Method for predicting spatial distribution of algal blooms in hour-scale lake based on stationary satellite

Abstract

The invention discloses an hour-scale lake algal bloom space distribution prediction method based on a stationary satellite, and the method comprises the steps: obtaining the hourly wind speed, wind direction, air temperature and other meteorological data and GOCI series satellite remote sensing data, and extracting the algal bloom coverage through an algal bloom recognition index; based on the priori knowledge of wind-driven algae bloom migration and accumulation, constructing a new characteristic index by using a wind direction and a remote sensing index; selecting representative algal bloom samples, extracting data such as wind speed, air temperature and remote sensing indexes of sample points in the same time period to construct a sample database, training a machine learning model and verifying the model precision; the model is applied to a typical eutrophicated lake to test the precision performance of the model, and the algal bloom distribution condition at the current moment and the environmental data at the subsequent moment are utilized to predict the algal bloom spatial distribution change of the lake hour by hour through model iteration. According to the method, prediction and early warning of the spatial position and the outbreak intensity of lake algal blooms in an hour scale can be realized, and prevention and control of eutrophic lake algal blooms are assisted.

Description

Technical Field

[0001] The present invention belongs to the technical field of satellite remote sensing and water environment analysis, and particularly relates to a method for predicting the spatial distribution of lake algal blooms on an hourly scale based on geostationary satellites. Background Art

[0002] Lakes around the world are generally facing ecological and environmental problems such as the intensification of water eutrophication and the outbreak of harmful algal blooms. The proportion of eutrophic lakes and reservoirs is increasing year by year, and the increasingly frequent outbreaks of algal blooms are seriously threatening water ecological security. The unique cell structure and community characteristics endow cyanobacteria with the ability to migrate up and down in the vertical water column to obtain light and food, making it the dominant species in the phytoplankton community. Under ideal meteorological and hydrological conditions, algal particles accumulate on the water surface to form cyanobacterial blooms, triggering a water use crisis.

[0003] The area of algal blooms changes greatly driven by environmental factors, and there is even a phenomenon that a large amount of algal blooms gather and then disappear quickly within a few hours, which is difficult to capture by conventional means. Remote sensing satellite data has the advantages of wide coverage, fast real-time, and periodicity, and has been widely used for long-term monitoring of algal blooms. Algal bloom prediction models can be divided into two categories: (1) Mechanism-driven prediction models, which simulate and predict the ecological changes of algal blooms by studying the hydrodynamic and algal bloom formation mechanisms. (2) Data-driven prediction models, which use a large amount of data to analyze the relationship between algal bloom indicators and various environmental factors for constructing algal bloom prediction models. Generally speaking, mechanism-driven prediction models intuitively present the changes of mechanism components and achieve the prediction of spatial variation of algal communities, but they are computationally complex and have numerous parameters, and different simulation models need to be constructed according to the characteristics of lakes. Data-driven prediction models do not require complex calculations, save time and effort, and can achieve algal bloom prediction by extracting potential patterns from historical monitoring data. However, the published research usually focuses on the time series prediction of algal bloom indicators at sampling points or the whole-lake mean in the study area, lacking research on spatial variation prediction. Therefore, it is very important to construct a simple and effective prediction model based on the prior knowledge of environmental factor-driven algal bloom outbreak and migration to achieve the prediction of the spatial occurrence location and outbreak intensity of algal blooms in the short term. Summary of the Invention

[0004] The purpose of the present invention is to provide a method for predicting the spatial distribution of lake algal blooms on an hourly scale based on geostationary satellites.

[0005] To achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0006] A method for predicting the spatial distribution of lake algal blooms on an hourly scale based on geostationary satellites, the method comprising:

[0007] Obtaining the wind direction data of the study area and the remote sensing image data of the geostationary satellite covering the study area;

[0008] Calculate the algal bloom identification index and algal bloom coverage of the study area based on the remote sensing image data;

[0009] Calculate the floating algae coverage index, distance index, and surrounding algae index of the image pixels using the wind direction data and the algal bloom coverage respectively, where:

[0010] The floating algae coverage index is the mean of the algal bloom coverage within the n×n matrix containing the target image pixel, and the position of the target image pixel in the matrix is determined based on the wind direction;

[0011] The distance index represents the distance between the target image pixel and the nearest land pixel along the wind direction of the target image pixel;

[0012] The surrounding algae index is the value of the central pixel of the matrix, and this central pixel value is characterized by the standard deviation of the algal bloom coverage of the pixels within the matrix;

[0013] Construct a machine learning model that includes the algal bloom identification index, algal bloom coverage, floating algae coverage index, distance index, and surrounding algae index as input features, train the machine learning model to obtain an algal bloom prediction model, and use the trained algal bloom prediction model to predict lake algal blooms.

