Three-dimensional prediction method for marine ranching environment based on three-dimensional environment field
By constructing a three-dimensional marine ranch model and combining it with high-precision shoreline data and multi-layer nested grids, the problems of low accuracy and efficiency in traditional marine ranch monitoring have been solved, enabling real-time high-precision forecasting and automated early warning of the marine environment.
Patent Information
- Application Number
- CN202411433297.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-15
- Publication Date
- 2024-11-12
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Traditional marine ranch monitoring methods are insufficient for real-time and comprehensive understanding of the environmental resources of the entire marine ranch area. They also suffer from low forecast accuracy and efficiency, and lack automation, resulting in significant errors, especially in areas with complex and tortuous coastlines.
A three-dimensional prediction method for marine ranching environment based on a three-dimensional environmental field is constructed. High-precision shoreline data and multi-layer nested grid model are used, combined with an automated forecasting system, to achieve real-time monitoring and high-precision forecasting of the marine environment.
It enables precise calculation and forecasting of high-resolution hydrodynamic characteristics of marine ranches in complex and varied coastal areas, allowing for early warning of ecological disasters in marine ranches and reducing losses.
Smart Images

Figure CN118940917A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of marine ranch temperature, salinity and flow velocity warning and forecasting, and in particular to a three-dimensional prediction method of a marine ranch environment based on a three-dimensional environmental field. Background Art
[0002] Traditional marine ranch monitoring mostly adopts a combination of fixed-point buoy monitoring and regular survey and cruise monitoring. The data obtained are mostly single-point data, which makes it difficult to timely and comprehensively grasp the environmental resources of the entire marine ranch area, and no effective early warning and forecasting system has been formed. The traditional marine ranch model is based on structured grid discretization, which is simple, but when the near-coastline is complex and tortuous, the structured grid is difficult to accurately fit the coastline, and the grid resolution is low, it will cause the traditional marine ranch forecast model to have large errors in the nearshore area.
[0003] Traditional forecasting models require manual collection of various real data, followed by numerical calculations, and waiting for data results for analysis. The whole process is cumbersome, time-consuming and labor-intensive. At the same time, in pursuit of the timeliness of the forecast, numerical forecasts are often started at night, taking up a lot of human resources.
[0004] Therefore, technical personnel in this field urgently need to solve existing problems such as low accuracy, low efficiency and low automation in forecasting the hydrological environment of marine ranches in waters with complex and tortuous coastlines. Summary of the invention
[0005] In view of the problems existing in the prior art, the purpose of the present invention is to construct a three-dimensional temperature-salinity flow forecast model to predict marine environmental changes with high precision, and on this basis, to build an automated forecasting system to realize the automated operation of the forecasting model.
[0006] To achieve the above object, the technical solution adopted by the present invention is: a three-dimensional prediction method of the marine ranch environment based on a three-dimensional environmental field, comprising the following steps: Step 1: Based on the diverse requirements of grid resolution for different sea areas and dynamic processes, the dynamic environment of the ranch sea area is simulated in three dimensions using real terrain data and forcing field conditions, the refined numerical values of the elements of the ranch sea area are analyzed, and a basic model of the ranch sea area is constructed; Step 2: In the constructed basic model of the ranch sea area, use make.inc in the source code to compile, select the code modules required for the ranch sea area to run, set variables to point to the location of the compiler, preprocessor, and message passing structure library, and compile to generate an executable fvcom; Step 3: Perform numerical calculations on the basic model, control the parameters during the numerical calculation of the model through the namelist file, generate a namelist file with the keyword tst, link the fvcom executable file compiled in step 2 to the working directory, and run fvcom according to the tst_run.nml parameter settings; Step 4: Based on fvcom, a multi-layer nested three-dimensional ocean high-resolution dynamic model is established in the ranch area. The real terrain and forcing field are used to make a stereoscopic forecast of the three-dimensional environment of the marine ranch and predict the changes in the temperature and salinity flow in the ranch area.
[0007] The above-mentioned three-dimensional prediction method of the marine ranch environment based on the three-dimensional environmental field, the step 1 comprises: Step 1-1: Obtain the real coastline, and extract the high-precision coastline of the marine ranch area based on the data source and high resolution of the land satellite; Step 1-2: Integrate the water depth data, combine the real seabed undulations, and the water depth data of the ranch sea chart, set multiple σ layers in the vertical direction, and use σ to describe the seabed topography undulations with the bottom coordinates to improve the accuracy of the basic model near the coast; Step 1-3: Establish a triangular grid, combine the real coastline and fused water depth data, focus on encrypting the core area of the marine ranch, establish a triangular unstructured model network, and accurately fit the irregular coastline near the ranch area; Step 1-4: Comprehensively evaluate the input files, compare various open boundary data with initial field data, run different temperature-salinity initial field and open boundary combination schemes, compare with measured data from multiple angles, and determine the temperature-salinity initial field and open boundary forcing field; Steps 1-5: Model debugging, building the operating environment, debugging various parameters, and ensuring the stability of the model.
[0008] In the above-mentioned three-dimensional prediction method of the marine ranch environment based on the three-dimensional environmental field, in the steps 1-3, the encryption processing adopts a multi-layer nested, gradually encrypted nested grid structure to form a nested model, thereby reducing the impact of the uncertainty of the open boundary on the inner area of the model.
[0009] In the above-mentioned three-dimensional prediction method of the marine ranch environment based on the three-dimensional environmental field, in steps 1-4, the nested model includes at least one surface runoff input, and according to the differences in the seabed structure, the bottom friction coefficient is parameterized using the accompanying assimilation results to describe the changes in bottom friction.
[0010] The above-mentioned three-dimensional prediction method of the marine ranch environment based on the three-dimensional environmental field, the nested model is composed of a number of grids and a number of nodes, and the grid division follows the principle of sparseness away from the coastline and denseness near the coast, so as to accurately capture the complex flow characteristics of the nearshore waters, use coarse grids in large areas, and use fine grids in key small areas.
