WO2018001338A1

WO2018001338A1 - Grey system-based pelagic squid resource richness forecasting method

Info

Publication number: WO2018001338A1
Application number: PCT/CN2017/090934
Authority: WO
Inventors: 陈新军; 汪金涛; 雷林; 余为
Original assignee: 上海海洋大学
Priority date: 2016-06-30
Filing date: 2017-06-29
Publication date: 2018-01-04
Also published as: CN106203686A; JP2019511783A; JP6630846B2

Abstract

A grey system-based pelagic squid resource richness forecasting method, characterized in comprising the following steps: step one: utilizing a method for grey relational analysis, calculating to acquire the grey correlation magnitudes of impact factors of pelagic squid resource richness; step two: selecting an impact factor having a high grey correlation as a factor for a resource richness forecast model; step three: utilizing a discrete GM model, employing the factor selected in step two to establish pelagic squid resource richness forecast models; step four: performing an effectiveness analysis with respect to the forecast models of step three, the effectiveness analysis comprising relative error and correlation coefficient analyses, and selecting the model of least relative error and greatest correlation coefficient as the most suitable forecast model. The method employs a grey system for forecasting squid resource richness to achieve an accuracy of 90% or more, and provides a reference value for fishery production practices and scientific management.

Description

Method for predicting abundance of large-scale squid resources based on grey system

Technical field

The invention relates to a method for predicting resource abundance of short life cycle species, in particular to a method for predicting abundance of large-scale squid resource based on gray system.

Background technique

The fishery forecast is a key link in fishery production. The forecasting of the resource abundance of the winter spring group of the northwest Pacific squid (Ommastrephes bartramii) is conducive to the scientific arrangement and scientific arrangement of the northwestern Pacific squid winter spring group. The general life cycle of squid is one year, and its spawning period is 1-4 months. The success of spawning in the current year directly determines the size of its supplement and determines the abundance of resources, and the environment of the spawning ground is good. Bad will inevitably lead to changes in its resources. External environmental factors include Sea surface temperature (SST), Chlorophyll-a concentration (Chl-a), El-nino index, and Pacific Decadal Oscillation (PDO). .

The Illus argentinus is an oceanic shallow sea species with short life span and rapid growth. The whole population is composed of almost a single generation and died after spawning. Usually distributed in the area of 22 ° ~ 54 ° S Patagonia continental shelf and continental slope 50 ~ 1000m water depth, especially 35 ° ~ 52 ° S, is currently one of the world's most important cephalopods, but also the southwest Atlantic An important part of the squid fishing industry.

The fishery forecast is the focus of fishery research. Accurate fishery forecast can guide enterprises to arrange fishery production reasonably, shorten the time for finding fishery, reduce costs and increase catch production. Generally, fishery forecasts include fishery forecasts and resource abundance forecasts, and the present invention relates to forecasting of resource abundance. In the 1950s, China began to fish forecasting the main economic fish species in the offshore area, and accumulated rich experience. Since the 1980s, the development of geographic information systems has provided strong research on fishery analysis and fishery forecasting. Large analytical tools. The popularity of marine satellite remote sensing technology enables people to quickly obtain large-scale sea state information. Real-time ship position monitoring and maritime satellite communication technology enable ocean-going fishing vessels to effectively receive real-time forecasts from fishery forecasting agencies. These high-tech developments are developed to develop efficient fishing. The situation forecast provides support. However, in the current resource abundance forecast, people often use a linear factor that affects a certain type of forecast, or several environmental factors, to use a linear model for forecasting. Although these forecasts have achieved good forecast results, they are usually forecasted. The accuracy is about 70%, but since the marine ecosystem is usually nonlinear, the influence factors interact and interact with each other, and the linear model may not be well simulated; in addition, the linear model often takes a long time to statistically. Sequences, in general, have more than 30 samples in order to meet statistical requirements. These conditions and requirements constrain the improvement of the accuracy of traditional resource abundance prediction models.

