Method for determining phytoplankton chlorophyll concentration values based on marine heatwaves

Abstract

The application discloses a method for determining phytoplankton chlorophyll concentration value based on marine heat wave, comprising: obtaining marine green ecological remote sensing image data and marine heat wave data of a target sea area; inputting the marine heat wave data and the marine green ecological remote sensing image data of the target sea area into a chlorophyll concentration determination model, so that the chlorophyll concentration determination model determines the marine heat wave grade of the target sea area according to the marine heat wave data, and inputs the marine green ecological remote sensing image data into a chlorophyll concentration determination sub-model according to the marine heat wave grade, so that the chlorophyll concentration determination sub-model extracts corresponding multispectral remote sensing image features and marine microwave remote sensing image features, and outputs the phytoplankton chlorophyll concentration value of the target sea area; and transmitting the phytoplankton chlorophyll concentration value of the target sea area to a marine ecological monitoring system, so as to assist the marine ecological monitoring system in sea area monitoring. The application can improve the accuracy and application reliability of the determination of the phytoplankton chlorophyll concentration value.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of marine ecological monitoring, and in particular to a method for determining phytoplankton chlorophyll concentration based on marine heat waves. BACKGROUND

[0002] Phytoplankton chlorophyll concentration is a key indicator reflecting the state of the marine ecosystem, and its changes are closely related to extreme climate events such as marine heat waves. Different levels of marine heat waves will affect the phytoplankton chlorophyll concentration in the ocean, because different levels of marine heat waves will supply different amounts of nutrients to phytoplankton, and different amounts of nutrients will change the community structure of phytoplankton, thereby leading to changes in phytoplankton chlorophyll concentration. However, in the prior art, only a single model is used to predict the phytoplankton chlorophyll concentration, which ignores the impact of marine heat waves on phytoplankton chlorophyll concentration, resulting in low accuracy of the predicted phytoplankton chlorophyll concentration, and unable to provide reliable data support for the correlation between marine heat waves and marine ecosystems. SUMMARY

[0003] The present application provides a method for determining the phytoplankton chlorophyll concentration based on marine heat waves, which can solve the problem of low accuracy of the predicted phytoplankton chlorophyll concentration caused by ignoring the impact of marine heat waves on phytoplankton chlorophyll concentration in the prior art, and improve the accuracy of determining the phytoplankton chlorophyll concentration, thereby improving the reliability of data support when applying the phytoplankton chlorophyll concentration to the marine ecological monitoring system for sea area monitoring.

[0004] An embodiment of the present application provides a method for determining the phytoplankton chlorophyll concentration based on marine heat waves, comprising:

[0005] Obtaining marine green ecological remote sensing image data and marine heat wave data of a target sea area; wherein the marine heat wave data includes historical sea surface temperature data in a preset monitoring period, current monitoring period sea surface temperature data and current monitoring period calendar day; the marine green ecological remote sensing image data includes multispectral remote sensing image and marine microwave remote sensing image;

[0006] inputting marine heat wave data and marine green ecological remote sensing image data of a target sea area into a chlorophyll concentration determination model, so that the chlorophyll concentration determination model determines a marine heat wave level of the target sea area according to the marine heat wave data, and inputs the marine green ecological remote sensing image data into a chlorophyll concentration determination sub-model according to the marine heat wave level, so that the chlorophyll concentration determination sub-model extracts corresponding multispectral remote sensing image features and marine microwave remote sensing image features, and outputs a phytoplankton chlorophyll concentration value of the target sea area according to the multispectral remote sensing image features and the marine microwave remote sensing image features;

[0007] transmitting the phytoplankton chlorophyll concentration value of the target sea area to a marine ecological monitoring system to assist the marine ecological monitoring system in sea area monitoring.

[0008] Further, the marine heat wave level includes a mild heat wave level, a moderate heat wave level, a severe heat wave level, and an extreme heat wave level.

[0009] The marine heat wave level of the target sea area is determined according to the marine heat wave data, including:

[0010] Grouping historical sea surface temperature data in a preset monitoring period by calendar day to obtain each historical sea surface temperature data under each calendar day:

[0011] For each calendar day, the average value of each historical sea surface temperature data under the current calendar day is calculated to obtain the marine heat wave threshold value under the current calendar day.

[0012] According to the marine heat wave threshold value of each calendar day, a marine heat wave threshold value matrix of the whole year is constructed.

[0013] According to the calendar day of the current monitoring period, the marine heat wave threshold value of each calendar day in the current monitoring period is obtained from the marine heat wave threshold value matrix.

[0014] According to the sea surface temperature data of the current monitoring period and the marine heat wave threshold value of each calendar day in the current monitoring period, the marine heat wave intensity and the marine heat wave duration of the target sea area in the current monitoring period are determined, and the marine heat wave level of the target sea area is determined according to the marine heat wave intensity and the marine heat wave duration.

[0015] Further, the construction of the chlorophyll concentration determination model includes:

[0016] A plurality of marine heat wave data samples and a plurality of marine green ecological remote sensing image data samples of each marine heat wave level are obtained, and a training sample set is constructed according to each marine heat wave data sample and each marine green ecological remote sensing image data sample; wherein each marine heat wave data sample comprises: a historical sea surface temperature data sample in a preset monitoring period, a historical sea surface data sample of a selected monitoring period, and a calendar day of the selected monitoring period, and each marine green ecological remote sensing image data sample comprises: a multispectral remote sensing image sample and a marine microwave remote sensing image sample;

[0017] An initial chlorophyll concentration determination model is constructed; wherein the initial chlorophyll concentration determination model comprises: a classifier, a first chlorophyll concentration determination sub-model, a second chlorophyll concentration determination sub-model, a third chlorophyll concentration determination sub-model, and a fourth chlorophyll concentration determination sub-model, and the first chlorophyll concentration determination sub-model, the second chlorophyll concentration determination sub-model, the third chlorophyll concentration determination sub-model, and the fourth chlorophyll concentration determination sub-model correspond to the mild heat wave level, the moderate heat wave level, the severe heat wave level, and the extreme heat wave level, respectively;

[0018] The initial chlorophyll concentration determination model is trained with the training sample set until the initial chlorophyll concentration determination model converges, and the chlorophyll concentration determination model is generated.

[0019] Further, the initial chlorophyll concentration determination model is trained with the training sample set until the initial chlorophyll concentration determination model converges, and the chlorophyll concentration determination model is generated, comprising:

[0020] The network parameters of each chlorophyll concentration determination sub-model are frozen, the classifier of the initial chlorophyll concentration determination model is trained with each marine heat wave data sample, the network parameters of the classifier of the initial chlorophyll concentration determination model are frozen when the classifier of the initial chlorophyll concentration determination model converges, and the network parameters of each chlorophyll concentration determination sub-model are unfrozen.

[0021] Each chlorophyll concentration determination sub-model is trained with each marine green ecological remote sensing image data sample, and the network parameters of the classifier are unfrozen when each chlorophyll concentration determination sub-model converges.

[0022] The classifier and each chlorophyll concentration determination sub-model are trained with the training sample set, and the initial chlorophyll concentration determination model is generated when the initial chlorophyll concentration determination model as a whole converges.