[0011] In the above-mentioned three-dimensional prediction method of the marine ranch environment based on the three-dimensional environmental field, step 2 comprises: Step 2-1: Edit make.inc, select options, uncomment the desired value for each option; Step 2-2: Set up the executable to accommodate the variable compiler locations and preprocessor and compiler debug flags; Steps 2-3: Execute make clean to remove all object files, compile the code with make, watch the screen output during compilation to ensure that the resulting preprocessor tags are being used, generate the executable fvcom, and copy and / or link the executable to the run directory.
[0012] In the above-mentioned three-dimensional prediction method of the marine ranch environment based on the three-dimensional environmental field, step 3 comprises: Step 3-1: Input the model file in the basic model and perform numerical calculation process; Step 3-2: After the numerical calculation is completed, the forecast results of the temperature and salinity flow in the ocean ranch area are obtained.
[0013] In the above-mentioned three-dimensional prediction method of the marine ranch environment based on the three-dimensional environmental field, step 4 comprises: Step 4-1: Determine the model simulation time. The start time of the current report is the end time of the previous current report. The end time of the current report simulation is used as the forecast start time. After the current report ends, find the latest file time from the current report simulation result as the forecast start time and set the forecast end time. Step 4-2: Import namelist and read the file content, including the start time, end time, working directory of large area model and small area model, input file and output file path, simulation result storage path, and matlab code storage path; Step 4-3: Make the input file of the large-area model, crawl the file data, make it into the input file, import the settings of each path in the namelist, do time processing, process the obtained date into a year-month-day string, delete the file simulated last time, link or copy the input file that has been made, and complete the production of the input file of the large-area model; Step 4-4: Modify the namelist file, based on the year-month-day character string in step 4-3, output the simulation results of the start time, end time, and modification processing time interval, and write them into the running file of the large-area model; Step 4-5: Set the running core and submit the task, import the namelist and read the file content, determine the number of core calculations, and determine whether to activate the large area model; Step 4-6: Monitor the model running status and print the latest restart file time; Step 4-7: Rename the model output results and save them to the specified directory, placing them in folders according to time; Step 4-8: After the large-area model automatic forecast is completed, the small-area model automatic forecast is started. The open boundary of the small-area model is replaced with the open boundary simulated by the large-area model. The hourly simulated data in the large area is used as the open boundary of the small-area model. Step 4-9: Create a small area model input file, import namelist, process the time, process the obtained date into a year-month-day string, delete the previous data file, use matlab to process the output of the large area model into the nested open boundary of the small area model, determine whether the file exists, and complete the creation of the small area model input file; Step 4-10: Modify the namelist and update the open boundary file of the small area model in real time with the results of the large area model; Step 4-11: Set up the running core and submit tasks; Step 4-12: Monitor the running status of the small area model and print the latest restart file time; Step 4-13: Rename the output results of the small area model and save them to the specified directory.
[0014] In the above-mentioned three-dimensional prediction method of the marine ranch environment based on the three-dimensional environmental field, in the steps 4-6, if the running state is not in the running or waiting state, that is, the running has ended, the info.dat file is used to determine whether the model ends normally. If DATA appears, the model ends normally. If it does not end normally, the crash processing program will be entered.
[0015] The above-mentioned three-dimensional prediction method of the marine ranch environment based on the three-dimensional environmental field, the collapse processing procedure includes: When the model crashes and stops, the automated control system will enter the crash handling procedure. The system will try to restart the model, keeping the original parameters unchanged. When submitting the running script, select the crash restart option, and the model will continue to run from the point where it stopped. If the model resumes running and is in normal status, the crash handling procedure will be exited. If the model still cannot run normally after restarting and crashes again, the system will enter the second stage of processing, replace the atmospheric forcing field, replace the previously used ERA5 atmospheric forcing field with the GFS atmospheric forcing field, and continue to run from the crash point. If the model runs stably, the GFS atmospheric forcing field will be switched back to the original ERA5 field and continue to execute. If this process goes smoothly, exit the crash handling program; If the above steps still fail to solve the problem, the system will take the third stage measures, use the wind field and heat flux data in the atmospheric forcing field of the previous day to replace the data at the stop time point, and resubmit the run. If the model operation returns to normal, the data will be replaced with the wind field and heat flux at the original time point, continue to run, and exit the crash handler.
[0016] The beneficial effects of the three-dimensional prediction method of the marine ranch environment based on the three-dimensional environmental field of the present invention are: by acquiring relevant data of the marine environment in real time, a high-resolution three-dimensional hydrodynamic model of the marine ranch in the sea area with complex coastline tortuosity is established, and a series of scripts are used to automatically control the model, which is started regularly or according to the data situation, to predict the high temperature and low oxygen disasters that will occur in the future, to issue forecast and warning information on marine ranch ecological disasters in advance, to take timely response measures, to minimize the disaster losses, and to use large-area and small-area multi-layer nesting technology to achieve accurate calculation of the hydrodynamic characteristics of the sea area with complex coastline tortuosity. The multi-layer nesting technology is also helpful in processing marine phenomena of different scales. By simulating processes of different scales on different grids, it helps us to have a deeper understanding of the interactions and influences between them. BRIEF DESCRIPTION OF THE DRAWINGS
[0017] Figure 1 It is a forecast flow chart of the present invention; Figure 2 Flowchart for automated forecasting of large or small area models. DETAILED DESCRIPTION
[0018] In order to enable those skilled in the art to better understand the technical solution of the present invention, the technical solution of the present invention is described below in conjunction with specific implementation methods and drawings.
