CDOM prediction method based on multi-source driving factor and related device

Abstract

The invention discloses a CDOM prediction method based on a multi-source driving factor and a related device, and relates to the technical field of water environment governance, and the method comprises the steps: carrying out the training based on an actually-measured CDOM concentration value and a remote sensing monitoring image corresponding to the time-space of the actually-measured CDOM concentration value, and obtaining a trained CDOM inversion model, according to CDOM data obtained through long-sequence remote sensing monitoring image inversion and long-sequence CDOM inversion data, a trained CDOM prediction model is obtained through training on the basis of the long-sequence CDOM inversion data and historical multi-source driving factors, and the CDOM prediction model can predict the target water body catchment area according to the multi-source driving factors of the target water body catchment area in the historical time period. And predicting and outputting CDOM prediction data of the target water catchment area in the prediction time period. According to the method, the LSTM model is utilized to learn the nonlinear coupling influence of the multi-element complex driving factors on the CDOM, so that the prediction of the CDOM under the influence of the multi-element complex driving factors can be accurately realized.

Description

Technical Field

[0001] The present application relates to the technical field of water environment management, and in particular to a CDOM prediction method and related devices based on multi-source driving factors. Background Art

[0002] With the development of the economy and society, the scale of traditional industry, agriculture, aquaculture and urban activities has continued to expand, which has led to a significant increase in the terrestrial input of colored dissolved organic matter (CDOM). CDOM mainly originates from the decomposition of terrestrial vegetation, the leaching of soil organic matter, and the discharge of industrial and agricultural wastewater and domestic sewage. It flows into rivers, lakes and offshore areas through surface runoff and groundwater infiltration, significantly changing the optical properties and biogeochemical cycle processes of the water body. At the same time, as an important carbon carrier, the spatiotemporal dynamics of CDOM have a profound impact on regional carbon budget assessment and global climate change research. Therefore, understanding the long-term evolution of CDOM and its driving forces, and achieving accurate prediction of the spatiotemporal changes of CDOM under different driving factors, provide key technical support for watershed pollution load early warning, dynamic assessment of carbon flux and optimization of "dual carbon" pathways.

[0003] Currently, research on the monitoring of colored dissolved organic matter (CDOM) in lakes mainly focuses on analyzing its spatiotemporal distribution characteristics and historical changes through methods such as remote sensing inversion, in-situ sensor networks, or periodic sampling. For example, CDOM concentrations are inverted based on spectral characteristics, and terrestrial input fluxes are traced in combination with hydrological data. However, under the dual pressures of future climate change and human activities, forward-looking prediction methods for the dynamic evolution of lake CDOM are still insufficient. This limitation mainly stems from two aspects: first, the generation, migration, and transformation of CDOM involve complex biogeochemical processes (such as leaching of terrestrial organic matter, microbial degradation, photochemical reactions, etc.), and its future behavior is affected by the nonlinear coupling of multiple factors such as changes in precipitation patterns, land use transitions, and carbon cycle disturbances. Existing mechanistic models have uncertainties in describing these complex physical processes; second, the lack of long-term, high-resolution monitoring data limits the in-depth exploration of the dynamic laws of CDOM, resulting in insufficient generalization of prediction models.

[0004] Therefore, it is particularly necessary to achieve accurate prediction of CDOM under the influence of multiple complex driving factors and provide key technical support for watershed pollution load early warning, dynamic assessment of carbon flux and optimization of "dual carbon" pathways. Summary of the Invention

[0005] The purpose of this application is to provide a CDOM prediction method and related devices based on multi-source driving factors, which can accurately predict CDOM under the influence of multiple and complex driving factors.

[0006] To achieve the above objectives, this application provides the following solutions:

[0007] In a first aspect, the present application provides a CDOM prediction method based on multi-source driving factors, comprising the following steps:

[0008] Obtain the multi-source driving factors of the target water body catchment area in the historical period; the multi-source driving factors include temperature data, rainfall data, wind speed data, runoff data and pollution source data.

