A method, system and computer program for monitoring net primary productivity of phytoplankton in polar winter

Abstract

This invention belongs to the field of marine remote sensing monitoring technology, specifically relating to a method, system, and computer program for dynamically monitoring the net primary productivity of phytoplankton in polar winter. The method is suitable for monitoring biogeochemical processes under low-light conditions in polar waters such as the Antarctic and Arctic. The method includes: multi-source data acquisition and spatiotemporal matching; signal correction of spaceborne lidar data, including transient response elimination and polarization crosstalk correction; quality control of spaceborne lidar data; construction of a physical bio-optical parameter database; dual-branch deep learning model training; step-by-step parameter reconstruction; vertical profile light adaptation modeling; carbon-based productivity calculation; and multi-source data verification, thereby obtaining accurate net primary productivity of phytoplankton in polar winter waters.

Description

Technical Field

[0001] The present invention belongs to the technical field of ocean remote sensing monitoring, and specifically relates to a method, system and computer program for dynamic monitoring of the net primary productivity of phytoplankton in polar winter based on spaceborne lidar, which is suitable for monitoring biogeochemical processes under weak light conditions in polar waters such as the Antarctic and the Arctic. Background Art

[0002] As an important component of the global carbon cycle, the ocean's phytoplankton net primary production (NPP) plays a vital role in global climate change and carbon sink assessment. Polar waters, in particular, have a significant impact on global climate regulation and marine biogeochemical processes due to their special climatic conditions and unique ecosystems. Polar waters not only exhibit high phytoplankton productivity in the summer, but may also maintain certain ecological functions in the winter. Despite extremely weak light levels, organisms in extreme environments may adopt special survival strategies, allowing polar phytoplankton to continue to play an ecological role under polar night conditions. However, due to the harsh polar winter environment and the continuous extremely low light conditions day and night, monitoring of phytoplankton NPP is extremely difficult.

[0003] Currently, global monitoring of marine phytoplankton NPP relies primarily on two methods: passive remote sensing and in situ observations from ships or buoys. Passive remote sensing methods indirectly infer phytoplankton productivity by acquiring ocean water color reflectance signals, such as chlorophyll concentration, particle backscatter coefficient, and water diffuse attenuation coefficient. A Chinese invention patent application (publication number: CN109490270A, publication date: March 19, 2019) discloses a device and method for measuring phytoplankton primary productivity based on chlorophyll fluorescence. Under light-shielded conditions, algae samples are illuminated with a simulated light source for at least 60 seconds. Leveraging the algae's dependence on light history, the device measures single-cycle and relaxation fluorescence kinetics under illumination, with the simulated light briefly off for 50 milliseconds. Fluorescence kinetic parameters under light-adapted conditions are obtained. A laser diode array serves as the induced excitation light source, and a photomultiplier tube serves as the fluorescence detector. An underwater photosynthetically active radiation measurement unit is designed using a crown-shaped optical collector and a multi-band photodetector array to measure the natural environmental spectrum corresponding to the algal absorption characteristics. This method facilitates data acquisition and provides high spatial coverage under bright daylight conditions. However, due to the extremely low solar altitude angle in the polar winter, the sea surface reflection signal is weak, and the effective data coverage of passive remote sensing products is usually less than 20%, which is not even enough to provide a continuous spatiotemporal data series, and thus cannot truly reflect the ecological dynamic changes in the polar waters.

[0004] On the other hand, while shipborne observations and in situ buoy observations can provide detailed local data in polar waters, these methods have very limited spatial coverage across the vast polar region due to the extremely harsh climate and ice distribution. Furthermore, data collection is constrained by factors such as weather, resulting in low temporal resolution. Furthermore, optical and biochemical parameters collected using in situ equipment often only provide local representative information and are unable to reflect the true conditions of the entire polar waters. Therefore, achieving large-scale, continuous, and high-temporal-resolution phytoplankton NPP monitoring during the polar winter has become a technical challenge urgently needed to be addressed in the fields of marine science and atmospheric environmental research.

[0005] Against this backdrop, the scientific community has recently begun exploring the feasibility of using spaceborne lidar (LiDAR) technology for ocean observations. Spaceborne LiDAR utilizes active light detection, independent of solar radiation, and can effectively acquire target information even in polar night or low-light conditions. Spaceborne LiDAR systems, such as CALIOP (Cloud-Aerosol Lidar with Orthogonal Polarization), operate continuously day and night, detecting scattered signals from the sea surface and its subsurface using laser emission and echo reception. This technology offers new possibilities for filling the data gap in polar winter and has attracted widespread attention from researchers both domestically and internationally. Summary of the Invention

[0006] To address these issues, the present invention provides a novel method for dynamically monitoring polar winter phytoplankton net primary productivity (NPP) using spaceborne lidar. This method, which incorporates a two-branch, two-step deep learning model, accurately inverts multiple ocean water optical and chlorophyll parameters from spaceborne lidar signals. This model then reconstructs phytoplankton NPP in high-latitude waters, enabling dynamic monitoring of phytoplankton NPP in polar winter. This provides more reliable data support for marine scientific research and polar carbon sink assessment.

