancy project scheduling data; the calibrated watershed water quality model is used to generate current prediction data of water environment quality indicators based on currently input point source pollution load, non-point source pollution load, and the current prediction data of hydrodynamic conditions; the selected water ecological prediction basic model is used to generate current prediction data of aquatic biological indicators based on currently input associated land use data, the current prediction data of hydrodynamic conditions, and the current prediction data of water environment quality indicators.

2. The method according to claim 1, characterized in that Calculating the aquatic biological index of the collected water sample according to the environmental DNA data of the collected water sample includes: calculating M aquatic biological indexes of the collected water sample according to the environmental DNA data of the collected water sample, wherein M is an integer greater than 1, Using the water environment quality index of the collected water sample, the hydrodynamic conditions at the sampling point and the associated land use data, and the aquatic biological index of the collected water sample to construct a training set and a validation set includes: using the water environment quality index of the collected water sample, the hydrodynamic conditions at the sampling point and the associated land use data, and the M aquatic biological indexes of the collected water sample to respectively construct M groups of training sets and validation sets, wherein each of the M groups of training sets and validation sets respectively contains a corresponding one of the M aquatic biological indexes, The initial machine learning model is trained using the training set and the validation set to obtain the selected water ecology prediction basic model including: Using the M groups of training sets and validation sets, N different initial machine learning models are trained respectively, where N is an integer greater than 1, to obtain M*N trained candidate water ecology prediction basic models and their objective functions, wherein the input of the candidate water ecology prediction basic model includes hydrodynamic conditions, water environment quality indicators and associated land use data, and the output of the candidate water ecology prediction basic model includes one aquatic biological indicator among the M aquatic biological indicators corresponding to the training set and validation set used for training the model; The model with the best objective function among the M*N trained candidate water ecology prediction basic models is selected as the selected water ecology prediction basic model.

3. The method according to claim 1, characterized in that: Calculating the aquatic biological index of the collected water sample according to the environmental DNA data of the collected water sample includes: calculating M aquatic biological indexes of the collected water sample according to the environmental DNA data of the collected water sample, wherein M is an integer greater than 1, Using the water environment quality index of the collected water sample, the hydrodynamic conditions at the sampling point and the associated land use data, and the aquatic biological index of the collected water sample to construct a training set and a validation set includes: using the water environment quality index of the collected water sample, the hydrodynamic conditions at the sampling point and the associated land use data, and the M aquatic biological indexes of the collected water sample to respectively construct M groups of training sets and validation sets, wherein each of the M groups of training sets and validation sets respectively contains a corresponding one of the M aquatic biological indexes, The initial machine learning model is trained using the training set and the validation set to obtain the selected water ecology prediction basic model including: Using the M groups of training sets and validation sets, N different initial machine learning models are trained respectively, where N is an integer greater than 1, to obtain M*N trained candidate water ecology prediction basic models and their objective functions, wherein the input of the candidate water ecology prediction basic model includes hydrodynamic conditions, water environment quality indicators and associated land use data, and the output of the candidate water ecology prediction basic model includes one aquatic biological indicator among the M aquatic biological indicators corresponding to the training set and validation set used for training the model; For each of the M aquatic biological indicators, the N trained alternative water ecological prediction basic models corresponding to the M aquatic biological indicators are selected as the selected water ecological prediction basic model corresponding to the aquatic biological indicator, thereby obtaining M selected water ecological prediction basic models.

4. The method according to claim 1, characterized in that: The aquatic biological index is selected from the group consisting of: dominant species, indicator species, sensitive species, richness index, Shannon index, Simpson index, pielou evenness index, Chao1 index, ACE index, fish integrity index F-IBI, plankton integrity index P-IBI, benthic animal integrity index B-IBI, and / or The initial machine learning model is selected from the group consisting of: support vector machine, random forest, gradient boosting machine, artificial neural network, decision tree, elastic network, XGBoost, LightGBM, CatBoost.