[0009] The multi-source driving factors are input into the trained CDOM prediction model to output the CDOM prediction data of the target water body catchment during the prediction period; the trained CDOM prediction model is an LSTM model trained based on long-sequence CDOM inversion data and historical multi-source driving factors; the long-sequence CDOM inversion data is the CDOM data obtained by inverting the trained CDOM inversion model based on the long-sequence remote sensing monitoring images; the trained CDOM inversion model is a linear regression model trained based on the measured CDOM concentration values ​​and the remote sensing monitoring images corresponding to the measured CDOM concentration values ​​in time and space.

[0010] Optionally, the method further comprises the following steps:

[0011] Several detection points were selected in the target water catchment area, and the actual CDOM concentration value, corresponding detection time and point spatial coordinates of each detection point were measured.

[0012] For any detection point, the remote sensing water surface reflectance corresponding to the measured CDOM concentration value is obtained in the remote sensing monitoring image near the detection time according to the spatial coordinates of the detection point.

[0013] Based on the measured CDOM concentration values ​​at each detection point and the corresponding remote sensing water surface reflectance, an inversion training sample set and an inversion verification sample set are constructed; both the inversion training sample set and the inversion verification sample set include several inversion data pair samples; the inversion data pair sample includes the measured CDOM concentration value at a detection point and the corresponding remote sensing water surface reflectance.

[0014] Based on the inversion training sample set, the remote sensing water surface reflectance was used as input and the corresponding CDOM concentration value was used as output, and several linear regression models were trained respectively; the linear regression models included stepwise linear regression model, back propagation neural network and support vector machine model.

[0015] Each linear regression model was evaluated based on the inversion verification sample set, and the linear regression model with the best inversion effect was used as the trained CDOM inversion model.

[0016] Optionally, before obtaining the remote sensing water surface reflectance corresponding to the measured CDOM concentration value in the remote sensing monitoring image near the detection time based on the point spatial coordinates of the detection point, the following steps are also included:

[0017] Atmospheric correction is performed on all acquired remote sensing monitoring images to eliminate the effects of atmospheric absorption and scattering on spectral reflectance.

[0018] Optionally, the coefficient of determination and the relative root mean square error are selected to evaluate each linear regression model, and the linear regression model with the largest coefficient of determination and the smallest relative root mean square error among several linear regression models is used as the trained CDOM inversion model.

[0019] Optionally, the method further comprises the following steps:

[0020] Obtain long-sequence remote sensing monitoring images of the target water body's catchment area at various historical dates.

[0021] Through the trained CDOM inversion model, the CDOM inversion data of the target water body catchment area on various historical dates are obtained based on long-sequence remote sensing monitoring images.

[0022] For the CDOM inversion data of the target water catchment area on any historical date, the historical multi-source driving factors of several dates before the date are obtained to construct a sample of predicted data pairs.

[0023] Based on the prediction data of the target water body catchment area on various historical dates, a prediction training sample set and a prediction verification sample set are constructed.

[0024] Based on the prediction training sample set, the historical multi-source driving factors are used as input and the CDOM inversion data is used as the target output to train the LSTM model and obtain a trained LSTM model.

[0025] The trained LSTM model is evaluated based on the prediction verification sample set. If the prediction accuracy meets the preset conditions, the trained CDOM prediction model is obtained.

[0026] If the prediction accuracy does not meet the preset conditions, use the Bayesian optimization algorithm to adjust the hyperparameters of the LSTM model and jump to the step "Based on the prediction training sample set, use the historical multi-source driving factors as input and the CDOM inversion data as the target output to train the LSTM model and obtain a trained LSTM model."

[0027] Optionally, the LSTM model structure includes an input layer, a hidden layer and an output layer; the hidden layer module is composed of multiple neural units; each neural unit includes a forget gate, an input gate and an output gate; the hyperparameters of the LSTM model include the number of hidden layers, the number of neurons and the learning rate.

