Satellite-borne laser radar ocean water color product quality evaluation and calibration method, system and program

Abstract

The invention relates to the technical field of ocean remote sensing and satellite monitoring, in particular to a satellite-borne laser radar ocean water color product quality evaluation and calibration method, system and program, and particularly aims at improving the inversion precision of laser radar ocean water body parameters. The method comprises the steps of multi-source data acquisition and standardization, dynamic space-time matching window optimization, abnormal data filtering, seasonal conversion factor dynamic calculation, multi-source cross validation, consistency analysis, regional interpolation modeling, full-link uncertainty quantification, day and night difference correction and dynamic feedback optimization. By constructing a self-adaptive calibration mechanism and an error control process, the inversion precision and stability of a laser radar water color product are remarkably improved, and high-credibility data support is provided for marine ecological monitoring and carbon flux evaluation.

Description

Technical Field

[0001] The present invention relates to the technical field of ocean remote sensing and satellite monitoring, and particularly relates to a method, system and program for evaluating and calibrating the quality of spaceborne lidar ocean color products, especially for improving the inversion accuracy of lidar ocean water parameters. Background Art

[0002] Ocean color remote sensing is an important means for ocean ecological environment monitoring and carbon cycle research. For a long time, it has mainly relied on passive optical remote sensing technology based on the principle of sunlight reflection. However, passive remote sensing technology is severely restricted by solar illumination conditions, weather conditions and day-night cycles in practical applications. In particular, it cannot provide continuous and effective observation data at night or in cloud-covered areas, and it is difficult to achieve all-weather and all-day monitoring of ocean ecological parameters.

[0003] In recent years, spaceborne lidar technology, with its active remote sensing mode of actively emitting laser pulses and receiving target scattering signals, has provided new possibilities for breaking through the above limitations of passive remote sensing technology. Spaceborne lidar has the advantages of high spatial resolution, day-night observation ability and strong penetration ability for thin clouds, and has gradually become an important supplementary means for ocean color observation, attracting high attention in the field of ocean remote sensing at home and abroad.

[0004] However, currently, most spaceborne lidar systems are designed for atmospheric detection or land ecological monitoring, and their initial targets are not the accurate inversion of ocean color parameters, resulting in many challenges in data quality verification and calibration in actual ocean applications. Specifically, it is manifested as follows: (1) There is an obvious mismatch problem between the spatial scale and revisit period of lidar observations and the rapidly changing ocean ecological processes. Due to the small laser spot footprint of lidar, limited spatial coverage and discrete data sampling, it is insufficient to match large-scale and high-frequency ocean ecological processes, making it difficult to effectively apply traditional evaluation methods based on in-situ measured data and large-scale passive remote sensing data.

[0005] (2) Existing evaluation methods usually lack effective spatio-temporal matching optimization strategies, resulting in the evaluation of lidar data quality being highly sensitive to the selection of artificially set spatio-temporal matching windows. Different window strategies are likely to produce very different evaluation results, seriously affecting the reliability and stability of product accuracy evaluation.

[0006] (3) Due to the interference of complex factors such as atmospheric aerosols and bubble noise caused by wind speed on ocean color signals, traditional evaluation methods have not effectively identified and filtered such abnormal data, resulting in relatively large deviations in product quality evaluation results and making it difficult to effectively guide the optimization and operational application of lidar ocean color products.

[0007] (4) In existing methods, fixed values or rough partitioning are usually adopted for the conversion factors involved in the inversion of ocean color parameters, failing to effectively consider the seasonal variations and regional differences in ocean bio-optical properties, resulting in insufficient accuracy of inversion products and unable to provide reliable data support for ocean ecological monitoring and carbon flux estimation.

[0008] Therefore, there is an urgent need to develop a multi-source data fusion evaluation and dynamic calibration method for the characteristics of spaceborne lidar ocean color products, so as to effectively improve the quality and stability of lidar ocean color products, solve the above technical problems, and give full play to the potential advantages of lidar in the field of ocean ecological environment monitoring. Summary of the Invention

[0009] In view of the above problems, the present invention provides a novel method for evaluating and calibrating the quality of spaceborne lidar ocean color products. A dynamic multi-source collaborative calibration framework is proposed, which combines lidar, buoy and passive remote sensing data to realize the optimization of ocean color products with high precision, full-time and global coverage, providing key technical support for ocean ecological research and climate change monitoring.

