A multi-source observation data-based multi-method marine element fusion and assimilation system

Abstract

The application relates to the field of element fusion, and discloses a multi-method marine element fusion assimilation system based on multi-source observation data, which is used for solving the problems of limited adaptability, rigid fusion strategy and insufficient physical consistency of an existing single assimilation method. The system comprises the following steps: performing bias correction and quality control on multi-source original observation data to form a standardized observation set; independently generating corresponding single-method analysis fields and uncertainty fields by parallel operation of at least two assimilation methods; for each spatial grid point, generating a consensus field based on the numerical values of all single-method analysis fields, and calculating dynamic fusion weights according to the deviation degrees of the analysis values and the uncertainty of each method and the consensus field; and weighting and fusing the single-method analysis fields by using the weights. The application realizes adaptive fusion of the advantages of multiple methods, and significantly improves the precision and physical consistency of the marine analysis field.

Description

Technical Field

[0001] This invention relates to the field of element fusion, and in particular to a multi-method marine element fusion and assimilation system based on multi-source observation data. Background Technology

[0002] As a key component of the Earth's climate system, the ocean's physical, chemical, and biological processes have a profound impact on global climate, the ecological environment, and human socio-economic activities. Accurately understanding the spatial distribution and dynamic changes of ocean elements (temperature, salinity, current velocity, etc.) is crucial for marine scientific research, marine resource development, marine disaster early warning, and climate prediction.

[0003] With the continuous development of marine observation technologies, multi-source observation data has become increasingly abundant, encompassing various methods such as satellite remote sensing, buoy observation, shipborne observation, and profiler observation. These observation data monitor ocean conditions from different angles and scales, providing a massive amount of data for a deeper understanding of the ocean. However, observation data from different sources exhibit significant differences in data quality, spatiotemporal resolution, and observation error characteristics. How to effectively integrate these multi-source observation data, fully leverage their respective advantages, and improve the accuracy and reliability of marine element analysis has become a key issue that urgently needs to be addressed in the field of marine data assimilation.

[0004] When integrating results from multiple assimilation methods, existing techniques often employ simple averaging or linear combination, failing to fully consider the uncertainties and relative advantages of each method at different spatial locations. This simplistic fusion approach cannot dynamically adjust the weights of each method according to actual conditions, resulting in poor spatial consistency and accuracy of the fusion results.

[0005] Current technologies do not adequately consider the physical equilibrium relationships between oceanographic variables when generating fusion analysis fields. Oceanographic elements such as temperature, salinity, and current velocity are interconnected and follow specific dynamic and thermodynamic laws. Fusion analysis fields lacking physical equilibrium constraints may violate these laws, resulting in analytical results that lack scientific validity and rationality, and fail to accurately reflect the true state of the ocean.

[0006] Existing technologies are not comprehensive or accurate enough in assessing the uncertainty of fusion results. During multi-method fusion, the uncertainty information of each method is not effectively integrated to generate a fusion uncertainty field that accurately reflects the uncertainty of the fusion analysis field, which is detrimental to the scientific evaluation of the reliability and accuracy of the fusion results.

[0007] Therefore, we propose a multi-method marine element fusion and assimilation system based on multi-source observation data to solve the above problems. Summary of the Invention

[0008] This invention provides a multi-method marine element fusion and assimilation system based on multi-source observation data, which addresses the problems of limited adaptability, rigid fusion strategies, and insufficient physical consistency of existing single assimilation methods.

[0009] The first aspect of this invention provides a multi-method marine element fusion and assimilation system based on multi-source observation data. The system includes: a processing module for bias correction and quality control of the multi-source raw observation data to form a standardized observation dataset; a parallel module for assimilating the standardized observation dataset independently and in parallel using at least two assimilation methods to generate a single-method analysis field and a single-method uncertainty field corresponding to each method; a weighting module for generating a consensus field value for each spatial grid point using the values ​​of all the single-method analysis fields at that grid point, and calculating the dynamic fusion weight for that grid point based on the deviation of the values ​​of each single-method analysis field, the corresponding single-method uncertainty field value, and the consensus field value; a setting module for weighted fusion of the corresponding single-method analysis fields at all spatial grid points using the dynamic fusion weight to generate a preliminary fusion analysis field; and a coordination module for coordinating the marine element variables in the preliminary fusion analysis field using physical equilibrium relationships as constraints to generate a marine element fusion analysis field.

[0010] Optionally, in a first implementation of the first aspect of the present invention, the method includes: assimilating the standardized observation dataset to generate a first analysis field and obtaining a first uncertainty field corresponding to the first analysis field; assimilating the standardized observation dataset to generate a second analysis field and obtaining a second uncertainty field corresponding to the second analysis field; the first uncertainty field is obtained by using the diagonal elements of the analysis error covariance matrix of the three-dimensional variational assimilation method, and the second uncertainty field is obtained by using the statistical characteristics of the analysis set samples of the ensemble Kalman filter assimilation method.

[0011] Optionally, in a second implementation of the first aspect of the present invention, the method includes: for each spatial grid point, calculating a multi-method consensus field value for that grid point based on the values ​​of all single-method analysis fields aggregated to that grid point; calculating an initial weight based on uncertainty according to the values ​​of each single-method uncertainty field at that grid point; calculating the relative deviation between the values ​​of each single-method analysis field and the values ​​of the multi-method consensus field, and calculating a consistency-based weight factor based on the relative deviation; and processing the initial weight based on uncertainty and the consistency-based weight factor to obtain a dynamic fusion weight.

[0012] Optionally, in a third implementation of the first aspect of the present invention, the method includes: for each spatial grid point, obtaining the values ​​of each single-method analysis field at that grid point and the dynamic fusion weight of that grid point; using the dynamic fusion weight to process the values ​​of each single-method analysis field to obtain the preliminary fusion analysis value of that grid point; and integrating the preliminary fusion analysis values ​​of all spatial grid points to form a spatially continuous preliminary fusion analysis field.