The prediction is actually to use the discussion of the past to speculate and understand the future development trend. The grey system theory is a new and erected discipline founded by Professor Deng Julong, a famous scholar in China in 1982. It is based on the "small sample" and "poor information" uncertain system of "some information is known, some information is unknown". Through the generation and development of "partial" known information, valuable information is extracted to achieve effective control of system operation. Grey prediction, through the processing of raw data and the establishment of gray model, discovers and masters the law of system development, and makes scientific quantitative prediction of the future state of the system. However, the same prediction can use different grey prediction models. Therefore, it is appropriate and accurate to choose which structure of the grey prediction model is needed. Choosing the correct forecasting model can effectively improve the production efficiency of fishing vessels and provide a reference for the company's annual planning.

Summary of the invention

The technical problem to be solved by the present invention is to provide a method for predicting the abundance of large-scale squid resources based on the gray system, and to study the understanding of the marine environmental factors for the replenishment of oceanic economic soft fish resources. The impact is to find out the marine environmental factors that have the most significant impact on the replenishment of oceanic economic soft fish resources for medium and long-term fishery forecasting.

Technical solutions

A method for predicting the abundance of large-scale squid resource based on gray system, which comprises the following steps:

Step 1: Using the grey correlation analysis method, calculate the gray correlation degree of the influence factor of the resource abundance of the oceanic squid;

Step 2: Select a factor with a large gray correlation degree as a factor of the resource abundance prediction model;

Step 3: Using the discrete GM model, the factor selected in step 2 is used to establish a prediction model for the abundance of oceanic carp.

Step 4: Perform a validity analysis on the predictive model of step three. The effectiveness analysis includes relative error and correlation coefficient analysis. The relative error is calculated by using the factor data to calculate the CPUE value and comparing with the real CPUE value to obtain the absolute value. Correlation coefficient analysis The correlation coefficient between the simulated CPUE value sequence and the actual CPUE value sequence obtained by the model in step 3 is selected, and the model with the smallest relative error and the largest correlation coefficient is selected as the most suitable prediction model.

For the northwest Pacific winter spring squid, the factors affecting the grey correlation degree include sea surface temperature SST, PDO, nino3.4 anomaly and chlorophyll concentration chl-a.

The resource abundance prediction for the northwest Pacific winter spring squid includes the following steps: (1) obtaining nino3.4 anomalies, PDO data, sea surface temperature SST and chlorophyll concentration chl-a through remote sensing satellites, 4 marine environments and climatic factors (2) Grey correlation analysis of four marine environments and climatic factors, that is, the CPUE of the current year is the mother sequence, and the spawning field environmental indicators and climate indicators for each month of the annual spawning time are subsequences. Calculate the absolute degree of gray correlation between each subsequence and the parent sequence, and evaluate the importance of each index by the magnitude of the gray absolute correlation degree; (3) the knot based on the gray correlation analysis The four factors with the highest correlation degree were: the average sea surface temperature SST of the spawning ground in March, the PDO of the Pacific Ocean inter-annual oscillation index in January, the nino3.4 anomaly of April, and the average chlorophyll concentration of chl-a in April; (4) Based on the selected four factors, eight grey-system-based resource abundance prediction models for the northwest Pacific squid winter spring population were established: GM(O,1), GM(O,2), GM(O,3) , GM(O,4), GM(1,1), GM(1,2), GM(1,3), GM(1,4); (5) Validation of eight prediction models, model The GM(1,4) structure is the most suitable predictive model for the abundance of winter spring resources of the Pacific Northwest squid.

For the Atlantic Atlantic squid in the Southwest Atlantic, the factors affecting the gray correlation include the SST of 45°~66°W and 32°~43°S in each month of June, July and August.

