A multi-scale assimilation method suitable for high-resolution ocean models

Abstract

The application discloses a multiscale assimilation method suitable for a high-resolution ocean model and belongs to the field of ocean observation numerical simulation.The application adopts a multi-grid interpolation method to realize conversion of high-resolution model grid data and low-resolution model grid data in view of defects of a single-scale assimilation method.On this basis, a two-step assimilation idea is adopted, that is, large-scale information is first assimilated based on low-resolution model data, and then medium and small scale assimilation is carried out based on the high-resolution model after the large-scale information is corrected.The application has the advantages that the conversion of high-resolution model grid data and low-resolution model grid data is realized, the shortcomings of the single-scale assimilation method in the assimilation of the high-resolution ocean model are overcome, the scientific combination of observation data and the ocean model is realized, a new multiscale assimilation method is formed, and the depiction capability of the multiscale process of the ocean model is improved.

Description

Technical Field

[0001] This invention relates to a multi-scale assimilation method for the ocean, belonging to the field of ocean observation numerical simulation. Background Technology

[0002] With advancements in ocean observation techniques, ocean data now contains more small- and mesoscale information, rendering existing assimilation methods ineffective in utilizing all the scales of information contained within the data. For example, the SWOT satellite, a new data source, provides greater coverage of mesoscale and sub-mesoscale fields. While model assimilation of SWOT data using existing single-scale assimilation methods has significantly improved assimilation efficiency, the assimilated model still doesn't include the smaller-scale information found in the SWOT satellite data. Current Glider data has high spatiotemporal resolution at the subsurface level; conventional assimilation methods can only correct for large-scale information, meaning high-resolution observational information cannot be effectively utilized. Furthermore, because high-resolution ocean observation data is relatively scarce, there is a need to develop methods that maximize the utilization of this data.

[0003] As the resolution of ocean numerical models increases, the information scales contained in these models become richer. Some observational data have low spatial resolution, and using conventional assimilation methods to assimilate high-resolution ocean models constitutes a spatial filtering process that destroys the small- and medium-scale information contained within the high-resolution model. Therefore, single-scale assimilation can lead to the problem of information confusion across different scales.

[0004] In summary, developing multi-scale assimilation methods can help to effectively utilize high-resolution observation data while avoiding the problem of information confusion at different scales caused by single-scale assimilation, thus further enhancing the data assimilation capability of high-resolution ocean models. Summary of the Invention

[0005] To overcome the shortcomings of single-scale assimilation methods in high-resolution ocean model assimilation studies, the present invention aims to provide a more reasonable multi-scale assimilation method suitable for high-resolution ocean models. This method overcomes the problems of single-scale methods in high-resolution ocean model assimilation studies, such as the inability to effectively utilize all scale information contained in high-resolution observations and the destruction of small and medium-scale information contained in high-resolution ocean models. The goal is to achieve a scientific combination of observational data and model results.

[0006] This invention improves upon single-scale assimilation methods. Addressing the shortcomings of single-scale assimilation, it employs a multi-grid interpolation method to convert high-resolution model grid data into low-resolution model grid data. Building upon this, a two-step assimilation approach is adopted: first, large-scale information is assimilated from the low-resolution model data; then, small-to-medium-scale assimilation is performed on the high-resolution model after correcting the large-scale information. This achieves a scientific integration of observational data and oceanographic models, forming a new multi-scale assimilation method.

[0007] The specific technical solution of the present invention is as follows:

[0008] (1) Based on the N-year high-resolution historical data obtained by the time-integration of the ocean model without assimilation, I seasons are randomly selected each year, and J days are randomly selected from the selected seasons to construct annual set data. At the same time, this annual set data is interpolated onto a low-resolution grid to construct low-resolution historical set data.

