Method and system for calculating upper-layer seawater particle organic carbon profile structure

Abstract

The invention provides an upper-layer seawater particle organic carbon profile structure calculation method and system, rich field data is provided for construction of a neural network model through high-frequency particle attenuation coefficient observation data, and on the basis, model precision and water color satellite remote sensing and numerical mode accessibility of input parameters are comprehensively considered, so that the accuracy of the model is improved. Using the sea surface chlorophyll concentration, the seawater temperature profile data, the seawater salinity profile data and the particle organic carbon profile type as input parameters to construct a water particle attenuation coefficient prediction model; and finally, according to the conversion relation between the water body particle organic carbon concentration and the particle attenuation coefficient, obtaining the three-dimensional distribution of the water body particle organic carbon concentration in the sea basin area. Therefore, dynamic monitoring and historical change evaluation of the three-dimensional field of the particle organic carbon concentration of the water body in the sea basin area are realized by means of the observation advantage of remote sensing large-range quasi-real-time long time sequence.

Description

Technical Field

[0001] The invention belongs to the field of remote sensing technology, and in particular relates to a method and system for calculating the organic carbon profile structure of upper seawater particles. Background Art

[0002] The ocean carbon pump refers to the process by which the ocean regulates the absorption of atmospheric carbon dioxide and stores it in the ocean or buries it in seafloor sediments. The biological pump process driven by photosynthesis of euphotic phytoplankton converts inorganic carbon in seawater into organic carbon and exports it to the deep sea in the form of particulate organic carbon (POC), thereby achieving carbon sequestration in the ocean. Although phytoplankton only accounts for 0.2%-0.3% of the marine biomass, due to its high turnover rate, this process can explain about 70% of the vertical gradient changes in dissolved inorganic carbon. The effective operation of the biological pump has a significant impact on atmospheric carbon dioxide ( ) concentration plays an important role in regulating the atmospheric The concentration will increase by nearly 200 ppm, equivalent to an increase of about 40% over current levels. Therefore, improving the understanding of the distribution and composition of POC in the euphotic zone and its changes under different climate conditions is one of the key areas of climate change research.

[0003] At present, ocean color satellite remote sensing is the only feasible means to meet the needs of continuous observation of POC in upper water bodies at global or regional scales. However, the observation capability of traditional ocean color satellites is theoretically limited to the first optical depth, that is, the upper 20% of the euphotic layer, which limits the observation of POC profiles in the entire euphotic layer. Existing studies mainly use two types of methods to evaluate the distribution of POC in upper water bodies: one is to establish a statistical relationship between the sea surface POC concentration and the integral or average concentration of the water layer (first optical depth, euphotic layer depth or mixed layer depth); the other is to combine sea surface satellite observations to construct a typical profile model. Early studies found that for seawater dominated by phytoplankton, the integrated reserves of chlorophyll Chla in the euphotic layer can be estimated by the power function relationship of sea surface Chla, and further constructed a Gaussian profile distribution model of 7 types of euphotic layer Chla, thereby realizing three-dimensional remote sensing observation of Chla in upper water bodies at a global scale. On this basis, the structural characteristics of the euphotic layer POC profile were studied, and the empirical power function relationship between the integrated POC reserves in the euphotic layer and the POC concentration on the sea surface was constructed under stratified and uniformly mixed conditions, respectively, to achieve global-scale POC remote sensing estimation.

[0004] However, when the global-scale empirical method is applied in the marginal sea area, the model constants need to be re-tuned, and even the model structure (linear or power function) itself may not be applicable to the regional change characteristics. In addition, the patent (CN202110272580.X) proposes a method for distinguishing POC profile distribution categories that depends on sea surface POC concentration, mixed layer depth and water depth. However, this method is based on fixed seasonal inputs and is suitable for monthly average scale observations. There may be jumps on the intra-month scale, and fixed constants are used within the season, which makes it difficult to meet the dynamic observation needs of remote sensing images pixel by pixel.