[0007] In order to achieve the above-mentioned purpose, the present invention adopts the following technical solutions:

[0008] A method for monitoring the net primary productivity of phytoplankton in polar winter based on spaceborne lidar comprises the following steps:

[0009] S1: Multi-source remote sensing data acquisition and spatiotemporal matching: registering spaceborne lidar data with ocean color remote sensing data through set time and space windows to ensure spatial coverage and temporal consistency;

[0010] S2: Perform signal correction on the raw echo data of the spaceborne lidar, including transient response correction and polarization crosstalk correction;

[0011] S3: LiDAR data quality control is performed based on water depolarization ratio, sea ice concentration, and wind speed data to eliminate sea ice interference and specular reflection anomalies;

[0012] S4: Construct a parameter database containing multiple physical-bio-optical characteristics and normalize the characteristic variables in the database. The multiple physical-bio-optical characteristics include the return signal after correction of space-borne lidar, water depolarization ratio, water backscatter coefficient, and ocean color remote sensing products;

[0013] S5: training a two-branch deep neural network model based on the database, wherein the first branch extracts echo waveform features based on a convolutional neural network, and the second branch fuses optical parameters to perform physical feature learning, and outputs optical chlorophyll concentration after splicing;

[0014] S6: Parameter reconstruction is achieved based on a two-step modeling strategy, which first outputs the monthly mean climate state and then reconstructs the outliers through interpolation to improve the accuracy of polar inversion;

[0015] S7: Based on the reconstructed parameters, vertical profile light adaptation modeling is performed to obtain the photosynthetically active radiation distribution in the depth direction;

[0016] S8: Calculate the net primary productivity of phytoplankton based on a carbon-based model using reconstructed optical parameters and light distribution;

[0017] S9: Use buoy data and passive remote sensing data for multi-source cross-validation to evaluate and optimize the productivity inversion accuracy.

[0018] Preferably, the transient response correction in step S2 includes deconvolving the original echo signal β(z) using the detector response function F to obtain a corrected signal:

[0019] β′(z)=F - ¹β(z)

[0020] Where β′(z) is the signal after the detector transient response correction, β(z) is the lidar measurement signal, and F is the detector transient response function;

[0021] Polarization crosstalk correction is achieved by the following formula:

[0022] ,

[0023] ,

[0024] Where CT is the crosstalk coefficient, and are the signals of the parallel channel and the vertical channel after polarization crosstalk correction, and These are the signals of the parallel channel and the vertical channel after transient response correction.

[0025] As an example, the total depolarization ratio δ of the water body in step S3 is T It is calculated as follows:

[0026] ,

[0027] Among them, δ T is the total depolarization ratio of the water body, and p is the peak position.

[0028] Preferably, the water backscatter coefficient γ in step S4 is calculated according to the following formula:

[0029] ,

[0030] Where γ is the backscattering coefficient of water body, β s is the sea surface backscatter coefficient:

[0031] ,

[0032] Where θ is the nadir angle of the laser beam, σ 2 is the mean square error of the wave slope:

[0033] ,

[0034] Where W is the sea surface wind speed;

[0035] Normalized processing is:

[0036] ,

[0037] Among them, X norm is the normalized parameter variable, X is the original variable, is the mean of the variable, and SD(X) is the standard deviation of the variable.

[0038] Preferably, the photosynthetically active radiation PAR (z) in the vertical direction in step S7 is estimated according to the following model:

[0039] PAR(z) = PAR(0) × exp(-2K d z),

[0040] PAR(z) is the photosynthetically active radiation at depth z, and PAR(0) is the photosynthetically active radiation at the sea surface K. d is the diffuse attenuation coefficient obtained by inversion.

[0041] Preferably, the net primary productivity NPP of phytoplankton in step S8 is calculated according to the following formula:

[0042] NPP = C × μ × Z eu ,

[0043] Where NPP is the net primary productivity of phytoplankton, C is the carbon biomass of phytoplankton, μ is the growth rate, and Z eu is the depth of the true light layer;

[0044] The carbon biomass C is calculated according to the following formula:

[0045] C = 13000 × (b bp - 0.00035),

[0046] where b bp is the particle backscattering coefficient;

[0047] The growth rate μ is calculated according to the following formula:

[0048] ,

[0049] Where Chl is the chlorophyll concentration and PAR(z) is the photosynthetically active radiation.