[0013] Optionally, in a fourth implementation of the first aspect of the present invention, the method includes: generating a local smoothing intensity parameter for each spatial grid point based on the spatial distribution of the uncertainty field of each single method; processing the preliminary fusion analysis value of each spatial grid point according to the local smoothing intensity parameter to obtain a smoothed preliminary fusion analysis value; and combining all the smoothed preliminary fusion analysis values ​​to form a spatially continuous preliminary fusion analysis field.

[0014] Optionally, in a fifth implementation of the first aspect of the present invention, the adaptive smoothing radius of the target grid point is calculated using the following formula. :

[0015] ;

[0016] in, Set the lower limit for background smoothing; This is the spatial expansion adjustment coefficient; The preset threshold for uncertainty in the marine environment; This is the quantification value of the uncertainty of the 3D VAR analysis field at the target grid point i; The contribution percentage of the 3DVAR analysis field results at target grid point i; This is the quantification value of the uncertainty of the EnKF analysis field at the target grid point i. This represents the percentage of the EnKF analysis field results contributed to the target grid point i.

[0017] Optionally, in the sixth implementation of the first aspect of the present invention, the method includes: constructing a physical equilibrium constraint relationship between marine element variables based on a preset marine dynamics and thermodynamics relationship; setting a balance adjustment target for at least two marine element variables in the preliminary fusion analysis field according to the physical equilibrium constraint relationship; performing coordinated iterative adjustment on at least two marine element variables to make the adjusted variable relationship approach the balance adjustment target, while minimizing the overall adjustment range of the original values ​​of each variable; stopping the iteration when the variable relationship meets the preset convergence criterion, and outputting the adjusted results of all marine element variables.

[0018] Optionally, in the seventh implementation of the first aspect of the present invention, a collaborative module is further included, which is used to: calculate the fusion uncertainty field corresponding to the marine element fusion analysis field based on the uncertainty fields of each single method and the dynamic fusion weight.

[0019] Optionally, in the eighth implementation of the first aspect of the present invention, the method includes: for each spatial grid point, processing the values ​​of each single-method uncertainty field at that grid point using the dynamic fusion weight of that grid point to obtain the fusion uncertainty value of that grid point; and integrating the fusion uncertainty values ​​of all spatial grid points to form a fusion uncertainty field.

[0020] The mechanism of this invention is as follows: it achieves autonomous selection and fusion of the advantages of each algorithm without human intervention, ensuring the dynamic and thermodynamic consistency of the final analysis field, and has the ability to provide an uncertain field after fusion;

[0021] Beneficial effects: The uncertainty fields of each method are obtained separately. The uncertainty is quantified by analyzing the diagonal elements of the error covariance matrix using 3DVAR and the statistical characteristics of the sample set using EnKF. The quantification of uncertainty provides a key basis for the subsequent calculation of dynamic fusion weights, enabling the fusion process to fully consider the reliability of each method.

[0022] For each spatial grid point, consensus field values ​​are generated using single-method analysis field values. Dynamic fusion weights are calculated based on the deviations of each single-method analysis field value, the corresponding uncertain field value, and the consensus field value. The dynamic fusion weights are obtained by comprehensively considering the initial weights based on uncertainty and the weight factors based on consistency through reasonable processing.

[0023] Complex physical equilibrium relationships exist among oceanographic variables, and preliminary fusion analysis may violate these laws. By adjusting physical equilibrium constraints, the relationships between variables in the fusion analysis field are ensured to conform to the basic principles of ocean dynamics and thermodynamics, thereby improving the scientific rigor and rationality of the analysis results and enabling them to more accurately reflect the true state of the ocean.

[0024] The fusion uncertainty field comprehensively assesses the reliability of the fusion results, providing users with important information about the accuracy of the fusion analysis field. Users can understand the credibility of the fusion results for different regions and different elements based on the fusion uncertainty field, thereby making more reasonable decisions in practical applications and improving the application value of the fusion results. Attached Figure Description

[0025] Figure 1 This is a schematic diagram of an embodiment of a multi-method marine element fusion and assimilation system based on multi-source observation data in this invention.

[0026] Figure 2This is a schematic diagram of an embodiment of the multi-method marine element fusion and assimilation device based on multi-source observation data in this invention.

[0027] Figure 3 This is a schematic diagram of an embodiment of a multi-method marine element fusion and assimilation device based on multi-source observation data in this invention. Detailed Implementation

[0028] This invention provides a multi-method oceanographic element fusion and assimilation system based on multi-source observation data, addressing the problems of limited adaptability, rigid fusion strategies, and insufficient physical consistency in existing single assimilation methods. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0029] For ease of understanding, the specific process of the embodiments of the present invention is described below. Please refer to [link / reference]. Figure 1 One embodiment of the multi-method ocean element fusion and assimilation system based on multi-source observation data in this invention includes:

[0030] 101. Processing module, used to perform bias correction and quality control on multi-source raw observation data to form a standardized observation dataset.

[0031] It is understood that the executing entity of this invention can be a multi-method ocean element fusion and assimilation device based on multi-source observation data, or it can be a terminal or a server; the specific implementation is not limited here. This embodiment of the invention will be described using a server as an example.

[0032] It should be noted that this process involves sea surface temperature observation data from the Northwest Pacific (20°N to 30°N) during the summer.

[0033] The system received area data from the polar-orbiting satellite's infrared sensor and point data from the Argo profiling buoy. Based on oceanographic knowledge, the satellite measures skin temperature, which is only a few millimeters above the sea surface, while the Argo buoy measures body temperature at a depth of approximately two meters. At one observation point, the satellite measured a skin temperature of 28.1 degrees Celsius, while Argo measured a body temperature of 28.6 degrees Celsius. The system calibrated the satellite data using a skin-to-body temperature physical conversion model, calculating the corresponding body temperature to be 28.3 degrees Celsius. This value was then compared to the absolute true value from Argo (28.6 degrees Celsius), revealing a systematic instrument bias of -0.3 degrees Celsius in this batch of satellite observations. The system subsequently adjusted all calibrated satellite data for the area upwards by 0.3 degrees Celsius for overall compensation.

[0034] Background field verification was performed on the compensated data. At a certain satellite grid point, the infrared signal was severely attenuated due to interference from unidentified local thick clouds, resulting in a temperature reading of 15.2 degrees Celsius. The system retrieved the summer climatological background field for that sea area (normal water temperature is between 26 and 31 degrees Celsius) and determined that this value constituted an anomalous cold jump that severely violated natural laws, triggering the control mechanism to directly remove it.