[0009] (2) Based on N years of historical data obtained from the unassimilated time integration of the ocean model, M days are randomly selected from each season of each year of the historical data to construct seasonal set data;

[0010] (3) Based on the high-resolution historical data and low-resolution historical data information in step (1), prepare high-resolution and low-resolution grid data land and sea mask information files;

[0011] (4) Add a multi-grid interpolation module, which includes three functions: reading in the high-resolution and low-resolution grid data land and sea mask information files prepared in step (3); reading in the high-resolution grid data, using the two land and sea mask information files, and converting the high-resolution grid data to low-resolution grid data based on the nearest neighbor interpolation method; reading in the low-resolution grid data, using the two land and sea mask information files, converting the low-resolution grid data to high-resolution grid data based on the inverse distance weighting method, and using the nine-point smoothing method to perform spatial smoothing on the high-resolution grid data obtained after conversion;

[0012] (5) Run the high-resolution ocean model to obtain the data to be assimilated at a certain assimilation time. Use the land-sea mask information file in step (3) and the multi-grid interpolation module in step (4) to interpolate the data to be assimilated onto the low-resolution grid.

[0013] (6) Using the low-resolution historical set data in step (1), the low-resolution data to be assimilated in step (5), and the observation data at the current assimilation time, perform assimilation analysis based on the EAKF assimilation method to obtain the low-resolution adjustment amount of the current running model.

[0014] (7) Using the land and sea mask information file in step (3) and the multi-grid interpolation module in step (4), the low-resolution adjustment amount of the current running model obtained in step (6) is interpolated onto the high-resolution model grid to obtain the large-scale high-resolution adjustment amount of the current running model.

[0015] (8) Add the high-resolution data to be assimilated in step (5) and the large-scale high-resolution adjustment in step (7) to obtain the high-resolution data to be assimilated after large-scale information correction.

[0016] (9) Using the high-resolution seasonal ensemble data obtained in step (2) and the high-resolution data to be assimilated after large-scale information correction obtained in step (8), as well as the observation data at the current assimilation time, the assimilation analysis step is carried out based on the EAKF assimilation method to obtain the high-resolution adjustment amount of small and medium-scale information of the current running model.

[0017] (10) Finally, input the large-scale high-resolution adjustment amount in step (7) and the small-scale information high-resolution adjustment amount in step (9) into the currently running high-resolution ocean model to complete a multi-scale assimilation.

[0018] (11) During the operation of the ocean model, repeat steps (5)-(10) at each assimilation time to finally complete the multi-scale assimilation of the high-resolution ocean model.

[0019] In steps (1) and (2), the historical data includes sea surface height, ocean temperature, and ocean salinity.

[0020] In step (1), N≥5, I≥3, J≥4.

[0021] The annual set of data constructed in step (2) is N*I*J≥60, but considering the computational cost, it should be N*I*J≤120.

[0022] In step (2), the data to be assimilated includes daily average sea surface height, daily average ocean temperature, and daily average ocean salinity.

[0023] In step (2), M = I*J.

[0024] This invention adds a multi-grid interpolation module based on different interpolation methods to the assimilation study of ocean models, realizing the conversion between high-resolution model grid data and low-resolution model grid data. It overcomes the shortcomings of single-scale assimilation methods in high-resolution ocean model assimilation, such as high computational requirements and destruction of small and medium-scale information in high-resolution models. It realizes the scientific combination of observation data and ocean models, forms a new multi-scale assimilation method, and improves the ability of ocean models to characterize multi-scale processes. Attached Figure Description

[0025] Figure 1 This is a flowchart of the present invention.

[0026] Figure 2 This is a spatial distribution map of the root mean square error of the time-averaged sea surface temperature relative to OISST in Example 1 (115°E-130°E, 30°N-40°N).

[0027] Figure 3 This is a spatial distribution map of the root mean square error of the time-averaged sea surface salinity relative to OISSS in Example 1 (115°E-130°E, 30°N-40°N). Detailed Implementation

[0028] To further understand the invention's content, features, and technical effects, the invention will be described in detail below with reference to the accompanying drawings, comparative examples, and specific embodiments.