[0010] In order to achieve the above object, the present invention adopts the following technical solutions: A method for evaluating and calibrating the quality of spaceborne lidar ocean color products, characterized by comprising the following steps: Step S1 Multi-source data acquisition and standardization: Integrate spaceborne lidar ocean color products, in-situ measured data, passive remote sensing ocean color products and auxiliary data, and perform time alignment, spatial registration and unit standardization on data from different sources; Step S2 Dynamic spatio-temporal matching window optimization: Based on the spatio-temporal heterogeneity of the buoy and lidar, construct multi-scale matching windows, and automatically screen the optimal matching strategy through comprehensive scoring; Step S3 Outlier data filtering and preprocessing: Eliminate outliers, and filter out atmospheric and bubble noises based on the aerosol optical depth (AOD) and wind speed thresholds; Step S4 Dynamic calculation of seasonal conversion factors: Based on the particulate backscattering coefficient bbp of the measured data and the 180° scattering coefficient βp(π) retrieved by lidar, calculate the regional-specific conversion factor χp(π) seasonally and generate a global seasonal lookup table; Step S5 Multi-source cross-validation and consistency analysis: Cross-compare the lidar ocean color products, passive ocean color products and measured data, generate a consistency report using the weighted average method, identify systematic biases and mark low-confidence regions; Step S6 Construction of spatio-temporal interpolation models for regional conversion factors Based on the differences in ocean dynamic zonation and bio-optical properties, the global sea area is divided into several sea areas, and the seasonal spatio-temporal prediction of the conversion factor χp(π) is carried out using the interpolation algorithm; Step S7 Full-link uncertainty quantification: Quantify the errors in each link of data acquisition, spatio-temporal matching, band conversion and inversion algorithm through Monte Carlo simulation, and generate the confidence interval map of the lidar water color product; Step S8 Day-night difference correction: Statistically correct the day-night differences of the conversion factor χp(π) for daytime and nighttime observation data respectively; Step S9 Dynamic feedback and product optimization: Real-time feedback the calibrated conversion factor χp(π) and error parameters to the lidar inversion process, update the matching logic and parameter priorities, and output the final water color product quality report containing indicators such as R², RMSE and confidence interval.

[0011] Preferably, in step S1, for the time asynchrony problem of different observation means of spaceborne lidar water color products, passive remote sensing water color products and in-situ measured data, the cubic spline interpolation method is used to align the buoy profile data with the satellite overpass time; In the spatial registration link, through the WGS84 coordinate system conversion and bilinear interpolation technology, the kilometer-level grid data of MODIS is coupled with the sub-meter-level lidar light spots in position, ensuring that the spatial offset between the buoy in-situ measurement points and the satellite observation trajectory is controlled within a hundred-meter error; Unit standardization focuses on solving the physical quantity differences between sensors. Based on the radiative transfer model, the buoy backscattering coefficient is unified to the lidar wavelength, and the auxiliary parameters including wind speed, aerosol optical depth AOD and sea surface temperature are dimensionless processed to form a multi-source data set that can be directly compared, providing data input with consistent benchmark for subsequent dynamic matching and calibration.

[0012] Preferably, the comprehensive scoring calculation formula of the multi-scale matching window in step S2 is: Comprehensive score = 0.4×R 2  + 0.3×(1 - RMSE) + 0.3×(1 - MAPE) where, R 2 is the coefficient of determination, RMSE is the root mean square error, and MAPE is the mean absolute percentage error; select the matching window with the highest comprehensive score as the optimal strategy.

[0013] Preferably, in step S3, the observation samples with AOD>0.3 and wind speed ≥9m / s are excluded or marked to prevent inversion anomalies caused by atmospheric interference and sea surface bubble noise.

[0014] Preferably, the calculation formula of the conversion factor χp(π) in the step S4 is as follows: χp(π)=bbp / (2πβp(π)), And it is statistically analyzed by region for the four seasons of spring, summer, autumn, and winter respectively to obtain a globally seasonal lookup table with regional specificity.