[0019] Traditional marine ranch models mostly use ocean model parameter settings, but there are certain differences between the nearshore dynamic background and the ocean dynamics, which causes the numerical results to deviate significantly from the true values. A series of targeted optimizations of the parameters of the marine ranch model are needed to make it suitable for marine ranches near the coast of Shandong.
[0020] The influence of the open sea is transmitted to the hydrodynamic model through the open boundary. The open boundary of the traditional marine ranch model only gives tidal information and ignores non-tidal signals. The use of multi-layer nesting technology can realize the accurate calculation of the hydrodynamic characteristics of Shandong coastal waters.
[0021] In order to achieve a high-precision forecasting effect of a three-dimensional forecasting model, the present invention optimizes the model coastline, unstructured grid and model parameters. At the same time, a control system is written to automatically control the operation process of the model to achieve automatic data collection, automatic processing and automatic operation. Real-time acquisition of marine environment-related data, a high-resolution three-dimensional hydrodynamic model of the marine ranch near the coast of Shandong is established, and a series of scripts are used to automatically control the model, which is started regularly or according to the data situation, predicting that high temperature and low oxygen disasters will occur in the future, issuing forecast and warning information for marine ranch ecological disasters in advance, and taking timely countermeasures to minimize disaster losses.
[0022] Example 1 A three-dimensional prediction method for a marine ranch environment based on a three-dimensional environmental field comprises the following steps.
[0023] Step 1: According to the diverse requirements of grid resolution in different sea areas and dynamic processes, the dynamic environment of the ranch area is simulated in three dimensions using real terrain data and forcing field conditions, the refined numerical values of the elements of the ranch area are analyzed, and a basic model of the ranch area is constructed.
[0024] Step 2: In the constructed basic model of the ranch sea area, use make.inc in the source code to compile, select the code modules required for the ranch sea area to run, set variables to point to the location of the compiler, preprocessor, and message passing structure library, and compile to generate an executable fvcom.
[0025] Step 3: Perform numerical calculations on the basic model, control the parameters during the numerical calculation of the model through the namelist file, generate a namelist file with the keyword tst, link the fvcom executable file compiled in step 2 to the working directory, and run fvcom according to the tst_run.nml parameter settings.
[0026] Step 4: Based on fvcom, a multi-layer nested three-dimensional ocean high-resolution dynamic model is established in the ranch area. The real terrain and forcing field are used to make a stereoscopic forecast of the three-dimensional environment of the marine ranch and predict the changes in the temperature and salinity flow in the ranch area.
[0027] Example 2 like Figure 1-Figure 2 As shown, a three-dimensional prediction method for marine ranch environment based on a three-dimensional environmental field is shown. The specific implementation steps can be divided into four parts: model construction, code compilation and numerical calculation, and are controlled by an automated system.
[0028] Step 1: Collect data and build a model.
[0029] 1. Automatic data collection.
[0030] The European Centre for Medium-Range Weather Forecasts (ECMWF) reanalysis data ERA5 data are collected automatically.
[0031] This part of the download is completed using a python script, which relies on the cdsapi library package and uses crontab for timing management. The latest ERA5 data is downloaded every day, including atmospheric temperature, atmospheric pressure, heat flux and other atmospheric data, which are used for the current report of the marine ranch model.
[0032] GFS data is collected automatically.
[0033] This part of the download is completed using a shell script, which relies on the wget function and uses crontab for timing management. The GFS data forecasted at 00:00 is downloaded regularly every day, including atmospheric data such as atmospheric temperature, atmospheric pressure, and heat flux, which are used for marine ranch model forecasts.
[0034] glory data is collected automatically.
[0035] This part of the download is completed using a Python script, which relies on the copernicusmarine library package and uses crontab for timing management. The latest glory data is downloaded every day, including non-tidal water level, temperature, salinity, current and other ocean data, which are used to open the boundary of the marine ranch model.
[0036] River data is collected automatically.
[0037] This part of the download is completed using a python script, and crontab is used for timing management to obtain river data from the hydrological website on a daily basis, including flow and water level data of major rivers, which are used for runoff input of the marine ranch model.
[0038] (2) Automated data processing.
[0039] ERA5 data processing.
[0040] The original ERA5 data are distributed in an orthogonal grid of 0.25×0.25°. The position format meets the requirements of the fvcom input file. The temperature unit needs to be converted from Kelvin to Celsius. At the same time, the relative humidity is calculated based on the atmospheric temperature and dew point temperature. The above process is completed using a MATLAB script.
[0041] GFS data processing.
[0042] First, wgrib2 is used to convert the GFS data in grib2 format into NetCDF format, and then the files are merged in chronological order. Matlab is used for further processing, mainly to complete unit conversion and Julian time setting.
[0043] glory data processing.
[0044] The original glory data is global in scope and needs to be restricted to the vicinity of the open boundary. The cropped data is then interpolated into the nested open boundary as the non-tidal signal input of the open boundary. The above process is completed using a MATLAB script.
[0045] River data processing.
[0046] The original river data is saved in an Excel spreadsheet. It is necessary to extract the flow data of major rivers such as the Yellow River, Yangtze River, and Liao River, and set the Julian time to meet the needs. The above process is completed using a MATLAB script.
[0047] Other data processing.
[0048] The input files of the model also include basic files, such as grid files, water depth files, etc. These files do not need to be modified, and the hot start files only need to be linked to the specified folder.
[0049] The above two parts are all prior art.
[0050] (III) Model construction
[0051] The Finite Volume Community Ocean Model (FVCOM) was jointly developed by the University of Massachusetts and the Woods Hole Oceanographic Institution. It uses the finite volume method to discretize the Navior-Stokes (NS) primitive integral equation. It uses unstructured triangular grids in the horizontal direction to better fit the complex coastline of shallow seas and estuaries. It uses σ with bottom coordinates in the vertical direction to describe the undulating seabed terrain. In addition, the FVCOM model supports dry-wet grid technology to solve floodplain problems, and internal and external model separation technology to accelerate calculations. Based on FVCOM, a multi-layer nested three-dimensional forecast model is constructed, and the three-dimensional environment of the ranch is predicted in three dimensions using real terrain and forcing fields, so as to achieve a refined forecast of the temperature, salinity and ocean currents in the core area of the ranch in the next three days.