[0029] Comparative Example 1: Assimilation-Free Method for Ocean Models

[0030] The ocean model is run directly to obtain numerical simulation results without introducing model adjustment variables through assimilation methods; the entire process only requires the ocean model.

[0031] Comparative Example 2: Single-Scale Assimilation Method for Ocean Models

[0032] The ocean model runs, and at the assimilation time, the model adjustment is obtained based on the EAKF assimilation method and then input into the ocean model. The overall process consists of two parts: the ocean model and the EAKF assimilation module.

[0033] Example 1: Multi-scale assimilation method for ocean models

[0034] like Figure 1 As shown, a multi-scale assimilation method suitable for high-resolution ocean models is proposed. This method employs a two-step assimilation process, which further assimilates the model adjustment quantities obtained from the two-step assimilation into the ocean numerical model to update the model state, thereby improving the ability of the ocean model to characterize multi-scale processes. The overall process consists of three parts: the ocean model, the EAKF assimilation module, and the multi-grid interpolation module.

[0035] The specific implementation below is a numerical simulation of a global ocean model, using the ocean model multi-scale assimilation method.

[0036] The observational data used in Comparative Examples 2 and 3 were OISST, OISSS, and Argo; the simulation period for all three experiments was from January to March 2011, and the assimilation time window for the assimilation experiment was 1 day.

[0037] The multi-scale assimilation method of this invention specifically includes the following steps:

[0038] Random set data preparation:

[0039] (1) Based on the 9-year historical data (including sea surface height, ocean temperature and ocean salinity) obtained by the time integration of the ocean model without assimilation, three seasons are randomly selected each year, and four days are randomly selected from the selected seasons to construct annual ensemble data. At the same time, this annual ensemble data is interpolated onto a 1° grid to construct historical ensemble data with a 1° spatial resolution.

[0040] (2) Based on the 9-year historical data (sea surface height, ocean temperature, and ocean salinity) obtained by the time integration of the ocean model without assimilation, 12 days are randomly selected from each season of each year of the historical data to construct seasonal ensemble data.

[0041] Multi-grid interpolation module development:

[0042] (3) Based on the high-resolution historical data and 1° spatial resolution historical data information in step (1), prepare high-resolution and 1° spatial resolution land and sea mask information files;

[0043] (4) Add a multi-grid interpolation module, which includes three functions: reading in the high-resolution and 1° spatial resolution land and sea mask information files prepared in step (3); reading in the high-resolution grid data, using the two land and sea mask information files, and converting the high-resolution grid data to low-resolution grid data based on the nearest neighbor interpolation method; reading in the low-resolution grid data, using the two land and sea mask information files, and converting the low-resolution grid data to high-resolution grid data based on the inverse distance weighting method, and using the nine-point smoothing method to perform spatial smoothing on the high-resolution grid data obtained after conversion.

[0044] Multiscale assimilation completed in one step:

[0045] (5) Run the high-resolution ocean model OGCTM to obtain the data to be assimilated at a certain assimilation time (daily average sea surface height, daily average ocean temperature, daily average ocean salinity). Use the land-sea mask information file in step (3) and the multi-grid interpolation module in step (4) to interpolate the data to be assimilated onto a grid with a spatial resolution of 1°.

[0046] (6) Using the historical set data with 1° spatial resolution in step (1), the data to be assimilated with 1° spatial resolution in step (5), and the observation data at the current assimilation time (the observation data information is shown in Table 1), the assimilation analysis step is carried out based on the EAKF assimilation method to obtain the adjustment amount of 1° spatial resolution of the current running model.

[0047] Table 1 Observational Data Information

[0048]

[0049] (7) Using the land and sea mask information file in step (3) and the multi-grid interpolation module in step (4), the adjustment amount of the current running model with 1° spatial resolution obtained in step (6) is interpolated onto the high-resolution model grid to obtain the large-scale high-resolution adjustment amount of the current running model.