[0005] In recent years, the research on POC has gradually focused on the intrinsic optical quantities of water bodies with high vertical resolution to better characterize the profile characteristics of traditional discrete sampling POC. However, there is currently a lack of dynamic observation methods for the profile structure of intrinsic optical quantities of water bodies based on satellite remote sensing. Related research still mainly relies on field observation data or directly uses optical profile data measured by field instruments as model input. Therefore, it is urgent to develop remote sensing dynamic observation technology based on intrinsic optical quantities of water bodies to improve the three-dimensional remote sensing observation capabilities of POC in upper water bodies. Summary of the invention

[0006] The present invention provides a method and system for calculating the organic carbon profile structure of upper seawater particles, so as to solve the problem that the prior art lacks a dynamic observation method for the profile structure of inherent optical quantities of water bodies based on remote sensing, and mainly relies on field observation data or directly uses optical profile data measured by field instruments. In order to solve the above technical problems, the embodiments of the present invention disclose the following technical solutions: One aspect of the present invention provides a method for calculating the organic carbon profile structure of upper seawater particles, which is applied to sea basins, and the method comprises: Acquire seawater characteristic data from multiple observation points in the sea basin area, where the seawater characteristic data from each observation point at least include sea surface chlorophyll concentration, seawater temperature profile data, seawater salinity profile data, particulate organic carbon profile type, and particulate attenuation coefficient profile data; The neural network is trained based on the seawater characteristic data of all observation points to obtain the particle attenuation coefficient profile prediction model; Identify the sea basin pixels in the target remote sensing image, and determine the particulate organic carbon profile type of each sea basin pixel by a preset method; For each sea basin pixel, the acquired sea surface chlorophyll concentration, seawater temperature profile data, seawater salinity profile data, and the corresponding particulate organic carbon profile type are input into the prediction model to obtain the particle attenuation coefficient profile data; Obtain in advance the conversion relationship between the particle organic carbon concentration and the particle attenuation coefficient in the sea basin area; Based on the conversion relationship, the particulate organic carbon profile data of each sea basin pixel is obtained according to the corresponding particle attenuation coefficient profile data, and a three-dimensional field of particulate organic carbon in the upper seawater of the target sea basin is constructed.

[0007] Optionally, the neural network is trained based on the seawater characteristic data of all observation points to obtain a particle attenuation coefficient profile prediction model, including: A training data set is constructed based on the seawater characteristic data of all observation points, and each sample includes the sea surface chlorophyll concentration, seawater temperature profile data, seawater salinity profile data and particle attenuation coefficient profile data observed at any observation point in the same period, as well as the particle organic carbon profile type at the observation point; The sea surface chlorophyll concentration, seawater temperature profile data, seawater salinity profile data and particulate organic carbon profile type in the sample were used as input data, and the particle attenuation coefficient profile data was used as output data. The neural network model was trained and verified using the 5-fold crossover method, and finally the particle attenuation coefficient profile prediction model was obtained.

[0008] Optionally, before executing the step of training a neural network based on the seawater characteristic data of all observation points to obtain a particle attenuation coefficient profile prediction model, the method further includes: The seawater characteristic data of all observation points are preprocessed, and the preprocessing at least includes outlier removal and standardization.

[0009] Optionally, identifying sea basin pixels in the target remote sensing image and determining the particulate organic carbon profile type of each sea basin pixel in a preset manner includes: Obtain the water depth at each pixel in the target remote sensing image; The pixels whose water depth exceeds the preset depth threshold are determined as sea basin pixels; Calculate the true light layer depth corresponding to each sea basin pixel, and determine whether the true light layer depth is greater than the preset classification threshold; If so, determining that the particulate organic carbon profile type corresponding to the pixel is a Gaussian-like type; If not, it is determined that the particulate organic carbon profile type corresponding to the pixel is an exponential decay type.

[0010] Optionally, the true light layer depth corresponding to each ocean basin pixel is calculated using the following formula: kvis=k1+k2 / (1+Z)^0.5 TE=kvis×Z Zeu=min(abs(TE-4.605)) Wherein, Z is the water depth at the pixel; kvis is the diffuse attenuation coefficient of the water body; TE is the optical depth; k1 and k2 are calculated from the inherent optical quantities of the water body by the following formula: k1=(x0+x1×at490^0.5+x2×bbt490) ×(1+a0×sin( )); k2=(c0+c1×at490+c2×bbt490) ×(a1+a2×cos( )); Among them, x0=-0.057; x1=0.482; x2=4.221; c0=0.183; c1=0.702; c2=-2.567; a0=0.090; a1=1.465; a2=-0.667; at490 and bbt490 are the total absorption coefficient and backscattering coefficient at a wavelength of 490nm below the sea surface, respectively, which are obtained from the remote sensing image metadata; is the solar zenith angle.