[0052] In order to meet the diverse requirements of grid resolution in different sea areas and dynamic processes, the dependence between tidal wave wavelength and water depth, as well as the specific requirements of hydrodynamic processes for resolution, are considered in the numerical discretization process to form unstructured triangular grids with differentiated resolutions in different sea areas. This aims to meet the scale requirements of the dynamic processes of interest to the greatest extent, reduce the complexity of calculations to a certain extent on the basis of successful simulation, and achieve an optimal balance between grid resolution and computing resources.
[0053] A multi-layer nested three-dimensional hydrodynamic model was constructed based on FVCOM. Using real terrain data and forcing field conditions, a comprehensive three-dimensional simulation of the dynamic environment of the Bohai and Yellow Seas was carried out, and a refined numerical analysis of key factors such as temperature, salinity and ocean currents in the Shandong coastal area was achieved. The model construction process mainly includes the following key steps.
[0054] 1-1: Get the real coastline.
[0055] Affected by global sea level changes and nearshore engineering construction, China's offshore coastline changes year by year, especially in the Bohai Sea. As the boundary between land and sea, the constant changes of the coastline shape an increasingly complex marine environment. The real coastline is of key significance for the high-precision prediction of thermohaline currents. The Landsat satellite operated by NASA in the United States has been widely recognized for its advantages such as years of data sources and high resolution. With the help of the satellite images recently released by Landsat 8, we successfully extracted the Shandong Peninsula coastline with an accuracy of 50~300m.
[0056] 1-2: Fusion of water depth data.
[0057] As the main factor affecting the accuracy of forecasts, the real seabed fluctuations have a positive effect on the model's reproduction and prediction of ocean dynamic processes at all scales. The depth of the Shandong coastal water dynamic model is based on the ETOPO1 data, combined with the depth data of 8 nautical charts of the Bohai and Yellow Seas (CN201301, CN201305, CN313001, CN323001, CN201302, CN201306, CN322001, CN333001) to optimize the coastal topographic data. The depth of the forecast model uses ETOPO1 data, combined with the depth data of the Bohai and Yellow Sea charts, and is divided into 30 σ layers in the vertical direction to improve the accuracy of the model near the coast.
[0058] 1-3: Create a triangular mesh.
[0059] Based on the real coastline and integrated water depth, and focusing on the densification of the core area of the pasture along the Shandong Peninsula, a triangular unstructured model grid is established to accurately fit the irregular coastline near the coast. In order to reduce the impact of the uncertainty of the open boundary on the inner area of the model, a multi-layer nested and gradually encrypted grid structure is adopted.
[0060] 1-4: Various input files.
[0061] The solution of the differential equation is subject to its boundary conditions. Similarly, the quality of the numerical results also depends on the selected temperature-salinity initial field or open boundary forcing field. In order to explore the influence of different temperature-salinity initial fields on the model output results. In the experiment, the open boundary forcing and other basic configurations were kept consistent, and 7 sets of different temperature-salinity initial field data were input to drive the model for simulation. Finally, it was determined that the optimal temperature-salinity initial field of the ECS3-SDP nested model came from the GDEM temperature-salinity data.
[0062] The importance of atmospheric forcing fields to ocean models is self-evident. As the main force driving ocean movement, the atmospheric forcing field has an important impact on the ocean's flow field, temperature field, and salinity field by transferring key variables such as momentum, heat, and freshwater flux. In order to further explore the impact of different atmospheric forcing fields on the simulation results of ocean models, two sets of experiments were designed, using ERA5 and CFSR meteorological field data as inputs of the nested model to compare their simulation effects. In the end, it was considered that the numerical results of the CFSR-driven model were slightly inferior to those driven by the ERA5 meteorological field, but the difference between the two was not large.
[0063] Bottom friction coefficient. In the past, the bottom friction coefficient of the hydrodynamic model was often simplified as a constant for the entire sea area, or taken as different constants in different sea areas. Although this processing method simplifies the process of model debugging, for areas such as the Shandong coastal area, which are significantly affected by the Yellow River sediment transport, have complex and changeable terrain, and strong tidal mixing, its simulation accuracy is often unsatisfactory and difficult to fully meet actual needs. In order to more accurately simulate the hydrodynamic process of the Shandong coastal area and fully consider the differences in bottom friction effects in different regions, a new method is to assume that the bottom friction coefficient has spatial distribution characteristics, that is, a series of representative points are selected in the calculation sea area as the setting points of independent bottom friction coefficients, and the bottom friction coefficient of any point in the calculation area is obtained by linear interpolation through these independent bottom friction coefficients. This processing method enables the model to read in and fully consider the bottom friction coefficients with regional heterogeneity, thereby further improving the accuracy and reliability of the hydrodynamic model.
[0064] The bottom stress satisfies: , The above formula is the Von Karman formula, which is used to calculate the bottom friction coefficient The empirical formula, and are the east-west and north-south seawater velocity components, and are the components of the bottom stress in the east-west direction and the north-south direction, is the Karman constant, is the water depth between the grid closest to the bottom layer and the seabed, is the seafloor roughness, To obtain the maximum value function, is the natural logarithm function.