[0050] (8) Add the high-resolution data to be assimilated in step (5) and the large-scale high-resolution adjustment in step (7) to obtain the high-resolution data to be assimilated after large-scale information correction.

[0051] (9) Using the high-resolution seasonal ensemble data obtained in step (2) and the high-resolution data to be assimilated after large-scale information correction obtained in step (8), as well as the observation data in Table 1, the assimilation analysis step is carried out based on the EAKF assimilation method to obtain the high-resolution adjustment amount of small and medium-scale information of the current running model.

[0052] (10) Finally, input the large-scale high-resolution adjustment amount in step (7) and the small-scale information high-resolution adjustment amount in step (9) into the currently running high-resolution ocean model to complete a multi-scale assimilation.

[0053] Multiscale assimilation experiments completed:

[0054] (11) During the operation of the ocean model, repeat steps (5)-(10) at each assimilation time to complete the multi-scale assimilation of the high-resolution ocean model.

[0055] In this embodiment, the seawater temperature and salinity data output by the model for February and March within the range of 66°S-66°N are compared with the seawater temperature and salinity data in the EN4.2.1 dataset (https: / / www.metoffice.gov.uk / hadobs / en4 / download-en4-2-1.html). The experimental results are evaluated, and the obtained deviations are averaged monthly. Taking seawater temperature and salinity as examples, the percentage improvement in temperature and salinity after assimilation using the single-scale method and after assimilation using the method of this invention is shown in Table 2.

[0056] Table 2. Seawater temperature and temperature deviation, and percentage reduction in deviation, averaged at shallow depths up to 100 meters using different assimilation methods.

[0057]

[0058] As shown in Table 2, in February, the reduction percentages of seawater temperature and salinity deviations obtained using the method of this invention were 38.30% and 23.01%, respectively, representing improvements of 0.72% and 1.42% compared to the single-scale assimilation method. In March, the reduction percentages of seawater temperature and salinity deviations obtained using the method of this invention were 24.85% and 33.21%, respectively, with salinity deviation improving by 1.17% and temperature deviation increasing by 0.94% compared to the single-scale assimilation method. In summary, the improvement in seawater salinity characterization by the method of this invention compared to the single-scale method is approximately 1.3%, improving the characterization of seawater salinity. However, the improvement in seawater temperature characterization is not significant from a deviation perspective.

[0059] The following section will select the 100-meter layer with the largest temperature deviation to further analyze the impact of the method of this invention on temperature characterization from the perspective of the spatial distribution of sea surface temperature before and after assimilation.

[0060] The mean temperature difference in the regions of 100°E-120°E and 15°N-40°N was calculated, as shown in Table 3. Table 3 shows that the mean sea surface temperature difference at a depth of 100 meters in the model after assimilation using the method of this invention is less than that in Comparative Example 2, demonstrating the improved assimilation effect of the method of this invention compared to single-scale assimilation methods.

[0061] Table 3. Absolute mean temperature difference (°C) at a depth of 100m in February and March (100°E-120°E, 15°N-40°N).

[0062] Comparative Example 1 1.4631 1.1796 Comparative Example 2 1.0140 0.8788 Example 1 0.9852 0.8736

[0063] Figure 2 and Figure 3 The spatial distribution of the root mean square error of sea surface temperature and salinity relative to the time averages of OISST and OISSS for the three experiments are shown. It can be found that the assimilation method using the single-scale assimilation method increases the root mean square error near the coast to a certain extent. The assimilation method of this invention can solve this problem and significantly reduce the root mean square error of sea surface temperature and salinity in the Yellow Sea Warm Current region.

[0064] Finally, the root mean square errors of sea surface temperature and salinity in the three experimental regions (100°E-150°E, 0°N-50°N) were calculated, as shown in Table 4. The method of this invention reduces the root mean square error of sea surface temperature by up to 61% and the root mean square error of sea surface salinity by up to 35%, which is an improvement of about 2% compared to single-scale assimilation methods.