[0011] Optionally, the conversion relationship between the particle organic carbon concentration and the particle attenuation coefficient in the sea basin area is determined according to the following formula: POC=10 ^ ( ( log(cp660)+1.71 ) / 1.27 ) Among them, cp660 is the particle attenuation coefficient; POC is the concentration of particulate organic carbon in the upper seawater.

[0012] Optionally, the method further includes: Preprocess the target remote sensing image; The sea surface chlorophyll concentration, seawater temperature profile data and seawater salinity profile data corresponding to each pixel are obtained respectively.

[0013] Optionally, respectively obtaining the sea surface chlorophyll concentration, seawater temperature profile data, and seawater salinity profile data corresponding to each pixel includes: Use publicly available remote sensing products to obtain sea surface chlorophyll concentration; The seawater temperature profile data and seawater salinity profile data are obtained based on the HYCOM global ocean model.

[0014] Optionally, construct a three-dimensional field of particulate organic carbon in the upper seawater of the target sea basin in the following manner: The particulate organic carbon profile data of each sea basin pixel are subjected to sliding average processing to obtain the vertical distribution data of the particulate organic carbon concentration of the sea basin pixel; The plane data composed of the vertical distribution data of the particle organic carbon concentration of all the sea basin pixels are smoothed by using median filtering; Construct three-dimensional fields based on the processed particulate organic carbon concentration data.

[0015] Another aspect of the present invention discloses a system for calculating the organic carbon profile structure of upper seawater particles, wherein the system executes the method for calculating the organic carbon profile structure of upper seawater particles as described in any one of the aforementioned aspects.

[0016] The present invention discloses a method and system for calculating the profile structure of particulate organic carbon in upper seawater. Through high-frequency observation data of particle attenuation coefficient, rich field data is provided for the construction of neural network model. On this basis, the accuracy of the model and the availability of water color satellite remote sensing and numerical models of input parameters are comprehensively considered. The sea surface chlorophyll concentration, seawater temperature profile data, seawater salinity profile data, and particulate organic carbon profile type are used as input parameters to construct a prediction model for the particle attenuation coefficient of water bodies. Finally, the three-dimensional distribution of the concentration of particulate organic carbon in water bodies in the sea basin area is obtained based on the conversion relationship between the concentration of particulate organic carbon in water bodies and the particle attenuation coefficient. Thus, with the advantage of remote sensing's large-scale, quasi-real-time and long-time series observation, the dynamic monitoring and historical change evaluation of the three-dimensional field of the concentration of particulate organic carbon in water bodies in the sea basin area can be realized.

[0017] This Summary is provided to introduce a selection of concepts in a simplified form that are further described below in the Detailed Description. This Summary is not intended to identify key features or essential features of the disclosure, nor is it intended to limit the scope of the disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The above and other objects, features and advantages of the present disclosure will become more apparent through a more detailed description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, wherein like reference numerals generally represent like components throughout the exemplary embodiments of the present disclosure.

[0019] Figure 1 A schematic diagram of a flow chart of a method for calculating the organic carbon profile structure of upper seawater particles provided by an embodiment of the present invention; Figure 2 An implementation provided for an embodiment of the present invention Figure 1 Schematic diagram of the process of step S200; Figure 3 An implementation provided for an embodiment of the present invention Figure 1 Schematic diagram of the process of step S300; Figure 4 A schematic flow chart of another method for calculating the organic carbon profile structure of upper seawater particles provided in an embodiment of the present invention.

[0020] Figure 5 An implementation provided for an embodiment of the present invention Figure 1 Schematic diagram of the process of step S600. DETAILED DESCRIPTION

[0021] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to make the present disclosure more thorough and complete, and to fully convey the scope of the present disclosure to those skilled in the art.

[0022] As used herein, the term "including" and its variations mean open inclusion, i.e., "including but not limited to". Unless otherwise stated, the term "or" means "and / or". The term "based on" means "based at least in part on". The terms "an example embodiment" and "an embodiment" mean "at least one example embodiment". The term "another embodiment" means "at least one additional embodiment". The terms "first", "second", etc. may refer to different or the same objects. Other explicit and implicit definitions may also be included below.