[0023] Further, the classifier of the initial chlorophyll concentration determination model is trained with each marine heat wave data sample, comprising:

[0024] inputting the marine heat wave data sample into the classifier, so that the classifier determines a marine heat wave intensity and a marine heat wave duration corresponding to the current marine heat wave data sample;

[0025] outputting a predicted marine heat wave grade of the current marine heat wave data sample according to the marine heat wave intensity and the marine heat wave duration corresponding to the current marine heat wave data sample.

[0026] Further, the training of the chlorophyll concentration determination sub-model for each marine green ecological remote sensing image data sample comprises:

[0027] inputting the marine ecological remote sensing image data sample into the first chlorophyll concentration determination sub-model, so that the first chlorophyll concentration determination sub-model extracts spectral reflectance of a blue light band and spectral reflectance of a green light band of the multispectral remote sensing image sample, determines a first spectral reflectance ratio according to the spectral reflectance of the blue light band and the spectral reflectance of the green light band, extracts a first microwave radiation feature of the marine microwave remote sensing image sample, determines a mixed layer depth according to the first microwave radiation feature, and outputs a predicted phytoplankton chlorophyll concentration value corresponding to the marine ecological remote sensing image data sample according to the mixed layer depth and the first spectral reflectance ratio;

[0028] inputting the marine ecological remote sensing image data sample into the second chlorophyll concentration determination sub-model, so that the second chlorophyll concentration determination sub-model extracts spectral reflectance of a red light band, spectral reflectance of a green light band and a chlorophyll fluorescence peak value of the multispectral remote sensing image sample, determines a second spectral reflectance ratio according to the spectral reflectance of the red light band and the spectral reflectance of the green light band, extracts a second microwave radiation feature of the marine microwave remote sensing image sample, determines a sea surface height anomaly value according to the second microwave radiation feature, and outputs a predicted phytoplankton chlorophyll concentration value corresponding to the marine ecological remote sensing image data sample according to the second spectral reflectance ratio, the sea surface height anomaly value and the chlorophyll fluorescence peak value;

[0029] inputting the marine ecological remote sensing image data sample into the third chlorophyll concentration determination sub-model, so that the third chlorophyll concentration determination sub-model extracts spectral reflectance of a red light band, spectral reflectance of a near-infrared band and a chlorophyll fluorescence signal of the multispectral remote sensing image sample, determines a third spectral reflectance ratio according to the spectral reflectance of the red light band and the spectral reflectance of the near-infrared band, determines a photosynthetic efficiency of the target sea area according to the chlorophyll fluorescence signal, and outputs a predicted phytoplankton chlorophyll concentration value corresponding to the marine ecological remote sensing image data sample according to the third spectral reflectance ratio and the photosynthetic efficiency;

[0030] input the marine ecological remote sensing image data sample into the fourth chlorophyll concentration determination sub-model, so that the fourth chlorophyll concentration determination sub-model extracts the spectral reflectance of the near-infrared band and the thermal infrared signal of the multispectral remote sensing image sample; determine the thermal infrared sea surface temperature peak value of the target sea area according to the thermal infrared signal; and output the predicted phytoplankton chlorophyll concentration value corresponding to the marine ecological remote sensing image data sample according to the thermal infrared sea surface temperature peak value, the spectral reflectance of the near-infrared band, and the preset chlorophyll background value.

[0031] Further, the training of the classifier and each chlorophyll concentration determination sub-model with the training sample set is performed until the initial chlorophyll concentration determination model converges as a whole, and the chlorophyll concentration determination model is generated, comprising:

[0032] The marine heat wave data samples and the marine green ecological remote sensing image data samples in each training sample set are grouped based on the marine heat wave level, and a training sample subset under each marine heat wave level is obtained; wherein the training sample subset includes marine heat wave data samples and marine green ecological remote sensing image data samples under the current marine heat wave level.

[0033] The classifier and the chlorophyll concentration determination sub-model corresponding to the marine heat wave level of each training sample subset are trained with each training sample subset, and the chlorophyll concentration determination model is generated when the initial chlorophyll concentration determination model converges as a whole.

[0034] Further, the output of the predicted phytoplankton chlorophyll concentration value corresponding to the marine ecological remote sensing image data sample according to the second spectral reflectance ratio, the sea surface height anomaly value, and the chlorophyll fluorescence peak value comprises:

[0035] A first correlation relationship between the second spectral reflectance ratio and the chlorophyll concentration is constructed, a second correlation relationship between the sea surface height anomaly value and the chlorophyll concentration is constructed, and a third correlation relationship between the chlorophyll fluorescence peak value and the chlorophyll concentration is constructed.

[0036] A spatial rectangular coordinate system is established with the first correlation relationship as the horizontal axis, the second correlation relationship as the vertical axis, and the third correlation relationship as the vertical axis; wherein the plane formed by the horizontal axis and the vertical axis is the first plane, the plane formed by the horizontal axis and the vertical axis is the second plane, and the plane formed by the vertical axis and the vertical axis is the third plane.

[0037] A first correlation vector between the second spectral reflectance ratio, the sea surface height anomaly value, and the chlorophyll concentration is constructed, a second correlation vector between the second spectral reflectance ratio, the chlorophyll fluorescence peak value, and the chlorophyll concentration is constructed, and a third correlation vector between the sea surface height anomaly value, the chlorophyll fluorescence peak value, and the chlorophyll concentration is constructed.

[0038] The first correlation vector, the second correlation vector and the third correlation vector are projected on a first plane, a second plane and a third plane respectively to obtain a first projection of the first correlation vector on the first plane, a second projection on the second plane and a third projection on the third plane; a fourth projection of the second correlation vector on the first plane, a fifth projection on the second plane and a sixth projection on the third plane; a seventh projection of the third correlation vector on the first plane, an eighth projection on the second plane and a ninth projection on the third plane;

[0039] The first projection, the fourth projection and the seventh projection are superimposed to obtain a first superimposed projection, the second projection, the fifth projection and the eighth projection are superimposed to obtain a second superimposed projection, and the third projection, the sixth projection and the ninth projection are superimposed to obtain a third superimposed projection; the first superimposed projection, the second superimposed projection and the third superimposed projection are converted into vector form to obtain a first superimposed vector, a second superimposed vector and a third superimposed vector; and a predicted phytoplankton chlorophyll concentration value corresponding to the ocean ecological remote sensing image data sample is output through the first superimposed vector, the second superimposed vector and the third superimposed vector.

[0040] Another embodiment of the present application provides a terminal device, comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor implements the method for determining a phytoplankton chlorophyll concentration value based on a marine heat wave according to the above-mentioned embodiments of the present application when executing the computer program.

[0041] Another embodiment of the present application provides a storage medium, comprising a stored computer program, wherein the storage medium controls a device where the storage medium is located to execute the method for determining a phytoplankton chlorophyll concentration value based on a marine heat wave according to the above-mentioned embodiments of the present application when the computer program is running.