The resource abundance prediction of the squid in the southwest Atlantic includes the following steps: 1) Calculate the gray correlation degree between the subsurface sequence of the marine surface temperature in the spawning sea area from June to August and the CPUE parent sequence of the next year, and select the spawning ground. The SST of the grey correlation degree greater than 0.9 in the sea is the influencing factor of the resource abundance prediction model of the squid squid in the southwest Atlantic; 2) the GM(0,N) grey prediction models are constructed, respectively: GM (0, 3) The model, which contains two factors, the average of the SSTs of all points where the gray absolute correlation is greater than 0.90 for each month in June and July; the GM (0,3) model, which contains two factors, June and August each month The average of the SSTs of all points where the gray absolute correlation is greater than 0.90; the GM (0,3) model, which contains two factors, the average of the SSTs of all points where the gray absolute correlation is greater than 0.90 for each month in July and August; The GM(0,4) model consists of three factors, including the average of the SSTs of all points with gray absolute correlations greater than 0.90 for each month from June to August; 3) validity test for four forecast models, model GM (0,4) structure is the most suitable forecast for the abundance of squid squid in the southwest Atlantic model.

Beneficial effect

The invention uses the gray correlation analysis and the gray prediction modeling method in the grey system theory to analyze the environmental factors affecting the abundance of the soft fish in the spawning period and the spawning sea area, thereby obtaining the importance evaluation of each index and selecting The most relevant impact factor, and use this to establish a grey prediction model, and Comparing the forecasting accuracy, the most suitable discrete model structure is selected as the prediction model of resource abundance. The prediction model obtained by this method can predict the resource abundance of oceanic carp more than 90%, which is better than the prediction accuracy of traditional forecasting model. (70%) has a significant improvement.

DRAWINGS

Figure 1 is a comparison and comparison of four abundance models of Argentine squid resource abundance.

detailed description

The invention is further illustrated below in conjunction with the specific embodiments and the accompanying drawings.

Example 1

Squid is one of the important economic cephalopods in the Pacific Northwest. It can usually be divided into winter spring group and autumn group, and winter and spring group is the main fishing object of China's offshore fishing. Aiming at the method for predicting the abundance of winter squid in the northwest Pacific squid, the statistical data of the Pacific Northwest squid production and the environmental data of the spawning field are used, including the following steps:

(1) Obtaining four oceanic environmental and climatic factors such as nino3.4 anomaly, PDO data, sea surface temperature SST and chlorophyll concentration chl-a through remote sensing satellites;

(2) Grey correlation analysis of four marine environments and climatic factors, that is, the CPUE of the current year is taken as the parent sequence, and the spawning field environmental indicators and climate indicators of each month should be sub-sequences. The gray absolute degree of association between each subsequence and the parent sequence is evaluated by the importance of the gray absolute correlation degree; the gray absolute correlation degree is calculated as follows:

It is assumed that there are a mother sequence X0 and subsequences X1, X2, X3, X4 and Xs, and the absolute degree of association between the parent sequence and the subsequences is obtained.

The first step: to start zeroing

by

Can be obtained;

Similarly, other zero points can be obtained, as follows

Step 2: Find |s ₀ |, |s ₁ | and |s _i -s ₀ |

The third step: seeking absolute relevance

The same can be obtained separately:

ε ₀₂ =0.50; ε ₀₃ =0.50; ε ₀₄ =0.54; ε ₀₅ =0.63

(3) According to the results of the grey correlation analysis, the four factors with the highest degree of relevance are selected: the average sea surface temperature SST of the spawning ground in March, the Pacific inter-annual oscillation index PD0 in January, the nino3.4 anomaly in April, and the average chlorophyll in April. Concentration chl-a;

(4) Based on the selected four factors, eight kinds of gray-scale system-based northwest Pacific squid winter spring The population resource abundance prediction models are: GM(0,1), GM(0,2), GM(0,3), GM(0,4), GM(1,1), GM(1, 2), GM (1, 3), GM (1, 4); GM (1, 1) model as an example for analysis.