[0050] Furthermore, the present invention also provides a system for monitoring the net primary productivity of phytoplankton in polar winter, which implements the method described above, including:

[0051] Multi-source remote sensing data acquisition module, used to receive satellite-borne lidar data and ocean color remote sensing data;

[0052] The spatiotemporal matching module is used to accurately align multi-source remote sensing data based on preset spatial distance windows and time windows to ensure the consistency of data in spatial coverage and time series;

[0053] The signal correction module is used to perform the following two corrections on the satellite-borne lidar data:

[0054] Transient response correction: Deconvolution of the original echo signal β(z) using the detector transient response function F to obtain the corrected signal;

[0055] Polarization crosstalk correction is performed by converting the parallel channel and perpendicular channel signals after correction;

[0056] Data quality control module, used to detect and filter abnormal data based on total water deflection ratio, sea ice concentration and wind speed information;

[0057] A physical bio-optical parameter database construction module is used to integrate the corrected lidar echo signal, depolarization ratio, seawater backscatter coefficient, and remote sensing products, and normalize the integrated parameters;

[0058] Dual-branch deep learning model module, including:

[0059] The first branch uses a convolutional neural network (CNN) to extract features from the lidar echo waveform;

[0060] The second branch integrates the physical parameters such as water depolarization ratio and backscattering coefficient output by the data quality control module and the parameter database construction module.

[0061] The dual-branch features are spliced through a fully connected regression network to output the optical chlorophyll parameters of the water body, realizing waveform-environment collaborative modeling under physical constraints;

[0062] The parameter reconstruction module is used to further reconstruct the inversion parameters using a two-step optimization strategy: first, the monthly average climatological parameters are output, and then the outliers are reconstructed based on the differences between the observed values and the climatological state, thereby improving the parameter inversion accuracy;

[0063] A vertical profile light adaptation modeling module is used to calculate the photosynthetically active radiation (PAR) in the depth direction based on the reconstructed water optical parameters;

[0064] The carbon-based productivity calculation module is used to calculate the net primary productivity (NPP) of phytoplankton based on the results of light adaptation modeling and water parameters. The multi-source data verification module is used to verify the optical parameters and NPP results output by the above modules using buoy data, ocean color remote sensing products and other measured data, thereby ensuring the accuracy and stability of the entire system operation.

[0065] Preferably, the transient response correction and polarization crosstalk correction formulas used in the signal correction module are implemented as preset modules in the system hardware or software to automatically correct the satellite-borne lidar echo data; and / or the data quality control module can automatically identify and filter abnormal data affected by sea ice, mirror reflection and foam interference based on the water body depolarization ratio δT and auxiliary sea ice concentration and wind speed data.

[0066] Furthermore, the present invention also provides a computer device, comprising:

[0067] processor;

[0068] Memory; and

[0069] A computer program stored in the memory, executed under the control of the processor, implements the method.

[0070] Furthermore, the present invention also provides a computer-readable storage medium having program instructions stored thereon, which are used to implement the method when executed by a computer.

[0071] By adopting the above-mentioned technical solution, the present invention integrates the active detection technology of spaceborne lidar with multi-source data fusion, deep learning algorithms and physical model constraints to achieve full automation, precision and continuity of the monitoring of net primary productivity (NPP) of phytoplankton in polar winter, and achieves the following significant technical effects:

[0072] 1. Greatly improve data coverage and spatiotemporal continuity: Traditional passive remote sensing methods have extremely low effective data coverage in the polar winter due to low light levels, polar day / polar night, and other factors, and can only cover very small areas or perform intermittent sampling. However, this invention utilizes the active laser detection capability of spaceborne lidar, which is not limited by solar radiation and can obtain ocean echo data in real time even under polar night conditions. By setting a strict spatiotemporal matching mechanism, it achieves precise temporal and spatial alignment of multi-source data, thereby greatly improving the coverage and spatiotemporal continuity of polar data collection, and effectively filling the blind spots of polar winter observations.

[0073] 2. Precision Correction and Reliable Signal Recovery: To address common issues such as detector transient response and polarization crosstalk in lidar signals, this paper proposes a correction scheme based on deconvolution and crosstalk correction formulas, effectively restoring the true physical properties of the original echo signal. By removing interference from non-ideal detector response and polarization signals, the accuracy and stability of the data are improved, providing a reliable basis for subsequent parameter inversion and ultimately enhancing the accuracy of ocean optical parameter and NPP estimation.

[0074] 3. Enhanced Data Quality and Intelligent Parameter Inversion: This invention addresses various interference factors in polar environments, such as sea ice, specular reflection, and foam. A multidimensional data quality control module, based on total water depolarization ratio, sea ice concentration, and wind speed data, is designed to effectively identify and eliminate abnormal data, ensuring high-quality data input. Furthermore, a physical-bio-optical parameter database construction module is used to normalize the fused multi-source data, improving its distribution within the neural network model and providing solid data support for subsequent deep learning parameter inversion.

[0075] 4. A breakthrough deep learning model enables multi-parameter collaborative inversion: Using a dual-branch deep learning model structure, the present invention fully utilizes convolutional neural networks to extract the spatiotemporal characteristics of the lidar echo waveform, while integrating physical parameters such as the water body depolarization ratio and backscattering coefficient to establish a waveform-environment collaborative inversion model under physical constraints. Through feature splicing and fully connected regression, this model effectively improves the accuracy of inverted chlorophyll concentration and other water body optical parameters, laying the foundation for the accurate calculation of phytoplankton NPP. In addition, through a step-by-step parameter reconstruction strategy, the climate state output is first used as the initial estimate, and then the outliers are corrected, further overcoming the seasonal bias problem caused by the sparsity of polar data.

