A method for predicting unit catch of fisheries based on space-borne lidar data

Abstract

This application discloses a method for predicting fishery unit catch based on spaceborne lidar data. The method is based on chlorophyll data for the Atlantic region retrieved from spaceborne lidar and a generalized linear prediction model. The method primarily includes the following steps: Step 1: Calculating historical fishery unit catch data; Step 2: Preprocessing CALIPSO lidar data; Step 3: Inverting chlorophyll from CALIPSO lidar data using an ANN model in combination with MODIS data; Step 4: Extracting sea surface temperature data; Step 5: Temporally and spatially matching the fishery unit catch data, the retrieved chlorophyll data, and the sea surface temperature data; and Step 6: Constructing a generalized linear prediction model for fishery unit catch. This application provides a fast, effective, and more comprehensive technology for predicting fishery unit catch data, improving standardized methods for fishery resource assessment.

Description

Technical Field

[0001] The present invention relates to a prediction method for a satellite remote sensing fishery resource assessment application, and in particular to a method for predicting the unit catch of fisheries based on satellite-borne laser radar remote sensing data. Background Art

[0002] Unit catch is a crucial metric for fishery resource assessment and management. While a linear relationship between unit catch and abundance is often assumed during assessments, commercial production data are strongly influenced by temporal, spatial, environmental, and fishing capacity factors. To ensure more reliable and representative unit catch data are used during assessments, it is necessary to standardize nominal unit catch data using statistical models to improve data quality and the predictability of consistent and stable models, providing better support for fishery resource assessment and management. Marine environmental data obtained from satellite remote sensing are often crucial for fishery efficiency assessment and standardization.

[0003] Marine scientists began studying fishery data standardization at the end of the last century. Among unit catch standardization models, the generalized linear model (GLM) is the primary and most commonly used. GLMs assume a linear relationship between the response variable and the explanatory variables. Their application relies on simple model manipulation for calculations, making them a good choice for unit catch standardization in fishery resource assessments.

[0004] Currently, remote sensing data for fishery resource assessments worldwide mainly come from passive remote sensing technology, such as the Moderate Resolution Imaging Spectroradiometer (MODIS). However, due to various factors such as low spatial resolution, sunlight angle and sunlight duration, passive remote sensing cannot obtain high-resolution chlorophyll data at night and in high latitudes. In addition, passive remote sensing can only detect information on environmental elements on the ocean surface or in the shallow ocean layer, and cannot ascertain the structural data of the spatial distribution of elements within the ocean. In contrast, lidar, as an active remote sensing technology, can characterize the vertical structure of water bodies with high spatial resolution under favorable weather conditions, with fast speed and high resolution. In addition, lidar has a wide range of applications, such as (1) fish detection, (2) plankton layer detection, (3) water depth measurement, (4) bubble detection in water and ocean surface roughness measurement. According to the literature, the cloud aerosol lidar launched by NASA and the remote sensing images provided by the infrared pathfinding satellite CALIPSO support the description of the vertical structure of plankton near the sea surface. The orthogonally polarized cloud-aerosol lidar (CALIOP) onboard CALIPSO is the first dual-polarization lidar, providing a global vertical profile of elastic backscatter all day. However, there are currently no reports on using this technology to standardize marine fishery stock assessments. Summary of the Invention

[0005] In response to the shortcomings of existing technologies, the present invention utilizes active remote sensing concepts to provide a method for predicting unit catch of fisheries based on satellite-borne lidar data. This method uses a generalized linear prediction model (GLM) based on chlorophyll data inverted from satellite-borne lidar data to predict the unit catch of a certain fishery product. Passive remote sensing technology has been improved with the supplementation of satellite-borne lidar remote sensing data, enhancing understanding of diurnal variations in the upper ocean. This method also fills the gap in current nighttime passive remote sensing observation data in the equatorial region, providing a fast, effective, and more comprehensive technology for predicting unit catch of fisheries, and improving standardized means of fishery resource assessment.

[0006] The purpose of the present invention is achieved through the following technical solutions:

[0007] A method for predicting unit catch of fisheries based on space-borne laser radar data, characterized in that the method comprises the following steps:

[0008] Step 1: Calculate the catch data of fisheries per year;

[0009] Step 2: CALIPSO lidar data preprocessing;

[0010] Step 3: Combine MODIS data with the ANN model to invert chlorophyll from CALIPSO lidar data;

[0011] Step 4: Extract sea surface temperature data;

[0012] Step 5: Temporal and spatial matching of fishery unit catch data, inverted chlorophyll data, and sea surface temperature data;

[0013] Step 6: Construct a generalized linear prediction model (GLM) for fishery unit catch.

[0014] In practical application of this application, time, longitude, latitude, environmental factor chlorophyll and sea surface temperature data are input into the GLM model, and the predicted unit catch of fisheries is output through the GLM model.