[0042] The present application has the following beneficial effects:

[0043] The present invention provides a method for determining the chlorophyll concentration value of phytoplankton based on ocean heat waves. When confirming the chlorophyll concentration value of phytoplankton in the target sea area, the marine green ecological remote sensing image data and the marine heat wave data of the target sea area are obtained, and then the marine heat wave data and the marine green ecological remote sensing image data of the target sea area are input into a chlorophyll concentration determination model, so that the chlorophyll concentration determination model determines the marine heat wave level of the target sea area according to the marine heat wave data, and then according to the determined marine heat wave level, the chlorophyll concentration determination sub-model of the corresponding level in the chlorophyll concentration determination model is selected, and the chlorophyll concentration determination sub-model extracts the corresponding multi-light wave level. The multispectral remote sensing image characteristics and the ocean microwave remote sensing image characteristics are used to output the phytoplankton chlorophyll concentration value of the target sea area according to the multispectral remote sensing image characteristics and the ocean microwave remote sensing image characteristics. By first determining the ocean heat wave level and then determining the chlorophyll concentration value using the chlorophyll concentration determination sub-model under the corresponding ocean heat wave level, the problem that the existing technology ignores the impact of ocean heat waves on phytoplankton chlorophyll concentration, resulting in low accuracy of the phytoplankton chlorophyll concentration prediction results, is solved. The accuracy of the determination of phytoplankton chlorophyll concentration value is improved, thereby improving the data support reliability when the phytoplankton chlorophyll concentration value is applied to the marine ecological monitoring system for sea area monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 This is a flow chart of a method for determining phytoplankton chlorophyll concentration values ​​based on ocean heat waves provided by one embodiment of the present invention.

[0045] Figure 2 Schematic diagram of the structure of a chlorophyll concentration determination model provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0046] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments 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 shall fall within the scope of protection of the present invention.

[0047] like Figure 1 FIG. 1 is a method for determining phytoplankton chlorophyll concentration based on ocean heat waves according to an embodiment of the present invention, comprising:

[0048] Step S1: obtaining marine green ecological remote sensing image data and marine heat wave data of a target sea area; wherein the marine heat wave data includes historical sea surface temperature data in a preset monitoring period, sea surface temperature data in a current monitoring period, and a calendar day in the current monitoring period; the marine green ecological remote sensing image data includes multispectral remote sensing image and marine microwave remote sensing image;

[0049] Step S2: inputting the marine heat wave data and the marine green ecological remote sensing image data of the target sea area into a chlorophyll concentration determination model, so that the chlorophyll concentration determination model determines the marine heat wave grade of the target sea area according to the marine heat wave data, and inputs the marine green ecological remote sensing image data into a chlorophyll concentration determination sub-model according to the marine heat wave grade, so that the chlorophyll concentration determination sub-model extracts corresponding multispectral remote sensing image features and marine microwave remote sensing image features, and outputs the phytoplankton chlorophyll concentration value of the target sea area according to the multispectral remote sensing image features and the marine microwave remote sensing image features;

[0050] Step S3: transmitting the phytoplankton chlorophyll concentration value of the target sea area to a marine ecological monitoring system to assist the marine ecological monitoring system in sea area monitoring.

[0051] For step S1, the marine green ecological remote sensing image data and the marine heat wave data of the target sea area are obtained, wherein the marine heat wave data includes historical sea surface temperature data in a preset monitoring period, sea surface temperature data in a current monitoring period, and a calendar day in the current monitoring period; the marine green ecological remote sensing image data includes multispectral remote sensing image and marine microwave remote sensing image. For example, in the present application, the end time point of the preset monitoring period is usually the starting time point of the current monitoring period, the length of the preset monitoring period is 30 years, the starting time point of the current monitoring period is set to January 2, 2025, and the length of the current monitoring period is 7 days. Therefore, the current monitoring period is from January 2, 2025 to January 8, 2025, the calendar day of the current monitoring period includes January 2, January 3, January 4, January 5, January 6, January 7, and January 8, and the preset monitoring period is from January 2, 1995 to January 2, 2025.

[0052] For step S2, the marine heat wave data and the marine green ecological remote sensing image data of the target sea area are input into the chlorophyll concentration determination model. The chlorophyll concentration determination model determines the marine heat wave grade of the target sea area according to the marine heat wave data. According to the marine heat wave grade, the marine green ecological remote sensing image data is input into the chlorophyll concentration determination sub-model, so that the chlorophyll concentration determination sub-model extracts corresponding multispectral remote sensing image features and marine microwave remote sensing image features, and outputs the phytoplankton chlorophyll concentration value of the target sea area according to the multispectral remote sensing image features and the marine microwave remote sensing image features.

[0053] In a preferred embodiment, the marine heat wave level comprises: a mild heat wave level, a moderate heat wave level, a severe heat wave level and an extreme heat wave level; the marine heat wave level of the target sea area is determined according to the marine heat wave data, comprising: grouping the historical sea surface temperature data in a preset monitoring period in a calendar day to obtain each historical sea surface temperature data under each calendar day; for each calendar day, the average value of each historical sea surface temperature data under the current calendar day is calculated to obtain the marine heat wave threshold value under the current calendar day; the marine heat wave threshold matrix of the whole year calendar day is constructed according to the marine heat wave threshold value of each calendar day; the marine heat wave threshold value of each calendar day in the current monitoring period is obtained from the marine heat wave threshold matrix according to the calendar day of the current monitoring period; the marine heat wave intensity and the marine heat wave duration of the target sea area in the current monitoring period are determined according to the sea surface temperature data of the current monitoring period and the marine heat wave threshold value of each calendar day in the current monitoring period, and the marine heat wave level of the target sea area is determined according to the marine heat wave intensity and the marine heat wave duration of the target sea area.

[0054] Specifically, the classification of marine heat wave level is based on the internationally recognized classification system formed according to the sea surface temperature anomaly amplitude (i.e. marine heat wave intensity) and the sea surface temperature anomaly duration (i.e. marine heat wave duration), which is an objective classification standard. When the marine heat wave intensity of the target sea area is between (1, 2] and the marine heat wave duration is between 5-10 days, the marine heat wave level is considered to be a mild heat wave level (I level); when the marine heat wave intensity of the target sea area is between (2, 3] and the marine heat wave duration is between 11-30 days, the marine heat wave level is considered to be a moderate heat wave level (II level); when the marine heat wave intensity of the target sea area is between (3, 4] and the marine heat wave duration is between 31-90 days, the marine heat wave level is considered to be a severe heat wave level (III level); when the marine heat wave intensity of the target sea area is greater than 4 and the duration is greater than 90 days, the marine heat wave level is considered to be an extreme heat wave level (IV level).

[0055] It needs to be supplemented that when the marine heat wave level is a mild heat wave level, the mixed layer is shallow, leading to short-term replenishment of nutrients, and the phytoplankton community fluctuates slightly; when the marine heat wave level is a moderate heat wave level, nutrient limitation appears, diatoms are converted to dinoflagellates, and the chlorophyll concentration decreases by 20%-40%; when the marine heat wave level is a severe heat wave level, the photosynthetic efficiency decreases significantly, the amount of chlorophyllide increases, and the ecosystem function is damaged; when the marine heat wave level is an extreme heat wave level, the phytoplankton community structure collapses, the proportion of microalgae is greater than 50%, and the coral bleaching rate is greater than 70%. According to the ecological response characteristics of phytoplankton under different marine heat wave levels, it can be known that with the increase of the marine heat wave level, the chlorophyll concentration of the phytoplankton in the target sea area decreases, and the marine heat wave level will affect the chlorophyll concentration value of the target sea area.

