A Remote Sensing Synergistic Inversion Method for Multi-Component Concentrations of Suspended Particulate Matter in Offshore Waters

Abstract

The present invention discloses a method for synergistic inversion of multi-component concentrations of suspended particulate matter in offshore waters, comprising the following steps: collecting measured data sets, including hyperspectral remote sensing reflectance data, total suspended particulate matter concentration TSM, organic suspended particulate matter concentration OSM, and inorganic suspended particulate matter concentration ISM; and performing preprocessing; calculating spectral feature quantities characterizing the concentrations of ISM and OSM according to MODIS band remote sensing reflectance data; inputting the ISM and OSM concentration data and their corresponding spectral feature quantities into the XGBoost algorithm for training to construct a remote sensing inversion model for the concentrations of ISM and OSM; evaluating the model accuracy using an independent validation sample set and a satellite-ground matching data set; the present invention provides an important technical means and data support for the research on marine water quality, ecological environment assessment, and related biogeochemical processes.

Description

Technical Field

[0001] The present invention relates to the technical field of seawater environment, and particularly relates to a method for synergistic remote sensing inversion of multi-component concentrations of suspended particulate matter in coastal waters. Background Art

[0002] Suspended particulate matter is a general term for all suspended substances in seawater, including phytoplankton, suspended sediment, and organic debris. As an important component of coastal waters, it plays a key role in maintaining the marine ecological balance, regulating the material and energy cycles, and environmental monitoring. The suspended particulate matter in water mainly includes organic particulate matter (such as phytoplankton and zooplankton, microorganisms, and organic debris, etc.) and inorganic particulate matter (such as inorganic debris, clay minerals, etc.). Among them, inorganic particulate matter is an important carrier of nutrients, affecting water transparency, sediment transport, and the aquatic ecosystem, while organic particulate matter reflects the nutrient level in seawater and provides an important reference for evaluating the changes in the marine ecological environment. Therefore, mastering the distribution of suspended particulate matter in coastal waters and the characteristics of its organic and inorganic components has important scientific value for in-depth understanding of marine ecological environment research and material cycling.

[0003] The concentrations of total suspended particulate matter (TSM), inorganic suspended particulate matter (ISM), and organic suspended particulate matter (OSM) are respectively used to characterize the content status of total suspended particulate matter and its components (organic and inorganic) in water. Affected by multiple factors such as tides, ocean currents, and human activities, the TSM, ISM, and OSM in coastal waters show significant spatio-temporal variability. Field surveys usually use the ignition gravimetric method to measure the concentrations of total, organic, and inorganic suspended particulate matter. However, this method is time-consuming and laborious, and the data is discrete, making it difficult to depict their spatio-temporal distribution characteristics. In contrast, optical satellite remote sensing can provide large-scale and long-term observation data, and has unique advantages in monitoring the concentrations of suspended particulate matter and its components. At present, many researchers have used satellite remote sensing technology to carry out remote sensing detection research on TSM and ISM, and have proposed a series of remote sensing inversion models. These models are mainly based on the correlation between the spectral characteristics of satellite remote sensing reflectance (R rs ) and the concentration of suspended particulate matter, and use empirical models or machine learning methods for inversion. Currently, optical satellite remote sensing technology has been widely used in the remote sensing inversion research of the concentration of total suspended particulate matter (TSM) and inorganic suspended particulate matter (ISM) in ocean waters, but there are few reports on the inversion algorithm for the concentration of marine organic suspended particulate matter (OSM). In addition, most of the existing remote sensing monitoring algorithms focus on a single parameter (such as TSM or ISM), and there is a lack of remote sensing methods that can synergistically invert TSM, ISM, and OSM. Summary of the Invention

[0004] Objective of the Invention: The objective of the present invention is to provide a method for synergistic retrieval of multi-component concentrations of suspended particulate matter in coastal waters, to solve the technical problems such as the lack of research on OSM retrieval and insufficient multi-parameter synergistic retrieval ability in existing methods, and to achieve the synchronous acquisition of the concentrations of suspended particulate matter and its components (organic and inorganic) in coastal waters.

[0005] Technical Solution: A method for synergistic retrieval of multi-component concentrations of suspended particulate matter in coastal waters according to the present invention includes the following steps:

[0006] (1) Collect measured data sets, including hyperspectral remote sensing reflectance data, total suspended particulate matter concentration TSM, organic suspended particulate matter concentration OSM, and inorganic suspended particulate matter concentration ISM; and perform preprocessing.

[0007] (2) Calculate spectral characteristic quantities representing the concentrations of ISM and OSM based on MODIS band remote sensing reflectance data.