[0015] Preferably, in the multi-source cross-validation and consistency analysis in the step S5, the spaceborne lidar water color product, passive remote sensing water color product, and in-situ measured data are cross-compared, and the weighted average method is used with the weight being the data confidence level to generate a consistency report, identify systematic biases, and mark low-confidence regions. The weighted average calculation is as follows: , where, bbp 融合 is the fused water body particulate matter backscattering product, W i is the weight of the water color product, passive water color product, and measured data, and bbp i is the backscattering coefficient of the water color product, passive water color product, and measured data; If the absolute deviation between the single-source data and the fused value exceeds 2 times its own uncertainty, it is determined that the data has a systematic bias and is marked as an abnormal source; meanwhile, the weighted variance of the multi-source data within the calculation unit is calculated: , When >0.2×10 -3 or the effective data volume is less than 2 groups, the region is marked as a low-confidence region, and its spatial position is marked in the output result. Finally, a consistency report including the fused value, deviation distribution, and confidence level is generated, providing a quantitative basis for the optimization of the conversion factor and the iteration of the product.

[0016] Preferably, when partitioning the global sea area in the step S6, at least 26 marine dynamic regions are divided according to factors such as ocean circulation, water mass characteristics, and bio-optical properties, and Kriging interpolation or other geostatistical methods are used to perform spatio-temporal interpolation prediction on the conversion factor χp(π).

[0017] Preferably, the specific method of the step S7 is as follows: First, define the probability distribution of the error sources in each link: The lidar signal noise follows a Poisson distribution, and its standard deviation is determined by the signal-to-noise ratio SNR, ; The measurement error of the buoy in-situ measured data is set to a uniform distribution based on the calibration certificate, ∈ buoy~ U(-0.02bbp, +0.02bbp); The spatial offset error of spatio-temporal matching is modeled as a Gaussian distribution, ∈space~ N(0, 100 2 m 2 ); The uncertainty of the band conversion coefficient γ is assumed to be a normal distribution, γ ~ N(0.78, 0.05 2 ); The deviation of the inversion algorithm parameters is simulated by Gaussian noise, ∈ algo~ N(0, 0.1 2 ); The Monte Carlo simulation performs 1000 independent iterations. Each iteration randomly extracts the above error parameters and perturbs the input data, and then performs signal - noise superposition in turn: LiDAR βp ( π ) noisy = βp ( π ) + , The bbp of the buoy data band conversion 532 = bbp 700 × (532 / 700) -γ+δγ , where δ, γ is the perturbation term, Space - time position offset correction, the matching position correction is x new = x + ∈ space , and the perturbation of the inversion algorithm parameters βp ( π ) final = βp ( π ) noisy × (1 + ∈ algo ); Finally, the bbp value with perturbation is generated through the inversion model; Statistics the probability distribution of all iteration results, and calculates the 95% confidence interval: , where μ is the mean value, σ is the standard deviation; And further quantify the contribution degree i of each link to the total uncertainty through variance decomposition, i = 1, 2, 3, 4 correspond to data acquisition, space - time matching, band conversion, and inversion algorithm respectively. Finally, output the spatial error hot - spot map and the link contribution degree report to support product quality grading and calibration strategy optimization.

[0018] Preferably, in step S9, by dynamically adjusting the data source weights and adaptively shrinking the spatio-temporal matching window, the algorithm parameter selection logic is optimized, and a timeliness weighting factor is introduced to ensure that recent observation data dominates the inversion process.

[0019] Furthermore, the present invention also provides an on-board lidar ocean color product quality evaluation and calibration system, which executes the method, including: a) A data acquisition and management module for multi-source data acquisition and standardization in; b) A spatio-temporal matching optimization module for constructing and screening the optimal matching window according to the heterogeneity between the buoy and the lidar; c) An abnormal data processing module for filtering data with AOD and wind speed thresholds exceeding the preset range; d) A seasonal conversion calculation module for calculating and storing the conversion factor χp(π); e) A partition interpolation and uncertainty evaluation module for spatio-temporal interpolation of the conversion factor and outputting a confidence evaluation; f) A day-night correction and dynamic feedback module for correcting for day-night differences and real-time feedback of the correction results to the inversion process to generate a final quality report.