[0056] Based on the above division basis, when determining the marine heat wave level of the target sea area, first, the historical sea surface temperature data in the preset monitoring period is grouped in calendar days. In the present application, the calendar day unit contains all calendar days in a complete year of a non-leap year, i.e. 365 calendar days corresponding to the period from January 1 to December 31. Through the above grouping, each historical sea surface temperature data under each calendar day is obtained, for example, when the preset monitoring period is 30 years, the obtained calendar day January 1 contains the historical sea surface temperature data under January 1 of each year in the preset monitoring period of 30 years. For each calendar day, the average value of the historical sea surface temperature data under the corresponding calendar day is calculated, i.e. the marine heat wave threshold value under the current calendar day is obtained. The marine heat wave threshold value matrix of the whole year calendar day is constructed according to the marine heat wave threshold value of each calendar day. According to the calendar day of the current monitoring period, the marine heat wave threshold value of each calendar day in the current monitoring period is obtained from the marine heat wave threshold value matrix, and the sea surface temperature data under the current monitoring period calendar day is compared with the marine heat wave threshold value of the corresponding calendar day to determine the marine heat wave intensity, and the marine heat wave duration in the current monitoring period can be determined according to the marine heat wave intensity of each calendar day in the current monitoring period.

[0057] For example, assuming that the current monitoring period is from January 2, 2025 to January 8, 2025, the sea surface temperatures from January 2, 2025 to January 8, 2025 are a, b, c, d, e, f, g, and h in turn, the marine heat wave thresholds A, B, C, D, E, F, G, and H from January 2 to January 8 are obtained from the marine heat wave threshold matrix, the difference between a and A is calculated to obtain the marine heat wave intensity σ1 on January 2, 2025, and similarly, the marine heat wave intensities from January 3, 2025 to January 8, 2025 are σ2, σ3, σ4, σ5, σ6, σ7, and σ8 in turn. The duration of the marine heat wave intensity is determined according to the continuity of the values of σ1-σ8. Here, the current monitoring period is only an example, and in actual application, the length of the current monitoring period is usually 365 days to more accurately determine the marine heat wave intensity and the duration of the marine heat wave. After determining the marine heat wave intensity and the duration of the marine heat wave, the corresponding marine heat wave grade can be obtained by comparing with the marine heat wave intensity and the duration of the marine heat wave range of each marine heat wave grade described above.

[0058] The above steps are mainly processed in the classifier of the chlorophyll concentration determination model. After the classifier outputs the marine heat wave grade, the chlorophyll concentration determination sub-model corresponding to the marine heat wave grade is indexed, and the marine green ecological remote sensing image data is input into the chlorophyll concentration determination sub-model, so that the chlorophyll concentration determination sub-model extracts the corresponding multispectral remote sensing image features and marine microwave remote sensing image features, and outputs the phytoplankton chlorophyll concentration value of the target sea area according to the multispectral remote sensing image features and the marine microwave remote sensing image features.

[0059] In a preferred embodiment, the construction of the chlorophyll concentration determination model comprises:

[0060] A plurality of marine heat wave data samples and a plurality of marine green ecological remote sensing image data samples of each marine heat wave grade are obtained, and a training sample set is constructed according to each marine heat wave data sample and each marine green ecological remote sensing image data sample. Each marine heat wave data sample includes historical sea surface temperature data samples in a preset monitoring period, historical sea surface data samples of a selected monitoring period, and calendar days of the selected monitoring period. Each marine green ecological remote sensing image data sample includes a multispectral remote sensing image sample and a marine microwave remote sensing image sample.

[0061] Constructing an initial chlorophyll concentration determination model; wherein the initial chlorophyll concentration determination model includes: a classifier, a first chlorophyll concentration determination sub-model, a second chlorophyll concentration determination sub-model, a third chlorophyll concentration determination sub-model, and a fourth chlorophyll concentration determination sub-model, wherein the first chlorophyll concentration determination sub-model, the second chlorophyll concentration determination sub-model, the third chlorophyll concentration determination sub-model, and the fourth chlorophyll concentration determination sub-model correspond to a mild heat wave level, a moderate heat wave level, a severe heat wave level, and an extreme heat wave level, respectively;

[0062] The initial chlorophyll concentration determination model is trained using the training sample set until the initial chlorophyll concentration determination model converges, thereby generating the chlorophyll concentration determination model.

[0063] Specifically, after obtaining the training sample set, construct Figure 2 The initial chlorophyll concentration determination model shown in the figure includes a classifier and four chlorophyll concentration determination sub-models, wherein the first chlorophyll concentration determination sub-model corresponds to the mild heat wave level, and is used to specifically determine the chlorophyll concentration of phytoplankton in the target sea area under the mild heat wave level; the second chlorophyll concentration determination sub-model corresponds to the moderate heat wave level, and is used to specifically determine the chlorophyll concentration of phytoplankton in the target sea area under the moderate heat wave level; the third chlorophyll concentration determination sub-model corresponds to the severe heat wave level, and is used to specifically determine the chlorophyll concentration of phytoplankton in the target sea area under the severe heat wave level; the fourth chlorophyll concentration determination sub-model corresponds to the extreme heat wave level, and is used to specifically determine the chlorophyll concentration of phytoplankton in the target sea area under the extreme heat wave level; the role of the classifier is to determine the ocean heat wave level according to the ocean heat wave data, and then index the ocean green ecological remote sensing image into the chlorophyll concentration determination sub-model corresponding to the ocean heat wave level for processing. After completing the construction of the initial chlorophyll concentration determination model, the initial chlorophyll concentration determination model is trained using the constructed training sample set until the loss function of the initial chlorophyll concentration determination model is minimized or the maximum number of iterations is reached, that is, when the initial chlorophyll concentration determination model converges, the above-mentioned chlorophyll concentration determination model is generated.

[0064] In a preferred embodiment, the initial chlorophyll concentration determination model is trained using the training sample set until the initial chlorophyll concentration determination model converges, thereby generating the chlorophyll concentration determination model, including:

[0065] Freezing network parameters of each chlorophyll concentration determination sub-model, training a classifier of the initial chlorophyll concentration determination model with each ocean heat wave data sample until the classifier of the initial chlorophyll concentration determination model converges, freezing the network parameters of the classifier of the initial chlorophyll concentration determination model, and unfreezing the network parameters of each chlorophyll concentration determination sub-model;

[0066] training the classifier and each chlorophyll concentration determination sub-model with the training sample set until the initial chlorophyll concentration determination model converges as a whole, to generate the chlorophyll concentration determination model.

[0067] training the classifier and each chlorophyll concentration determination sub-model with the training sample set until the initial chlorophyll concentration determination model converges as a whole, to generate the chlorophyll concentration determination model.

[0068] Specifically, when training the initial chlorophyll concentration determination model, the training is divided into preliminary training of the classifier, preliminary training of each chlorophyll concentration determination sub-model, and overall fine-tuning of the initial chlorophyll concentration determination model. This training method can reduce the dependence on the number of training samples in the training process of the initial chlorophyll concentration determination model, reduce the amount of training data, and improve the training efficiency by fine-tuning after determining the network parameters through step-by-step training.

[0069] In a preferred embodiment, the training of the classifier of the initial chlorophyll concentration determination model with the marine heat wave data samples comprises: inputting the marine heat wave data samples into the classifier to enable the classifier to determine the marine heat wave intensity and the marine heat wave duration corresponding to the current marine heat wave data sample; and outputting the predicted marine heat wave grade of the current marine heat wave data sample according to the marine heat wave intensity and the marine heat wave duration corresponding to the current marine heat wave data sample.