[0035] All retained observations were standardized to the unit of Celsius and the time to Coordinated Universal Time. Spatial objective interpolation was used to smoothly map the point and area data onto a standard latitude and longitude grid with a spatial resolution of 0.25 degrees, forming a standardized observation dataset.

[0036] 102. Parallel module, used to assimilate a standardized observation dataset in parallel and independently using at least two assimilation methods, generating a single-method analysis field and a single-method uncertainty field corresponding to each method.

[0037] It should be noted that a standardized sea surface temperature dataset is obtained at this point. At the standard grid coordinates (25°N, 135°E), the ocean model's background predicted temperature is 28.0 degrees Celsius, while the standardized observed temperature provided in step 101 is 28.6 degrees Celsius. The system inputs this data in parallel into the following two independent methods:

[0038] The three-dimensional variational assimilation method (3DVAR) relies on the historical statistical error covariance for spatial weight allocation. After variational iterative solution, 3DVAR calculates the temperature of the analysis field at this grid point to be 28.4 degrees. Simultaneously, based on its static error model, the system assesses the single-method uncertainty (i.e., error variance estimate) of this result to be 0.3 degrees.

[0039] The Ensemble Kalman Filter (EnKF) method tracks transient characteristics of ocean fluids in real time using dozens of dynamic model samples. The system detected the passage of a branch of the Kuroshio Current, and EnKF's dynamic samples accurately captured this change, thus giving the observational data higher reliability. Through independent solution, EnKF calculated the analytical field temperature at this grid point to be 28.5 degrees Celsius. Due to its excellent dynamic tracking performance, the system assesses its single-method uncertainty to be low, at only 0.2 degrees Celsius.

[0040] For the same grid point, two independent temperature analysis fields with error assessment were successfully generated in parallel.

[0041] 103. Weighting module: For each spatial grid point, it generates a consensus field value using the values ​​of all single-method analysis fields at that grid point, and calculates the dynamic fusion weight of that grid point based on the values ​​of each single-method analysis field, the degree of deviation between the corresponding single-method uncertainty field value and the consensus field value.

[0042] It should be noted that at the grid point (25 degrees north latitude, 135 degrees east longitude), two independent results from step 102 are read: 3DVAR (value 28.4 degrees, uncertainty 0.3 degrees) and EnKF (value 28.5 degrees, uncertainty 0.2 degrees).

[0043] The arithmetic mean of 28.4 degrees and 28.5 degrees is used to establish a consensus field reference benchmark of 28.45 degrees for the current grid point. The system quantifies the overall performance of each method by constructing a statistical error penalty function. For the 3DVAR method, its absolute deviation from the consensus benchmark is 0.05 degrees. The system inputs this deviation, along with its own uncertainty of 0.3 degrees, into the penalty function, calculating its overall error penalty index to be 0.35. For the EnKF method, its absolute deviation is also 0.05 degrees, but because its own uncertainty is only 0.2 degrees, the overall error penalty index obtained after inputting it into the function is only 0.25.

[0044] Based on the inverse mapping principle that smaller error penalties result in higher weights, the system adds the reference values ​​of the two methods together (0.60). The EnKF method, which performs better overall, obtains approximately 58% of the dynamic fusion weights through cross-inverse calculation; while the slightly less effective 3DVAR method obtains the remaining approximately 42% of the weights. This precise set of dynamic weights will directly serve subsequent fusion calculations.

[0045] 104. The configuration module is used to perform weighted fusion of the corresponding single-method analysis fields on all spatial grid points using dynamic fusion weights to generate a preliminary fused analysis field.

[0046] It should be noted that global weighted calculations were initially performed on the 0.25-degree resolution grid of the Northwest Pacific.

[0047] At (25°N, 135°E), the system retrieves data and performs multiplication and addition operations. Multiplying the 28.4 degrees from 3DVAR by its 42% weight yields a contribution value of 11.928 degrees; multiplying the 28.5 degrees from EnKF by its 58% weight yields a contribution value of 16.53 degrees. The system adds the two contribution values ​​together, obtaining a preliminary fusion value of 28.458 degrees for this grid point, which is recorded as 28.46 degrees with two decimal places.

[0048] At the adjacent location (25.25°N, 135°E), due to changes in the zone's dynamics, the weight allocation was reversed: 3DVAR received 60% of the weight (value 28.2 degrees), and EnKF received 40% of the weight (value 28.1 degrees). The system then performed the same calculations: 28.2 multiplied by 60% equals 16.92 degrees, and 28.1 multiplied by 40% equals 11.24 degrees. Summing these two values ​​yields a preliminary fusion value of 28.16 degrees for grid point B.

[0049] The aforementioned weighted accumulation operation was performed on tens of thousands of grid points one by one in the background. Finally, these independent grid points carrying fusion characteristics were stitched together by a spatial coordinate matrix, formally generating a continuous preliminary fused sea surface temperature analysis field.

[0050] 105. The Collaboration Module is used to collaboratively adjust the marine element variables in the preliminary fusion analysis field by taking the physical equilibrium relationship as a constraint, and generate the marine element fusion analysis field.

[0051] It should be noted that the current simultaneous processing of the preliminary merged sea surface temperature (SST) and sea surface salinity (SSS) data fields is as follows: At (25 degrees North latitude, 135 degrees East longitude), the system reads the preliminary merged SST calculated in step 104 as 28.46 degrees and the preliminary merged SSS as 34.20 practical salinity units.

[0052] Substituting these temperature and salinity values ​​into the seawater state equation, the surface density was calculated. Combined with the historical temperature and salinity structure of the subsurface at this point, the system found that the current surface seawater density is abnormally low due to the high temperature. This top-heavy stratification state violates the physical stability characteristics of the Kuroshio Current in summer and could trigger false convective overturning in the model.

[0053] Using the high-confidence temperature-salinity (TS) characteristic curve of the sea area as a constraint, the temperature and salinity are simultaneously and bidirectionally fine-tuned using a minimum cost function to restore density stability with minimal correction.