[0023] Figure 1 The present invention provides a flow chart of a method for calculating the organic carbon profile structure of upper seawater particles, which is applied to sea basins, which are low-lying areas in the ocean terrain. Figure 1 As shown, the method comprises the following steps: Step S100: Acquire seawater characteristic data of multiple observation points in a sea basin area.

[0024] In the disclosed embodiment of the present invention, the seawater characteristic data of each observation point at least includes sea surface chlorophyll concentration, seawater temperature profile data, seawater salinity profile data, particulate organic carbon profile type and particle attenuation coefficient profile data.

[0025] The sea surface chlorophyll concentration, seawater temperature profile data, seawater salinity profile data and particle attenuation coefficient profile data at the observation point can be obtained through remote sensing data or field measurements.

[0026] For example, sea surface chlorophyll concentration can be obtained by collecting samples with water samplers and then measuring them in the laboratory using fluorescence methods, or it can be obtained from public remote sensing products; Seawater temperature profile data and seawater salinity profile data can be obtained by measuring on-site CTD detectors or through the HYCOM model. HYCOM is a global ocean numerical model that provides temperature profile data with high temporal and spatial resolution.

[0027] The particle attenuation coefficient profile data can be obtained by profile detection using the AC-S water body measurement equipment. AC-S can measure the light absorption and scattering of water bodies.

[0028] Step S200: training a neural network based on the seawater characteristic data of all observation points to obtain a particle attenuation coefficient profile prediction model.

[0029] In one embodiment disclosed in the present invention, Figure 2 As shown, step S200 may be implemented in the following manner: Step S201: construct a training data set based on the seawater characteristic data of all observation points, each sample includes the sea surface chlorophyll concentration, seawater temperature profile data, seawater salinity profile data and particle attenuation coefficient profile data observed at any observation point in the same period, and the particulate organic carbon profile type at the observation point.

[0030] In one embodiment disclosed in the present invention, the seawater characteristic data of each observation point are composed of matching samples according to the observation time period. Taking a sample as an example, the sample includes the sea surface chlorophyll concentration, seawater temperature profile data, seawater salinity profile data and particle attenuation coefficient profile data observed at a certain observation point in a certain period of time, as well as the particulate organic carbon profile type at the observation point.

[0031] All samples are divided into a training data set and a test data set. For example, 70% of the samples are used to construct the training data set, and the remaining 30% of the samples are used to construct the test data set.

[0032] Step S202: Using the sea surface chlorophyll concentration, seawater temperature profile data, seawater salinity profile data and particulate organic carbon profile type in the sample as input data and the particle attenuation coefficient profile data as output data, the neural network model is trained and verified using a 5-fold crossover method, and finally a particle attenuation coefficient profile prediction model is obtained.

[0033] In one embodiment disclosed in the present invention, before training the prediction model, the seawater characteristic data of all observation points are first preprocessed, for example: Detect outliers and remove abnormal samples; use Min-Max standardization or Z-score standardization to normalize samples to ensure that data of different dimensions can be input into the neural network.

[0034] The network architecture of the neural network model includes an input layer, a hidden layer, an activation function and an output layer. The input layer includes 4 input variables (sea surface chlorophyll concentration, seawater temperature profile data, seawater salinity profile data and particulate organic carbon profile type); the hidden layer can use 2 to 3 layers, with 16 to 64 neurons in each layer (adjusted according to actual needs); the activation function ReLU is used to enhance nonlinear modeling capabilities; the output layer includes 1 output variable (particle attenuation coefficient profile data).

[0035] The neural network model can use mean square error (MSE) as the loss function and Adam optimizer as the optimization algorithm.

[0036] In addition, the embodiment of the present invention adopts 5-fold cross validation for the neural network model to improve the stability of the model and prevent overfitting. The validation process is: divide the training data into 5 parts, select 4 parts as training sets and 1 part as validation set each time, train 5 times, and take the average of the final performance.

[0037] After completing the training of the model, the model performance is tested using the test data set. For example, the following performance indicators are calculated: mean square error (MSE), root mean square error (RMSE), determination coefficient, etc.

[0038] Step S300: identifying sea basin pixels in the target remote sensing image, and determining the particulate organic carbon profile type of each sea basin pixel using a preset method.