[0014] As a preferred embodiment, the input features of the algal bloom prediction model further include environmental data and remote sensing indices related to algal bloom changes screened by correlation analysis.

[0015] As a preferred embodiment, the input features of the algal bloom prediction model further include hourly environmental data, as well as land surface temperature, water depth, and algal bloom frequency data;

[0016] The environmental data includes wind speed, air temperature, and solar radiation data.

[0017] Furthermore, in the input features, the environmental data takes the average value of the current moment and the next moment as the input value.

[0018] As a preferred embodiment, the determination of the position of the target image pixel in the matrix based on the wind direction includes:

[0019] Divide the wind direction into eight intervals, with each interval centered on due east, due west, due south, due north, northeast, northwest, southeast, and southwest;

[0020] Obtain the average wind direction of the target pixel at the current moment and the next moment, and confirm the wind direction of the target pixel based on the wind direction interval where the average wind direction is located;

[0021] The target pixel is located downwind of the central pixel in the matrix.

[0022] As a preferred embodiment, the matrix does not contain land pixels or pixels removed due to masking.

[0023] As a preferred embodiment, the distance index is characterized based on a normalization index.

[0024] As a preferred embodiment, during model training, images with wind speeds lower than a critical threshold are selected as training data; the critical threshold is the wind speed threshold for algal bloom occurrence.

[0025] As a preferred embodiment, the geostationary satellite remote sensing image data includes geostationary satellite remote sensing image data from different sources;

[0026] The geostationary satellite remote sensing image data from different sources is used to jointly construct a prediction model after passing consistency verification.

[0027] Furthermore, the method further includes, for geostationary satellite remote sensing image data from different sources that exhibit a linear relationship, selecting the geostationary satellite remote sensing image data from one source as standard data, and after correcting the geostationary satellite remote sensing image data from other sources using the linear relationship, using it to jointly construct a prediction model.

[0028] As a preferred embodiment, the machine learning model is a random forest model. In the present invention, the newly constructed indicators have high feature importance in the random forest model, and the trained prediction model also has higher accuracy.

[0029] Based on the characteristics of algal bloom migration with the wind, the present invention constructs characteristic indicators reflecting algal bloom migration under different conditions using the algal bloom index and wind direction, including the floating algae coverage index related to the wind direction, the distance index considering that algal blooms are likely to form near the shore and including information on wind direction and offshore distance, and the surrounding algae index related to changes in surrounding algal blooms. The newly constructed indicators are used as input features for prediction model training, and have good prediction effects on short-term spatial prediction of algal blooms. The data-driven algal bloom prediction model constructed by the method of the present invention can efficiently predict the outbreak location and intensity of lake algal blooms several hours in advance, and is an important technology for assisting in the treatment of eutrophic lakes and the prevention and control of algal blooms, with great scientific significance and practical value. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] The drawings are to clearly and intuitively show the process from model construction to model application, help readers better understand this invention patent, enhance the information transmission effect and readability of the document, and corresponding explanations are also given for the components and markings in the drawings. Now, each step of the present invention will be described through examples and with reference to the drawings, where:

[0031] Figure 1 is the annual average statistical result of the FAC data calculated from each scene of GOCI images.

[0032] Figure 2 It is a schematic diagram for calculating FACI and DI. The left figure is the schematic diagram for calculating FACI, and the right figure is the schematic diagram for calculating DI.

[0033] Figure 3 It is the ranking result of the importance of input features in the random forest model and the deep neural network model.

[0034] Figure 4 It is the model verification result of the random forest model. (a) is the FAC data at the initial moment, (c) and (d) are the true value and predicted value of FAC at the next moment, and (b) is the scatter verification after synthesis by the 3*3 mean matrix.

[0035] Figure 5 It is the display of the spatial prediction accuracy of pixel-level algal blooms in Lake Taihu.

[0036] Figure 6 It is the prediction accuracy of the FAC mean value of the whole lake. (a), (b), and (c) respectively show the FAC mean values after 1, 4, and 7 hours. Specific implementation manners

[0037] In order to more effectively explain the technical solution of the present invention, specific embodiments are specifically given and illustrated with the accompanying drawings.