[0054] The specific collaborative adjustment data is shown in Table 1 below:

[0055] Table 1

[0056]

[0057] After fine-tuning in accordance with physical principles (salinity increased by only 0.02, avoiding unreasonable drastic jumps; temperature decreased by 0.11 degrees), the grid point regained a stable state consistent with natural laws. After completing such constraint checks and adjustments across the entire grid, the system officially outputs the fused analysis field of marine elements.

[0058] Please see Figure 2 Another embodiment of the multi-method ocean element fusion and assimilation system based on multi-source observation data in this invention includes:

[0059] 101. Processing module, used to perform bias correction and quality control on multi-source raw observation data to form a standardized observation dataset.

[0060] Specifically, bias correction is performed on at least two of the satellite remote sensing data, buoy profile data, and ship survey data to obtain corrected observation data; quality control based on climatological range and spatiotemporal consistency is performed on the corrected observation data to remove outlier data and obtain valid observation data; the valid observation data is matched to a preset assimilation analysis grid in time and space.

[0061] It should be noted that the sea surface temperature (SST) assimilation task in the winter Kuroshio Extension (center coordinates: 35.0°N, 145.0°E) will be used as an example for detailed explanation.

[0062] Within the pre-defined assimilation time window (00:00 to 06:00 UTC), two types of raw observation data with completely different structures were received: the first type was large-area sea surface infrared remote sensing temperature data retrieved by the Advanced Very High Resolution Radiometer (AVHRR) carried by a polar-orbiting meteorological satellite; the second type was high-precision point data of surface water temperature along the shipping route continuously measured by the thermistor in the intake pipe of a merchant ship passing through the area. Multi-source data bias correction was initiated for both types of data. At the observation time, when the merchant ship reached (35.0°N, 145.0°E), the absolute sea surface temperature measured by the thermistor in the intake pipe was 15.50 degrees Celsius. However, the infrared channel signal of the satellite remote sensing data passing through the area at the same time was attenuated due to interference from a specific aerosol layer frequently occurring in this sea area during winter and the white-hat effect caused by strong winds at the sea surface; the temperature retrieved at the same coordinate point was only 15.10 degrees Celsius. The system retrieved all merchant ship-satellite spatiotemporal matching sample databases for the sea area over the past three months. Statistical regression analysis using the least squares method confirmed that this batch of satellite infrared data generally exhibited a systematic cold bias of -0.40 degrees. Based on this, the system automatically performed a global bias correction, forcibly increasing the overall accuracy of all satellite remote sensing observation data within the region by 0.40 degrees, thereby generating corrected observation data that eliminated systematic errors from instruments and the environment.

[0063] The corrected data underwent rigorous quality control based on climatological range and spatiotemporal consistency. At a nearby location with spatial coordinates (35.1°N, 145.1°E), the system detected a satellite observation point with a temperature of 5.20°C. A climatological range check was performed: historical climatological background data from the past twenty years for this sea area during winter was retrieved. This confirmed that the three-standard-deviation confidence interval for normal winter surface water temperature at this location should strictly fall between 12.00°C and 18.00°C; 5.20°C significantly deviated from common physical norms. A spatiotemporal consistency check (BuddyCheck) was performed: dozens of other valid observation data within a 10-kilometer radius of this anomaly were examined, revealing that adjacent water temperatures were consistently distributed between 15.20°C and 15.60°C. In the absence of support from extreme mesoscale strong cold vortex weather systems, such a 10-degree spatial temperature jump over such a short distance completely violates the spatial continuity of oceanic fluids. The 5.20-degree data was determined to be caused by the satellite sensor misidentifying an unidentified low-altitude cold cloud top as the sea surface. This triggered the anomaly removal mechanism, which completely deleted the data. The remaining data that passed the double verification was marked as valid observation data.

[0064] The preset assimilation analysis grid has a spatial resolution of 0.1 degrees. The system extracts and retains the measured data of the merchant ship (15.50 degrees) and the effective satellite data from nearby locations (15.40 degrees and 15.60 degrees). Based on their spatial distance from the central grid point and the observation time difference, the basic interpolation weights are calculated using an objective analysis method. After multi-dimensional distance inverse weighting calculation, all scattered data are smoothly and accurately mapped to a standard analysis grid with a central coordinate of (35.0 degrees North latitude, 145.0 degrees East longitude). A unique and representative standardized observation value of 15.50 degrees is generated at this grid point, and a standardized observation dataset with high signal-to-noise ratio and a unified spatial benchmark is constructed.

[0065] 102. Parallel module, used to assimilate a standardized observation dataset in parallel and independently using at least two assimilation methods, generating a single-method analysis field and a single-method uncertainty field corresponding to each method.

[0066] Specifically, a three-dimensional variational assimilation method is run to assimilate the standardized observation dataset, generate a first analysis field, and obtain a first uncertainty field corresponding to the first analysis field; an ensemble Kalman filter assimilation method is run to assimilate the standardized observation dataset, generate a second analysis field, and obtain a second uncertainty field corresponding to the second analysis field; wherein, the first uncertainty field is obtained by the diagonal elements of the analysis error covariance matrix of the three-dimensional variational assimilation method, and the second uncertainty field is obtained by the statistical characteristics of the analysis set samples of the ensemble Kalman filter assimilation method.

[0067] It should be noted that for the grid point (35.0°N, 145.0°E) generated in the previous step, its standardized observation value of 15.50 degrees is extracted. Simultaneously, the system reads the background field prediction value for this grid point at this time from the current high-resolution ocean numerical prediction model (ROMS model), assumed to be 16.00 degrees. The observed value of 15.50 degrees and the background value of 16.00 degrees are paired and simultaneously distributed to two completely independent assimilation calculations.