A GM (1,1) gray sequence prediction model was established for marine fishing production from 1983 to 1994. The specific calculation is as follows:

The output was selected from 1983 to 1994 as the original data column.

X(0)=(46.54,52.50,53.20,59.94,71.76,80.98,89.93,103.27,113.84,138.40,155.57,160.82)

One cumulative generation sequence is

X(1)=(46.54,99.04,152.24,212.18,283,94,364.92,454.85,558.12,671.96,810.36,965.93,ll26.75)

Constructing the cumulative matrix B and the constant term vector Yn

Gray parameter a’

a'=(B ^T B) ^-1 B ^T Y _n ,

It is calculated that a = one 0.1202055, u = 41.3759.

Substitute the corresponding equation of time and restore it to:

X'(t+1)=390.7498e ^0.1202055t -344.2097

Calculation results and test

From the calculation results and the back-test of the following table, the average relative error between the actual value of the above model and the fitted value is 2.19%. It indicates that the accuracy and reliability of the calculation results are high, so it is considered that the gray sequence prediction model is applicable to the prediction of marine fishing production.

Calculation and Test of Yield from 1983 to 1994 by Grey Prediction Model

(5) Validation of eight forecast models, GM(O,1), GM(O,2), GM(O,3), GM(O,4), GM(1,1), GM( The accuracy of the 1,2) and GM(1,3) model predictions is between 60-80%, and the average accuracy is 75%. The relative accuracy of the GM(1,4) model is above 90%. Therefore, the gray correlation model GM(1,4) structure is selected as the prediction method for the resource abundance of the northwest Pacific squid winter spring population.

Example 2

The study uses the unit catching effort (CPUE) to characterize the west. The abundance of squid in the South Atlantic Argentina is calculated as:

CPUE=C/B

Where C is the annual catch (t), B is the annual fishing total (vessel), and the unit of CPUE is t·vessel ^-1 .

Generally, the spawning ground of the Atlantic Atlantic squid in the Southwest Atlantic is 40°W～65°W, 30°S～45°S, and the spawning month of Argentine squid is June-August, due to the Argentinian squid. The growth cycle is short, generally an annual fish species. Therefore, the gray correlation degree between the time series value of the logarithm of the next year's logarithm and the time series of the SST of the spawning field at 1°×1° is calculated, and the spawning ground is selected. The SST with a gray correlation degree greater than 0.9 in the sea is the influencing factor of the abundance prediction model of the squid in the southwest Atlantic.

The gray absolute correlation analysis results of the marine surface temperature subsequences at each point in the spawning sea area from June to August and the CPUE parent sequence in the next year are shown in Table 1 below. In June, the gray absolute correlation degree is greater than 0.90, and there are 25 points in July. There are 10 points in the point where the gray absolute correlation is greater than 0.90, and 7 points in the gray absolute correlation degree greater than 0.90 in August. Therefore, these points with a large gray absolute correlation are selected as the environmental factors for establishing the model.

Table 1 gray absolute correlation analysis results

The resource abundance of Argentine squid is predicted by the discrete GM(0,N) model, and the point with higher gray absolute correlation between the sequence value of the selected SST and the sequence value of the next year's CPUE is used as the point. Factor, construct a variety of GM (0, N) gray prediction models. They are:

Scenario 1: The GM (0,3) model, which contains two factors, the average of the SSTs of all points where the gray absolute correlation is greater than 0.90 for each month in June and July.

Scenario 2: The GM (0,3) model, which contains two factors, the average of the SSTs of all points where the gray absolute correlation is greater than 0.90 for each month in June and August.

Option 3: GM (0,3) model, containing two factors, the gray absolute correlation degree in July and August The average of the SST at all points of 0.90.

Scenario 4: The GM(0,4) model consists of three factors, including the average of the SSTs of all points with gray absolute correlations greater than 0.90 for each month from June to August.