[0020] Preferably, the system further includes a multi-source cross-validation unit for automatically comparing the lidar ocean color product, passive remote sensing data, and in-situ measurement data during the product generation stage, outputting the confidence interval and consistency score of each index, and issuing a calibration requirement instruction when a deviation exceeds the threshold, triggering the dynamic feedback process of the system.

[0021] Furthermore, the present invention also provides a computer-readable storage medium, on which a computer program or instruction is stored, and when the computer program or instruction is executed by a processor, the method is implemented.

[0022] Furthermore, the present invention also provides a computer program product, including a computer program or instruction, and when the computer program or instruction is executed by a processor, the method is implemented.

[0023] Due to the adoption of the above technical solutions, the technical effects of the present invention are as follows: (1) The present invention first proposes a dynamically optimized multi-scale spatio-temporal matching strategy, which can adaptively screen the optimal matching window between the lidar and in-situ measurement data, solves the problems of unstable verification conclusions and insufficient accuracy caused by the traditional manual experience setting of the window, and significantly improves the objectivity and accuracy of the on-board lidar ocean color product quality evaluation.

[0024] (2) The present invention designs a method based on multi-source data cross-validation and consistency analysis, integrating lidar water color products, passive remote sensing products, and in-situ measured data, effectively identifying and marking systematic biases and low-confidence regions through confidence-weighted averaging, improving the reliability of ocean water color parameter inversion results and the stability of data products.

[0025] (3) The present invention innovatively proposes a dynamic calculation and regional interpolation model for seasonal conversion factors, fully considering the spatio-temporal variation laws of ocean bio-optical properties, effectively reducing the inversion errors caused by fixed conversion factors, significantly improving the accuracy of inverted water color parameters, and enhancing the applicability and long-term monitoring ability of lidar water color products.

[0026] (4) By establishing a day-night difference correction module, the present invention solves for the first time the systematic difference problem between spaceborne lidar day and night observations, realizes continuous, consistent, and high-precision processing of day and night observation data, and ensures the continuity and accuracy of all-weather ocean ecological environment monitoring.

[0027] (5) The present invention constructs a full-link uncertainty quantification method, comprehensively evaluating the errors in each link of data acquisition, spatio-temporal matching, band conversion, and inversion algorithm through Monte Carlo simulation technology, and realizing for the first time the quantitative evaluation of the quality of lidar ocean water color products, providing an accurate credibility basis for the scientific application and decision support of data products.

[0028] In summary, the present invention significantly improves the accuracy and stability of spaceborne lidar ocean water color products, breaks through various bottlenecks and deficiencies existing in traditional verification and calibration methods, can effectively support the high-reliability data requirements for ocean ecological environment monitoring, carbon flux assessment, and climate change research, and has important scientific research and business application values. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 It is a flowchart of the quality evaluation and calibration method for spaceborne lidar ocean water color products.

[0030] Figure 2 Evaluation scores of different matching windows.

[0031] Figure 3 Comparison difference diagram before and after correction. DETAILED DESCRIPTION OF THE INVENTION

[0032] The following combines the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0033] As Figure 1 shown, the method for evaluating and calibrating the quality of spaceborne lidar ocean color products of the present invention specifically includes the following steps: Step 1: Multi-source data acquisition and standardization; Step 2: Optimization of dynamic spatio-temporal matching windows; Step 3: Filtering and preprocessing of abnormal data; Step 4: Dynamic calculation of seasonal conversion factors; Step 5: Multi-source cross-validation and consistency analysis; Step 6: Construction of spatio-temporal interpolation models for conversion factors in sub-regions; Step 7: Quantification of uncertainty in the entire link; Step 8: Day-night difference correction module; Step 9: Dynamic feedback and product optimization. The following will elaborate on each step in detail.

[0034] Step 1: Multi-source data acquisition and standardization; In multi-source data acquisition and standardization, spaceborne lidar ocean color products, in-situ measured data, passive remote sensing ocean color products, and auxiliary data (wind speed, aerosol optical depth AOD, sea surface temperature) are integrated; the data is time-aligned, spatially registered, and unit-standardized. Here, taking the spaceborne lidar ocean particulate backscattering coefficient bbp product as an example, in-situ measured data from buoys and the passive water color remote sensing MODIS product are used for verification.