[0015] Furthermore, the fishery unit catch mentioned in step 1 is defined as the catch per hour, year i, month j, longitude k, and latitude l, where the longitude × latitude resolution is 1° × 1°. The nominal unit catch corresponding to each 1° × 1° longitude and latitude grid is calculated as follows:

[0016]

[0017] Where: ∑Catch i,j,k,l is the total catch in year i, month j, longitude k, and latitude l, ∑E i,j,k,l The corresponding working time.

[0018] Furthermore, the CALIPSO lidar data preprocessing described in step 2 includes correcting the CALIPSO measurement signal:

[0019] β′(z)=[F] -1 β(z) (2)

[0020] Where: β′(z) is the true backscattered signal, β(z) is the output signal of the receiver, and [F] is the matrix form of the transient function;

[0021] Calculate the transient function [F]:

[0022]

[0023] Eliminate the crosstalk effects of parallel (‖) and perpendicular (⊥) signals:

[0024]

[0025] β ⊥,c =β ⊥,m -CT×β ||,c (5)

[0026] Where: CT is crosstalk, β ⊥,m , β ⊥,c are the measured and true vertical signals, β ||,m , β ||,c are the measured and true parallel signals, respectively.

[0027] After eliminating the effects of transient response and crosstalk in the CALIPSO receiver subsystem, the particle backscattering coefficient (b bp ). First, considering the influence of the atmosphere, the vertical parallel ratio is used to calculate the underwater column integrated backscattering of the vertical component:

[0028]

[0029] β W+ is the vertical component of the underwater column integrated backscatter, where: T is the total depolarization ratio, δ w is the column-integrated surface depolarization rate, β S is the lidar surface backscatter.

[0030] Then, calculate the particle backscattering coefficient at a scattering angle of 180°:

[0031]

[0032] Among them: K d is the ocean diffusion attenuation coefficient, t is the ocean surface transmittance, δp is the particle depolarization rate.

[0033] Furthermore, the chlorophyll inversion of CALIPSO lidar data using the ANN model in combination with MODIS data as described in step 3 is as follows:

[0034] According to b(π) and b bp Calculate the relationship between b bp :

[0035]

[0036] According to MODIS chlorophyll and CALIPSO b bp The relationship between is modeled using an artificial neural network (ANN) and chlorophyll is inverted. The overall ANN model is given by the following formula:

[0037]

[0038] Where: w j,i represents the weight of the connection from node j to i, o is the output node, and f is the logic function

[0039] Furthermore, the spatial resolution of the fishery unit catch data, marine environmental factor chlorophyll data, and sea surface temperature data in this application is consistent (the same), and is all based on a 1°×1° (longitude×latitude) resolution scale.

[0040] Furthermore, the generalized linear prediction model (GLM) of unit catch of bigeye tuna is constructed as described in step 6.

[0041] The data source for model construction uses the unit catch dataset that has been successfully matched in step 5 above;

[0042] The generalized linear prediction model GLM equation is described as follows:

[0043]

[0044] μ i =E(Y i ) (11)

[0045] Where: g is the link function, X i is the explanatory variable of the i-th response variable, Y i is the i-th random variable and β is the vector of parameters.

[0046] Assuming that the unit catch follows a log-normal distribution, the GLM model is expressed as:

[0047] Ln(CPUE i,j,k,l +1)=k+α1yeari +α2month i +α3lon i +α4lat i +α5SST i +

[0048] α6chl_a i +α1interactions+ε i,j,k,l (12)

[0049] Where: α1~α6 are model parameters; ε is the residual, which is assumed to have a normal distribution.

[0050] In actual application, the predicted time, longitude, latitude, marine environmental factor chlorophyll and sea surface temperature data are input into the GLM model, and the predicted unit catch of fisheries during the day and night is output through the GLM model.

[0051] Beneficial effects

[0052] The present invention has the beneficial effects of replacing traditional passive remote sensing data with chlorophyll data derived from active remote sensing, providing rapid modeling and prediction using generalized linear models. This data, acquired through nighttime active remote sensing, enhances understanding of diurnal variations in the upper ocean, while also filling the gap in passive nighttime remote sensing data for the equatorial region. This provides a fast, effective, and more comprehensive technology for predicting unit catches in fisheries, improving standardized methods for fishery resource assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 It is a basic framework diagram of an embodiment of the present application;

[0054] Figure 2 This is a schematic block diagram of a specific application of an embodiment of the present application;

[0055] Figure 3 This is a flow chart for producing chlorophyll data from spaceborne lidar data;

[0056] Figure 4 This is the distribution map of the predicted unit catch of bigeye tuna using the existing traditional method;

[0057] Figure 5 This is a distribution map of unit catch prediction of bigeye tuna using the method of this application. DETAILED DESCRIPTION

[0058] The present invention will be further described in detail below with reference to the accompanying drawings and examples, which will make its objectives and effects more apparent. It should be understood that the specific embodiments described herein are intended solely to illustrate the present invention and are not intended to limit it. This example utilizes CALIPSO Level 1B V4.10 satellite-borne lidar product data, supplemented by MODIS-Aqua Level 3 chlorophyll data and sea surface temperature data to invert equatorial diurnal chlorophyll data. The fishery product is specifically Atlantic bigeye tuna.