After the prediction model is established, the model needs to be analyzed for effectiveness. That is, the evaluation of the model prediction results is divided into two aspects: relative error and correlation analysis. 1) Relative error: Firstly, the CPUE value is calculated by using the factor data, and compared with the real CPUE value, so that the absolute value of the relative error is obtained, and the relative error of all factors is compared. In the process of model construction, data on the abundance of Argentine squid in the Southwest Atlantic Ocean from 2000 to 2015 was used as model construction. In 2016, the abundance data of Argentine squid resource was used for model verification, and the relative error was absolute. Value to evaluate the quality of the model. 2) Correlation coefficient, calculate the correlation coefficient between the simulated CPUE value sequence and the actual CPUE value sequence. The larger the correlation coefficient, the better the prediction effect of the model.

See Figure 1 for the results of the CPUE predicted by the model and Table 2 and Table 3 for the model effect analysis. From the relative error, the GM(0,4) model, which contains the average of the SSTs of all points with gray absolute correlations greater than 0.90 for each month from June to August, is better than the models constructed by other months. . From the results of the 2016 model verification, the prediction accuracy of Model 4 is also the highest, and the relative error is only 0.04, or 4%, followed by Model 1, Model 3 and Model 2. The correlation coefficients between the CPUE sequence and the real value sequence obtained from the prediction model of the Atlantic Atlantic squid in the Southwest Atlantic are model 4 (0.73), model 1 (0.72), model 3 (0.70), and model 2 (0.15). From the different aspects of model validity, the importance of factors is different. However, from the overall perspective, the error between the CPUE and its true value predicted by Model 4 is relatively small, and the correlation coefficient between the predicted CPUE sequence value and the true value is also large. Model 4 can be used as a prediction. The most accurate model to predict the resource abundance of Argentine squid in the Southwest Atlantic.

Table 2 Correlation coefficients of the predictive model of Argentine squid

Table 3 Relative error of Argentine squid prediction model

In this study, the SST sequence values at various points of 1°×1° in the study area were taken as environmental factors, and the resource abundance of Argentinian squid in the Southwest Atlantic was predicted by grey correlation analysis and grey model prediction. The results predicted by the grey model indicate (Table 3) that the environmental factor contains the model of the average of the SSTs of all points with gray absolute correlations greater than 0.90 for each month from June to August, ie the accuracy of model 4 is the highest, the model for 2016 The prediction accuracy is over 95%. However, we can also see that there is a slight difference between the trends of CPUE in 2003-2004, 2006, 2008-2009, and 2014-2015, which is related to the Southwest Atlantic developed by Wu Yumei and others in 2004-2008. The SST has undergone major changes. The resource abundance of Argentinian squid in 2004-2008 is significantly negatively correlated with SST, probably due to the difference in migration routes and migratory time of Argentinian squid population. Caused. In addition, through the selection of environmental factors in the model and the effect of the model, the environmental factors including the SST model of all points with gray absolute correlation greater than 0.9 in June and July are better than Model 2 and Model 3, which indicates that The SST of the spawning ground in the two months of July and July has a greater impact on the CPUE of the following year. These two months are more important for the construction of the abundance model of the squid resource in the southwest Atlantic. In Model 2, the environmental factors do not include the July SST. The model predicts a relative error of 354% in resource abundance in 2016 (Table 3), indicating that July is the main spawning month for Argentine squid in the Southwest Atlantic.

In summary, the good prediction results of the model show that the gray system can have less sample data and no information. It is a well-known situation that can also be better predicted, suitable for the short life cycle type of Argentine squid. The time series taken by this study is relatively long, which makes the predicted results convincing. The gray system is used to predict the resource abundance of Argentine squid in the Southwest Atlantic, which provides a reference value for fishery production practices.