[0008] (3) Input the ISM and OSM concentration data and their corresponding spectral characteristic quantities into the XGBoost algorithm for training to construct remote sensing inversion models for the concentrations of ISM and OSM.

[0009] (4) Evaluate the model accuracy using an independent validation sample set and a satellite-ground matching data set.

[0010] Further, in step (1), the preprocessing is specifically as follows: Based on the spectral response function of the MODIS sensor, convert the measured hyperspectral remote sensing reflectance data into remote sensing reflectance data corresponding to the MODIS bands. ; Through the sampling time and longitude and latitude information, match the converted data with the TSM, OSM, and ISM concentration data.

[0011] Further, the MODIS bands include 10 bands of 412, 443, 469, 488, 531, 547, 555, 645, 667, and 678 nm, and the conversion of the remote sensing reflectance data is through the formula:

[0012] ;

[0013] In the formula, R rs (λ M ) is the remote sensing reflectance at the λ M band of the MODIS sensor; λ1 and λ2 are the band ranges of this band; R rs (λ) is the measured hyperspectral remote sensing reflectance; SRF(λ) is the spectral response function of the MODIS sensor.

[0014] Further, in step (2), the ISM feature quantities include the reflectance values and their band ratios in the bands of 412, 443, 469, 488, 531, 547, 555, 645, 667, and 678; the OSM feature quantities include: the values after logarithmic transformation, their band differences, and their band combination ratios in the bands of 412, 443, 469, 488, 531, 547, 555, 645, 667, and 678.

[0015] Further, in step (3), the total suspended matter concentration TSM is obtained by the following formula:

[0016] ;

[0017] where TSM, OSM, and ISM are the total, organic, and inorganic suspended matter concentrations of the water body, respectively.

[0018] Further, in step (3), the training parameters of the XGBoost model are optimized by cross-validation, including the learning rate, tree depth, and regularization coefficient.

[0019] Further, in step (4), the model is applied to long-time-series MODIS remote sensing images to generate a remote sensing product set of TSM, ISM, and OSM concentrations, and analyze their spatio-temporal variation characteristics.

[0020] Further, in step (4), the indicators for model accuracy verification include the root mean square error RMSE, mean absolute error MAE, and determination coefficient R 2 .

[0021] An electronic device according to the present invention includes a memory, a processor, and a computer program stored on the memory. When the processor executes the program, the steps of any one of the methods are implemented.

[0022] A computer-readable storage medium according to the present invention stores a computer program. When the program is executed by a processor, the steps of any one of the methods are implemented.

[0023] Beneficial effects: Compared with the prior art, the present invention has the following remarkable advantages: The present invention combines the XGBoost method to develop a remote sensing collaborative inversion method for suspended particulate matter and its component concentrations in complex optical waters near the sea, and evaluates the model accuracy. In addition, this method has been successfully applied to long-term MODIS remote sensing images to generate a remote sensing product set of total suspended matter (TSM), inorganic suspended matter (ISM), and organic suspended matter (OSM) in coastal waters, and reveals their spatio-temporal variation characteristics. The method of the present invention can be extended and applied to other ocean regions and other optical satellite sensors, and can provide high-quality remote sensing thematic products with long time series and multiple parameters, providing important technical means and data support for marine water quality, ecological environment assessment, and related biogeochemical process research. Description of the Drawings

[0024] Figure 1 is a flowchart of the present invention;

[0025] Figure 2 is the construction and verification of the suspended particulate matter and its component concentration model using the measured data set of the present invention; among them, Figure 2 in (a) is the TSM model training; Figure 2 in (b) is the ISM model training; Figure 2 in (c) is the OSM model training; Figure 2 in (d) is the TSM model verification; Figure 2 in (e) is the ISM model verification; Figure 2 in (f) is the OSM model verification;

[0026] Figure 3 is the satellite remote sensing inversion accuracy evaluation of the suspended particulate matter and its component concentrations of the present invention; among them, Figure 3 in (a) is the satellite-ground matching verification of TSM; Figure 3 in (b) is the satellite-ground matching verification of ISM; Figure 3 in (c) is the satellite-ground matching verification of OSM. Detailed Embodiments

[0027] The technical solution of the present invention will be further described below in conjunction with the drawings.

[0028] As Figure 1 shown, an embodiment of the present invention provides a remote sensing collaborative inversion method for multi-component concentrations of suspended particulate matter in coastal waters, including the following steps:

[0029] Step 1: Organize and collect the measured survey data set in coastal waters, and the parameters include remote sensing reflectance spectrum, total suspended particulate matter concentration, organic suspended particulate matter concentration, and inorganic suspended particulate matter concentration. According to the spectral response function of the MODIS sensor, based on the measured hyperspectral remote sensing reflectance data, obtain the corresponding Data. According to the sampling time and the longitude and latitude of the site, the measured R rs (λ M ) is matched with the concentrations of suspended particulate matter and its components (TSM, ISM, and OSM).

