CN115080905A - Remote sensing inversion method for chlorophyll a concentration of plateau lake - Google Patents

Abstract

The invention discloses a remote sensing inversion method for chlorophyll a concentration of a plateau lake, comprising: respectively acquiring the chlorophyll a concentration of the plateau lake and the remote sensing reflectivity data of the plateau lake; performing atmospheric correction on the remote sensing reflectivity data of the plateau lake to obtain the reflectivity of the bottom layer of the atmosphere According to the intrinsic optical properties of chlorophyll a and the reflectance data of the bottom atmosphere, the chlorophyll a inversion spectral index was constructed respectively; according to the inversion spectral index of chlorophyll a, a decision tree model was established to determine the spectral index that had the greatest impact on the chlorophyll a concentration result, as Optimizing the spectral index; performing linear regression on the optimal spectral index and the chlorophyll-a concentration of the plateau lakes to establish a chlorophyll-a inversion model; according to the chlorophyll-a inversion model and the atmospheric bottom reflectance data, realize the remote sensing inversion of the chlorophyll-a concentration of the plateau lakes. This method can quickly optimize the most effective model for inversion of chlorophyll a concentration, and further improve the inversion accuracy of chlorophyll a.

Description

A remote sensing retrieval method for chlorophyll a concentration in plateau lakes

technical field

The invention relates to the technical field of lake chlorophyll a concentration inversion, in particular to a remote sensing inversion method of lake chlorophyll a concentration.

Background technique

The eutrophic condition of lake water quality in plateau lakes is a topic of concern in today's society. Chlorophyll a is one of the main components of phytoplankton organisms, and it is an important indicator to reflect the nutritional status of inland water bodies and monitor the outbreak of cyanobacterial blooms. Monitoring the distribution of chlorophyll a concentration in lakes is helpful for measuring phytoplankton biomass and evaluating the nutritional status of water bodies. Plateau lakes are affected by topographic factors, so traditional lake sampling and monitoring methods are not suitable. With the in-depth study of water spectral characteristics and the improvement of lake chlorophyll a concentration inversion models, remote sensing methods can more accurately obtain lake chlorophyll a concentration information. , which can effectively discover eutrophic water bodies and water quality trends.

There are mainly two traditional methods for effectively extracting the chlorophyll a concentration of plateau lakes: one is to take the samples of plateau lake water back to the laboratory for extraction through manual field measurement. Lake chlorophyll a concentration information; another commonly used chlorophyll a extraction method is based on lake spectral measurement. According to the characteristics of the measured spectral information, the corresponding wavelength band of the reflectivity of the remote sensing satellite is selected for fitting to achieve the effect of chlorophyll a inversion. Although this method is a common method for remote sensing inversion, but there are many plateau lakes, and the spectral characteristics of plateau lakes are complex.

Therefore, on the basis of the existing lake chlorophyll a concentration inversion technology, how to not only overcome the one-sidedness of manual field measurement, but also solve the difficulty of measuring the spectrum of plateau lakes and improve the work efficiency and inversion accuracy has become an urgent need for those skilled in the art. solved problem.

SUMMARY OF THE INVENTION

In view of the above problems, the present invention proposes a remote sensing inversion method for chlorophyll a concentration in plateau lakes that solves at least some of the above-mentioned technical problems. Improve work efficiency and inversion accuracy.

An embodiment of the present invention provides a remote sensing inversion method for chlorophyll a concentration in a plateau lake, comprising the following steps:

S1. Obtain the chlorophyll a concentration of the plateau lake and the remote sensing reflectance data of the plateau lake respectively; the chlorophyll a concentration of the plateau lake is obtained from the ground measuring point of the plateau lake;

S2, performing atmospheric correction on the remote sensing reflectivity data of the plateau lakes to obtain bottom atmospheric reflectivity data;

S3. According to the intrinsic optical properties of chlorophyll a and the reflectivity data of the bottom layer of the atmosphere, respectively construct a chlorophyll a inversion spectral index; the chlorophyll a inversion spectral index includes: a single-band index, a ratio index, a normalized chlorophyll a index and three-band index;

S4, inverting the spectral index according to the chlorophyll a, establishing a decision tree model, and determining the spectral index that has the greatest impact on the chlorophyll a concentration result, as the preferred spectral index;

S5, perform linear regression on the preferred spectral index and the chlorophyll a concentration of the plateau lake, establish a chlorophyll a inversion model, and obtain the correlation coefficient of the regression model;

S6. According to the chlorophyll a inversion model and the atmospheric bottom reflectance data, the remote sensing inversion of the chlorophyll a concentration of the plateau lake is realized.

Further, the step S2 further includes: extracting the plateau lake water area by using the improved normalized differential water index according to the green light band and the short-wave infrared band in the atmospheric bottom reflectivity data.

Further, the plateau lake water body area is extracted by the following formula:

In the above formula, MNDWI represents the improved normalized difference water index; ρ _Green represents the reflectivity in the green light band in the bottom atmospheric reflectivity data; ρ _SWIR represents the reflectivity in the short-wave infrared band in the bottom atmospheric reflectivity data .