[0076] 5. Build a scientifically rigorous and universal NPP calculation model: Based on the inversion of water optical parameters, this paper constructs an NPP estimation scheme centered on a carbon-based model. By precisely calculating phytoplankton carbon biomass, chlorophyll growth rate, and euphotic depth, this scheme ensures accurate assessment of phytoplankton net primary productivity even under polar low-light conditions. This scheme leverages the complementary advantages of LiDAR data and auxiliary water color remote sensing data, overcoming the underestimation of phytoplankton production in polar environments by traditional models, providing more comprehensive and accurate data support for global carbon cycle research and climate change prediction.

[0077] 6. Multi-source verification ensures system stability and reliability: This system incorporates a multi-source data verification module, utilizing buoy observations, passive water color remote sensing products, and other measured data to cross-compare and verify the optical parameters and NPP results derived from the lidar, thereby comprehensively evaluating and optimizing the overall performance of the monitoring system. This multi-source verification approach not only enhances the stability and reliability of the system but also provides sufficient experimental and data support for practical applications, making the system suitable for long-term continuous monitoring in complex polar environments.

[0078] 7. Promoting Advances in Polar Ocean Remote Sensing and Carbon Cycle Science: This invention not only overcomes the limitations of passive remote sensing data acquisition in the low-light conditions of polar winter, but also achieves multi-parameter collaborative inversion in polar environments by deeply integrating lidar technology with advanced data processing and deep learning methods. This technological advancement significantly improves the accuracy and temporal resolution of polar ocean phytoplankton NPP monitoring, has significant theoretical and practical significance for strengthening global carbon sink research and accurately assessing the impacts of climate change. It also provides new ideas and platforms for the subsequent development and application of related technologies.

[0079] In summary, the present invention, through the collaborative work of multiple modules and innovative data processing, deep learning models, and physical constraint mechanisms, has formed a complete, rationally structured, and stable polar winter phytoplankton net primary productivity monitoring system. This system not only achieves a technical breakthrough in active sensing and multi-source data fusion, but also significantly improves the data quality and spatial coverage of polar ecological monitoring in practical applications. It provides reliable and advanced technical support for global carbon cycle modeling, climate change prediction, and polar ecological environmental protection, and has broad application prospects and far-reaching scientific significance. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] Figure 1 This is a flow chart of the method for monitoring net primary productivity of phytoplankton in polar winter based on spaceborne lidar.

[0081] Figure 2 It is a two-branch two-step neural network structure.

[0082] Figure 3 It is a comparison between the inversion parameters of spaceborne lidar and passive water color remote sensing products.

[0083] Figure 4 It is a comparison between the inversion parameters of the spaceborne lidar and the measured results.

[0084] Figure 5 It is a comparison of the net primary productivity of spaceborne lidar and the net primary productivity of passive water color remote sensing.

[0085] Figure 6 It is a comparison of the net primary productivity of spaceborne lidar and buoy lidar.

[0086] Figure 7 It is a comparison of the winter coverage of spaceborne lidar and the coverage of passive water color remote sensing. DETAILED DESCRIPTION

[0087] The following is a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts are within the scope of protection of the present invention.

[0088] like Figure 1 As shown, the present invention's method for monitoring polar winter phytoplankton net primary productivity using spaceborne lidar includes the following steps: Step 1: Multi-source remote sensing data acquisition and spatiotemporal matching; Step 2: Spaceborne lidar data signal correction; Step 3: Spaceborne lidar data quality control; Step 4: Construction of a physical bio-optical parameter database; Step 5: Dual-branch deep learning model training; Step 6: Step-by-step parameter reconstruction; Step 7: Vertical profile light adaptation modeling; Step 8: Carbon-based productivity calculation; Step 9: Multi-source data verification. Each step will be described in detail below.

[0089] Step 1: Multi-source remote sensing data collection and spatiotemporal matching

[0090] In multi-source remote sensing data collection and spatiotemporal matching, spaceborne lidar data and passive ocean color remote sensing products are acquired and matched within specific spatial and temporal windows to ensure strict consistency in spatial coverage and temporal series. This example uses CALIOP spaceborne lidar data and MODIS chlorophyll (Chl), particle backscatter coefficient (bbp), and diffuse attenuation coefficient remote sensing products. The spatial range window is 25 km and the temporal window is 12 hours; data from the same day within a 25 km radius are considered valid matching points. This example uses the Antarctic waters as an example to detail the method for monitoring polar winter phytoplankton net primary productivity using spaceborne lidar.

[0091] Step 2: Satellite-borne LiDAR data signal correction

[0092] Correct the satellite-borne lidar data signal to remove the impact of the detector's non-ideal characteristics on the data, including the detector's transient response correction:

[0093] β′(z)=F - ¹β(z),

[0094] Where β′(z) is the signal after the detector transient response correction, β(z) is the lidar measurement signal, and F is the detector transient response function.