[0035] For the time asynchrony problem of different observation means such as spaceborne lidar, in-situ buoys, and passive remote sensing, the cubic spline interpolation method is used to align the buoy profile data with the satellite overpass time. In the spatial registration process, through the WGS84 coordinate system transformation and bilinear interpolation technology, the kilometer-level grid data of MODIS is position-coupled with the sub-hundred-meter-level laser radar light spots, ensuring that the spatial offset between the in-situ measured points of the buoy and the satellite observation trajectory is controlled within a hundred-meter error. Unit standardization focuses on solving the physical quantity differences between sensors. Based on the radiative transfer model, the backscattering coefficient of the buoy is unified to the laser radar wavelength, and dimensionless processing is performed on auxiliary parameters such as wind speed and aerosol, forming a multi-source data set that can be directly compared, providing data input with consistent benchmarks for subsequent dynamic matching and calibration.

[0036] Step 2: Optimization of dynamic spatio-temporal matching windows. Based on the spatio-temporal heterogeneity between buoys and lidar, multi-scale matching windows are constructed, and the optimal window is automatically selected through statistical scoring. The calculation formula for statistical scoring is: Comprehensive score = 0.4×R 2  + 0.3×(1 - RMSE) + 0.3×(1 - MAPE) where R 2 is the coefficient of determination, RMSE is the root mean square error, and MAPE is the mean absolute percentage error; the window with the highest score is selected as the optimal matching strategy.

[0037] Here, the time window is 3 to 24 hours, and the spatial window is 9 - 50 kilometers.Figure 2 Evaluate scores for different matching windows.

[0038] Step 3: Anomaly data filtering and preprocessing: Remove outliers, and filter out atmospheric interference and bubble noise by combining the thresholds of AOD > 0.3 and wind speed ≥ 9 m / s.

[0039] Step 4: Dynamically calculate the seasonal conversion factor: Based on the particulate backscattering coefficient bbp of the measured data and the 180° scattering coefficient βp(π) obtained by lidar inversion, calculate the regional-specific conversion factor χp(π) = bbp / (2πβp(π)) seasonally, and generate a global seasonal lookup table.

[0040] In Step 5: Multi-source cross-validation and consistency analysis, cross-compare lidar water color products, passive water color products and measured data, and generate a consistency report using the weighted average method (weight is the data confidence level), identify systematic biases and mark low-confidence regions; The weighted average calculation is as follows: , where, bbp 融合 is the fused water particulate backscattering product, W i is the weight of water color products, passive water color products and measured data, and bbp i is the backscattering coefficient of water color products, passive water color products and measured data; If the absolute deviation between the single-source data and the fused value exceeds 2 times its own uncertainty (such as the lidar bbp error is ±0.5×10 -³ m -¹ when, the deviation threshold is set to 1.0×10 -³ m -¹ ), then it is determined that the data has a systematic bias and is marked as an abnormal source; at the same time, calculate the weighted variance of multi-source data within the calculation unit: , When > 0.2×10 -3 or the number of valid data is less than 2 groups, mark the area as a low-confidence area, mark its spatial position in the output result, and finally generate a consistency report including the fused value, deviation distribution and confidence level, providing a quantitative basis for conversion factor optimization and product iteration.

[0041] In Step 6: Construct a spatio-temporal interpolation model for the conversion factor by region. Based on the differences in ocean dynamic zoning and bio-optical characteristics, divide the global sea area into 26 sea areas, and establish a seasonal spatial distribution model of χp(π) by combining the Kriging interpolation algorithm to predict parameters in un-sampled areas.