[0039] In one embodiment disclosed in the present invention, Figure 3 As shown, the following sub-steps may be used to complete step S300: Step S301: Obtain the water depth corresponding to each pixel in the target remote sensing image.

[0040] The water depth at each pixel can be obtained through the ETOPO1 global topography dataset, which records global topography and ocean depth datasets, covering all land and ocean areas in the world.

[0041] According to the geographic coordinate information of the remote sensing image and the row and column numbers of the pixel in the image, the longitude and latitude corresponding to the pixel are determined, and then the water depth at the pixel is obtained through the ETOPO1 global landform dataset.

[0042] Step S302: Determine pixels whose water depth exceeds a preset depth threshold as sea basin pixels.

[0043] In the embodiment disclosed in the present invention, the preset depth threshold is 200 meters, and pixels with a water depth exceeding 200 meters are determined as pixels representing the sea basin area.

[0044] Step S303: Calculate the true photosphere depth corresponding to each sea basin pixel, and determine whether the true photosphere depth is greater than a preset classification threshold.

[0045] In one embodiment disclosed in the present invention, the true photosphere depth corresponding to each sea basin pixel is determined in the following manner.

[0046] kvis=k1+k2 / (1+Z)^0.5 TE=kvis×Z Zeu=min(abs(TE-4.605)) Among them, Z is the water depth at the pixel; kvis is the diffuse attenuation coefficient of the water body; TE is the optical depth; Zeu is obtained by the depth when TE is closest to 4.605; k1 and k2 are calculated from the inherent optical quantities of the water body by the following formula: k1=(x0+x1×at490^0.5+x2×bbt490) ×(1+a0×sin( )); k2=(c0+c1×at490+c2×bbt490) ×(a1+a2×cos( )); Among them, x0=-0.057; x1=0.482; x2=4.221; c0=0.183; c1=0.702; c2=-2.567; a0=0.090; a1=1.465; a2=-0.667; at490 and bbt490 are the total absorption coefficient and backscattering coefficient at a wavelength of 490nm below the sea surface, respectively, which are obtained from the remote sensing image metadata; is the solar zenith angle.

[0047] For each sea basin pixel, it is determined whether the true light layer depth is greater than the preset classification threshold.

[0048] In a specific embodiment disclosed in the present invention, through the research and analysis of the true light layer depth and particulate organic carbon profile types of water bodies for many years, the preset classification threshold is set to 80 meters, that is, to determine whether the true light layer depth at the pixel is greater than 80 meters.

[0049] If so, determine that the particulate organic carbon profile type corresponding to the pixel is Gaussian-like; If not, it is determined that the particulate organic carbon profile type corresponding to the pixel is exponential decay type.

[0050] The present invention is applied to sea basin areas, and the particulate organic carbon profile types in sea basin areas are mainly Gaussian-like and exponential decay types, so other types are not distinguished in the present invention.

[0051] Step S400: For each sea basin pixel, the acquired sea surface chlorophyll concentration, seawater temperature profile data, seawater salinity profile data, and the corresponding particulate organic carbon profile type are input into the prediction model to obtain the particle attenuation coefficient profile data.

[0052] In one embodiment disclosed in the present invention, Figure 4 As shown, the method disclosed in the present invention also includes the following steps: Step S010: pre-processing the target remote sensing image.

[0053] For example, remote sensing images can be preprocessed in the following ways.

[0054] (1) Radiation correction: convert the digital value (DN) of the image into physical units (such as radiance, reflectivity, etc.).

[0055] (2) Atmospheric correction: for example, Rayleigh scattering correction to remove the influence of atmospheric molecular scattering; or aerosol correction to reduce the interference of suspended particles on water reflectivity.

[0056] (3) Geometric correction: Project the image to a unified coordinate system (WGS84, UTM) to correct the geometric distortion caused by terrain and sensor tilt.

[0057] (4) Noise removal: For example, cloud removal, using a cloud mask to remove cloud-covered areas in the image; or water masking to exclude water body edges and areas severely affected by sediment.

[0058] Step S020: respectively obtaining the sea surface chlorophyll concentration, sea water temperature profile data and sea water salinity profile data corresponding to each pixel.

[0059] In one embodiment disclosed in the present invention, the sea surface chlorophyll concentration at each pixel can be obtained using a published remote sensing product, or the sea surface chlorophyll concentration at the pixel can be calculated using a remote sensing inversion algorithm.