[0028] In a second aspect, the present application provides a CDOM prediction system based on multiple driving factors, including the following modules:

[0029] The multi-source driving factor acquisition module is used to obtain the multi-source driving factors of the target water body catchment area in the historical period; the multi-source driving factors include temperature data, rainfall data, wind speed data, runoff data and pollution source data.

[0030] The CDOM prediction module is used to input multi-source driving factors into a trained CDOM prediction model and output CDOM prediction data for the target water body catchment area during the prediction period; the trained CDOM prediction model is an LSTM model trained based on long-sequence CDOM inversion data and historical multi-source driving factors; the long-sequence CDOM inversion data is CDOM data obtained by inverting the trained CDOM inversion model based on long-sequence remote sensing monitoring images; the trained CDOM inversion model is a linear regression model trained based on measured CDOM concentration values ​​and remote sensing monitoring images corresponding to the measured CDOM concentration values ​​in time and space.

[0031] In a third aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the CDOM prediction method based on multi-source driving factors described above.

[0032] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the CDOM prediction method based on multi-source driving factors described above.

[0033] In a fifth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the steps of the CDOM prediction method based on multi-source driving factors described above.

[0034] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0035] The present application provides a CDOM prediction method based on multi-source driving factors and a related device. In this method, a trained CDOM inversion model is obtained by training based on measured CDOM concentration values ​​and remote sensing monitoring images corresponding to the measured CDOM concentration values ​​in time and space. Long-sequence CDOM inversion data is obtained based on CDOM data inverted from long-sequence remote sensing monitoring images. A trained CDOM prediction model is obtained by training based on the long-sequence CDOM inversion data and historical multi-source driving factors. The CDOM prediction model can predict and output CDOM prediction data for a target water body catchment area in a prediction period based on the multi-source driving factors of the target water body catchment area in a historical period. This application uses an LSTM model to learn the nonlinear coupling effects of multiple complex driving factors on CDOM based on temperature data, rainfall data, wind speed data, runoff data and pollution source data for historical periods, thereby accurately predicting CDOM under the influence of multiple complex driving factors. Moreover, since the CDOM prediction model is trained based on a long sequence of CDOM inversion data, it can deeply explore the dynamic laws of CDOM, improve the generalization ability of the CDOM prediction model, and provide key technical support for watershed pollution load warning, carbon flux dynamic assessment and "dual carbon" path optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0037] Figure 1 A flowchart of a CDOM prediction method based on multi-source driving factors provided in one embodiment of the present application.

[0038] Figure 2 This is a flowchart for constructing and training a CDOM inversion model in a CDOM prediction method based on multiple driving factors provided in one embodiment of the present application.

[0039] Figure 3 This is a regional schematic diagram of Chagan Lake in a CDOM prediction method based on multi-source driving factors provided in one embodiment of the present application.

[0040] Figure 4 This is a scatter diagram of three linear regression models SVR, SLR, and BP obtained by training in a CDOM prediction method based on multi-source driving factors provided in one embodiment of the present application.

[0041] Figure 5This is a flowchart of constructing and training a CDOM prediction model in a CDOM prediction method based on multiple driving factors provided in one embodiment of the present application.

[0042] Figure 6 A schematic diagram of a CDOM prediction model based on LSTM in a CDOM prediction method based on multi-source driving factors provided in one embodiment of the present application.

[0043] Figure 7 This is a schematic diagram of the functional modules of a CDOM prediction system based on multi-source driving factors provided in another embodiment of the present application.

[0044] Figure 8 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0045] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0046] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0047] The present application provides a CDOM prediction method based on multiple driving factors. In an exemplary embodiment, Figure 1 As shown, the following steps are included:

[0048] A1. Obtain the multi-source driving factors for the target water catchment area over a historical period. These factors include temperature data, rainfall data, wind speed data, runoff data, and pollution source data. Pollution source data include agricultural fertilization, livestock and poultry breeding, and sewage treatment plant discharge.