[0059] Figure 1 The basic framework diagram for the method of the present invention is as follows

[0060] The specific method for predicting fishery unit catch is as follows:

[0061] Step 1: Obtain and calculate the unit catch data of bigeye tuna in previous years;

[0062] This example uses Atlantic bigeye tuna statistics published by the International Commission for the Conservation of Atlantic Bigeye Tuna (ICCAT). Production statistics include data from 2008 to 2015, including fishing date, location (longitude and latitude), catch, and operation time. The spatial resolution of production data statistics is 1°×1°. The unit catch of bigeye tuna is defined as the catch (tons) per hour, year i, month j, longitude k, and latitude l. The corresponding nominal unit catch in each 1°×1° grid is calculated as follows:

[0063]

[0064] Where: ∑Catch i,j,k,l is the total catch in year i, month j, longitude k, and latitude l, ∑E i,j,k,l The corresponding working time.

[0065] Step 2: CALIPSO lidar data preprocessing;

[0066] (1) Correction of CALIPSO measurement signals:

[0067] Correction of CALIPSO measurement signals:

[0068] β′(z)=[F] -1 β(z) (2)

[0069] Where: β′(z) is the true backscattered signal, β(z) is the output signal of the receiver, and [F] is the matrix form of the transient function;

[0070] (2) Calculate the transient function [F]:

[0071]

[0072] (3) Eliminate the crosstalk effects of parallel (‖) signals and vertical (⊥) signals:

[0073]

[0074] β ⊥,c =β ⊥,m -CT×β ||,c (5)

[0075] Where: CT is crosstalk, β ⊥,m , β ⊥,c are the measured and true vertical signals, β ||,m , β ||,c are the measured and true parallel signals, respectively.

[0076] (4) Calculate the vertical component of the underwater column integrated backscatter using the vertical-parallel ratio:

[0077]

[0078] Where: β W+ is the vertical component of the underwater column integrated backscatter, δ T is the total depolarization ratio, δ w is the column-integrated surface depolarization rate, β S is the lidar surface backscatter.

[0079] (5) Calculate the particle backscattering coefficient at a scattering angle of 180°:

[0080]

[0081] Among them: K d is the ocean diffusion attenuation coefficient, t is the ocean surface transmittance, δ p is the particle depolarization rate.

[0082] According to b(π) and b bp Calculate the relationship between b bp :

[0083]

[0084] Step 3: Combine MODIS data with the ANN model to invert chlorophyll from CALIPSO lidar data;

[0085] This implementation uses the MATLAB neural network ANN toolbox. The structure of the ANN toolbox includes an input layer, a hidden layer, and an output layer. The network is a two-layer feedforward network. The hidden layer has a sigmoid transfer function and the output layer has a linear transfer function. The overall ANN model is given by the following formula

[0086]

[0087] Where: w j,i represents the weight of the connection from node j to i, o is the output node, and f is the logic function

[0088] Step 4: Extract sea surface temperature data;

[0089] Step 5: Spatiotemporally matching the bigeye tuna unit catch data with the inverted chlorophyll data and sea surface temperature data; wherein the spatial resolution of the matching of the bigeye tuna unit catch data with the marine environmental factor chlorophyll and sea surface temperature data is based on a 1°×1° (longitude×latitude) resolution scale;

[0090] Step 6: Construct a generalized linear prediction model (GLM) for bigeye tuna unit catch. The characteristics are:

[0091] The data source for model construction uses the unit catch dataset that has been successfully matched in step 5 above;

[0092] The GLM equation is described as follows:

[0093]

[0094] μ i =E(Y i ) (11)

[0095] Where: g is the link function, X i is the explanatory variable of the i-th response variable, Y i is the i-th random variable and β is the vector of parameters.

[0096] Assuming that the unit catch follows a lognormal distribution, the GLM is expressed as:

[0097]

[0098] Where: α1~α6 are model parameters; ε is the residual, which is assumed to have a normal distribution.

[0099] The ratio of model training data is set as follows: 80% for training data and 20% for validation data.

[0100] Step 1 of this application can be completed before step 5 where the data is used.