[0060] And, seawater temperature profile data and seawater salinity profile data are obtained based on the HYCOM global ocean model. HYCOM is a global three-dimensional ocean numerical model. After obtaining the longitude and latitude corresponding to the pixel, the seawater temperature profile data and seawater salinity profile data at the pixel can be determined through this model.

[0061] Step S500: pre-acquire the conversion relationship between the concentration of particulate organic carbon and the particle attenuation coefficient in the sea basin area.

[0062] Multiple sets of historical data matching the concentration of particulate organic carbon and the particle attenuation coefficient in the sea basin area at multiple observation points were obtained in advance, and regression analysis was performed on the historical data of the two to obtain the following conversion relationship: POC=10 ^ ( ( log(cp660)+1.71 ) / 1.27 ) Among them, cp660 is the particle attenuation coefficient; POC is the concentration of particulate organic carbon in the upper seawater.

[0063] Step S600: Based on the conversion relationship, the particulate organic carbon profile data of each sea basin pixel is obtained according to the corresponding particle attenuation coefficient profile data, and a three-dimensional field of particulate organic carbon in the upper seawater of the target sea basin is constructed.

[0064] In one embodiment disclosed in the present invention, Figure 5As shown in the figure, the three-dimensional field of particulate organic carbon in the upper seawater of the target sea basin is constructed in the following way: Step S601: For each sea basin pixel, the particulate organic carbon profile data is subjected to sliding average processing to obtain the vertical distribution data of the particulate organic carbon concentration of the sea basin pixel.

[0065] MATLAB can be used to implement sliding average processing of particulate organic carbon profile data to obtain vertical distribution data. For example, vertical distribution data can be obtained at intervals of 1 meter to reduce fluctuations, eliminate extreme outliers, make the profile smoother, and at the same time, make the hierarchical structure of the profile data clear, which is helpful for analyzing the distribution pattern of particulate organic carbon concentration at different depths.

[0066] Step S602: The plane data composed of the vertical distribution data of the particle organic carbon concentration of all the sea basin pixels is smoothed by using median filtering.

[0067] In order to improve data quality, make the vertical distribution of particulate organic carbon concentration smoother, and avoid the influence of extreme values, the present invention uses MATLAB to perform median filtering and smoothing on the plane data composed of the vertical distribution data of particulate organic carbon concentration of all sea basin pixels to reduce noise and maintain the distribution trend of particulate organic carbon concentration.

[0068] Step S603: constructing a three-dimensional field based on the processed particulate organic carbon concentration data.

[0069] Based on the data of all depth layers, a complete three-dimensional field of particulate organic carbon concentration is constructed, and the distribution of particulate organic carbon concentration data at different latitudes, longitudes and depths is intuitively output.

[0070] The embodiment of the present invention further provides a system for calculating the organic carbon profile structure of particles in the upper layer of seawater, which can implement the method for calculating the organic carbon profile structure of particles in the upper layer of seawater disclosed in the aforementioned embodiment.

[0071] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or technical improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A method for calculating the profile structure of particulate organic carbon in upper seawater, characterized in that: Applied to sea basins, the method comprises: Acquire seawater characteristic data from multiple observation points in the sea basin area, where the seawater characteristic data from each observation point at least include sea surface chlorophyll concentration, seawater temperature profile data, seawater salinity profile data, particulate organic carbon profile type, and particulate attenuation coefficient profile data; The neural network is trained based on the seawater characteristic data of all observation points to obtain the particle attenuation coefficient profile prediction model; Identify the sea basin pixels in the target remote sensing image, and determine the particulate organic carbon profile type of each sea basin pixel by a preset method; For each sea basin pixel, the acquired sea surface chlorophyll concentration, seawater temperature profile data, seawater salinity profile data, and the corresponding particulate organic carbon profile type are input into the prediction model to obtain the particle attenuation coefficient profile data; Obtain in advance the conversion relationship between the particle organic carbon concentration and the particle attenuation coefficient in the sea basin area; Based on the conversion relationship, the particulate organic carbon profile data of each sea basin pixel is obtained according to the corresponding particle attenuation coefficient profile data, and a three-dimensional field of particulate organic carbon in the upper seawater of the target sea basin is constructed.