[0095] And polarization crosstalk correction of polarization beamsplitter:

[0096] ,

[0097] ,

[0098] in, and are the signals of the parallel channel and the vertical channel after polarization crosstalk correction, and These are the signals of the parallel channel and the vertical channel after transient response correction.

[0099] Step 3: Spaceborne LiDAR data quality control is as follows:

[0100] The total deflection ratio of the water body is used to remove the influence of sea ice, the ice-water mixing zone is removed based on the sea ice concentration data, and the wind speed is limited to avoid mirror reflection and foam interference. Among them, the total deflection ratio of the water body is:

[0101] ,

[0102] Among them, δ T is the total depolarization ratio of the water body, and p is the peak position.

[0103] Step 4: Constructing a physical-bio-optical parameter database

[0104] Integrate the calibrated echo signal of the spaceborne lidar, the total depolarization ratio of water bodies, the backscatter coefficient of water bodies, and the ocean color remote sensing products, and normalize all parameters to improve the stability of model training. Among them, the backscatter coefficient of water bodies is:

[0105] ,

[0106] in, is the water backscattering coefficient, is the sea surface backscatter coefficient:

[0107] ,

[0108] in, is the laser beam nadir angle, is the mean square error of the wave slope:

[0109] ,

[0110] Where W is the sea surface wind speed.

[0111] Normalized processing is:

[0112] ,

[0113] Among them, X norm is the normalized parameter variable, X is the original variable, is the mean of the variable, and SD(X) is the standard deviation of the variable.

[0114] Step 5: Dual-branch deep learning model training

[0115] The dual-branch deep learning model is trained to construct a dual-branch network combining waveforms and physical parameters through deep learning (such as Figure 2 As shown in Figure 2, CNN is the convolutional layer and FC is the fully connected layer).

[0116] 1. Branch 1 - LiDAR echo waveform feature extraction (CNN branch):

[0117] Input layer: Receives one-dimensional (or two-dimensional, depending on the data preprocessing results) lidar echo waveform data.

[0118] Convolutional layers (CNN): Design at least 2-3 convolutional layers, using different filter sizes to extract local features. Each convolution layer can be followed by batch normalization and a nonlinear activation function (such as ReLU) to accelerate convergence.

[0119] Pooling layer: Maximum pooling or average pooling is used after some convolutional layers to reduce the size of the feature map and make the model more robust to noise.

[0120] Global pooling / flattening layer: Finally, the feature map output by the convolutional layer is flattened or global average pooling is used to generate a fixed-size feature vector.

[0121] 2. Branch 2 - Physical parameter fusion branch:

[0122] Input layer: Receives a set of physical parameter vectors, mainly including the total depolarization ratio of water body δ T and backscatter coefficient γ (other related parameters can be expanded if necessary).

[0123] Fully connected layer (FC): Design 1-2 fully connected layers (Dense layers). In each layer, use nonlinear activation (such as ReLU) to transform the input vector, extract physical features, and output a feature vector of fixed dimension.

[0124] 3. Feature splicing and regression output

[0125] Feature fusion:

[0126] The waveform feature vector obtained by branch 1 and the physical parameter feature vector obtained by branch 2 are concatenated to form a joint feature representation that combines waveform information and environmental physical parameters.

[0127] Fully connected regression layer:

[0128] The concatenated joint features are regressed through one or more fully connected layers until the target parameter, the water body optical chlorophyll parameter, is output. A linear activation function is used in the output layer to directly predict the continuous value of the target.

[0129] 4. Loss Function and Training Strategy

[0130] 1) Loss function:

[0131] The mean squared error (MSE) loss function is used to measure the difference between the predicted and true chlorophyll parameters. End-to-end training can be performed using either the Adam or SGD optimizer, combined with an appropriate learning rate (e.g., 1e-3 or lower, depending on the data characteristics).

[0132] 2) Training process:

[0133] Divide the dataset into training, validation, and test sets to ensure model generalization performance. Implement early stopping or learning rate decay during training to prevent overfitting and improve convergence. Data augmentation techniques (such as adding slight noise to the echo data) can also be used to enhance model robustness.

[0134] 5. Model output:

[0135] After training, the model can output accurate water optical chlorophyll parameters based on the input lidar waveform and corresponding physical parameters, providing accuracy guarantee for subsequent NPP inversion and parameter reconstruction.

[0136] Step 6: Step-by-step parameter reconstruction into two-step optimization inversion:

[0137] To address the problems of polar data sparsity and seasonal bias, a two-step optimization inversion strategy is adopted, dividing the inversion process into the "climatological state estimation stage" and the "outlier correction stage".