[0049] A2. Input the multi-source driving factors into the trained CDOM prediction model and output the CDOM prediction data for the target water body catchment during the prediction period. The trained CDOM prediction model is an LSTM model trained based on long-sequence CDOM inversion data and historical multi-source driving factors. The long-sequence CDOM inversion data is CDOM data inverted from long-sequence remote sensing monitoring images by the trained CDOM inversion model. The trained CDOM inversion model is a linear regression model trained based on measured CDOM concentration values ​​and remote sensing monitoring images corresponding to the measured CDOM concentration values ​​in time and space.

[0050] In an exemplary embodiment, the method further includes a process of constructing and training a CDOM inversion model, such as Figure 2 As shown, the following steps are included:

[0051] B1. Select several detection points in the target water catchment area and measure the actual CDOM concentration value, corresponding detection time and point spatial coordinates of each detection point. Specifically, Figure 3 As shown in the figure, Chagan Lake was selected as the research object in this study. Monitoring points were set at intervals of 500-1000 meters and 1 km from the shore. Three sampling sessions were conducted between May and November 2021 (covering spring, summer, and autumn), with 60 groups collected in each session, resulting in a total of 180 sets of field-measured CDOM data, meeting the requirements for establishing the CDOM inversion model.

[0052] B2. For any detection point, based on the detection point's spatial coordinates, obtain the remotely sensed water surface reflectance corresponding to the measured CDOM concentration value in the remote sensing monitoring imagery near the detection time. Specifically, based on the detection time and point spatial coordinate information, download OLCI and MODIS remote sensing data covering the corresponding time period (±3 hours) for Chagan Lake. Before extracting the reflectance data, to reduce external influences, the method of this embodiment further includes the following steps:

[0053] C1. Atmospheric correction was performed on all acquired remote sensing monitoring images to eliminate the effects of atmospheric absorption and scattering on spectral reflectance. In this example, SeaDAS software was used to perform atmospheric correction on both data sets to obtain the true water surface reflectance after eliminating the effects of atmospheric absorption and scattering. Ultimately, 180 matching data pairs were obtained, comparing measured CDOM concentrations with corresponding temporal and spatially matched OLCI and MODIS imagery for water surface reflectance.

[0054] B3. Construct an inversion training sample set and an inversion validation sample set based on the measured CDOM concentration values ​​at each detection point and the corresponding remotely sensed water surface reflectance. Each inversion training sample set and the inversion validation sample set include several inversion data pairs. An inversion data pair consists of the measured CDOM concentration value at a detection point and the corresponding remotely sensed water surface reflectance. 70% (126 sets) of the 180 "remotely sensed reflectance-ground CDOM concentration" data sets were used as the inversion training sample set, and 30% (54 sets) were used as the inversion validation sample set.

[0055] B4. Based on the inversion training sample set, the remote sensing water surface reflectance is used as input and the corresponding CDOM concentration value is used as output. Several linear regression models are trained respectively; the linear regression models include stepwise linear regression model, back propagation neural network and support vector machine model. Specifically, the obtained blue, green, red and near-infrared band remote sensing reflectance is used as input parameters, and the measured CDOM concentration value is the target parameter. CDOM is fitted by stepwise linear regression (SLR), back propagation neural network (BP) and support vector machine (SVR) models respectively, such as Figure 4 As shown, three CDOM inversion models are obtained.

[0056] B5. Evaluate each linear regression model based on the inversion validation sample set, and use the linear regression model with the best inversion effect as the trained CDOM inversion model. 2 and rRMSE), and found that the SVR model showed the best model evaluation indicators in both the training and validation periods (training set: R 2 =0.88, rRMSE=23%, validation set: R 2 =0.86, rRMSE=24%). Therefore, SVR was selected as the final CDOM inversion model.

[0057] Specifically, in step B5, the determination coefficient and relative root mean square error are selected to evaluate each linear regression model, and the linear regression model with the largest determination coefficient and the smallest relative root mean square error among several linear regression models is used as the trained CDOM inversion model.