[0038] In the present disclosure, for various aspects of the present invention, its embodiments will be described with reference to the accompanying drawings. However, the embodiments and their accompanying drawings do not cover all instances to which the present invention can be applied. Because, the above-mentioned proposed technical solutions and implementation processes, as well as the more detailed concepts and specific steps below, can be flexibly implemented in various aspects, because the technical content disclosed by the present invention is not limited by any specific implementation manner. In addition, according to the specific characteristics of the lake, some content in the technical process can be added or deleted to build a specific prediction model for different lakes.

[0039] The data sources of the environmental data used in the embodiments are as follows:

[0040] Download the hourly wind speed, wind direction, air temperature, and solar radiation data of the research area from 2011 to 2024 from the European Centre for Medium-Range Weather Forecasts (ECMWF, https: / / cds.climate.copernicus.eu / ).

[0041] Download the land surface temperature dataset (MODIS_LST, https: / / doi.org / 10.5067 / MODIS / MOD11A1.061) from 2011 to 2024 on the Google Earth Engine (GEE) platform, with a spatial resolution of 1000 m and a temporal resolution of 1 day. Based on the GEE platform MODIS Aqua (MODIS / 061 / MYD09GA) dataset, resample the spatial resolution to 250 m, and use FAC≥5 as the discrimination condition for algal blooms. Calculate the algal bloom frequency (MODIS_Fre) for the pixel location in a month.

[0042] The water depth data is calculated by subtracting the lake bottom elevation from the average water level.

[0043] The geostationary satellite remote sensing data used in the embodiment is from the GOCI data from May 2011 to March 2021 and the GOCI-II data from January 2021 to October 2024 downloaded from the website (http: / / kosc.kiost.ac.kr / eng / ).

[0044] The spatial resolutions of GOCI and GOCI-II are 500 meters and 250 meters respectively, and the temporal resolution is 1 hour, providing 8 to 10 images per day. Use the SeaDAS (7.5.3) software to remove water vapor, ozone absorption, and Rayleigh scattering signals from the GOCI data, and calculate the Rayleigh reflectance R of each band. rc ,. For GOCI-II, R rc The data product is processed by G2AC (GOCI-II atmospheric correction). Using the threshold combination of the red, green, and blue band R rc values, remove the interference of clouds, solar flares (R rc (490)>0.14 & R rc (555)>0.16 & R rc (660)>0.15) and cloud shadows (R rc (555)<0.07), select the images with better quality and less cloud cover. Finally, obtain 6731 images, and use the images with wind speed lower than the critical threshold (3 m / s) and obvious algal bloom changes as the sample data for model training and verification, totaling 365 images.

[0045] Example 1

[0046] This example illustrates the method for predicting the spatial distribution of lake algal blooms at the hourly scale based on geostationary satellites in the present invention.

[0047] This example predicts the algal blooms in Lake Taihu based on machine learning and GOCI observations, and the implementation method is as follows:

[0048] Taking Taihu Lake as the study area, historical monitoring data such as wind speed, wind direction, and air temperature in the lake area were collected, and the GOCI data with high temporal resolution was used to extract algal blooms; a new characteristic index was constructed based on the wind direction and remote sensing index, and a database of typical algal bloom samples was outlined for training a random forest model. After verifying the model accuracy, the trained model was used for the spatial prediction of algal blooms in Taihu Lake.

[0049] The implementation process of the present invention will be gradually exemplified below and illustrated with drawings.

[0050] 1) Calculate the FAI index or AFAI index based on the geostationary satellite image, and then substitute the FAI index or AFAI index as an independent variable into the FAC calculation formula to obtain the FAC value. The FAC index is the algal bloom coverage, which can intuitively represent the proportion of algal blooms in a single pixel.

[0051] In this embodiment, since the GOCI satellite lacks the short-wave infrared band, the AFAI index is used instead of the FAI index. The calculation formulas of the AFAI index and the FAC index are as follows:

[0052] (1)

[0053] (2)

[0054] In the formula, λ1 = 660 nm, λ2 = 745 nm, λ3 = 865 nm, R rc is the Rayleigh reflectance of the corresponding band. Figure 1 is the annual average statistical result of the FAC data calculated from each scene of GOCI images.

[0055] The GOCI and GOCI-II data were verified for consistency, and the two have high consistency. Preferably, based on the linear relationship between GOCI AFAI and GOCI-II AFAI, the AFAI value calculated based on GOCI-II data was linearly adjusted, and the adjusted GOCI-II AFAI was used for subsequent calculations.