[0138] 1. First step: Estimation of climate state parameters

[0139] Objective: Using historical data and training samples, train the model to output monthly average climatological parameters as a baseline. Climatological parameters typically refer to the average state over a longer time scale (e.g., monthly), encompassing basic optical properties of the ocean environment, such as monthly average chlorophyll concentration, particle backscatter coefficient (bbp), and water diffuse attenuation coefficient (Kd).

[0140] Implementation:

[0141] 1) Data segmentation: Use a longer time period (e.g., multi-year data) to calculate the average value of each parameter each month and establish the target output data set.

[0142] 2) Model structure: Based on the dual-branch model trained in step 5, its output layer is expanded to estimate the monthly average climate state parameters.

[0143] 3) Training Process: The input data consists of preprocessed CALIOP lidar data and corresponding physical parameters, and the target output is the historical average (climatological state) for the corresponding month. The mean squared error loss function is also used to fit the model output to the monthly average climatological state parameters. After training, the model can provide a baseline estimate for a given month that is highly stable and statistically representative.

[0144] 2. Step 2: Outlier Reconstruction

[0145] Objective: Based on the baseline climate state parameters, by modeling the deviation between the current observations and the climate state, correcting abnormal deviations caused by data sparsity and extreme environmental changes, and achieving more refined parameter inversion.

[0146] Implementation:

[0147] 1) Difference calculation:

[0148] For each actual observation point, calculate the difference (residual / anomaly) between the observed value and the monthly average climate state parameter output in the first step, recorded as Δ, that is: Δ=y obs- y climate ; where y obs is the current observation value, and y climate is the climatological parameter of the same month.

[0149] 2) Reconstruction model design:

[0150] Construct an outlier reconstruction network, which can use relatively simple fully connected or convolutional layers. Its inputs are the original input data (or intermediate features from the first two-branch model) and the calculated difference Δ, and its output is the corrected anomaly component. This network design can leverage residual learning, allowing it to focus on fitting the nonlinear relationship between the observed values and the baseline climate state. The MSE loss function is also used during training to ensure that the output outliers closely match the true differences (provided by high-quality observations).

[0151] 3) Integrated output:

[0152] The final inversion result is: y final =y climate+ Δp redy ; where Δ pred The network output is reconstructed to correct the outliers. This not only maintains the stability of the monthly average climate state, but also effectively compensates for seasonal and local anomalies, thus effectively solving the problems of data sparsity and seasonal bias.

[0153] 3. Joint training strategy

[0154] Cascade training: To improve overall inversion accuracy, a cascade training strategy can be used. This involves first training the climate state estimation model separately, freezing some weights after convergence, and then training the outlier reconstruction model. This approach stabilizes the baseline output while enabling detailed optimization of the outliers.

[0155] Joint Optimization: Another approach is to design an end-to-end joint network structure, where the output of the first stage is directly connected in series with the anomaly correction component, with a joint loss function guiding the parameter updates of both stages. The joint loss function can be composed of a climatological loss and an outlier reconstruction loss, with the weights of each component adjusted based on experimental results.

[0156] 4. Model Validation and Parameter Monitoring

[0157] Cross-validation and multi-source data (such as buoy data and passive water color remote sensing data) are used to ensure the prediction accuracy of the climate state and anomaly correction models, as well as the overall prediction accuracy. Predictions at each stage are compared with actual observations, and model performance is quantitatively evaluated using metrics such as correlation coefficient (R), mean absolute percentage deviation (MAPD), and root mean square deviation (RMSD). Training strategies are dynamically adjusted during model updates.

[0158] 5. Final output parameters

[0159] After step six, the model is able to simultaneously invert multiple key water optical parameters, including chlorophyll concentration (Chl), backscatter coefficient of water particles (bbp), and diffuse attenuation coefficient (Kd). These highly accurate inversions will serve as the basis for subsequent NPP calculations and will be further optimized and refined through multi-source data validation to ensure the reliability and accuracy of the overall monitoring system.

[0160] Step 7: Vertical Profile Light Adaptation Modeling

[0161] The vertical profile light adaptation model in step 7 is to calculate the vertical photosynthetically active radiation profile based on the inverted water optical parameters:

[0162] PAR(z) = PAR(0) × exp(-2K d z),

[0163] Where PAR(z) is the photosynthetically active radiation at depth z, and PAR(0) is the photosynthetically active radiation at the sea surface. d is the diffuse attenuation coefficient obtained by inversion.

[0164] Step 8: Carbon-based productivity calculation

[0165] Calculation of phytoplankton net primary productivity based on carbon-based productivity:

[0166] NPP = C × μ × Z eu ,

[0167] Where NPP is the net primary productivity of phytoplankton, C is the carbon biomass of phytoplankton, μ is the growth rate, and Z eu is the depth of the true light layer;

[0168] The carbon biomass C is calculated according to the following formula:

[0169] C = 13000 × (b bp - 0.00035),

[0170] where b bp is the particle backscattering coefficient;

[0171] The growth rate μ is calculated according to the following formula:

[0172] ,

[0173] Where Chl is the chlorophyll concentration and PAR(z) is the photosynthetically active radiation.