2. The method according to claim 1, characterized in that: The neural network is trained based on the seawater characteristic data of all observation points to obtain a particle attenuation coefficient profile prediction model, including: A training data set is constructed based on the seawater characteristic data of all observation points, and each sample includes the sea surface chlorophyll concentration, seawater temperature profile data, seawater salinity profile data and particle attenuation coefficient profile data observed at any observation point in the same period, as well as the particle organic carbon profile type at the observation point; The sea surface chlorophyll concentration, seawater temperature profile data, seawater salinity profile data and particulate organic carbon profile type in the sample were used as input data, and the particle attenuation coefficient profile data was used as output data. The neural network model was trained and verified using the 5-fold crossover method, and finally the particle attenuation coefficient profile prediction model was obtained.

3. The method according to claim 2, characterized in that Before executing the step of training the neural network based on the seawater characteristic data of all observation points to obtain the particle attenuation coefficient profile prediction model, the method further includes: The seawater characteristic data of all observation points are preprocessed, and the preprocessing at least includes outlier removal and standardization.

4. The method according to claim 1, characterized in that: The identifying of the sea basin pixels in the target remote sensing image and determining the particulate organic carbon profile type of each sea basin pixel in a preset manner include: Obtain the water depth at each pixel in the target remote sensing image; The pixels whose water depth exceeds the preset depth threshold are determined as sea basin pixels; Calculate the true light layer depth corresponding to each sea basin pixel, and determine whether the true light layer depth is greater than the preset classification threshold; If so, determining that the particulate organic carbon profile type corresponding to the pixel is a Gaussian-like type; If not, it is determined that the particulate organic carbon profile type corresponding to the pixel is an exponential decay type.

5. The method according to claim 4, characterized in that: The true light layer depth corresponding to each ocean basin pixel is calculated using the following formula: kvis=k1+k2 / (1+Z)^0.5 TE=kvis×Z Zeu=min(abs(TE-4.605)) Wherein, Z is the water depth at the pixel; kvis is the diffuse attenuation coefficient of the water body; TE is the optical depth; k1 and k2 are calculated from the inherent optical quantities of the water body by the following formula: k1=(x0+x1×at490^0.5+x2×bbt490) ×(1+a0×sin( )); k2=(c0+c1×at490+c2×bbt490) ×(a1+a2×cos( )); Among them, x0=-0.057; x1=0.482; x2=4.221; c0=0.183; c1=0.702; c2=-2.567; a0=0.090; a1=1.465; a2=-0.667; at490 and bbt490 are the total absorption coefficient and backscattering coefficient at a wavelength of 490nm below the sea surface, respectively, which are obtained from the remote sensing image metadata; is the solar zenith angle.

6. The method according to claim 1, characterized in that The conversion relationship between the particle organic carbon concentration and the particle attenuation coefficient in the sea basin area is determined according to the following formula: POC=10 ^ ( ( log(cp660)+1.71 ) / 1.27 ) Among them, cp660 is the particle attenuation coefficient; POC is the concentration of particulate organic carbon in the upper seawater.

7. The method according to claim 1, characterized in that The method further comprises: Preprocess the target remote sensing image; The sea surface chlorophyll concentration, seawater temperature profile data and seawater salinity profile data corresponding to each pixel are obtained respectively.

8. The method according to claim 7, characterized in that The step of respectively obtaining the sea surface chlorophyll concentration, seawater temperature profile data and seawater salinity profile data corresponding to each pixel includes: Use publicly available remote sensing products to obtain sea surface chlorophyll concentration; The seawater temperature profile data and seawater salinity profile data are obtained based on the HYCOM global ocean model.

9. The method according to claim 1, characterized in that: The three-dimensional field of particulate organic carbon in the upper seawater of the target sea basin is constructed in the following way: The particulate organic carbon profile data of each sea basin pixel are subjected to sliding average processing to obtain the vertical distribution data of the particulate organic carbon concentration of the sea basin pixel; The plane data composed of the vertical distribution data of the particle organic carbon concentration of all the sea basin pixels are smoothed by using median filtering; Construct three-dimensional fields based on the processed particulate organic carbon concentration data.

10. A system for calculating the profile structure of upper seawater particles organic carbon, characterized in that: The system executes the method for calculating the upper seawater particulate organic carbon profile structure according to any one of claims 1-9.

Images