[0042] Step 7: In the full-link uncertainty quantification, the errors in each link of data acquisition, spatio-temporal matching, band conversion, and inversion algorithm are quantified through Monte Carlo simulation to generate a 95% confidence interval map of the lidar water color product; First, define the probability distributions of the error sources in each link: The lidar signal noise follows a Poisson distribution, and its standard deviation is determined by the signal-to-noise ratio SNR, ; The measurement error of the buoy measured data is assumed to be uniformly distributed based on the calibration certificate, ∈ buoy~ U(-0.02bbp, +0.02bbp); The spatial offset error of spatio-temporal matching is modeled as a Gaussian distribution, ∈ space~ N(0, 100 2 m 2 ); The uncertainty of the band conversion coefficient γ is assumed to be a normal distribution, γ~ N(0.78, 0.05 2 ); The deviation of the inversion algorithm parameters is simulated with Gaussian noise, ∈ algo~ N(0, 0.1 2 ); The Monte Carlo simulation is performed with 1000 independent iterations. In each iteration, the above error parameters are randomly sampled and the input data is perturbed, and the signal noise superposition is performed in sequence: Lidar βp ( π ) noisy = βp ( π ) + , Buoy data band conversion bbp 532 = bbp 700 × (532 / 700) -γ+δγ , where δ, γ is the perturbation term, Spatio-temporal position offset correction, the matching position is corrected to x new = x + ∈ space , And the inversion algorithm parameter perturbation βp ( π ) final = βp ( π ) noisy × (1 + ∈ algo ); Finally, the perturbed bbp value is generated through the inversion model; Statistically analyze the probability distribution of all iteration results and calculate the 95% confidence interval: , where μ is the mean value, σ is the standard deviation; And further quantify the contribution degree i of each link to the total uncertainty through variance decomposition, i i = 1, 2, 3, 4 correspond to data acquisition, spatio-temporal matching, band conversion, and inversion algorithm respectively. Finally, output the spatial error hotspot map and the link contribution degree report to support the product quality grading and calibration strategy optimization.

[0043] Step eight: In the day-night difference correction module, separate the day and night observation data, statistically analyze the day-night difference of χp(π), design the time-sharing calibration coefficient and embed it into the inversion algorithm.

[0044] Step nine: In dynamic feedback and product optimization, the calibrated χp(π) and error parameters are real-time fed back to the lidar inversion process. By dynamically adjusting the data source weight (such as increasing the priority of measured data to 0.9 in the buoy dense area) and adaptively shrinking the spatio-temporal matching window (the spatial window in the high-dynamic sea area is shrunk to 5 km and the time window is compressed to ±3 hours), optimizing the algorithm parameter selection logic (preferably using the core parameters with an error contribution degree < 20%), and introducing a timeliness weighting factor ( α ( t ) = e- 0.1Δ t ) to ensure that the recent observation data dominates the inversion process. Finally, generate the global bbp product and the supporting quality report, covering the R², RMSE, 95% confidence interval, and the spatial distribution map of the low-confidence area, and support the data credibility grading for scientific research and operational applications.

[0045] Figure 3 is the comparison difference map before and after calibration. After calibration, the RMSE drops from 0.009 to 0.003, and the inversion accuracy is significantly improved.

[0046] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0047] This application is described with reference to the flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, and combinations of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processors of general-purpose computers, special-purpose computers, embedded processors, or other programmable data processing devices to generate a machine, such that the instructions executed by the processors of the computer or other programmable data processing devices produce means for implementing the functions specified in one or more of the flows Figure 1 one or more of the flows and / or blocks Figure 1 or means for implementing the functions specified in one or more of the blocks.

[0048] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the computer-readable memory produce a manufacture including instruction means that implement the functions specified in one or more of the flows Figure 1 one or more of the flows and / or blocks Figure 1 or means for implementing the functions specified in one or more of the blocks.

[0049] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are performed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more of the flows Figure 1 one or more of the flows and / or blocks Figure 1 or means for implementing the functions specified in one or more of the blocks.

[0050] In a typical configuration, a computing device includes one or more processors (CPUs), an input / output interface, a network interface, and memory.

[0051] The memory may include non-permanent memory in the form of computer-readable media, random access memory (RAM), and / or non-volatile memory such as read-only memory (ROM) or flash memory (flash RAM). The memory is an example of computer-readable media.

[0052] A computer-readable medium includes both permanent and non-permanent, removable and non-removable media that can store information by any method or technology. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices, or any other non-transitory medium that can be used to store information accessible by a computing device. As defined herein, a computer-readable medium does not include transitory computer-readable media such as modulated data signals and carrier waves.