[0174] Step 9: Multi-source data verification

[0175] Multi-source data validation uses a variety of field-measured data to verify the optical chlorophyll parameters derived from lidar and compare them with the NPP calculated from buoy parameters to ensure model accuracy. The lidar-derived net primary productivity (NPP) in Antarctic waters during winter is calculated to be 54.5 Tg C, six times the 8.5 Tg C estimated from passive ocean color remote sensing.

[0176] Figure 3 This comparison shows the inversion parameters of spaceborne lidar and passive water color remote sensing products. The correlation coefficients (R) are all above 0.8, the mean absolute percentage error (MAPD) is within 30%, and the root mean square error (RMSD) is also very low, indicating the accuracy of the inversion of Sydney Harbour.

[0177] Figure 4 It is a comparison between the inversion parameters of the spaceborne lidar and the measured results. For marine environment monitoring, a MAPD within 40% is acceptable.

[0178] Figure 5 This figure compares Net Primary Productivity (NPP) from spaceborne LiDAR and passive ocean color remote sensing. It shows monthly NPP (NPP) at different latitudes in the Southern Hemisphere's high-latitude oceans. While summer values are largely consistent, winter LiDAR data are significantly higher than MODIS, demonstrating that LiDAR compensates for the lack of winter data.

[0179] Figure 6 It is a comparison of the net primary productivity of spaceborne lidar and buoy lidar. The MAPD is within 20%, indicating the accuracy of the results.

[0180] Figure 7Comparison of winter coverage by spaceborne lidar and passive ocean color remote sensing. This map shows the effective remote sensing observation coverage of the passive ocean color remote sensing MODIS and spaceborne lidar CALIOP during the Antarctic winter. The ratio of remote sensing coverage to ice-free sea area is shown, as well as the distribution of climatological net primary productivity in June, July, and August. Comparisons of CALIOP (blue) and MODIS (red) data are shown for the entire winter month and for June, July, and August. Effective coverage percentages are as follows: winter (CALIOP 80.7%, MODIS 12.5%), June (CALIOP 61.85%, MODIS 0.2%), July (CALIOP 87.6%, MODIS 0.3%), and August (CALIOP 92.6%, MODIS 37.1%). The solid gray line on the map represents the sea ice edge. It shows that by collaborating with the CALIOP lidar and the physically constrained deep learning model, the polar winter data coverage can be increased from 12.5% of the traditional method to 80.7%, thereby achieving all-weather continuous monitoring of the polar night period and the sea ice edge area, filling the winter NPP data gap in the global carbon cycle model.

[0181] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0182] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes 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 a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0183] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0184] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0185] The above is a description of the embodiments of the present invention. The above description of the disclosed embodiments will enable professionals in the field to implement or use the present invention. Various modifications to these embodiments will be apparent to professionals in the field. 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 should conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for monitoring the net primary productivity of phytoplankton in polar winter, characterized in that: The steps include: S1: Multi-source remote sensing data acquisition and spatiotemporal matching: registering spaceborne lidar data with ocean color remote sensing data through set time and space windows to ensure spatial coverage and temporal consistency; S2: Perform signal correction on the raw echo data of the spaceborne lidar, including transient response correction and polarization crosstalk correction; S3: LiDAR data quality control is performed based on the total depolarization ratio of the water body, sea ice concentration, and wind speed data to eliminate sea ice interference and specular reflection anomalies; S4: Construct a parameter database containing multiple physical-bio-optical characteristics and normalize the characteristic variables in the database. The multiple physical-bio-optical characteristics include the return signal after correction of space-borne lidar, water depolarization ratio, water backscatter coefficient, and ocean color remote sensing data; S5: Based on the database, a two-branch deep neural network model is trained. The first branch extracts echo waveform features based on a convolutional neural network, and the second branch fuses optical parameters to learn physical features. After splicing, the optical chlorophyll concentration and particle backscattering coefficient b are output. bp and water body attenuation coefficient K d ; S6: Parameter reconstruction is achieved based on a two-step modeling strategy. The first step is to output the monthly average climatological parameters, which include the monthly average chlorophyll concentration, particulate matter backscatter coefficient, and water body diffuse attenuation coefficient. The outliers are then reconstructed by calculating the difference between the observed values and the monthly average climatological parameters output in the first step, thereby improving the accuracy of polar retrieval. S7: Based on the reconstructed parameters, vertical profile light adaptation modeling is performed to obtain the photosynthetically active radiation distribution in the depth direction; S8: Calculate the net primary productivity of phytoplankton based on a carbon-based model using reconstructed optical parameters and photosynthetically active radiation distribution; S9: Use buoy data and passive remote sensing data for multi-source cross-validation to evaluate and optimize the inversion accuracy of the productivity.

2. The method according to claim 1, characterized in that The transient response correction in step S2 includes deconvolving the original echo signal β(z) using the detector response function F to obtain a corrected signal: β′(z)=F⁻¹β(z) Where β′(z) is the signal after the detector transient response correction, and F is the transient response function of the detector; Polarization crosstalk correction is achieved by the following formula: ， ， Where CT is the crosstalk coefficient, and are the signals of the parallel channel and the vertical channel after polarization crosstalk correction, and These are the signals of the parallel channel and the vertical channel after transient response correction.