[0065] The atmospheric forcing field uses the European Center for Medium-Range Weather Forecasts (ERA5) reanalysis data with high accuracy near the coast of China. The data includes 0.25° horizontal resolution, 10m wind speed per hour, heat flux, atmospheric pressure, evaporation and precipitation. Multi-year ERA5 data from 1980 to the present have been collected. Multiple land surface runoff inputs are considered in the nested model, including the Yangtze River, Yellow River and Liaohe River, and the data comes from the United States Geological Survey. Considering the differences in the submarine structures of the Bohai and Yellow Seas, the same parameter is not enough to describe the changes in bottom friction. We use the assimilation results to parameterize the bottom friction coefficient. We compare a variety of open boundary data and initial field data, and run different temperature-salinity initial field and open boundary combination schemes, and make multi-angle comparisons with measured data. After comprehensive evaluation, GDEM and glory data are used as the temperature-salinity initial field and open boundary forcing field, respectively.
[0066] 1-5: Model debugging: In order to run the forecast model successfully, it is necessary to build the corresponding operating environment and debug various parameters to ensure the stability of the model. Physical processes at different scales correspond to different parameterization schemes in the numerical model. We debug this to determine the various parameters applicable to the Bohai and Yellow Seas.
[0067] Specifically, the grid coverage of the East China Sea Large Area Model (ECS3) is extensive, from Liaodong Bay in the Bohai Sea in the north to 21°N in the south, spanning 20 degrees of latitude. Its western boundary extends from Liaoning to the Chinese coastline of Fujian, and its eastern boundary reaches 138°E, covering part of the Japanese coastline including the Korean Peninsula and Kyushu, spanning 21 degrees of longitude. The computational domain of this model contains three open boundaries. The ECS3 model consists of 136,612 grids and 70,479 nodes. Its grid division follows the principle of sparse away from the coastline and dense near the coast to accurately capture the complex flow characteristics of the nearshore waters.
[0068] The grid range of the Shandong offshore small-area model (SDP) starts from the southern Yellow Sea (about 34°N) in the south and ends at Liaodong Bay in the Bohai Sea in the north. It is bordered by China in the west and the Korean Peninsula in the east, covering the Bohai Sea and the Yellow Sea. Only one open boundary is retained in the computational domain of the model, which is consistent with the grid of the ECS3 model at 34°N in the southern Yellow Sea. During the numerical simulation, the hourly temperature-salinity flow data of the SDP open boundary grid output by the ECS3 model are used as the input of the SDP model to drive its internal calculation. This process realizes the physical wave transmission from the East China Sea to the Shandong offshore, which is called the "nested process of the East China Sea-Shandong offshore". By using coarse grids in large areas and fine grids in key small areas, this nested model can realize the accurate calculation of the hydrodynamic characteristics of the Shandong offshore. In addition, this nested method is also helpful in dealing with marine phenomena of different scales. By simulating processes of different scales on different grids, it helps us to have a deeper understanding of their interactions and influences.
[0069] Table 1 lists the parameter configuration of the nested model in detail. The models all adopt the cold start method, that is, the initial temperature and salt field is given, and the flow velocity is simulated from zero. During the numerical calculation process, the data output interval is daily or hourly, and it stops once a year. The output results of the previous year will be used as the input file of the hot start of the next year until the simulation ends.
[0070] Table 1 Nested model parameter configuration .
[0071] Step 2: Compile the code.
[0072] The code is written primarily in Fortran and is compiled using make.inc included in the source code. Edit make.inc to select the desired code modules (parallel, water quality, etc.) and certain code options (double / single precision). Then, you must set make.inc variables to point to the Fortran compiler, the location of the c preprocessor (cpp), the location of the MPI (Message Passing Interface) library (if you plan to run FVCOM in parallel), and the location of the NetCDF library (if you plan to use NetCDF output). There are several steps to follow.
[0073] 2-1: Edit make.inc and select from the following options, uncommenting (removing the #) the desired value for each option. (The meaning of the various values is also outlined in make.inc).
[0074] For example: FLAG_1 controls the precision of floating-point numbers in the output results. The default is single precision. If double precision is selected, the result file doubles from 4 bytes to 8 bytes.
[0075] FLAG_2 is the switch that controls spherical coordinates. If you use longitude and latitude to indicate the location, you must use spherical coordinates. Turn on FLAG_2. If you use xy to indicate the location, you must use the rectangular coordinate system. Turn off FLAG_2.
[0076] FLAG_3 is the dry-wet grid switch. By setting a minimum water depth value, when the water depth at a certain moment is less than this value, it will be judged as land and will not participate in the calculation. There will be a difference between "dry points" and "wet points" in the results.
[0077] FLAG_4 is the switch for multi-core operation. Generally, we will choose multi-core parallel computing to speed up the process.
[0078] FLAG_5 is the water quality module switch.
[0079] FLAG_6 is the switch for linking Proj4 software and is used for projection conversion.
[0080] FLAG_7 is the switch between Nudging and OI.
[0081] FLAG_8 controls the form of the upwind scheme when it is discrete. To put it simply, the first-order upwind scheme has first-order accuracy, and the second-order upwind scheme has second-order accuracy. The higher the order, the better, because the stability and convergence of the equation also need to be considered. Therefore, we generally choose to use NO LIMITATION.
[0082] FLAG_9 is the switch for semi-hidden and semi-visible format.
[0083] FLAG_10 is to select the micro-element processing method for the land and sea boundary, that is, to choose whether to use ghost.
[0084] 2-2: Set up the executable to accommodate the variable compiler locations and c preprocessor (cpp) and compiler optimization / debugging flags.
[0085] 2-3: Execute "make clean" to remove all object files. Compile the code with "make". Watch the screen output during compilation to ensure that the preprocessor flags are being used. A successful compilation will produce the executable fvcom. Copy or link the executable to the run directory. Linking is recommended because code changes and subsequent recompilation are automatically picked up by the executable in the run directory.
[0086] Step 3: Numerical calculation.
[0087] The parameters during the numerical calculation of the model are controlled by the namelist file, which is usually named casename_run.nml, where casename is the keyword of this project. . / fvcom --CREATE_NAMELIST=tst A namelist file with the keyword tst is generated. This file sets the start and end time, data input, data output, and parameter values called during runtime.