[0058] In an exemplary embodiment, the method further includes a process of constructing and training a CDOM prediction model, such as Figure 5 As shown, the following steps are included:

[0059] D1. Obtain long-sequence remote sensing monitoring images of the target water body's catchment area at various historical dates. Download long-sequence remote sensing monitoring images covering Chagan Lake from 2014 to 2024. Specifically, obtain long-sequence OLCI and MODIS images for this period, one every 2-3 days.

[0060] D2. Using the trained CDOM inversion model, obtain CDOM inversion data for the target water catchment on various historical dates based on long-sequence remote sensing monitoring images. The trained CDOM inversion model, developed through steps B1-B5, can directly invert CDOM concentration data for Chagan Lake over the past decade from long-sequence remote sensing monitoring images. Since multiple pixels in the remote sensing images cover Chagan Lake, the average CDOM concentration across all pixels covering Chagan Lake is used as the lake's CDOM concentration for that day.

[0061] D3. For the CDOM inversion data for the target watershed at any historical date, obtain historical multi-source driving factors for several dates prior to that date to construct a sample of predicted data pairs. Specifically, compile and collect natural climate data such as temperature, rainfall, wind speed, and runoff within the Chagan Lake basin from 2014 to 2024, as well as human activity data in surrounding counties and cities, including agricultural fertilizer application, livestock breeding, and sewage treatment plant discharge data.

[0062] In an improved embodiment, in order to reduce the complexity of the LSTM model, a strong correlation is selected from the above multi-source driving factors through correlation analysis. The historical multi-source driving factors obtained in step D3 are screened using the mutual information method to determine the optimal historical multi-source driving factors as the input parameters of the model. Specifically, the mutual information value I(X) between the CDOM concentration value Y and each driving factor (such as temperature X1, rainfall X2, etc.) is calculated by the following formula i ; Y):

[0063]

[0064] Among them, Y is the CDOM concentration value, X i is the i-th driving factor, p(x,y) is the joint probability density function of X and Y, p(x) is the probability density function of X, and p(y) is the probability density function of Y. i ; Y)>0.3, the driving parameters closely related to the dynamic changes of CDOM were screened out, and irrelevant or redundant parameters were eliminated.

[0065] D4. Based on the prediction data of the target water body catchment area on various historical dates, a prediction training sample set and a prediction verification sample set are constructed.

[0066] D5. Based on the predicted training sample set, the LSTM model is trained using historical multi-source driving factors as input and CDOM inversion data as target output. Specifically, the LSTM model structure includes an input layer, a hidden layer, and an output layer. The hidden layer module consists of multiple neural units. Each neural unit includes a forget gate, an input gate, and an output gate. The hyperparameters of the LSTM model include the number of hidden layers, the number of neurons, and the learning rate. The hyperparameter settings include the number of hidden layers: 1 to 3, the number of neurons: 32 to 256, and the learning rate: 0.0001 to 0.01.

[0067] D6. Evaluate the trained LSTM model based on the prediction validation sample set. If the prediction accuracy does not meet the preset conditions, proceed to step D7. If the prediction accuracy meets the preset conditions, a trained CDOM prediction model is obtained.

[0068] D7. Use the Bayesian optimization algorithm to adjust the hyperparameters of the LSTM model and jump to step D5.

[0069] Specifically, R 2 As the primary indicator of model accuracy, rRMSE is used as a secondary indicator. 2 If the value is less than 0.85 and the rRMSE is greater than 25%, the Bayesian optimization algorithm is used to adjust the hyperparameters of the LSTM model, and the process jumps to step D5 to continue optimizing and adjusting the model until the final CDOM prediction model is obtained.

[0070] Specifically in this embodiment, Figure 6 As shown in the figure, the multi-source driving factors (rainfall, temperature, wind speed, sewage plant discharge, and agricultural fertilizer application) of the past three days are input into the LSTM-based trained CDOM prediction model, which can output the predicted CDOM concentration value of Chagan Lake for the next day. There may be multiple remote sensing data in a month. CDOM concentration prediction is performed for each remote sensing data, and the average of all prediction results for the month is taken as the final CDOM concentration prediction result for the month. It can be found that the LSTM-based CDOM prediction model trained in this application achieves good simulation accuracy, and the training period R 2 The rRMSE is 0.88 and 12.6%, which meets the accuracy requirements and can fully capture the dynamic changes of Chagan Lake CDOM.