[0065] The method of this invention completes multi-scale assimilation of ocean models without increasing computational requirements, and improves the assimilation effect compared with single-scale assimilation experiments.

[0066] Comparative Example 1 1.00 0.60 Comparative Example 2 0.41 0.40 Example 1 0.39 0.39

[0067] The preferred embodiments of the present invention have been described in detail above, but the present invention is not limited to the embodiments described. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention, and these equivalent modifications or substitutions are all included within the scope defined by the claims of this application.

Claims

1. A multi-scale assimilation method suitable for high-resolution ocean models, characterized in that, Includes the following steps: (1) Based on the N-year high-resolution historical data obtained by the time-integration of the ocean model without assimilation, I seasons are randomly selected each year, and J days are randomly selected from the selected seasons to construct annual set data. At the same time, this annual set data is interpolated onto a low-resolution grid to construct low-resolution historical set data. (2) Based on N years of historical data obtained from the unassimilated integral of the ocean model, M days are randomly selected from each season of each year of the historical data to construct seasonal set data; (3) Based on the high-resolution historical data and low-resolution historical data information in step (1), prepare high-resolution and low-resolution grid data land and sea mask information files; (4) Add a multi-grid interpolation module: read in the high-resolution and low-resolution grid data land and sea mask information files prepared in step (3); read in the high-resolution grid data, and use the two land and sea mask information files to convert the high-resolution grid data to low-resolution grid data based on the nearest neighbor interpolation method; Low-resolution grid data is read in, and two land and sea mask information files are used to convert the low-resolution grid data to high-resolution grid data based on the inverse distance weighting method. The high-resolution grid data obtained after conversion is then spatially smoothed using the nine-point smoothing method. (5) Run the high-resolution ocean model to obtain the data to be assimilated at a certain assimilation time. Use the land-sea mask information file in step (3) and the multi-grid interpolation module in step (4) to interpolate the data to be assimilated onto the low-resolution grid. (6) Using the low-resolution historical set data in step (1) and the low-resolution data to be assimilated in step (5) and the observation data at the current assimilation time, perform assimilation analysis based on the EAKF assimilation method to obtain the low-resolution adjustment amount of the current running model. (7) Using the land and sea mask information file in step (3) and the multi-grid interpolation module in step (4), the low-resolution adjustment amount of the current running model obtained in step (6) is interpolated onto the high-resolution model grid to obtain the large-scale high-resolution adjustment amount of the current running model. (8) Add the high-resolution data to be assimilated in step (5) and the large-scale high-resolution adjustment in step (7) to obtain the high-resolution data to be assimilated after large-scale information correction. (9) Using the high-resolution seasonal ensemble data obtained in step (2) and the high-resolution data to be assimilated after large-scale information correction obtained in step (8), as well as the observation data at the current assimilation time, the assimilation analysis step is carried out based on the EAKF assimilation method to obtain the high-resolution adjustment amount of small and medium-scale information of the current running model. (10) Finally, input the large-scale high-resolution adjustment amount in step (7) and the small-scale information high-resolution adjustment amount in step (9) into the currently running high-resolution ocean model to complete a multi-scale assimilation. (11) During the operation of the ocean model, repeat steps (5)-(10) at each assimilation time to finally complete the multi-scale assimilation of the high-resolution ocean model.

2. The multi-scale assimilation method as described in claim 1, characterized in that, In steps (1) and (2), the historical data includes sea surface height, ocean temperature, and ocean salinity.

3. The multi-scale assimilation method as described in claim 1, characterized in that, In step (1), N≥5, I≥3, J≥4.

4. The multi-scale assimilation method as described in claim 1, characterized in that, The size of the annual set data N*I*J constructed in step (2) is between 60 and 120.

5. The multi-scale assimilation method as described in claim 1, characterized in that, In step (2), the data to be assimilated includes daily average sea surface height, daily average ocean temperature, and daily average ocean salinity.

6. The multi-scale assimilation method as described in claim 1, characterized in that, In step (2), M = I * J.

Images