[0053] The foregoing is a description of embodiments of the present invention. By the above description of the disclosed embodiments, those skilled in the art can implement or use the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather will be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for evaluating and calibrating the quality of spaceborne lidar ocean color products, characterized in that, It includes the following steps: Step S1 Multi-source data collection and standardization: Integrate spaceborne lidar water color products, in-situ measured data, passive remote sensing water color products, and auxiliary data. The auxiliary data includes wind speed, aerosol optical depth AOD, and sea surface temperature. Perform time alignment, spatial registration, and unit standardization on the source heterogeneous data; Step S3 Dynamic spatio-temporal matching window optimization: Based on the spatio-temporal heterogeneity of buoys and lidar, construct multi-scale matching windows, and automatically select the optimal matching strategy through comprehensive scoring; Step S5 Abnormal data filtering and preprocessing: Remove outliers, and filter out atmospheric and bubble noise based on the aerosol optical depth AOD and wind speed thresholds; Step S7 Seasonal conversion factor dynamic calculation: Based on the particulate backscattering coefficient bbp of the measured data and the 180° scattering coefficient βp(π) inverted by lidar, calculate the regional-specific conversion factor χp(π) seasonally and generate a global seasonal lookup table; Step S9 Multi-source cross-validation and consistency analysis: Cross-compare lidar water color products, passive water color products, and measured data, generate a consistency report using the weighted average method, identify systematic biases, and mark low-confidence regions; Step S11 Regional conversion factor spatio-temporal interpolation model construction: Based on the differences in ocean dynamic zoning and bio-optical characteristics, divide the global sea area into several sea areas, and use the interpolation algorithm to perform seasonal spatio-temporal prediction on the conversion factor χp(π); Step S13 Full-link uncertainty quantification: Quantify the errors in each link of data collection, spatio-temporal matching, band conversion, and inversion algorithm through Monte Carlo simulation, and generate a confidence interval map of lidar water color products; Step S15 Day-night difference correction: Separate day and night observation data, statistically analyze the day-night differences of χp(π), design time-sharing calibration coefficients, and embed them into the inversion algorithm; Step S17 Dynamic feedback and product optimization: Real-time feedback the calibrated conversion factor χp(π) and error parameters to the lidar inversion process, update the matching logic and parameter priorities, and output the final water color product quality report containing indicators such as R², RMSE, and confidence intervals.

2. The method according to claim 1, wherein In step S1, for the time asynchrony problem of different observation means of spaceborne lidar water color products, passive remote sensing water color products, and in-situ measured data, the cubic spline interpolation method is used to align the buoy profile data with the satellite overpass time; In the spatial registration link, through the WGS84 coordinate system conversion and bilinear interpolation technology, the kilometer-level grid data of MODIS is position-coupled with the sub-hundred-meter-level lidar light spots, ensuring that the spatial offset between the buoy measured points and the satellite observation trajectory is controlled within a hundred-meter error; Unit standardization focuses on solving the physical quantity differences between sensors. Based on the radiative transfer model, the buoy backscattering coefficient is unified to the lidar wavelength, and the auxiliary parameters including wind speed, aerosol optical depth AOD, and sea surface temperature are made dimensionless, forming a multi-source data set that can be directly compared, providing data input with consistent benchmarks for subsequent dynamic matching and calibration.

3. The method according to claim 1, wherein The comprehensive scoring calculation formula for the multi-scale matching window in step S2 is: Comprehensive score = 0.4×R 2 + 0.3×(1 - RMSE)+ 0.3×(1 - MAPE) where R 2 is the coefficient of determination, RMSE is the root mean square error, and MAPE is the mean absolute percentage error; the matching window with the highest comprehensive score is selected as the optimal strategy; And / or, in step S3, the observation samples with AOD > 0.3 and wind speed ≥ 9 m / s are excluded or marked to prevent inversion anomalies caused by atmospheric interference and sea surface bubble noise.

4. The method according to claim 1, wherein In step S4, the calculation formula of the conversion factor χp(π) is: χp(π) = bbp / (2πβp(π)), And zonal statistics are carried out for each of the four seasons of spring, summer, autumn, and winter to obtain a globally seasonal lookup table with regional specificity; And / or, in the construction of the spatio-temporal interpolation model of the conversion factor χp(π) in step S6, based on the differences in ocean dynamic zonation and bio-optical characteristics, the global sea area is divided into 26 sea areas, and a seasonal spatial distribution model of χp(π) is established in combination with the Kriging interpolation algorithm to predict the parameters of the unsampled area.