[0101] Figure 2This is a schematic diagram for a specific application. In actual application, you only need to input the year, month, longitude, latitude, chlorophyll and sea surface temperature data of the predicted bigeye tuna data into the GLM model, and the model will output the unit catch data of bigeye tuna;

[0102] Figure 3 Flowchart for producing chlorophyll data from spaceborne lidar data;

[0103] Figure 4 This is the existing traditional method for predicting the distribution of bigeye tuna unit catch, which only has data for prediction during the day when there is chlorophyll data;

[0104] Figure 5 The following are the predicted distribution maps of bigeye tuna catches using the method presented in this invention. The left figure shows the predicted distribution map of bigeye tuna catches during the day, and the right figure shows the predicted distribution map of bigeye tuna catches during the night. Traditional methods provide no nighttime catch data for bigeye tuna, and similarly, no catch predictions can be obtained during the day if lighting conditions are insufficient to obtain chlorophyll data. However, using chlorophyll data retrieved from spaceborne lidar and the GLM model effectively predicts both daytime and nighttime catches of bigeye tuna. Using this method, the predicted catches are higher than those predicted by traditional methods, and the discrepancy with actual catches is smaller. Furthermore, the nighttime catches are higher than the daytime catches, consistent with the diurnal vertical migration of bigeye tuna, which sinks during the day and floats at night. The predicted catches of bigeye tuna using this method are more clustered than those predicted by traditional methods, primarily concentrated near the Gulf of Guinea, consistent with the schooling habits of bigeye tuna.

Claims

1. A method for predicting unit catch in fisheries based on spaceborne lidar data, the characteristics of which include: Step 1: Calculate the catch data of fisheries per year; Step 2: CALIPSO lidar data preprocessing; CALIPSO lidar data preprocessing includes: (1) Correction of CALIPSO measurement signals: Where: ′ (z) is the true backscattered signal, β(z) is the output signal of the receiver, and [F] is the matrix form of the transient function; (2) Calculate the transient function [F]: (3) Eliminate the crosstalk effects of parallel (||) signals and vertical (⊥) signals: b ⊥,c =b ⊥,m -CT×β ||,c (5) Where: CT is crosstalk, β ⊥,m , β ⊥c are the measured and true vertical signals, β ||,m , β ||,c are the measured and true parallel signals, respectively; (4) Calculate the vertical component of the underwater column integrated backscatter using the vertical-parallel ratio: Where: W+ is the vertical component of the underwater column integrated backscatter, δ T is the total depolarization ratio, δ w is the column-integrated surface depolarization rate, β S is the lidar surface backscatter; (5) Calculate the particle backscattering coefficient at a scattering angle of 180°: Among them: K d is the ocean diffusion attenuation coefficient, β W+ is the vertical component of the underwater column integrated backscatter, t is the ocean surface transmittance, δ p is the particle depolarization rate; Step 3: Combine MODIS data with the ANN model to invert chlorophyll from CALIPSO lidar data; The step three includes: (1) According to b(π) and b bp Calculate the relationship between b bp : (2) Use artificial neural network ANN to model and invert chlorophyll: According to MODIS chlorophyll and CALIPSO b bp The relationship between the two, the ANN overall model is given by the following formula: Where: w j,i represents the weight of the connection from node j to i, o is the output node, and f is the logic function Step 4: Extract sea surface temperature data; Step 5: Temporal and spatial matching of fishery unit catch data, inverted chlorophyll data, and sea surface temperature data; Step 6: Construct a generalized linear prediction model for fishery unit catch.

2. The method for predicting unit catch of fisheries based on space-borne lidar data according to claim 1, characterized in that: The fishery unit catch in step 1 is defined as the production per hour, year i, month j, longitude k, and latitude l, where the longitude × latitude resolution is 1° × 1°. The nominal unit catch corresponding to each 1° × 1° longitude and latitude grid is calculated as follows:

3. The method for predicting unit catch of fisheries based on space-borne laser radar data according to claim 1, characterized in that: The temporal and spatial resolutions of the fishery unit catch data, marine environmental factor chlorophyll data, and sea surface temperature data are consistent.

4. The method for predicting unit catch of fisheries based on space-borne laser radar data according to claim 1, characterized in that: The step six comprises: (1) The data source for constructing the model uses the unit catch dataset that has been successfully matched in step 5; (2) The GLM equation is described as follows: m i =E(Y i ) (11) Where: g is the link function, X i is the explanatory variable of the i-th response variable, Y i is the i-th random variable, β is the vector of parameters; (3) Assuming that the unit catch follows a log-normal distribution, the GLM model is expressed as: Ln(CPUE i,j,k,l +1 =k+α1year i +a2month i +α3lon i +α4lat i +α5SST i α6chl_a i +α1interactions+e i,j,k,l (12) Where: α1~α6 are model parameters; ε is the residual, which is assumed to have a normal distribution.

Images