5. The method according to claim 1, characterized in that Acquiring water environment quality indicators and environmental DNA data of water samples collected at sampling points in a specific river basin, as well as hydrodynamic conditions and associated land use data at the sampling points, comprises: acquiring water environment quality indicators and environmental DNA data of water samples collected at sampling points in the specific river basin, as well as specific time point data of each of the hydrodynamic conditions and associated land use data at the sampling points at a plurality of specific time points according to a predefined time interval, Using the water environment quality index of the collected water sample, the hydrodynamic conditions at the sampling point and the associated land use data, and the aquatic biological index of the collected water sample to construct a training set and a validation set includes: using the specific time point data of each of the water environment quality index of the collected water sample, the hydrodynamic conditions at the sampling point and the associated land use data, and the aquatic biological index of the collected water sample to respectively construct a training set and a validation set corresponding to the specific time point, Using the training set and the validation set to train the initial machine learning model to obtain the selected water ecology prediction basic model includes: using the training set and the validation set corresponding to the specific time point to train the initial machine learning model to obtain the selected water ecology prediction basic model corresponding to the specific time point.

6. A method for water ecology prediction based on a coupling mechanism model and a machine learning model, characterized in that: The method comprises: Obtaining a water ecology prediction model for a specific watershed generated according to any one of the methods of claims 1-5; Generate current forecast data of hydrodynamic conditions through the calibrated distributed hydrological model of the watershed based on the current input of associated land use data, soil cover data, meteorological data, digital elevation information, water diversion data, and water conservancy project scheduling data; Generate current forecast data of water environment quality indicators based on currently input point source pollution load, non-point source pollution load, and current forecast data of the hydrodynamic conditions through the calibrated watershed water quality model; By selecting a water ecological prediction basic model, current prediction data of aquatic biological indicators are generated based on currently input related land use data and current prediction data of the hydrodynamic conditions and current prediction data of the water environment quality indicators.

7. The method according to claim 6, characterized in that The method further comprises one or more of the following steps: Acquire future meteorological forecast data of the specific river basin, input the future meteorological forecast data into the water ecology forecast model, and acquire current forecast data of aquatic biological indicators in response to the future meteorological forecast data via the water ecology forecast model; Generate regulated water diversion data according to the water resource allocation information of the specific river basin, input the regulated water diversion data into the water ecology prediction model, and obtain current prediction data of aquatic biological indicators responsive to the water resource allocation information through the water ecology prediction model; Generate point source pollution load and non-point source pollution load control data according to the water pollution control information of the specific river basin, input the point source pollution load and non-point source pollution load control data into the water ecology prediction model, and obtain current prediction data of aquatic biological indicators in response to the water pollution control information through the water ecology prediction model; Updated water conservancy project scheduling data is obtained, the updated water conservancy project scheduling data is input into the water ecology prediction model, and current prediction data of aquatic biological indicators in response to the updated water conservancy project scheduling data is obtained via the water ecology prediction model.