[0070] Specifically, when the classifier is preliminarily trained, firstly, the network parameters of each chlorophyll concentration determination sub-model are frozen, and only the network parameters of the classifier are trained. Secondly, in the training, the marine heat wave data sample is taken as the input of the initial chlorophyll concentration determination model of the classifier, the predicted marine heat wave grade corresponding to the marine heat wave data sample is taken as the output of the initial chlorophyll concentration determination model of the classifier, the initial chlorophyll concentration determination model of the classifier is iteratively trained until the initial chlorophyll concentration determination model of the classifier converges, and the network parameters of the initial chlorophyll concentration determination model of the classifier at the time of convergence are frozen. In each iterative training process, the initial chlorophyll concentration determination model of the classifier determines the marine heat wave intensity and the marine heat wave duration corresponding to the current marine heat wave data sample according to the marine heat wave data sample and the above-mentioned method of determining the marine heat wave intensity and the marine heat wave duration, and then outputs the predicted marine heat wave grade according to the marine heat wave intensity and the marine heat wave duration corresponding to the current marine heat wave data sample and the above-mentioned marine heat wave grade determination rule, and determines the classification loss of the current iterative training according to the predicted marine heat wave grade and the true predicted marine heat wave grade of the current marine heat wave data sample, and adjusts the network parameters of the classifier based on the classification loss to improve the prediction accuracy of the next iteration; wherein the true predicted marine heat wave grade of the current marine heat wave data sample is determined according to historical monitoring data and labeled on the current marine heat wave data sample, and the above-mentioned training process of the classifier is a supervised learning process.

[0071] In a preferred embodiment, the training of each chlorophyll concentration determination sub-model with each marine green ecological remote sensing image data sample comprises:

[0072] The marine ecological remote sensing image data sample is input into the first chlorophyll concentration determination sub-model, so that the first chlorophyll concentration determination sub-model extracts the spectral reflectance of the blue light band and the spectral reflectance of the green light band of the multispectral remote sensing image sample, determines the first spectral reflectance ratio according to the spectral reflectance of the blue light band and the spectral reflectance of the green light band, extracts the first microwave radiation feature of the marine microwave remote sensing image sample, determines the mixed layer depth according to the first microwave radiation feature, and outputs the predicted phytoplankton chlorophyll concentration value corresponding to the marine ecological remote sensing image data sample according to the mixed layer depth and the first spectral reflectance ratio;

[0073] input the marine ecological remote sensing image data sample into the second chlorophyll concentration determination sub-model, so that the second chlorophyll concentration determination sub-model extracts the spectral reflectance of the red light band, the spectral reflectance of the green light band and the chlorophyll fluorescence peak value of the multispectral remote sensing image sample, determines the second spectral reflectance ratio according to the spectral reflectance of the red light band and the spectral reflectance of the green light band, extracts the second microwave radiation feature of the marine microwave remote sensing image sample, and determines the sea surface height abnormal value according to the second microwave radiation feature; and outputs the predicted phytoplankton chlorophyll concentration value corresponding to the marine ecological remote sensing image data sample according to the second spectral reflectance ratio, the sea surface height abnormal value and the chlorophyll fluorescence peak value.

[0074] input the marine ecological remote sensing image data sample into the third chlorophyll concentration determination sub-model, so that the third chlorophyll concentration determination sub-model extracts the spectral reflectance of the red light band, the spectral reflectance of the near-infrared band and the chlorophyll fluorescence signal of the multispectral remote sensing image sample, determines the third spectral reflectance ratio according to the spectral reflectance of the red light band and the spectral reflectance of the near-infrared band, and determines the photosynthetic efficiency of the target sea area according to the chlorophyll fluorescence signal; and outputs the predicted phytoplankton chlorophyll concentration value corresponding to the marine ecological remote sensing image data sample according to the third spectral reflectance ratio and the photosynthetic efficiency.

[0075] input the marine ecological remote sensing image data sample into the fourth chlorophyll concentration determination sub-model, so that the fourth chlorophyll concentration determination sub-model extracts the spectral reflectance of the near-infrared band and the thermal infrared signal of the multispectral remote sensing image sample; determines the thermal infrared sea surface temperature peak value of the target sea area according to the thermal infrared signal, and outputs the predicted phytoplankton chlorophyll concentration value corresponding to the marine ecological remote sensing image data sample according to the thermal infrared sea surface temperature peak value, the spectral reflectance of the near-infrared band and the preset chlorophyll background value.

[0076] Specifically, after the initial chlorophyll concentration determination model is completed, the network parameters of the classifier are frozen, and the network parameters of each chlorophyll concentration determination sub-model are unfrozen. Since each chlorophyll concentration determination sub-model is independent, each chlorophyll concentration determination sub-model can be trained in parallel.

[0077] When training each chlorophyll concentration determination sub-model, the marine ecological remote sensing image data sample is used as the input of the chlorophyll concentration determination sub-model, the predicted phytoplankton chlorophyll concentration value corresponding to the marine ecological remote sensing image data sample is used as the output of the chlorophyll concentration determination sub-model, and only the extracted features are differentiated and trained.

[0078] In the training of the first chlorophyll concentration determination sub-model, in each iteration training process, the marine ecological remote sensing image data sample is taken as the input of the first chlorophyll concentration determination sub-model, so that the first chlorophyll concentration determination sub-model extracts the spectral reflectance of the blue light band and the spectral reflectance of the green light band of the multispectral remote sensing image sample, calculates the ratio between the spectral reflectance of the blue light band and the spectral reflectance of the green light band according to the spectral reflectance of the blue light band and the spectral reflectance of the green light band (i.e. the first spectral reflectance ratio), and the ratio is the initial chlorophyll concentration value. In addition, the first chlorophyll concentration determination sub-model also extracts the first microwave radiation feature of the marine microwave remote sensing image sample, and then determines the mixed layer depth according to the first microwave radiation feature, corrects the initial chlorophyll concentration value through the mixed layer depth, and outputs the predicted phytoplankton chlorophyll concentration value corresponding to the marine ecological remote sensing image data sample. The first prediction loss of the first chlorophyll concentration determination sub-model is determined according to the predicted phytoplankton chlorophyll concentration value and the true phytoplankton chlorophyll concentration value of the marine ecological remote sensing image data sample, and the network parameters of the first chlorophyll concentration determination sub-model are adjusted according to the first prediction loss, so as to optimize the prediction accuracy of the next iteration training.

[0079] In the training of the second chlorophyll concentration determination sub-model, in each iteration training process, the marine ecological remote sensing image data sample is taken as the input of the second chlorophyll concentration determination sub-model, so that the second chlorophyll concentration determination sub-model extracts the spectral reflectance of the red light band, the spectral reflectance of the green light band and the chlorophyll fluorescence peak value of the multispectral remote sensing image sample, determines the second spectral reflectance ratio (i.e. the initial chlorophyll concentration under the spectrum) according to the spectral reflectance of the red light band and the spectral reflectance of the green light band; extracts the second microwave radiation feature of the marine microwave remote sensing image sample, and determines the sea surface height abnormal value according to the second microwave radiation feature; according to the second spectral reflectance ratio, the sea surface height abnormal value and the chlorophyll fluorescence peak value, the predicted phytoplankton chlorophyll concentration value corresponding to the marine ecological remote sensing image data sample is outputted; the second prediction loss of the second chlorophyll concentration determination sub-model is determined according to the predicted phytoplankton chlorophyll concentration value and the true phytoplankton chlorophyll concentration value of the marine ecological remote sensing image data sample, and the network parameters of the second chlorophyll concentration determination sub-model are adjusted according to the second prediction loss, so as to optimize the prediction accuracy of the next iteration training.