[0068] The 3D Variational Assimilation (3DVAR) method is used to find the optimal solution by minimizing the objective cost function. Its error perception relies on a pre-defined, time-constant background error covariance matrix. Upon receiving data, 3DVAR, based on its static historical statistical characteristics, assesses that the 16.00 degrees given by the forecast model and the 15.50 degrees given by the observation equipment each contain a certain proportion of random error. After multiple rounds of variational iteration, 3DVAR finds a static mathematical equilibrium point between the background field and the observation field, appropriately stretching the predicted value towards the observed value, generating the first analysis field value of 15.70 degrees for this grid point. Simultaneously, to quantify the reliability of this result, the system directly consults the pre-defined analysis error covariance matrix in the 3DVAR algorithm, accurately extracting the main diagonal elements corresponding to the spatial location (35.0 degrees North latitude, 145.0 degrees East longitude). The diagonal element physically represents the estimated variance of that point. The system analyzes it and obtains the first uncertainty field (standard deviation) value of 0.40 degrees for that grid point.

[0069] Concurrently, the Ensemble Kalman Filter Assimilation Method (EnKF) was run in parallel. Unlike the static framework of 3DVAR, EnKF, by introducing random perturbations, evolved in real time 50 three-dimensional ensemble samples of ocean models representing different hydrological probabilities. At this specific observation moment, the system performed statistical algorithm analysis on the temperature prediction values ​​of these 50 samples at (35.0°N, 145.0°E), discovering an unusually large dispersion among the samples. This large ensemble divergence conveyed a clear physical signal to the system: there is strong mesoscale eddy activity and fluid transients in the current Kuroshio Extension, resulting in extremely low prediction confidence for the model background field (16.00°). Based on this real-time dynamic feedback, EnKF automatically and significantly increased the confidence weight of the observed data (15.50°) in the filter update equation. After matrix gain calculation, the second analysis field value generated by EnKF was 15.60°, which is closer to the actual observation than the result of 3DVAR.

[0070] When obtaining the error quantification index, the system directly performs mathematical statistics on the 50 assimilated and updated analysis set samples, calculating the standard deviation of the temperature values ​​of these samples at the grid point. Because the samples converge rapidly to the observed values ​​after the update, the dispersion is significantly reduced. Using this statistical characteristic, the system accurately obtains the second uncertainty field value of only 0.25 degrees for this grid point. Thus, for the same spatial coordinates, the system not only outputs two sets of different temperature analysis values ​​(15.70 degrees and 15.60 degrees), but also outputs their respective rigorous error confidence indices (0.40 degrees and 0.25 degrees).

[0071] 103. Weighting module: For each spatial grid point, it generates a consensus field value using the values ​​of all single-method analysis fields at that grid point, and calculates the dynamic fusion weight of that grid point based on the values ​​of each single-method analysis field, the degree of deviation between the corresponding single-method uncertainty field value and the consensus field value.

[0072] Specifically, for each spatial grid point, based on the values ​​of all single-method analysis fields aggregated to that grid point, the multi-method consensus field value for that grid point is calculated. Based on the values ​​of the uncertainty fields of each single method at that grid point, a set of initial weights based on uncertainty is calculated. The relative deviation between the values ​​of each single-method analysis field and the values ​​of the multi-method consensus field is calculated, and a set of consistency-based weight factors is calculated based on the relative deviation. The initial weights based on uncertainty and the consistency-based weight factors are synthesized and normalized to obtain dynamic fusion weights. Further, the absolute difference between the values ​​of each single-method analysis field and the values ​​of the multi-method consensus field is divided by the standard deviation of all single-method analysis field values ​​to obtain the normalized deviation for each method. The normalized deviations are transformed using an exponential decay function to obtain a set of initial consistency weight factors. The initial consistency weight factors are then normalized to obtain consistency-based weight factors.

[0073] It should be noted that the two independently generated core data pairs in step 102 are retrieved: the analysis field value output by 3DVAR is 15.70 degrees, and the corresponding uncertainty value (standard deviation) is 0.40 degrees; the analysis field value output by EnKF is 15.60 degrees, and the corresponding uncertainty value is 0.25 degrees.

[0074] The multi-method consensus field value is calculated. The arithmetic mean of all single-method analysis field values ​​aggregated to this grid point is calculated as (15.70 degrees + 15.60 degrees) / 2, yielding a baseline value of 15.65 degrees for this grid point. Subsequently, the system performs an initial weight allocation strictly based on the inverse ratio of error variance, according to the uncertainty field values ​​of each single method. The system squares the uncertainty value of 3DVAR (0.40 degrees), obtaining an error variance of 0.1600; taking its reciprocal yields a baseline confidence index of 6.25. Similarly, the system squares the uncertainty value of EnKF (0.25 degrees), obtaining an error variance of 0.0625; taking its reciprocal yields a baseline confidence index of 16.00. To ensure the total weight sum is 100%, the system normalizes these two reciprocal indices: the total confidence index is (6.25 + 16.00) = 22.25. Calculations show that the initial weights obtained by 3DVAR based on uncertainty are approximately (6.25 / 22.25) ≈ 28.09%, which are retained in the system's internal high-precision calculations and are approximately equal to 28% for simplicity. The initial weights obtained by EnKF are approximately (16.00 / 22.25) ≈ 71.91%, approximately 72%. This allocation logic based on the inverse of the error variance mathematically maximizes the weight advantage of the low-error method.

[0075] The absolute difference between the numerical values ​​of each single-method analysis field and the consensus numerical value of the multi-method analysis field (15.65 degrees) was calculated. The absolute difference for 3DVAR (15.70 degrees) and EnKF (15.60 degrees) was both 0.05 degrees. Subsequently, the standard deviation of these two numerical values ​​(15.70 and 15.60) was calculated, and the result was also 0.05 degrees. Dividing the absolute difference of each method by this standard deviation yielded a normalized deviation of 1 for both 3DVAR and EnKF.

[0076] Next, the two normalized deviations are input into a preset exponential decay function for nonlinear transformation. Since the deviation values ​​are exactly the same, the initial consistency weight factors obtained after the transformation are also absolutely equal.

[0077] After final normalization, both methods obtain a 50% consistency-based weighting factor each. During the synthesis phase, the initial weight of 3DVAR (28%) is multiplied by its consistency weighting factor (50%), and the initial weight of EnKF (72%) is multiplied by its consistency weighting factor (50%). After normalizing these two products again, due to their perfectly equivalent consistency performance, the final dynamic fusion weight for this grid point is entirely dominated by the previous optimal variance assessment. The determined result is: for this specific spatial grid point, the 3DVAR method is assigned a 28% dynamic fusion weight, while the superior EnKF method is assigned an absolutely dominant 72% dynamic fusion weight.