[0056] 2) Construct new characteristic indicators:

[0057] Wind speed and wind direction are the key factors affecting the horizontal and vertical movement of algal blooms. The disturbance caused by the wind drives the migration of cyanobacteria and their aggregation in the upwind direction. Driven by the wind, algae are more likely to aggregate near the shore, resulting in algal blooms. Therefore, based on the above prior knowledge, a new characteristic index is constructed in the present invention, combining remote sensing indicators with environmental factors for modeling.

[0058] In this embodiment, three new characteristic indices, namely the Floating algae cover index (FACI), the Distance index (DI), and the Algae around index (AAI), are constructed using the wind direction and the FAC index. Their calculation methods are as follows:

[0059] ① Floating algae cover index FACI;

[0060] The wind direction range of 0° to 360° is divided into 8 equally spaced intervals, as shown in Table 1:

[0061] Table 1 Wind direction intervals

[0062]

[0063] According to the average wind direction at the current moment and the next moment, the corresponding statistical area is selected to obtain the average FAC in the 3×3 matrix, and it is assigned to the target pixel. For example Figure 2 as shown in Figure 2 (left), the average wind direction is northwest. Algal blooms are likely to migrate from the northwest to the southeast under the action of wind. Therefore, the corresponding statistical range is the 3*3 matrix in the upper left corner of the target pixel (that is, the target pixel is located in the lower right corner, or the southeast corner). The calculation formula for the floating algae cover index is as follows:

[0064] (3)

[0065] In the formula, n is the number of valid pixels included in the 3×3 matrix corresponding to the target pixel, is the FAC value of the i-th pixel. The matrix does not include land pixels and masked pixels.

[0066] ② Distance index DI;

[0067] The DI value represents the distance between the target pixel and the land. The judgment of the distance between the pixel and the land in the DI value is related to the wind direction (hereinafter referred to as the wind direction-land distance). The wind direction-land distance of the target pixel is characterized by the distance between the target pixel and the nearest land pixel in the wind direction. The normalized value is used in the formula to represent it. As Figure 2 shown in Figure 2 (right), the average wind direction is northeast. The distance between the target pixel and the nearest land pixel in the wind direction is the distance between the target pixel and the nearest land pixel in the due northeast direction. The calculation formula for DI is as follows:

[0068] (4)

[0069] In the formula, Distance iDistance is the distance between the target pixel i in the current image and the nearest land pixel along the wind direction. max Distance for all target pixels calculated from the current image i is the maximum value among the values.

[0070] ③ Surrounding Algae Index AAI;

[0071] Calculate the standard deviation of FAC within the 3×3 matrix to which the target pixel in (1) belongs, and use it as the AAI value of the central pixel of the matrix. The formula is as follows:

[0072] (5)

[0073] 3) Use environmental factors, AFAI index, AFAI gradient (AFAI_Fradient), FAC index, and the FACI, DI, and AAI characteristic indices constructed in (2) as input features, and the algal bloom index as the output feature for model training. In this embodiment, the FAC index is selected as the algal bloom index.

[0074] In this embodiment, two machine learning models, the random forest model (RF) and the deep neural network (DNN), which can handle high-dimensional data and capture complex non-linear relationships, are selected for the spatial prediction of cyanobacterial blooms in Lake Taihu.

[0075] The input environmental factors are wind speed, air temperature, solar radiation data, water depth, algal bloom occurrence frequency (MODIS_Fre), and land surface temperature (MODIS_LST). Among them, the wind speed, air temperature, and solar radiation take the average value of the current moment and the next moment, the water depth data remains fixed, the algal bloom occurrence frequency is the algal bloom outbreak frequency in the previous month of the current moment, and the land surface temperature is the value of the current moment.