[0080] In a preferred embodiment, the output of the predicted phytoplankton chlorophyll concentration value corresponding to the marine ecological remote sensing image data sample according to the second spectral reflectance ratio, the sea surface height abnormal value and the chlorophyll fluorescence peak value comprises:

[0081] constructing a first correlation between the second spectral reflectance ratio and the chlorophyll concentration, constructing a second correlation between the sea surface height anomaly and the chlorophyll concentration, and constructing a third correlation between the chlorophyll fluorescence peak value and the chlorophyll concentration;

[0082] establishing a space rectangular coordinate system with the first correlation as the horizontal axis, the second correlation as the vertical axis, and the third correlation as the vertical axis; wherein a plane formed by the horizontal axis and the vertical axis is a first plane, a plane formed by the horizontal axis and the vertical axis is a second plane, and a plane formed by the vertical axis and the vertical axis is a third plane;

[0083] constructing a first correlation vector between the second spectral reflectance ratio, the sea surface height anomaly, and the chlorophyll concentration, constructing a second correlation vector between the second spectral reflectance ratio, the chlorophyll fluorescence peak value, and the chlorophyll concentration, and constructing a third correlation vector between the sea surface height anomaly, the chlorophyll fluorescence peak value, and the chlorophyll concentration;

[0084] projecting the first correlation vector, the second correlation vector, and the third correlation vector on the first plane, the second plane, and the third plane respectively to obtain a first projection of the first correlation vector on the first plane, a second projection on the second plane, and a third projection on the third plane; a fourth projection of the second correlation vector on the first plane, a fifth projection on the second plane, and a sixth projection on the second plane; and a seventh projection of the third correlation vector on the first plane, an eighth projection on the second plane, and a ninth projection on the third plane;

[0085] superimposing the first projection, the fourth projection, and the seventh projection to obtain a first superimposed projection, superimposing the second projection, the fifth projection, and the eighth projection to obtain a second superimposed projection, and superimposing the third projection, the sixth projection, and the ninth projection to obtain a third superimposed projection; converting the first superimposed projection, the second superimposed projection, and the third superimposed projection into vector form to obtain a first superimposed vector, a second superimposed vector, and a third superimposed vector; and outputting a predicted phytoplankton chlorophyll concentration value corresponding to a marine ecological remote sensing image data sample through the first superimposed vector, the second superimposed vector, and the third superimposed vector.

[0086] Specifically, by constructing the correlation between the second spectral reflectance ratio, sea surface height anomaly, chlorophyll fluorescence peak value and chlorophyll concentration respectively, the core influence of each parameter is extracted separately, avoiding the one-sided reflection of a single parameter on complex ecological relationships; then the three-dimensional correlation is mapped to the spatial rectangular coordinate system and the plane projection, which intuitively presents the interaction between parameters (such as the synergistic effect of sea surface height anomaly and chlorophyll fluorescence peak value), breaking through the limitations of traditional linear models in expressing nonlinear relationships; by projecting and integrating vector information of different dimensions, the complementarity of each parameter (second spectral reflectance ratio, sea surface height anomaly, chlorophyll fluorescence peak value and chlorophyll concentration) is strengthened while the measurement error of a single parameter is weakened, and finally the prediction value carries the multi-dimensional characteristics of spectral reflectance ratio (second spectral reflectance ratio), physical environment (sea surface height anomaly value) and physiological activity (chlorophyll fluorescence peak value), which is more in line with the actual ecological mechanism of the synergistic effect of multiple factors on phytoplankton chlorophyll concentration, improving the prediction accuracy.

[0087] In the training of the third chlorophyll concentration determination sub-model, in each iteration training process, the marine ecological remote sensing image data sample is taken as the input of the third chlorophyll concentration determination sub-model, so that the third chlorophyll concentration determination sub-model extracts the spectral reflectance of the red light band, the spectral reflectance of the near-infrared band and the chlorophyll fluorescence signal of the multispectral remote sensing image sample, determines the third spectral reflectance ratio (i.e. the initial chlorophyll concentration under this spectrum) according to the spectral reflectance of the red light band and the spectral reflectance of the near-infrared band, and determines the photosynthetic efficiency of the target sea area according to the chlorophyll fluorescence signal; according to the third spectral reflectance ratio and the photosynthetic efficiency, the predicted phytoplankton chlorophyll concentration value corresponding to the marine ecological remote sensing image data sample is output; according to the predicted phytoplankton chlorophyll concentration value and the true phytoplankton chlorophyll concentration value of the marine ecological remote sensing image data sample, the third prediction loss of the third chlorophyll concentration determination sub-model is determined, and the network parameters of the third chlorophyll concentration determination sub-model are adjusted according to the third prediction loss, so as to optimize the prediction accuracy of the next iteration training.

[0088] In the training of the fourth chlorophyll concentration determination sub-model, in each iteration training process, the marine ecological remote sensing image data sample is taken as the input of the fourth chlorophyll concentration determination sub-model, so that the fourth chlorophyll concentration determination sub-model extracts the spectral reflectance of the near-infrared band (i.e. the initial chlorophyll concentration under the spectrum) and the thermal infrared signal of the multispectral remote sensing image sample; the thermal infrared sea surface temperature peak of the target sea area is determined according to the thermal infrared signal, and the predicted phytoplankton chlorophyll concentration value corresponding to the marine ecological remote sensing image data sample is output according to the thermal infrared sea surface temperature peak, the spectral reflectance of the near-infrared band and the preset chlorophyll background value; the fourth prediction loss of the fourth chlorophyll concentration determination sub-model is determined according to the predicted phytoplankton chlorophyll concentration value and the true phytoplankton chlorophyll concentration value of the marine ecological remote sensing image data sample, and the network parameters of the fourth chlorophyll concentration determination sub-model are adjusted according to the fourth prediction loss, so as to optimize the prediction accuracy of the next iteration training.

[0089] After the training of each chlorophyll concentration determination sub-model is completed, the network parameters of the classifier are unfrozen to perform overall fine-tuning of the initial chlorophyll concentration determination sub-model on the training sample set.

[0090] In a preferred embodiment, the training of the classifier and each chlorophyll concentration determination sub-model on the training sample set is performed until the initial chlorophyll concentration determination model converges as a whole, and the chlorophyll concentration determination model is generated, including:

[0091] The marine heat wave data samples and the marine green ecological remote sensing image data samples in the training sample set are grouped based on the marine heat wave level to obtain a training sample subset under each marine heat wave level; wherein the training sample subset includes marine heat wave data samples and marine green ecological remote sensing image data samples under the current marine heat wave level.

[0092] The classifier and the chlorophyll concentration determination sub-model corresponding to the marine heat wave level of each training sample subset are trained until the initial chlorophyll concentration determination model converges as a whole, and the chlorophyll concentration determination model is generated.