[0088] Next, link the compiled fvcom executable file to the working directory, using mpirun -np 2 . / fvcom casename=tst You can run fvcom according to the parameter settings in tst_run.nml.
[0089] Specifically include: 3-1: Input file.
[0090] The model files established in the previous article are required during the numerical calculation process, including grid files, water depth files, bottom friction coefficient files, Coriolis force files, open boundary position files, vertical stratification files, sponge layer files, etc. At the same time, to start the model, an initial field file is required. If a cold start is used, the initial field file only needs to provide temperature and salinity data; when considering the influence of the atmosphere, an atmospheric forcing field file is required, including atmospheric pressure files, heat flux files, evaporation precipitation and wind field files; the influence of the open sea is transmitted through the open boundary, and an open boundary forcing file is required. The collection and processing of the above input data are all automatically realized in the automation system.
[0091] 3-2: Output results.
[0092] After the numerical operation is completed, the forecast results of the ocean's temperature-salinity flow for the next three days are obtained. The automation system automatically processes it and draws a spatial distribution map of the temperature-salinity flow for the next three days. At the same time, the forecast results can be automatically sent to the designated mailbox as needed for easy viewing.
[0093] Step 4: Automated forecasting system.
[0094] In order to realize the daily automatic forecast of the changes in temperature and salinity of the ocean ranch in the next three days, we developed an "automatic forecast system" based on the CentOS system. Taking manual forecasting as an example, we first need to make the input files and settings of the large-area model, manually submit the running task, generate the simulation results, process the simulation results, process them into an open boundary of the small-area model, start the small-area model according to the open boundary, continue the simulation, and finally get the simulation results, which takes a lot of time and energy. The automatic forecast system can realize automatic operation every day through a series of shell scripts, automatically download global wind field data, heat flux data, ocean current data, etc., and after the download is completed, automatically call matlab to make input files such as atmosphere and open boundary to meet the needs of model calculation, and start the model forecast regularly.
[0095] Example 3 The model automated forecasting system is divided into two parts: reanalysis forecast and forecast. The reanalysis forecast uses the ERA5 atmospheric forcing field, and the running time is from the end of the last forecast to the current time minus 6 days. The forecast uses the GFS atmospheric forcing field, and the running time is from the current time minus 6 days to the next three days. The overall process is controlled by shell scripts, which are written in a modular way. Different scripts correspond to different functions and complete the entire forecasting system together. They rely on the transmission of parameters to start different scripts, and return signals at the end to execute the next script, thereby completing the automation of the entire system.
[0096] Based on the finite volume coastal numerical model fvcom, a three-dimensional ocean high-resolution dynamic model was established in the coastal waters of Shandong to predict the changes in temperature and salinity in the marine ranch in the next three days. Figure 1 and Figure 2 shown.
[0097] First, the large-area model of the Bohai, Yellow and East China Seas is calculated, and the hourly data is output as the opening boundary of the small-area model of the core area of the marine ranch, and finally the calculation is performed.
[0098] Step1: Determine the model simulation time (start time, end time).
[0099] The start time of the current report is the end time of the previous current report. The specific implementation can be to find the latest file time in the restart file directory as the start time of the current report; the time corresponding to the latest reanalysis ERA5 data is used as the end time (generally the current time minus six days).
[0100] The end time of the current report simulation is used as the start time of the forecast. After the current report ends, the restart file storage directory is updated in the current report simulation results. The latest file time can be found as the forecast start time, that is, the end time of the current report is connected. The forecast end time is set to the third day in the future.
[0101] Step2: Import namelist and read the file contents.
[0102] Write the settings of each parameter of the entire script in namelist, including the start time, end time, working directory of large area model and small area model, path of input file and output file, storage path of input files such as atmospheric forcing field, storage path of simulation results, and storage path of MATLAB code involved in the process.
[0103] Step3: Make input files for large area models.
[0104] The large-area model requires input files, such as atmospheric field, river, open boundary, hot start and other input files. When running the large-area model, the system has already prepared the input files. Take the automatic preparation of ERA5 input files by the system as an example: crawl ERA5 data, process ERA5 data through the system's scheduled execution every day and call matlab code to prepare input files. The preparation of open boundary, hot start and river input files is the same as that of ERA5 input files.
[0105] First, import the settings of each path in namelist, do a time processing, that is, process the obtained date into a year-month-day string, and delete the file simulated last time. Link or copy the input file that has been prepared. This completes the preparation of the large area model input file.
[0106] Check the linked input files, mainly the time. If the data obtained from ERA5 is less than the simulation time, then the model is rerun.
[0107] Step 4: Modify the namelist file, mainly modifying the file for running the large area model.
[0108] First, import the settings of each path in namelist and do a time processing, that is, process the obtained date into a year-month-day character string, output the start time, end time or the time interval for the simulation results, and write them into the running file of the large area model. No other changes are needed because other file names are fixed here.
[0109] Step5: Set up the running core and submit the task.
[0110] First, import namelist and read the file content. Then determine how many cores to use for calculation. Take the Bohai-Yellow-East China Sea large-area model as an example. Here, each node has 32 cores. If 20 nodes are used, there are 640 cores. Finally, give a judgment on whether to activate the large-area model.
[0111] Step 6: Monitor the running status of the model and print the latest restart file time.
[0112] After submitting the control script, the model will continue to run. During this process, there will be a script monitoring the running status of the model. When monitoring the running small area model to see if it is running in real time, if it is running normally, it will print the latest restart file time.
[0113] If it is not in the running or waiting state, that is, the operation has ended, the info.dat file will be used to determine whether the model has ended normally. If DATA appears, the model has ended normally. If it has not ended normally, the crash handling program will be entered. First of all, regarding crash handling, there are three progressive solutions. The first solution is to restart.