Claims

A method for predicting the abundance of large-scale squid resource based on gray system, which comprises the following steps:

Step 1: Using the grey correlation analysis method, calculate the gray correlation degree of the influence factor of the resource abundance of the oceanic squid;

Step 2: Select a factor with a large gray correlation degree as a factor of the resource abundance prediction model;

Step 3: Using the discrete GM model, the factor selected in step 2 is used to establish a prediction model for the abundance of oceanic carp.

Step 4: Perform a validity analysis on the predictive model of step three. The effectiveness analysis includes relative error and correlation coefficient analysis. The relative error is calculated by using the factor data to calculate the CPUE value and comparing with the real CPUE value to obtain the absolute value. Correlation coefficient analysis The correlation coefficient between the simulated CPUE value sequence and the actual CPUE value sequence obtained by the model in step 3 is selected, and the model with the smallest relative error and the largest correlation coefficient is selected as the most suitable prediction model.
The method for predicting resource abundance of large-scale squid based on gray system according to claim 1, characterized in that: for the northwest Pacific winter spring squid, the influence factors of large gray correlation degree include sea surface temperature SST, PDO, nino3. .4 anomalies and chlorophyll concentrations chl-a.
The method for predicting resource abundance of large-scale squid based on gray system according to claim 2, wherein the resource abundance prediction of the squid in the northwestern Pacific winter comprises the following steps: (1) obtaining nino3 through remote sensing satellites. .4 anomalies, PDO data, sea surface temperature SST and chlorophyll concentration chl-a 4 marine environments and climatic factors; (2) grey correlation analysis of 4 marine environments and climatic factors, ie the CPUE of the current year as the parent sequence According to the sub-sequences of the spawning field environmental indicators and climate indicators of each month of the annual spawning time, the gray absolute correlation degree of each sub-sequence and the parent sequence is calculated respectively, and the magnitude of the absolute correlation degree of gray is important to each index. (3) According to the results of the grey correlation analysis, the four factors with the highest degree of relevance are selected: the spawning field in March The surface temperature of the sea is SST, the Pacific inter-annual oscillation index PDO in January, the nino3.4 anomaly in April, and the average chlorophyll concentration in the month of chl-a; (4) 8 kinds of gray system-based Northwest Pacific softness based on the selected four factors The resource abundance prediction models of the fish winter spring group are: GM(O,1), GM(O,2), GM(O,3), GM(O,4), GM(1,1),GM (1,2), GM(1,3), GM(1,4); (5) Validation of eight prediction models, the model GM(1,4) structure is the most suitable northwest Pacific squid winter A predictive model of the abundance of spring resources.
The method for predicting resource abundance of large-scale squid based on gray system according to claim 1, characterized in that: for the Atlantic Atlantic squid in the southwest Atlantic, the influence factors of the gray correlation degree include the months of July, July and August. SST in the sea area of ° ~ 66 ° W, 32 ° ~ 43 ° S.
The gray system-based method for predicting resource abundance of large-scale squid according to claim 4, wherein the resource abundance prediction of the squid in the southwest Atlantic includes the following steps: 1) spawning from June to August The gray correlation degree between the subsurface sequence of ocean surface temperature and the CPUE parent sequence in the next year is selected, and the SST of the gray correlation degree greater than 0.9 in the sea of spawning ground is selected as the prediction model of the resource abundance of Argentine squid in the Southwest Atlantic. Factor; 2) Construct a variety of GM (0, N) gray prediction models, respectively: GM (0, 3) model, containing two factors, all points of gray absolute correlation greater than 0.90 in June and July each month The average of the SST; the GM(0,3) model, which contains two factors, the average of the SSTs of all points where the gray absolute correlation is greater than 0.90 for each month in June and August; the GM(0,3) model, which contains two The average value of SST for all points with gray absolute correlation greater than 0.90 in July and August; GM (0,4) model, containing three factors, including the gray absolute correlation of each month from June to August Average of SST for all points greater than 0.90; 3) for 4 forecast models Validation, model GM (0,4) structure is the most appropriate in the Southwest Atlantic Argentine Illex predictive models abundance of fish resources.