[0093] Specifically, the classifier and the chlorophyll concentration determination sub-model corresponding to the marine heat wave level of each training sample subset are iteratively trained, and a joint loss under the joint training of the classifier and each chlorophyll concentration determination sub-model is determined, a loss function of the initial chlorophyll concentration determination model is constructed according to the joint loss, the classification loss, the first prediction loss, the second prediction loss, the third prediction loss and the fourth prediction loss, and the initial chlorophyll concentration determination model converges when the sum of the loss function of the initial chlorophyll concentration determination model is minimized, and the chlorophyll concentration determination model is generated.

[0094] For step S3, after determining the phytoplankton chlorophyll concentration value of the target sea area, the phytoplankton chlorophyll concentration value of the target sea area is transmitted to the marine ecological monitoring system, preferably, the marine heat wave grade, the marine heat wave intensity and the marine heat wave duration of the target sea area are synchronously transmitted to the marine ecological monitoring system, so that the marine ecological monitoring system monitors the target sea area according to the received phytoplankton chlorophyll concentration value, marine heat wave grade, marine heat wave intensity and marine heat wave duration, and timely takes corresponding measures to maintain the fishery culture of the target sea area when the phytoplankton chlorophyll concentration value is lower than the preset threshold.

[0095] On the basis of the above-mentioned method embodiment, the application provides a terminal device embodiment.

[0096] An embodiment of the application provides a terminal device, which comprises a processor, a memory and a computer program stored in the memory and configured to be executed by the processor, and the processor executes the computer program to implement any one of the phytoplankton chlorophyll concentration value determination methods based on marine heat waves.

[0097] The terminal device can be a desktop computer, a notebook computer, a palm computer, a cloud server and the like, and can comprise, but is not limited to, a processor and a memory.

[0098] The processor can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor and the like, and the processor is a control center of the terminal device, and is connected with all parts of the terminal device through various interfaces and lines.

[0099] The memory can be used to store the computer program, and the processor realizes various functions of the terminal device by running or executing the computer program stored in the memory and calling data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required by a function, and the like; and the data storage area can store data created according to the use of the mobile phone and the like. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, for example, a hard disk, a memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state memory devices.

[0100] On the basis of the above-mentioned method embodiment, the application correspondingly provides a storage medium embodiment.

[0101] An embodiment of the application provides a storage medium, which comprises a stored computer program, wherein the computer program controls a device where the storage medium is located to perform a method for determining a phytoplankton chlorophyll concentration value based on a marine heat wave.

[0102] The storage medium is a computer-readable storage medium, and the computer program is stored in the computer-readable storage medium. When the computer program is executed by a processor, the steps of each method embodiment described above can be realized. The computer program comprises computer program code, which can be in a form of source code, object code, an executable file or some intermediate form. The computer-readable medium can include any entity or device capable of carrying the computer program code, a recording medium, a U disk, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal and a software distribution medium, etc.

[0103] The above-mentioned is the preferred embodiment of the application, and it should be pointed out that, for those skilled in the art, without departing from the principle of the application, a number of improvements and refinements can be made, and these improvements and refinements are also regarded as the protection scope of the application.

Claims

1. A method for determining phytoplankton chlorophyll concentration values based on marine heatwaves, characterized by, include: Acquire marine green ecological remote sensing image data and marine heat wave data of the target sea area; wherein the marine heat wave data includes: historical sea surface temperature data within a preset monitoring period, sea surface temperature data for the current monitoring period, and calendar days of the current monitoring period; the marine green ecological remote sensing image data includes: multispectral remote sensing images and marine microwave remote sensing images; Inputting the ocean heat wave data and the ocean green ecology remote sensing image data of the target sea area into a chlorophyll concentration determination model, so that the chlorophyll concentration determination model determines the ocean heat wave level of the target sea area according to the ocean heat wave data, and inputting the ocean green ecology remote sensing image data into a chlorophyll concentration determination sub-model according to the ocean heat wave level, so that the chlorophyll concentration determination sub-model extracts corresponding multispectral remote sensing image features and ocean microwave remote sensing image features, and outputs a phytoplankton chlorophyll concentration value of the target sea area according to the multispectral remote sensing image features and the ocean microwave remote sensing image features; Transmitting the phytoplankton chlorophyll concentration value of the target sea area to the marine ecological monitoring system to assist the marine ecological monitoring system in performing sea area monitoring; The marine heat wave levels include: mild heat wave level, moderate heat wave level, severe heat wave level and extreme heat wave level; Determining the ocean heat wave level of the target sea area according to the ocean heat wave data includes: The historical sea surface temperature data within the preset monitoring period are grouped by calendar day, and the historical sea surface temperature data for each calendar day are obtained: For each calendar day, the average of the historical sea surface temperature data for the current calendar day is calculated to obtain the ocean heat wave threshold for the current calendar day; Construct the ocean heatwave threshold matrix for calendar days throughout the year based on the ocean heatwave threshold for each calendar day; Obtaining, from the ocean heat wave threshold matrix according to the calendar day of the current monitoring period, the ocean heat wave threshold for each calendar day of the current monitoring period; Determine the ocean heatwave intensity and duration of the target sea area during the current monitoring period based on the sea surface temperature data of the current monitoring period and the ocean heatwave thresholds of each calendar day during the current monitoring period, and determine the ocean heatwave level of the target sea area based on the ocean heatwave intensity and duration of the target sea area; The construction of the chlorophyll concentration determination model includes: Acquire a number of ocean heatwave data samples and a number of ocean green ecological remote sensing image data samples for each ocean heatwave level, and construct a training sample set based on each ocean heatwave data sample and each ocean green ecological remote sensing image data sample; wherein each ocean heatwave data sample includes: a historical sea surface temperature data sample within a preset monitoring period, a historical sea surface data sample within a selected monitoring period, and a calendar day of the selected monitoring period; and each ocean green ecological remote sensing image data sample includes: a multispectral remote sensing image sample and an ocean microwave remote sensing image sample; The initial chlorophyll concentration determination model comprises a classifier, a first chlorophyll concentration determination sub-model, a second chlorophyll concentration determination sub-model, a third chlorophyll concentration determination sub-model and a fourth chlorophyll concentration determination sub-model, and the first chlorophyll concentration determination sub-model, the second chlorophyll concentration determination sub-model, the third chlorophyll concentration determination sub-model and the fourth chlorophyll concentration determination sub-model correspond to a mild heat wave grade, a moderate heat wave grade, a severe heat wave grade and an extreme heat wave grade, respectively. The initial chlorophyll concentration determination model is trained by using the training sample set until the initial chlorophyll concentration determination model converges, and the chlorophyll concentration determination model is generated.

2. The method for determining phytoplankton chlorophyll concentration based on ocean heat waves according to claim 1, wherein: The initial chlorophyll concentration determination model is trained by using the training sample set until the initial chlorophyll concentration determination model converges, and the chlorophyll concentration determination model is generated, including: The network parameters of each chlorophyll concentration determination sub-model are frozen, the classifier of the initial chlorophyll concentration determination model is trained by using each marine heat wave data sample until the classifier of the initial chlorophyll concentration determination model converges, the network parameters of the classifier of the initial chlorophyll concentration determination model are frozen, and the network parameters of each chlorophyll concentration determination sub-model are unfrozen. Each chlorophyll concentration determination sub-model is trained by using each marine green ecological remote sensing image data sample until each chlorophyll concentration determination sub-model converges, and the network parameters of the classifier are unfrozen. The classifier and each chlorophyll concentration determination sub-model are trained by using the training sample set until the initial chlorophyll concentration determination model as a whole converges, and the chlorophyll concentration determination model is generated.