[0114] When the model crashes and stops, the automated control system will enter the crash handling procedure. First, the system will try to restart the model, keeping the original parameters unchanged. When submitting the run script, select the "crash restart" option, and the model will continue to run from the point where it stopped. If the model resumes running and the status is normal, it means that the first stage of crash handling is successful, and the crash handling procedure is exited.
[0115] If the model still cannot run normally after restarting and crashes again, the system will enter the second stage of processing: replacing the atmospheric forcing field. Replace the previously used ERA5 atmospheric forcing field with the GFS atmospheric forcing field and continue running from the crash point. If the model runs stably, switch the GFS atmospheric forcing field back to the original ERA5 field and continue execution. If this process goes smoothly, the second stage of crash processing is successful and the crash handling program is exited.
[0116] If the above steps still fail to solve the problem, the system will take the third stage of measures: use the wind field and heat flux data in the atmospheric forcing field of the previous day to replace the data at the stop time point, and resubmit the run. If the model runs normally again, the data will be replaced with the wind field and heat flux at the original time point and continue to run. At this point, the third stage of crash handling is completed and the crash handling program is exited.
[0117] Step7: Rename the model output results and save them to the specified directory.
[0118] Because the output file does not correspond to a specific time, rename it and put it in the folder according to the time.
[0119] Step8: The large-area model automated forecast ends and the small-area model automated forecast begins.
[0120] The operation process of the small area model is the same as that of the large area model, except that the open boundary of the small area model is replaced by the open boundary simulated by the large area model, that is, the hourly simulated data in the large area is used as the open boundary of the small area model.
[0121] Step9: Create small area model input file.
[0122] As with the large-area model, import namelist and process the time, that is, process the obtained date into a year-month-day string, and delete the previous data file. Use matlab to process the output of the large-area model into the nested open boundary of the small-area model. Here, it will be determined whether the file exists. If not, submit the matlab processing program. Then link the ERA5 input file used in the large-area model, the nested open boundary input file just generated, the hot start file at the end of the last small-area model simulation, and the river input file to complete the production of the small-area model input file, and check the small-area model input file at the same time.
[0123] Step 10: Modify namelist.
[0124] Similar to the large area model, it should be noted that the open boundary file of the small area model is updated in real time with the results of the large area model.
[0125] Step11: Set up the running core and submit the task.
[0126] Similar to the large area model, it should be noted that the small area model has more triangle meshes and higher resolution, and requires more node computing resources during operation.
[0127] Step 12: Monitor the running status of the small area model.
[0128] Similar to the large-area model, if it runs normally, the latest restart file time will be printed. If it is not running or waiting, that is, the operation has ended, the info.dat file will be used to determine whether the model has ended normally. If DATA appears, the model has ended normally. If it has not ended normally, the crash handling program will be entered. Similar to the large-area model, there are three options: restart, change the wind field and heat flux, or replace today's heat flux and wind field data with the previous day's heat flux and wind field data.
[0129] Step13: Rename the output results of the small area model and save them to the specified directory.
[0130] The above embodiments are only for illustrating the inventive concept and features of the present invention, and their purpose is to enable those skilled in the art to understand the content of the present invention and implement it accordingly, and they cannot be used to limit the protection scope of the present invention. Any equivalent changes or modifications made based on the essence of the content of the present invention should be included in the protection scope of the present invention.
Claims
1. A three-dimensional prediction method for marine ranching environment based on a three-dimensional environmental field, It is characterized in that The following steps are involved: Step 1: Based on the diverse requirements of grid resolution for different sea areas and dynamic processes, the dynamic environment of the ranch sea area is simulated in three dimensions using real terrain data and forcing field conditions, the refined numerical values of the elements of the ranch sea area are analyzed, and a basic model of the ranch sea area is constructed; Step 2: In the constructed basic model of the ranch sea area, use make.inc in the source code to compile, select the code module required for the ranch sea area to run, set parameters to point to the location of the compiler, preprocessor, and message passing structure library, and compile to generate an executable fvcom; Step 3: Perform numerical calculations on the basic model, control the parameters during the numerical calculation of the model through the namelist file, generate a namelist file with the keyword tst, link the fvcom executable file compiled in step 2 to the working directory, and run fvcom according to the tst_run.nml parameter settings; Step 4: Based on fvcom, a multi-layer nested three-dimensional ocean high-resolution dynamic model is established in the ranch area. The real terrain and forcing field are used to make a stereoscopic forecast of the three-dimensional environment of the marine ranch and predict the changes in the temperature and salinity flow in the ranch area.
2. The three-dimensional prediction method of the marine ranch environment based on the three-dimensional environment field according to claim 1, It is characterized in that The step 1 comprises: Step 1-1: Obtain the real coastline, and extract the high-precision coastline of the marine ranch area based on the data source and high resolution of the land satellite; Step 1-2: Integrate the water depth data, combine the real seabed undulations, and the water depth data of the ranch sea chart, set multiple σ layers in the vertical direction, and use σ to describe the seabed topography undulations with the bottom coordinates to improve the accuracy of the basic model near the coast; Step 1-3: Establish a triangular grid, combine the real coastline and fused water depth data, focus on encrypting the core area of the marine ranch, establish a triangular unstructured model network, and accurately fit the irregular coastline near the ranch area; Step 1-4: Comprehensively evaluate the input files, compare various open boundary data with initial field data, run different temperature-salinity initial field and open boundary combination schemes, compare with measured data from multiple angles, and determine the temperature-salinity initial field and open boundary forcing field; Steps 1-5: Model debugging, building the operating environment, debugging various parameters, and ensuring the stability of the model.
3. The three-dimensional prediction method of the marine ranch environment based on the three-dimensional environment field according to claim 2, It is characterized in that In the steps 1-3, the encryption process uses a nested grid structure with multiple layers of nesting and gradual encryption to form a nested model, thereby reducing the impact of the uncertainty of the open boundary on the inner area of the model.