[0071] Based on the same inventive concept, the present application also provides a system for implementing the aforementioned CDOM prediction method based on multiple driving factors. The solution provided by the system is similar to the solution described in the aforementioned method. In an exemplary embodiment, Figure 7 As shown in the figure, a CDOM prediction system based on multi-source driving factors is provided, which includes the following modules:

[0072] The multi-source driving factor acquisition module is used to obtain the multi-source driving factors of the target water body catchment area in the historical period; the multi-source driving factors include temperature data, rainfall data, wind speed data, runoff data and pollution source data.

[0073] The CDOM prediction module is used to input multi-source driving factors into a trained CDOM prediction model and output CDOM prediction data for the target water body catchment area during the prediction period; the trained CDOM prediction model is an LSTM model trained based on long-sequence CDOM inversion data and historical multi-source driving factors; the long-sequence CDOM inversion data is CDOM data obtained by inverting the trained CDOM inversion model based on long-sequence remote sensing monitoring images; the trained CDOM inversion model is a linear regression model trained based on measured CDOM concentration values ​​and remote sensing monitoring images corresponding to the measured CDOM concentration values ​​in time and space.

[0074] certainly, Figure 7 The architecture shown is only exemplary and can be omitted according to actual needs when implementing different functions. Figure 7 One or at least two components of the system shown.

[0075] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 8 As shown. The computer device includes a processor, a memory, an input / output interface (I / O) and a communication interface. The processor, the memory and the input / output interface are connected via a system bus, and the communication interface is connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a CDOM prediction method based on multi-source driving factors provided in the aforementioned embodiment can be implemented.

[0076] Those skilled in the art will understand that Figure 8 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0077] In an exemplary embodiment, a computer device is further provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps in the above method embodiments when executing the computer program.

[0078] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0079] In an exemplary embodiment, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0080] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0081] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, database or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory may include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0082] The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processors involved in the various embodiments provided herein may include, but are not limited to, general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic units, data processing logic units based on quantum computing, and the like.

[0083] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0084] This document uses specific examples to illustrate the principles and implementation methods of this application. The description of the above examples is only intended to help understand the method and core concept of this application. At the same time, for those skilled in the art, based on the concept of this application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A CDOM prediction method based on multi-source driving factors, characterized in that: include: Obtain the multi-source driving factors of the target water catchment area in the historical period; The multi-source driving factors include temperature data, rainfall data, wind speed data, runoff data and pollution source data; The multi-source driving factors are input into a trained CDOM prediction model to output CDOM prediction data of the target water body catchment area during the prediction period; the trained CDOM prediction model is an LSTM model trained based on long-sequence CDOM inversion data and historical multi-source driving factors; the long-sequence CDOM inversion data is CDOM data obtained by inverting the trained CDOM inversion model based on long-sequence remote sensing monitoring images; the trained CDOM inversion model is a linear regression model trained based on measured CDOM concentration values ​​and remote sensing monitoring images corresponding to the measured CDOM concentration values ​​in time and space.

2. The CDOM prediction method based on multi-source driving factors according to claim 1 is characterized in that: Also includes: Select several detection points in the target water catchment area, measure the actual CDOM concentration value of each detection point, the corresponding detection time and the spatial coordinates of the point; For any detection point, based on the spatial coordinates of the detection point, obtaining the remote sensing water surface reflectance corresponding to the measured CDOM concentration value in the remote sensing monitoring image near the detection time; According to the measured CDOM concentration value and the corresponding remote sensing water surface reflectance of each detection point, an inversion training sample set and an inversion verification sample set are constructed; the inversion training sample set and the inversion verification sample set each include a plurality of inversion data pair samples; the inversion data pair sample includes the measured CDOM concentration value and the corresponding remote sensing water surface reflectance of a detection point; Based on the inversion training sample set, the remote sensing water surface reflectance is used as input and the corresponding CDOM concentration value is used as output, and several linear regression models are trained respectively; the linear regression models include a stepwise linear regression model, a back propagation neural network and a support vector machine model; Each linear regression model is evaluated based on the inversion verification sample set, and the linear regression model with the best inversion effect is used as the trained CDOM inversion model.