3. The method for determining phytoplankton chlorophyll concentration based on ocean heat waves according to claim 2, wherein: The classifier of the initial chlorophyll concentration determination model is trained by using each marine heat wave data sample, including: The marine heat wave data sample is input into the classifier, so that the classifier determines the marine heat wave intensity and the marine heat wave duration corresponding to the current marine heat wave data sample. The predicted marine heat wave grade of the current marine heat wave data sample is output according to the marine heat wave intensity and the marine heat wave duration corresponding to the current marine heat wave data sample.

4. The method for determining phytoplankton chlorophyll concentration based on ocean heat waves according to claim 3, wherein: Each chlorophyll concentration determination sub-model is trained by using each marine green ecological remote sensing image data sample, including: The marine ecological remote sensing image data sample is input into the first chlorophyll concentration determination sub-model, so that the first chlorophyll concentration determination sub-model extracts the spectral reflectivity of a blue light band and the spectral reflectivity of a green light band of the multispectral remote sensing image sample, determines a first spectral reflectivity ratio according to the spectral reflectivity of the blue light band and the spectral reflectivity of the green light band, extracts a first microwave radiation feature of the marine microwave remote sensing image sample, determines a mixed layer depth according to the first microwave radiation feature, and outputs the predicted phytoplankton chlorophyll concentration value corresponding to the marine ecological remote sensing image data sample according to the mixed layer depth and the first spectral reflectivity ratio; The marine ecological remote sensing image data sample is input into the second chlorophyll concentration determination sub-model, so that the second chlorophyll concentration determination sub-model extracts the spectral reflectance of the red light band, the spectral reflectance of the green light band and the chlorophyll fluorescence peak value of the multispectral remote sensing image sample, determines the second spectral reflectance ratio value according to the spectral reflectance of the red light band and the spectral reflectance of the green light band, extracts the second microwave radiation feature of the marine microwave remote sensing image sample, and determines the sea surface height abnormal value according to the second microwave radiation feature; and outputs the predicted phytoplankton chlorophyll concentration value corresponding to the marine ecological remote sensing image data sample according to the second spectral reflectance ratio value, the sea surface height abnormal value and the chlorophyll fluorescence peak value. The marine ecological remote sensing image data sample is input into the third chlorophyll concentration determination sub-model, so that the third chlorophyll concentration determination sub-model extracts the spectral reflectance of the red light band, the spectral reflectance of the near-infrared band and the chlorophyll fluorescence signal of the multispectral remote sensing image sample, determines the third spectral reflectance ratio value according to the spectral reflectance of the red light band and the spectral reflectance of the near-infrared band, and determines the photosynthetic efficiency of the target sea area according to the chlorophyll fluorescence signal; and outputs the predicted phytoplankton chlorophyll concentration value corresponding to the marine ecological remote sensing image data sample according to the third spectral reflectance ratio value and the photosynthetic efficiency. The marine ecological remote sensing image data sample is input into the fourth chlorophyll concentration determination sub-model, so that the fourth chlorophyll concentration determination sub-model extracts the spectral reflectance of the near-infrared band and the thermal infrared signal of the multispectral remote sensing image sample; determines the thermal infrared sea surface temperature peak value of the target sea area according to the thermal infrared signal, and outputs the predicted phytoplankton chlorophyll concentration value corresponding to the marine ecological remote sensing image data sample according to the thermal infrared sea surface temperature peak value, the spectral reflectance of the near-infrared band and a preset chlorophyll background value.

5. The method for determining phytoplankton chlorophyll concentration based on ocean heat waves according to claim 4, wherein: The training of the classifier and the chlorophyll concentration determination sub-models by using the training sample set is performed until the initial chlorophyll concentration determination model converges as a whole, and the chlorophyll concentration determination model is generated, including: The marine thermal wave data samples and the marine green ecological remote sensing image data samples in each training sample set are grouped based on marine thermal wave grades, to obtain training sample subsets under each marine thermal wave grade; wherein the training sample subset includes marine thermal wave data samples and marine green ecological remote sensing image data samples under the current marine thermal wave grade; The classifier and the chlorophyll concentration determination sub-models corresponding to the marine thermal wave grades of each training sample subset are trained by using each training sample subset, until the initial chlorophyll concentration determination model converges as a whole, and the chlorophyll concentration determination model is generated.

6. The method for determining phytoplankton chlorophyll concentration based on ocean heat waves according to claim 5, wherein: The output of the predicted phytoplankton chlorophyll concentration value corresponding to the marine ecological remote sensing image data sample according to the second spectral reflectance ratio value, the sea surface height abnormal value and the chlorophyll fluorescence peak value includes: A first correlation relationship between the second spectral reflectance ratio value and the chlorophyll concentration is constructed, a second correlation relationship between the sea surface height abnormal value and the chlorophyll concentration is constructed, and a third correlation relationship between the chlorophyll fluorescence peak value and the chlorophyll concentration is constructed. A space rectangular coordinate system is established with the first correlation relationship as the horizontal axis, the second correlation relationship as the vertical axis, and the third correlation relationship as the vertical axis; wherein, the plane formed by the horizontal axis and the vertical axis is the first plane, the plane formed by the horizontal axis and the vertical axis is the second plane, and the plane formed by the vertical axis and the vertical axis is the third plane; A first correlation vector between the second spectral reflectance ratio, the sea surface height anomaly value, and the chlorophyll concentration is constructed, a second correlation vector between the second spectral reflectance ratio, the chlorophyll fluorescence peak value, and the chlorophyll concentration is constructed, and a third correlation vector between the sea surface height anomaly value, the chlorophyll fluorescence peak value, and the chlorophyll concentration is constructed; The first correlation vector, the second correlation vector, and the third correlation vector are projected on the first plane, the second plane, and the third plane respectively to obtain a first projection of the first correlation vector on the first plane, a second projection on the second plane, and a third projection on the third plane; a fourth projection of the second correlation vector on the first plane, a fifth projection on the second plane, and a sixth projection on the second plane; and a seventh projection of the third correlation vector on the first plane, an eighth projection on the second plane, and a ninth projection on the third plane; The first projection, the fourth projection, and the seventh projection are superimposed to obtain a first superimposed projection, the second projection, the fifth projection, and the eighth projection are superimposed to obtain a second superimposed projection, and the third projection, the sixth projection, and the ninth projection are superimposed to obtain a third superimposed projection; the first superimposed projection, the second superimposed projection, and the third superimposed projection are converted into vector form to obtain a first superimposed vector, a second superimposed vector, and a third superimposed vector; and the marine ecological remote sensing image data sample corresponding to the predicted phytoplankton chlorophyll concentration value is output through the first superimposed vector, the second superimposed vector, and the third superimposed vector.

7. A terminal device, characterized by, The storage medium comprises a stored computer program, wherein when the computer program is running, the device where the storage medium is located is controlled to execute the marine heat wave-based phytoplankton chlorophyll concentration value determination method according to any one of claims 1 to 6.

8. A storage medium, characterized by The storage medium comprises a stored computer program, wherein when the computer program is running, the device where the storage medium is located is controlled to execute the marine heat wave-based phytoplankton chlorophyll concentration value determination method according to any one of claims 1 to 6.

Images