4. The three-dimensional prediction method of the marine ranch environment based on the three-dimensional environment field according to claim 3, It is characterized in that In the steps 1-4, the nested model includes at least one surface runoff input, and according to the difference in seabed structure, the bottom friction coefficient is parameterized using the accompanying assimilation results to describe the change in bottom friction.
5. The three-dimensional prediction method of the marine ranch environment based on the three-dimensional environment field according to claim 4, It is characterized in that The nested model consists of several grids and several nodes. The grid division follows the principle of sparse grids away from the coastline and dense grids near the coast to accurately capture the complex flow characteristics of the nearshore waters. Coarse grids are used in large areas, while fine grids are used in key small areas.
6. The three-dimensional prediction method of the marine ranch environment based on the three-dimensional environment field according to claim 1, It is characterized in that The step 2 comprises: Step 2-1: Edit make.inc, select options, uncomment the desired value for each option; Step 2-2: Set up the executable to accommodate the variable compiler locations and preprocessor and compiler debug flags; Steps 2-3: Execute make clean to remove all object files, compile the code with make, watch the screen output during compilation to ensure that the resulting preprocessor tags are being used, generate the executable fvcom, and copy and / or link the executable to the run directory.
7. The three-dimensional prediction method of the marine ranch environment based on the three-dimensional environment field according to claim 1, It is characterized in that The step 3 comprises: Step 3-1: Input the model file in the basic model and perform numerical calculation process; Step 3-2: After the numerical calculation is completed, the forecast results of the temperature and salinity flow in the ocean ranch area are obtained.
8. The three-dimensional prediction method of the marine ranch environment based on the three-dimensional environment field according to claim 1, It is characterized in that The step 4 comprises: Step 4-1: Determine the model simulation time. The start time of the current report is the end time of the previous current report. The end time of the current report simulation is used as the forecast start time. After the current report ends, find the latest file time from the current report simulation result as the forecast start time and set the forecast end time. Step 4-2: Import namelist and read the file content, including the start time, end time, working directory of large area model and small area model, input file and output file path, simulation result storage path, and matlab code storage path; Step 4-3: Make the input file of the large-area model, crawl the file data, make it into the input file, import the settings of each path in the namelist, do time processing, process the obtained date into a year-month-day string, delete the file simulated last time, link or copy the input file that has been made, and complete the production of the input file of the large-area model; Step 4-4: Modify the namelist file, based on the year-month-day character string in step 4-3, output the simulation results of the start time, end time, and modification processing time interval, and write them into the running file of the large-area model; Step 4-5: Set the running core and submit the task, import the large area namelist of the large area model and read the file content, determine the number of core calculations, and determine whether to activate the large area model; Step 4-6: Monitor the model running status and print the latest restart file time; Step 4-7: Rename the model output results and save them to the specified directory, placing them in folders according to time; Step 4-8: After the large-area model automatic forecast is completed, the small-area model automatic forecast is started. The open boundary of the small-area model is replaced with the open boundary simulated by the large-area model. The hourly simulated data in the large area is used as the open boundary of the small-area model. Step 4-9: Create a small area model input file, import the small area namelist of the small area model, process the time, process the obtained date into a year-month-day string, delete the previous data file, use MATLAB to process the output result of the large area model into the nested open boundary of the small area model, determine whether the file exists, and complete the creation of the small area model input file; Step 4-10: Modify the namelist and update the open boundary file of the small area model in real time with the results of the large area model; Step 4-11: Set up the running core and submit tasks; Step 4-12: Monitor the running status of the small area model and print the latest restart file time; Step 4-13: Rename the output results of the small area model and save them to the specified directory.
9. The three-dimensional prediction method of the marine ranch environment based on the three-dimensional environment field according to claim 8, It is characterized in that In the steps 4-6, if the running state is not in the running or waiting state, that is, the running has ended, the info.dat file is used to determine whether the model ends normally. If DATA appears, the model ends normally. If it does not end normally, the crash processing program will be entered.
10. The three-dimensional prediction method of the marine ranch environment based on the three-dimensional environment field according to claim 9, It is characterized in that The crash handling procedure includes, When the model crashes and stops, the automated control system will enter the crash handling procedure. The system will try to restart the model, keeping the original parameters unchanged. When submitting the running script, select the crash restart option, and the model will continue to run from the point where it stopped. If the model resumes running and is in normal status, the crash handling procedure will be exited. If the model still cannot run normally after restarting and crashes again, the system will enter the second stage of processing, replace the atmospheric forcing field, replace the previously used ERA5 atmospheric forcing field with the GFS atmospheric forcing field, and continue to run from the crash point. If the model runs stably, the GFS atmospheric forcing field will be switched back to the original ERA5 field and continue to execute. If this process goes smoothly, exit the crash handling program; If the above steps still fail to solve the problem, the system will take the third stage measures, use the wind field and heat flux data in the atmospheric forcing field of the previous day to replace the data at the stop time point, and resubmit the run. If the model operation returns to normal, the data will be replaced with the wind field and heat flux at the original time point, continue to run, and exit the crash handler.
Citation Information
Patent Citations
Micro-landform icing prediction method for power grid, and system
CA3185295A1
Marine environment information three-dimensional visualization method
CN102831644A
Jellyfish disaster early warning model construction method
CN112613239A
Method for predicting marine organisms in surrounding sea area of nuclear power plant
CN114492973A
Whole-water-area intelligent water detection management method for marine ranch construction
CN114596167A
Cited By
Construction method of earth system coupling device and earth system coupling device
CN119938130A
Data pre-processing device and method for domestic ocean mode earth coupling system
CN120297004A
Bay environment simulation forecasting method and device
CN120409326A
A bay environment simulation and forecasting method and device
CN120409326B
Intelligent management system and method for sheep flock pasture
CN120688720A