3. The method according to claim 2, characterized in that The total depolarization ratio δ of the water body in step S3 T It is calculated as follows: ， Among them, δ T is the total depolarization ratio of the water body, and p is the peak position.

4. The method according to claim 3, characterized in that In step S4, the water body backscatter coefficient γ is calculated according to the following formula: , Where γ is the backscattering coefficient of water body, β s is the sea surface backscatter coefficient: ， Where θ is the nadir angle of the laser beam, σ 2 is the mean square error of the wave slope: ， Where W is the sea surface wind speed; Normalized processing is: ， Among them, X norm is the normalized parameter variable, X is the original variable, is the mean of the variable, and SD(X) is the standard deviation of the variable.

5. The method according to claim 1, wherein The photosynthetically active radiation PAR (z) in the vertical direction in step S7 is estimated according to the following model: PAR(z) = PAR(0) × exp(−2K d z), PAR(z) is the photosynthetically active radiation at depth z, and PAR(0) is the photosynthetically active radiation at the sea surface K. d is the diffuse attenuation coefficient obtained by inversion.

6. The method according to claim 1, characterized in that In step S8, the net primary productivity (NPP) of phytoplankton is calculated according to the following formula: NPP = C × μ × Z eu , Where NPP is the net primary productivity of phytoplankton, C is the carbon biomass of phytoplankton, μ is the growth rate, and Z eu is the depth of the true light layer; The carbon biomass C is calculated according to the following formula: C = 13000 × (b bp - 0.00035), where b bp is the particle backscattering coefficient; The growth rate μ is calculated according to the following formula: ， Where Chl is the chlorophyll concentration and PAR(z) is the photosynthetically active radiation.

7. A polar winter phytoplankton net primary productivity monitoring system, characterized in that: The system implements the method according to any one of claims 1 to 6, comprising: a multi-source remote sensing data acquisition module for receiving spaceborne lidar data and ocean color remote sensing data; A spatiotemporal matching module is used to accurately align the multi-source remote sensing data based on a preset spatial distance window and time window to ensure the consistency of the data in terms of spatial coverage and time series; The signal correction module is used to perform the following two corrections on the satellite-borne lidar data: Transient response correction: Deconvolution of the original echo signal β(z) using the detector transient response function F to obtain the corrected signal; Polarization crosstalk correction is performed by converting the parallel channel and perpendicular channel signals after correction; Data quality control module, used to detect and filter abnormal data based on total water deflection ratio, sea ice concentration and wind speed information; A physical bio-optical parameter database construction module is used to integrate the corrected lidar echo signal, total water depolarization ratio, seawater backscattering coefficient, and ocean color remote sensing data, and normalize the integrated parameters; Dual-branch deep learning model module, including: The first branch uses a convolutional neural network to extract features from the lidar echo waveform; The second branch integrates the physical parameters of the total depolarization ratio and backscattering coefficient of the water body output by the data quality control module and the parameter database construction module. The dual-branch features are spliced through a fully connected regression network to output the optical chlorophyll parameters of the water body, realizing waveform-environment collaborative modeling under physical constraints; The parameter reconstruction module is used to further reconstruct the inversion parameters using a two-step optimization strategy: first, the monthly average climatological parameters are output, including the monthly average chlorophyll concentration, particulate matter backscatter coefficient, and water body diffuse attenuation coefficient; then, the outliers are reconstructed by calculating the difference between the observed values and the monthly average climatological parameters output in the first step, thereby improving the parameter inversion accuracy; The vertical profile light adaptation modeling module is used to calculate the photosynthetically active radiation (PAR) in the depth direction based on the reconstructed water optical parameters; Carbon-based productivity calculation module, which uses reconstructed optical parameters and photosynthetically active radiation distribution to calculate the net primary productivity of phytoplankton based on a carbon-based model; The multi-source data verification module is used to verify the optical parameters and NPP results output by the above modules using buoy data and ocean color remote sensing data, thereby ensuring the accuracy and stability of the entire system operation.

8. The system according to claim 7, characterized in that The transient response correction and polarization crosstalk correction formulas used by the signal correction module are implemented as preset modules in the system hardware or software to automatically correct the satellite-borne lidar echo data; and / or the data quality control module is based on the total depolarization ratio δ of the water body. T And auxiliary sea ice concentration and wind speed data, it can automatically identify and filter abnormal data affected by sea ice, mirror reflection and foam interference.

9. A computer device, characterized in that: include: processor; Memory; as well as A computer program stored in the memory, executed under the control of the processor, implements the method according to any one of claims 1 to 6.

10. A computer-readable storage medium having program instructions stored thereon, wherein when the instructions are executed by a computer, they are used to implement the method according to any one of claims 1 to 6.

Images