[0076] Manually outline typical algal bloom samples and water samples randomly distributed in the lake area, extract the characteristic data at the same positions to construct a sample database, and divide the dataset into a training set and a test set in a ratio of 9:1 for model training and verification. The training of the random forest model is completed based on the Scikit-Learn package in Python 3.7, and the DNN model is constructed by sequentially adding an input layer, a hidden layer, and an output layer through the Keras library. Use R 2 , RMSE, PRD, MAPE, and UPD to evaluate the model accuracy. The results are shown in Table 2:

[0077] Table 2 Performance evaluation of RF and DNN models for algal bloom spatial prediction

[0078]

[0079] It can be seen that the RF model has a better prediction effect. Figure 3The importance contributions of 12 input features in model training are shown. Among them, the FAC and AFAI indices, which are the algal bloom recognition indicators, predictably rank among the top two in terms of importance in both models. In the feature importance ratio of the RF model, FACI, DI, and AAI rank 4th, 5th, and 7th respectively, with a contribution rate of 27% to the RF model, which can effectively assist in the spatial prediction of pixel-level algal blooms in Lake Taihu. In the feature importance ratio of the DNN model, FACI, DI, and AAI rank 5th, 8th, and 12th respectively, with a contribution rate of 14.4% to the DNN model. As can be seen from Table 1, the RF model, to which the newly constructed index contributes more, has a better prediction effect than the DNN model, which also demonstrates the effectiveness of the newly constructed index.

[0080] 4) Apply the trained RF model to specific images to verify the accuracy performance of the model. Select the scene on September 18, 2017. Figure 4 (a) is the true value data of FAC at the current moment. Figure 4 (c) and (d) are the true value and predicted value of FAC at the next moment respectively. Figure 4 (b) is the scatter plot of the true value and the predicted value.

[0081] 5) Apply the trained model to GOCI and GOCI-II images, and through model iterative operations, predict the spatial location and outbreak intensity of algal blooms in Lake Taihu hour by hour. Figure 5 For pixel-level spatial prediction of algal blooms in Lake Taihu Figure 6 For the prediction accuracy of the overall FAC mean value.

Claims

1. A method for predicting the spatial distribution of lake algal blooms on an hourly scale based on geostationary satellites, characterized in that, The method includes: Obtaining wind direction data of the study area and geostationary satellite remote sensing image data covering the study area; Calculating the algal bloom identification index and algal bloom coverage of the study area based on the remote sensing image data; Calculating the floating algae coverage index, distance index, and surrounding algae index of the image pixels using the wind direction data and the algal bloom coverage respectively, where: The floating algae coverage index is the mean of the algal bloom coverage within the n × n matrix range of the pixel of the target image, and the position of the pixel of the target image in the matrix is determined based on the wind direction; The distance index represents the distance between the target image pixel and the nearest land pixel along the wind direction of the target image pixel; The surrounding algae index is the value of the central pixel of the matrix, and this central pixel value is characterized by the standard deviation of the algal bloom coverage of the pixels within the matrix; Constructing a machine learning model with the algal bloom identification index, algal bloom coverage, floating algae coverage index, distance index, and surrounding algae index as input features, training the machine learning model to obtain an algal bloom prediction model, and using the trained algal bloom prediction model to predict lake algal blooms.

2. The method according to claim 1, wherein The input features of the algal bloom prediction model further include environmental data and remote sensing indices related to algal bloom changes screened by correlation analysis.

3. The method according to claim 1, wherein The input features of the algal bloom prediction model further include hourly environmental data, as well as land surface temperature, water depth, and algal bloom frequency data; The environmental data includes wind speed, air temperature, and solar radiation data.

4. The method according to claim 1, wherein The position of the target image pixel in the matrix is determined based on the wind direction and includes: Dividing the wind direction into eight intervals, with each interval centered on due east, due west, due south, due north, northeast, northwest, southeast, and southwest; Obtaining the average wind direction of the target pixel at the current moment and the next moment, and confirming the wind direction of the target pixel based on the wind direction interval where the average wind direction is located; The target pixel is located downwind of the central pixel in the matrix.

5. The method according to claim 1, characterized in that The matrix does not contain land pixels or pixels removed by masking.

6. The method according to claim 1, wherein The distance index is characterized based on a normalized index.

7. The method according to claim 1, wherein When training the model, select images with a wind speed lower than the critical threshold as training data; the critical threshold is the wind speed threshold for algal bloom occurrence.

8. The method according to claim 1, characterized in that, The geostationary satellite remote sensing image data includes geostationary satellite remote sensing image data from different sources; The geostationary satellite remote sensing image data from different sources is used to jointly construct a prediction model after passing the consistency verification.

9. The method according to claim 8, wherein The method further includes, for geostationary satellite remote sensing image data from different sources that have a linear relationship, selecting the geostationary satellite remote sensing image data from one source as the standard data, and using the linear relationship to correct the geostationary satellite remote sensing image data from other sources before using it to jointly construct a prediction model.

10. The method according to claim 1, characterized in that, The machine learning model is a random forest model.

Images