[0078] 104. The configuration module is used to perform weighted fusion of the corresponding single-method analysis fields on all spatial grid points using dynamic fusion weights to generate a preliminary fused analysis field.

[0079] Specifically, for each spatial grid point, the numerical values ​​of each single-method analysis field at that grid point and the dynamic fusion weights of that grid point are obtained; the numerical values ​​of each single-method analysis field are weighted and summed using the dynamic fusion weights to obtain the preliminary fusion analysis value for that grid point; the preliminary fusion analysis values ​​of all spatial grid points are integrated to form a spatially continuous preliminary fusion analysis field. Further, based on the spatial distribution of the uncertainty fields of each single method, local smoothing intensity parameters are generated for each spatial grid point; according to the local smoothing intensity parameters, the preliminary fusion analysis values ​​of each spatial grid point are subjected to adaptive spatial smoothing processing to obtain smoothed preliminary fusion analysis values; all smoothed preliminary fusion analysis values ​​are combined to form a spatially continuous preliminary fusion analysis field.

[0080] It should be noted that at the target grid point (35.0°N, 145.0°E), various parameters were precisely read: the 3DVAR analysis field value was 15.70 degrees, with a corresponding dynamic fusion weight of 28%; the EnKF analysis field value was 15.60 degrees, with a corresponding dynamic fusion weight of 72%. The system strictly performed a weighted summation operation: multiplying 15.70 degrees by 28% yielded an absolute numerical contribution of 4.396 degrees to this point; multiplying 15.60 degrees by 72% yielded a numerical contribution of 11.232 degrees. The system precisely added the two contribution values, i.e., 4.396 degrees plus 11.232 degrees, to obtain the initial absolute value of the grid point after weighted fusion, which was 15.628 degrees. Following a uniform standard and retaining two decimal places, the initial fusion analysis value of this grid point without smoothing was determined to be 15.63 degrees.

[0081] Because data discontinuities often exist in ocean observation grids, directly stitching together weighted values ​​can cause non-physical spatial abrupt changes at grid boundaries. Therefore, the system extracts the spatial distribution characteristics of the uncertainty fields of each single method to generate local smoothing intensity parameters, adaptively controlling the scale of spatial filtering. The adaptive smoothing radius of the target grid point is calculated using the following formula:

[0082] ;

[0083] Among them, the lower limit of background smoothing Set at 10.00 km to ensure fundamental continuity of fluid dynamics; spatial expansion adjustment coefficient. The distance is set to 20.00 km; the preset background uncertainty threshold for this sea area. The degree is 0.50. The target grid point i's... The degree is 0.40, corresponding to a weight. It is 0.28; The degree is 0.25, corresponding to the weight. It is 0.72.

[0084] The uncertainty-weighted composite term in the numerator is calculated as follows: 0.40 multiplied by 0.28 yields 0.112, and 0.25 multiplied by 0.72 yields 0.180. Adding these two together gives a composite error of 0.292 degrees. Subsequently, 0.292 is divided by the threshold of 0.50, resulting in a dimensionless ratio of 0.584. The system squares 0.584 to obtain 0.341056. Multiplying this squared value by the extended adjustment coefficient of 20.00 km yields approximately 6.82 km. Finally, by adding the basic smoothing lower limit of 10.00 km, the adaptive smoothing radius for this specific grid point is calculated with extremely high precision. It is 16.82 kilometers.

[0085] After obtaining a dedicated smoothing radius of 16.82 km, a two-dimensional Gaussian spatial filter kernel is constructed with (35.0°N, 145.0°E) as the center. The system performs spatial collaborative filtering on the initial value of the central grid point at 15.63 degrees and the weighted values ​​of all neighboring effective grid points within a radius of 16.82 km. Since this radius just covers an adjacent grid affected by the upwelling of the Kuroshio subsurface cold water, the target grid point fully absorbs the surrounding physical cold water spatial trend during the filtering process.

[0086] Through this adaptive spatial smoothing process tailored to the intensity of local errors, the harsh stitching marks of the target grid points were perfectly smoothed out, and their values ​​were smoothly fine-tuned from 15.63 degrees to 15.60 degrees. The system was started in parallel on the background server cluster, and this adaptive smoothing was performed on tens of thousands of spatial grid points in the Kuroshio Extension area one by one. All smoothed values ​​were combined and reconstructed to form a preliminary fusion analysis field with continuous spatial distribution and physically soft edge transitions.

[0087] 105. The Collaboration Module is used to collaboratively adjust the marine element variables in the preliminary fusion analysis field by taking the physical equilibrium relationship as a constraint, and generate the marine element fusion analysis field.

[0088] Specifically, based on the preset ocean dynamics and thermodynamics relationship, a physical equilibrium constraint relationship is constructed between ocean element variables; according to the physical equilibrium constraint relationship, an equilibrium adjustment target is set for at least two ocean element variables in the preliminary fusion analysis field; the at least two ocean element variables are adjusted collaboratively and iteratively to make the adjusted variable relationship approach the equilibrium adjustment target, while minimizing the overall adjustment range of the original values ​​of each variable; when the variable relationship meets the preset convergence criterion, the iteration stops and the adjusted results of all ocean element variables are output.

[0089] It should be noted that although step 104 generates spatially continuous preliminary fused values ​​through Gaussian filtering, the independent piecing together of elements such as temperature and salinity in the ocean is highly susceptible to disrupting the internal temperature-salinity (TS) balance of the seawater because these elements were weighted and fused separately in the early stages. For the grid point (35.0°N, 145.0°E), the system simultaneously retrieved the smoothed sea surface temperature (SST) value of 15.60 degrees from step 104, and the preliminary smoothed sea surface salinity (SSS) value of 34.75 practical salinity units output through the same parallel business flow.