5. The method according to claim 1, wherein In the multi-source cross-validation and consistency analysis in step S5, the spaceborne lidar water color products, passive remote sensing water color products, and in-situ measured data are cross-compared, and the weighted average method is used with the weight being the data confidence level to generate a consistency report, identify systematic biases, and mark low-confidence areas, where the weighted average calculation is as follows: ， Among them, bbp 融合 is the backscattering product of the fused water particulate matter, W i is the weight of the spaceborne lidar water color product, in-situ measured data, and passive remote sensing water color product, bbp i is the backscattering coefficient of the water color product, passive water color product, and measured data; If the absolute deviation between the single-source data and the fusion value exceeds 2 times its own uncertainty, it is determined that the data has a systematic bias and is marked as an abnormal source; at the same time, the weighted variance of the multi-source data within the calculation unit is calculated: ， When > 0.2×10 -3 or the amount of valid data is less than 2 groups, mark this area as a low-confidence area, mark its spatial position in the output result, and finally generate a consistency report including the fusion value, deviation distribution, and confidence level, providing a quantitative basis for the optimization of conversion factors and product iteration.

6. The method according to claim 1, characterized in that, The specific method of step S7 is as follows: First, define the probability distributions of the error sources in each link: The lidar signal noise follows a Poisson distribution, and its standard deviation is determined by the signal-to-noise ratio SNR. ; The measurement error of the actual buoy data is assumed to be uniformly distributed based on the calibration certificate, ∈ buoy~ U(-0.02bbp, +0.02bbp); The spatial offset error of spatio-temporal matching is modeled as a Gaussian distribution, ∈ space~ N(0, 100 2 m 2 ); Band conversion coefficient γ The uncertainty of is assumed to be a normal distribution, γ ~ N(0.78, 0.05 2 ) The parameter deviation of the inversion algorithm is simulated by Gaussian noise, ∈ algo~ N(0, 0.1 2 ); The Monte Carlo simulation performs 1000 independent iterations. Each iteration randomly extracts the above error parameters and perturbs the input data, and sequentially performs signal noise superposition: LiDAR βp ( π ) noisy = βp ( π )+ , Buoy data band conversion bbp 532 = bbp 700 × (532 / 700) -γ+δγ , where δ, γ is the perturbation term Spatial-temporal position offset correction, the matching position is corrected to x new = x +∈ space , and inversion algorithm parameter perturbation βp ( π ) final = βp ( π ) noisy ×(1 + ∈ algo ); Finally, a perturbed bbp value is generated through the inversion model; Statistical analysis of the probability distributions of all iteration results is carried out to calculate the 95% confidence interval: , where μ is the mean value, σ is the standard deviation; And further quantify the contribution degree i of each link to the total uncertainty through variance decomposition. i = 1, 2, 3, 4 correspond to data acquisition, spatio-temporal matching, band conversion, and inversion algorithm respectively. Finally, a spatial error hotspot map and a link contribution degree report are output to support product quality grading and calibration strategy optimization.

7. The method according to claim 1, characterized in that In step S9, by dynamically adjusting the data source weights and adaptively shrinking the spatio-temporal matching window, the logic of algorithm parameter selection is optimized, and a timeliness weighting factor is introduced to ensure that recent observation data dominates the inversion process.

8. An on-orbit lidar ocean water color product quality evaluation and calibration system, characterized in that This system executes the method described in any one of claims 1-6, including: A data acquisition and management module for multi-source data acquisition and standardization; A spatio-temporal matching optimization module for constructing and screening the optimal matching window according to the heterogeneity of the buoy and lidar; An abnormal data processing module for filtering data with AOD and wind speed thresholds exceeding the preset range; A seasonal conversion calculation module for calculating and storing the conversion factor χp(π); A zonal interpolation and uncertainty evaluation module for spatio-temporal interpolation of the conversion factor and outputting confidence evaluation; A day-night correction and dynamic feedback module for correcting the day-night difference and real-time feedback of the correction result to the inversion process to generate a final quality report.

9. A computer-readable storage medium having a computer program or instructions stored thereon, characterized in that, When this computer program or instruction is executed by a processor, it implements the method described in any one of claims 1-7.

10. A computer program product comprising a computer program or instructions, characterized in that, When the computer program or instruction is executed by a processor, it implements the method according to any one of claims 1 to 7.

Images