3. The CDOM prediction method based on multi-source driving factors according to claim 2 is characterized in that: Before obtaining the remote sensing water surface reflectance corresponding to the measured CDOM concentration value in the remote sensing monitoring image near the detection time based on the point spatial coordinates of the detection point, the method further includes: Atmospheric correction is performed on all acquired remote sensing monitoring images to eliminate the effects of atmospheric absorption and scattering on spectral reflectance.

4. The CDOM prediction method based on multi-source driving factors according to claim 2 is characterized in that: The determination coefficient and relative root mean square error were selected to evaluate each linear regression model, and the linear regression model with the largest determination coefficient and the smallest relative root mean square error among several linear regression models was taken as the trained CDOM inversion model.

5. The CDOM prediction method based on multi-source driving factors according to claim 1 is characterized in that: Also includes: Obtain long-sequence remote sensing monitoring images of the target water catchment area at various historical dates; Using the trained CDOM inversion model, the CDOM inversion data of the target water catchment area on various historical dates are obtained according to the long sequence remote sensing monitoring images; Based on the CDOM inversion data of the target water catchment area on any historical date, the historical multi-source driving factors of several dates before the date are obtained to construct a sample of predicted data pairs; According to the prediction data of the target water body catchment area on various historical dates, a prediction training sample set and a prediction verification sample set are constructed; Based on the prediction training sample set, the historical multi-source driving factors are used as input, the CDOM inversion data is used as the target output, and the LSTM model is trained to obtain a trained LSTM model; Evaluate the trained LSTM model based on the prediction verification sample set, and if the prediction accuracy meets the preset conditions, obtain a trained CDOM prediction model; If the prediction accuracy does not meet the preset conditions, the Bayesian optimization algorithm is used to adjust the hyperparameters of the LSTM model, and the process jumps to step "Based on the prediction training sample set, the historical multi-source driving factors are used as input, the CDOM inversion data is used as the target output, and the LSTM model is trained to obtain a trained LSTM model." 6. The CDOM prediction method based on multiple driving factors according to claim 5 is characterized in that: The LSTM model structure includes an input layer, a hidden layer, and an output layer; the hidden layer module is composed of multiple neural units; each neural unit includes a forget gate, an input gate, and an output gate; the hyperparameters of the LSTM model include the number of hidden layers, the number of neurons, and the learning rate.

7. A CDOM prediction system based on multiple driving factors, characterized by: include: The multi-source driving factor acquisition module is used to obtain the multi-source driving factors of the target water catchment area in the historical period; The multi-source driving factors include temperature data, rainfall data, wind speed data, runoff data and pollution source data; The CDOM prediction module is used to input the multi-source driving factors into a trained CDOM prediction model and output CDOM prediction data of the target water body catchment area during the prediction period; the trained CDOM prediction model is an LSTM model trained based on long-sequence CDOM inversion data and historical multi-source driving factors; the long-sequence CDOM inversion data is CDOM data obtained by inverting the trained CDOM inversion model based on long-sequence remote sensing monitoring images; the trained CDOM inversion model is a linear regression model trained based on measured CDOM concentration values ​​and remote sensing monitoring images corresponding to the measured CDOM concentration values ​​in time and space.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the CDOM prediction method based on multi-source driving factors according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the CDOM prediction method based on multi-source driving factors according to any one of claims 1 to 6 is implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the CDOM prediction method based on multi-source driving factors according to any one of claims 1 to 6 is implemented.

Images