[0090] Based on the pre-defined internationally accepted seawater thermodynamic equation (TEOS-10 standard), a physical equilibrium constraint detection mechanism among oceanographic variables was constructed. Substituting SST (15.60 degrees) and SSS (34.75 degrees) into the equation of state, the absolute density of the current surface seawater was precisely calculated. The background density of the subsurface layer at a depth of 50 meters at this location was retrieved for hydrostatic stability (by calculating the Brunt-Väisälä frequency). The detection results triggered a system alarm: the surface water, after independent mathematical smoothing, exhibited an abnormally high salinity and relatively low temperature, resulting in an abnormally high fluid density, even exceeding that of the subsurface water. This typical top-heavy negative stratification severely exceeded the physical tolerance limit of the mixed layer in this sea area during winter. Directly inputting this into the forecasting model would trigger a false and violent non-physical convective overturning.

[0091] To address this imbalance, high-confidence temperature-salinity (TS) characteristic fitting curves of the Kuroshio Extension water mass over the past thirty years in winter were extracted and set as the target for balance adjustment. A multidimensional cost function including temperature, salinity, and density penalty terms was constructed, and collaborative iterative adjustment was initiated on the two related variables, SST and SSS. The core principle of optimization is to find a standard solution in the multidimensional gradient descent space that enables stable convergence of density recovery, and this solution must minimize the overall modification to the previously fused data to preserve the true observational information.

[0092] In the initial iteration, a slight decrease in salinity and a slight increase in temperature were observed, but the calculated density remained on the critical edge. After multiple microsecond-level continuous collaborative iterations, the system found the perfect balance between the modification cost and the physical laws. The final collaborative adjustment results that meet the preset convergence criteria are detailed in Table 2 below:

[0093] Table 2

[0094]

[0095] After this adjustment, the ocean element field at this grid point not only possesses extremely high-precision observation characteristics, but also completely eliminates non-physical noise. After completing such rigorous constraint checks and iterative corrections for all grid points across the entire sea area, a fully integrated and logically consistent ocean element fusion analysis field is officially output.

[0096] 106. Dynamic module, used to calculate the fusion uncertainty field corresponding to the fusion analysis field of marine elements based on the uncertainty fields of each single method and the dynamic fusion weight.

[0097] Furthermore, for each spatial grid point, the weighted sum of squares of the individual method uncertainty fields at that grid point is calculated using the dynamic fusion weight of that grid point to obtain the fusion uncertainty value of that grid point; the fusion uncertainty values ​​of all spatial grid points are integrated to form the fusion uncertainty field.

[0098] It should be noted that in numerical weather prediction and marine environmental assessment, the scientific value of any analytical field is greatly reduced without a precise error confidence interval. This step, through algebraic derivation, strictly follows the error covariance propagation law in optimal estimation theory, utilizing the previously generated specific weights and single-method error field to calculate the final synthetic uncertainty field.

[0099] At the target spatial grid point (35.0 degrees North latitude, 145.0 degrees East longitude), the system comprehensively retrieves the underlying calculation evidence: First, the single-method uncertainty value estimated by the 3D Variational Assimilation Method (3DVAR) itself is 0.40 degrees, and its final dynamic fusion weight assigned to this point is 28%; Second, the single-method uncertainty value evaluated by the Ensemble Kalman Filter Assimilation Method (EnKF) itself is 0.25 degrees, and its absolute dominant dynamic fusion weight assigned is 72%.

[0100] The uncertainty values ​​(standard deviations) output by the two methods are converted into error variances that can be linearly superimposed. The 0.40-degree variance of 3DVAR is squared, yielding an error variance of 0.1600; the 0.25-degree variance of EnKF is squared, yielding an error variance of 0.0625. Subsequently, the system uses the corresponding dynamic fusion weights to perform a weighted attenuation calculation on this set of variances. The 3DVAR variance of 0.1600 is multiplied by the square of its weight (i.e., 0.28). 2 =0.0784), yielding a residual variance contribution of 0.012544 for this method in the final system. Multiplying the EnKF variance of 0.0625 by the square of its weights (0.72) 2 =0.5184), resulting in a variance contribution value of 0.0324 for this method. Next, the system combines and sums these two weighted independent contribution values ​​to obtain a systematic fusion variance sum of 0.044944.

[0101] To restore the dimension of the error from the square of temperature to the absolute physical unit of degree, which is completely consistent with the sea surface temperature itself, the system performs a rigorous square root operation on the fusion variance of 0.044944. Mathematically, the square root of 0.044944 is approximately 0.212 degrees. Following a globally consistent decimal place standard, the system retains two decimal places, ultimately determining the fusion uncertainty value for this grid point to be 0.21 degrees.

[0102] To visually demonstrate the error fusion derivation mechanism, the core data for the entire process is summarized in Table 3 below:

[0103] Table 3

[0104]

[0105] The derivation results show that although the initial error of 3DVAR is as high as 0.40 degrees, the system is heavily biased towards the low-error EnKF (0.25 degrees, accounting for 72%) during the fusion weight allocation. This effectively reduces the final fusion uncertainty (0.21 degrees), even surpassing the uncertainty assessment of any single method. This perfectly demonstrates the statistical advantage of leveraging the strengths of multiple methods in assimilation. The system rapidly traverses tens of thousands of spatial analysis grid points in the Kuroshio Extension region in the background, performing the rigorous weighted sum of squares and square root operations on each point. The results from all grid points are then combined into a spatial matrix, successfully generating a final fusion uncertainty field covering the entire sea area with continuous gradients.

[0106] Figure 3 This is a schematic diagram of a multi-method oceanographic element fusion and assimilation device based on multi-source observation data provided in an embodiment of the present invention. The multi-method oceanographic element fusion and assimilation device 200 based on multi-source observation data can vary considerably due to different configurations or performance. It may include one or more central processing units (CPUs) 210 (e.g., one or more processors) and a memory 220, and one or more storage media 230 (e.g., one or more mass storage devices) for storing application programs 233 or data 232. The memory 220 and storage media 230 can be temporary or persistent storage. The program stored in the storage media 230 may include one or more modules (not shown in the diagram), each module including a series of instruction operations on the multi-method oceanographic element fusion and assimilation device 200 based on multi-source observation data. Furthermore, the processor 210 may be configured to communicate with the storage media 230 and execute the series of instruction operations in the storage media 230 on the multi-method oceanographic element fusion and assimilation device 200 based on multi-source observation data.