[0030] (1);

[0031] In the formula, R rs (λ M ) is the remote sensing reflectance at the λ M band of the MODIS sensor; λ1 and λ2 are the band ranges of this band; R rs (λ) is the measured hyperspectral remote sensing reflectance; SRF(λ) is the spectral response function of the MODIS sensor.

[0032] Step 2: According to the measured data set, calculate the spectral characteristic quantities that can characterize the ISM concentration and the OSM concentration respectively, as shown in Table 1.

[0033] Table 1 Spectral characteristic quantities characterizing the ISM and OSM concentrations

[0034] ;

[0035] Step 3: Input the measured ISM and OSM concentrations and their corresponding spectral characteristic quantities in Table 1 into XGBoost for model training to obtain the remote sensing inversion methods for the ISM and OSM concentrations; calculate the TSM concentration according to formula (2).

[0036] (2);

[0037] Among them, TSM, OSM, and ISM are the total, organic, and inorganic suspended particulate matter concentrations in the water body respectively.

[0038] Step 4: According to the above steps of the XGBoost method, construct a remote sensing collaborative inversion method for the concentrations of suspended particulate matter and its components in the coastal sea, and use the measured data and the satellite-ground matching data set for accuracy verification, as Figure 2 and Figure 3 shown.

[0039] Step 5: Apply the developed remote sensing collaborative pan-Asian method for suspended particulate matter and its components to the long-term MODIS satellite remote sensing reflectance product, generate a remote sensing product set of the total, organic, and inorganic suspended particulate matter concentrations in the coastal sea, and reveal their spatio-temporal variation characteristics.

[0040] The present invention is applied to the coastal water body to obtain the long-term total, organic, and inorganic suspended particulate matter concentrations and conduct spatio-temporal variation characteristic analysis.

Claims

1. A remote sensing collaborative inversion method for multi-component concentration of suspended particulate matter in offshore waters, characterized in that: The following steps are involved: (1) Collect measured data sets, including hyperspectral remote sensing reflectance data, total suspended particulate matter concentration (TSM), organic suspended particulate matter concentration (OSM), and inorganic suspended particulate matter concentration (ISM); and perform preprocessing. The preprocessing is as follows: Based on the spectral response function of the MODIS sensor, convert the measured hyperspectral remote sensing reflectance data into remote sensing reflectance data corresponding to the MODIS band. ; By sampling time and longitude and latitude information, the converted The data are matched with TSM, OSM, and ISM concentration data; (2) Calculate the spectral characteristics of ISM and OSM concentrations based on MODIS band remote sensing reflectance data; MODIS bands include 10 bands of 412, 443, 469, 488, 531, 547, 555, 645, 667, and 678 nm, and the conversion of remote sensing reflectance data is through the formula: ； In the formula, R rs (λ M ) is the MODIS sensor λ M Remote sensing reflectance at the band; λ1 and λ2 are the band ranges of the current band; R rs (λ) is the measured hyperspectral remote sensing reflectance; SRF(λ) is the spectral response function of the MODIS sensor; ISM characteristic quantities: including Reflectance values ​​and band ratios in bands 412, 443, 469, 488, 531, 547, 555, 645, 667, and 678; OSM feature quantities include: The logarithmic transformed values ​​of bands 412, 443, 469, 488, 531, 547, 555, 645, 667, and 678, as well as their band differences and band combination ratios, are as follows: ； (3) The ISM and OSM concentration data and their corresponding spectral features are respectively input into the XGBoost algorithm for training to construct the remote sensing inversion model of ISM and OSM concentrations; the total suspended particulate matter concentration TSM is obtained by the following formula: Among them, TSM, OSM and ISM are the total, organic and inorganic suspended particulate matter concentrations in the water body, respectively; (4) The model accuracy is evaluated using an independent validation sample set and a satellite-ground matching dataset.

2. The method for remote sensing collaborative inversion of multi-component concentration of suspended particulate matter in offshore waters according to claim 1 is characterized in that: In step (3), the training parameters of the XGBoost model are optimized through cross-validation, including learning rate, tree depth, and regularization coefficient.

3. An electronic device comprising a memory, a processor and a computer program stored in the memory, characterized in that: When the processor executes the program, the steps of the method according to any one of claims 1 to 2 are implemented.

4. A computer-readable storage medium storing a computer program, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 2 are implemented.

Images