8. A device for generating a water ecology prediction model by coupling a mechanism model and a machine learning model, characterized in that: The device comprises: A sampling data acquisition module configured to acquire water environment quality indicators and environmental DNA data of water samples collected at sampling points in a specific river basin, as well as hydrodynamic conditions and associated land use data at the sampling points; A sampling aquatic biological index calculation module, configured to calculate the aquatic biological index of the collected water sample according to the environmental DNA data of the collected water sample; A training set construction module is configured to construct a training set and a validation set using the water environment quality index of the collected water sample, the hydrodynamic conditions and associated land use data at the sampling point, and the aquatic biological index of the collected water sample; A basic model training module is configured to train the initial machine learning model using the training set and the validation set to obtain a selected water ecology prediction basic model, wherein the input of the selected water ecology prediction basic model includes hydrodynamic conditions, water environment quality indicators and associated land use data, and the output of the selected water ecology prediction basic model includes aquatic biological indicators; A hydrological data acquisition module is configured to acquire current survey data of each of the associated land use data, soil cover data, meteorological data, digital elevation information, water diversion data, and water conservancy project scheduling data of the specific river basin, as well as hydrological station measured data of the hydrodynamic conditions measured by the hydrological stations deployed in the specific river basin; A hydrological model calibration module is configured to use the current survey data of the associated land use data, soil cover data, meteorological data, digital elevation information, water diversion data, and water conservancy project scheduling data of the specific river basin as input data to drive the basin distributed hydrological model of the specific river basin to simulate the hydrodynamic conditions, and use the measured data of the hydrological station of the hydrodynamic conditions as calibration data to calibrate the parameters of the basin distributed hydrological model of the specific river basin, so as to obtain the calibrated basin distributed hydrological model of the specific river basin, wherein the input of the calibrated basin distributed hydrological model includes the associated land use data, soil cover data, meteorological data, digital elevation information, water diversion data, and water conservancy project scheduling data, and the output of the calibrated basin distributed hydrological model includes the hydrodynamic conditions; A water quality related data acquisition module, configured to acquire current actual inspection data of point source pollution load and non-point source pollution load of the specific river basin, and monitoring data of water environment quality indicators monitored by water quality monitoring stations deployed in the specific river basin; A watershed hydrodynamic condition prediction module is configured to predict the watershed hydrodynamic conditions of the specific watershed through the calibrated watershed distributed hydrological model, based on the current actual data of the associated land use data, soil cover data, meteorological data, digital elevation information, water diversion data, and water conservancy project scheduling data of the specific watershed, and obtain prediction data for the driving water quality model of the watershed hydrodynamic conditions; A water quality model calibration module is configured to use the current actual inspection data of the point source pollution load and non-point source pollution load of the specific river basin and the prediction data of the driving water quality model of the watershed hydrodynamic conditions as input data to drive the watershed water quality model of the specific river basin to simulate the water environment quality index, and use the monitoring data of the water environment quality index as calibration data to calibrate the parameters of the watershed water quality model of the specific river basin to obtain the calibrated watershed water quality model of the specific river basin, wherein the input of the calibrated watershed water quality model includes the point source pollution load, the non-point source pollution load and the hydrodynamic conditions, and the output of the calibrated watershed water quality model includes the water environment quality index; The water ecology prediction model generation module is configured to generate the water ecology prediction model of the specific watershed based on the selected water ecology prediction basic model, the calibrated watershed distributed hydrological model and the calibrated watershed water quality model, wherein the calibrated watershed distributed hydrological model is used to generate current prediction data of hydrodynamic conditions based on currently input associated land use data, soil cover data, meteorological data, digital elevation information, water diversion data, and water conservancy project scheduling data; the calibrated watershed water quality model is used to generate current prediction data of water environment quality indicators based on currently input point source pollution load, non-point source pollution load, and the current prediction data of hydrodynamic conditions; the selected water ecology prediction basic model is used to generate current prediction data of aquatic biological indicators based on currently input associated land use data, the current prediction data of hydrodynamic conditions, and the current prediction data of water environment quality indicators.

9. A water ecology prediction device that couples a mechanism model and a machine learning model, characterized in that: The device comprises: a water ecology prediction model acquisition module, configured to acquire a water ecology prediction model of a specific watershed generated by the device according to claim 7; A hydrodynamic condition prediction module configured to generate current prediction data of hydrodynamic conditions based on currently input associated land use data, soil cover data, meteorological data, digital elevation information, water diversion data, and water conservancy project scheduling data via a calibrated basin distributed hydrological model; A water environment quality index prediction module is configured to generate current prediction data of water environment quality indexes based on currently input point source pollution load, non-point source pollution load, and current prediction data of the hydrodynamic conditions through a calibrated watershed water quality model; The aquatic biological indicator prediction module is configured to generate current prediction data of aquatic biological indicators based on currently input related land use data and current prediction data of the hydrodynamic conditions and current prediction data of the water environment quality indicators through selecting a water ecological prediction basic model.

10. A storage medium, characterized in that: Computer-readable instructions are stored, and when the instructions are executed by a processor, the method according to any one of claims 1 to 7 is executed.

Citation Information

Patent Citations

  • Method and device for simulating and predicting drainage basin water environment based on multi-model coupling

    CN114117848A

  • Lake ecological hydraulic regulation and control method and system based on hydrological-ecological response relationship

    CN116630122A

  • Accessing agriculture productivity and sustainability

    US20220061236A1

Cited By

  • Water habitat multi-granularity data alignment method and device based on machine learning, computer equipment and readable storage medium

    CN119961769A

  • Basin hydrological situation change attribution quantification method

    CN121365881A

  • Water ecological environment deep learning prediction method based on space-time heterogeneity nesting

    CN121561439A

  • Multi-input multi-output river water ecological integrity prediction method

    CN122452437A