[0107] The multi-method ocean element fusion and assimilation device 200 based on multi-source observation data may also include one or more power supplies 240, one or more wired or wireless network interfaces 250, one or more input / output interfaces 260, and / or one or more operating systems 231, such as Windows Server, MacOSX, Unix, Linux, FreeBSD, etc. Those skilled in the art will understand that... Figure 3 The structure of the multi-method ocean feature fusion and assimilation device based on multi-source observation data shown does not constitute a limitation on the multi-method ocean feature fusion and assimilation device based on multi-source observation data. It may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0108] The present invention also provides a multi-method marine element fusion and assimilation device based on multi-source observation data. The multi-method marine element fusion and assimilation device based on multi-source observation data includes a memory and a processor. The memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, the processor performs the steps of the multi-method marine element fusion and assimilation system based on multi-source observation data in the above embodiments.

[0109] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when the instructions are executed on a computer, cause the computer to perform the steps of the multi-method marine element fusion and assimilation system based on multi-source observation data.

[0110] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0111] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0112] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A multi-method ocean element fusion and assimilation system based on multi-source observation data, characterized in that, include: The processing module is used to perform bias correction and quality control on multi-source raw observation data to form a standardized observation dataset; A parallel module is used to assimilate the standardized observation dataset in parallel and independently using at least two assimilation methods, generating a single-method analysis field and a single-method uncertainty field corresponding to each method; The weighting module is used to generate a consensus field value for each spatial grid point using the values ​​of all the single-method analysis fields at that grid point, and to calculate the dynamic fusion weight of that grid point based on the values ​​of each single-method analysis field, the degree of deviation between the corresponding single-method uncertainty field value and the consensus field value, including: For each spatial grid point, the multi-method consensus field value of that grid point is calculated based on the values ​​of all single-method analysis fields aggregated to that grid point. Based on the values ​​of the uncertainty field of each single method at this grid point, the initial weights based on uncertainty are calculated; Calculate the relative deviation between the values ​​of each single-method analysis field and the values ​​of the multi-method consensus field, and calculate the consistency-based weighting factor based on the relative deviation. The initial weights based on uncertainty and the weight factors based on consistency are processed to obtain dynamic fusion weights; The configuration module is used to perform weighted fusion of the corresponding single-method analysis fields at all spatial grid points using the dynamic fusion weights to generate a preliminary fusion analysis field. The collaborative module is used to collaboratively adjust the marine element variables in the preliminary fusion analysis field by taking the physical equilibrium relationship as a constraint, and generate the marine element fusion analysis field.

2. The multi-method ocean element fusion and assimilation system based on multi-source observation data according to claim 1, characterized in that, include: Assimilate the standardized observation dataset to generate a first analysis field, and obtain the first uncertainty field corresponding to the first analysis field; Assimilate the standardized observation dataset to generate a second analysis field, and obtain the second uncertainty field corresponding to the second analysis field; The first uncertainty field is obtained by analyzing the diagonal elements of the error covariance matrix using the three-dimensional variational assimilation method, and the second uncertainty field is obtained by analyzing the statistical characteristics of the ensemble samples using the ensemble Kalman filter assimilation method.

3. The multi-method ocean element fusion and assimilation system based on multi-source observation data according to claim 1, characterized in that, include: For each spatial grid point, obtain the numerical values ​​of each single-method analysis field at that grid point and the dynamic fusion weights of that grid point; The dynamic fusion weights are used to process the values ​​of each single-method analysis field to obtain the preliminary fusion analysis value of the grid point; Integrate the preliminary fusion analysis values ​​of all spatial grid points to form a spatially continuous preliminary fusion analysis field.

4. The multi-method ocean element fusion and assimilation system based on multi-source observation data according to claim 3, characterized in that, include: Based on the spatial distribution of the uncertainty field of each single method, the local smoothing intensity parameter of each spatial grid point is generated; Based on the local smoothing intensity parameter, the preliminary fusion analysis value of each spatial grid point is processed to obtain the smoothed preliminary fusion analysis value. All smoothed preliminary fusion analysis values ​​are combined to form a spatially continuous preliminary fusion analysis field.

5. The multi-method ocean element fusion and assimilation system based on multi-source observation data according to claim 3, characterized in that, Calculate the adaptive smoothing radius of the target grid points using the following formula. : ； in, Set the lower limit for background smoothing; This is the spatial expansion adjustment coefficient; The preset threshold for uncertainty in the marine environment; This is the quantification value of the uncertainty of the 3D VAR analysis field at the target grid point i; The contribution percentage of the 3DVAR analysis field results at target grid point i; This is the quantification value of the uncertainty of the EnKF analysis field at the target grid point i. This represents the percentage of the EnKF analysis field results contributed to the target grid point i.

6. The multi-method ocean element fusion and assimilation system based on multi-source observation data according to claim 4, characterized in that, include: Based on the pre-defined relationship between ocean dynamics and thermodynamics, physical equilibrium constraints between ocean element variables are constructed. Based on the physical equilibrium constraint relationship, equilibrium adjustment targets are set for at least two ocean element variables in the preliminary fusion analysis field; At least two ocean element variables are adjusted collaboratively and iteratively to make the adjusted variable relationship approach the equilibrium adjustment target, while minimizing the overall adjustment magnitude of the original values ​​of each variable; When the variable relationships meet the preset convergence criteria, the iteration stops and the adjusted results for all ocean element variables are output.

7. The multi-method oceanographic element fusion and assimilation system based on multi-source observation data according to any one of claims 1-6, characterized in that, It also includes dynamic modules: Based on the uncertainty fields of each individual method and the dynamic fusion weights, the fusion uncertainty field corresponding to the fusion analysis field of the marine elements is calculated.

8. The multi-method ocean element fusion and assimilation system based on multi-source observation data according to claim 7, characterized in that, include: For each spatial grid point, the dynamic fusion weight of that grid point is used to process the numerical values ​​of the uncertainty fields of each single method at that grid point, so as to obtain the fusion uncertainty value of that grid point. The fusion uncertainty values ​​of all spatial grid points are integrated to form a fusion uncertainty field.

Images