Further, in the step S3, according to the atmospheric bottom reflectance data, the red light absorption peak band is selected to construct a single band index.

Further, in the step S3, according to the atmospheric bottom reflectance data, the red light absorption peak band and the fluorescence peak band are selected to construct a ratio index; The difference between the absorption trough and the reflection peak of chlorophyll a.

Further, in the step S3, according to the atmospheric bottom reflectance data, select the red light absorption peak band and the fluorescence peak band; further enhance the selected red light absorption peak band and the fluorescence peak band by means of nonlinear stretching. Contrast of reflectance to construct normalized chlorophyll a index.

Further, in the step S3, according to the inherent optical properties of chlorophyll a and the reflectivity data of the bottom layer of the atmosphere, the red light absorption peak band, the fluorescence peak band and the absorption peak band of pure water under ideal conditions are respectively selected to construct a three-band index. ; The selected fluorescence peak band is close to the selected red absorption peak band.

Further, the step S5 further includes: constructing a linear expression composed of each of the chlorophyll a inversion spectral indices and the chlorophyll a concentration of the plateau lake according to the reflection characteristics of the plateau lake water body and the atmospheric bottom reflectance data.

Further, the linear expression formed by the single-band index and the chlorophyll a concentration of the plateau lake is:

C _chl-a =A+B· _ρa

In the above formula, C _chl-a represents the chlorophyll a concentration of the plateau lake; A and B are the first correlation coefficients of the regression model; ρ _a represents the water body reflectance in the red light absorption peak band.

Further, the linear expression formed by the ratio index and the chlorophyll a concentration of the plateau lake is:

In the above formula, C _chl-a represents the concentration of chlorophyll a in the plateau lake; C and D are the second correlation coefficients of the regression model; ρ _a and ρ _b represent the water reflection in the red light absorption peak band and the fluorescence peak band, respectively Rate.

Further, the linear expression formed by the normalized chlorophyll a index and the plateau lake chlorophyll a concentration is:

In the above formula, C _chl-a represents the concentration of chlorophyll a in the plateau lake; J and K are the third correlation coefficients of the regression model; ρ _a and ρ _b represent the water reflection in the red light absorption peak band and the fluorescence peak band, respectively Rate.

Further, the linear expression formed by the three-band index and the chlorophyll a concentration of the plateau lake is:

C _chl-a =M+N·(ρ _a ^-1 -ρ _b ^-1 )·ρ _c

In the above formula, C _chl-a represents the concentration of chlorophyll a in the plateau lake; M and N represent the fourth correlation coefficient of the regression model; ρ _a ^-1 and ρ _b ^-1 represent the red light absorption peak band and the fluorescence peak, respectively The reciprocal of the water body reflectivity in the wavelength band; ρ _c represents the water body reflectivity in the absorption peak band.

Further, the step S4 includes:

Inverting the spectral index according to the chlorophyll a, establishing a decision tree model, selecting the corresponding out-of-bag data to calculate the out-of-bag data error, and generating the first error;

adding random noise, respectively changing the value of the out-of-bag data corresponding to each of the chlorophyll-a inversion spectral indices, and calculating the out-of-bag data error again to generate a second error;

According to the first error and the second error, the feature importance of each of the chlorophyll a inversion spectral indices is calculated respectively, and the spectral index corresponding to the highest feature importance is determined;

The spectral index corresponding to the highest feature importance is taken as the spectral index that has the greatest influence on the result of chlorophyll a concentration, as the preferred spectral index.

Further, the feature importance of each of the chlorophyll a inversion spectral indices is calculated by the following formula:

In the above formula, V represents feature importance; err ₁ represents the first error; err ₂ represents the second error; N represents the number of decision trees.

The beneficial effects of the above technical solutions provided by the embodiments of the present invention include at least:

A method for remote sensing inversion of chlorophyll-a concentration in a plateau lake provided by an embodiment of the present invention includes: respectively acquiring the chlorophyll-a concentration of the plateau lake and the remote sensing reflectance data of the plateau lake; performing atmospheric correction on the remote sensing reflectance data of the plateau lake to obtain the bottom layer of the atmosphere Reflectance data; according to the inherent optical properties of chlorophyll a and the reflectance data of the bottom atmosphere, respectively construct the chlorophyll a inversion spectral index; according to the chlorophyll a inversion spectral index, establish a decision tree model to determine the spectral index that has the greatest impact on the chlorophyll a concentration results , as the optimal spectral index; perform linear regression on the optimal spectral index and the chlorophyll a concentration of the plateau lakes to establish a chlorophyll a inversion model; according to the chlorophyll a inversion model and the atmospheric bottom reflectance data, realize the remote sensing inversion of the chlorophyll a concentration of the plateau lakes. play. The method can effectively overcome the one-sidedness of manual field measurement, and solve the difficulty of measuring the spectrum of plateau lakes, and can realize rapid, accurate, and large-area monitoring of plateau lakes. Based on feature selection, various inversion spectral indices can be compared, and the most effective model for inversion of chlorophyll a concentration can be quickly selected, which further improves the inversion accuracy of chlorophyll a.

Other features and advantages of the present invention will be set forth in the description which follows, and in part will be apparent from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention may be realized and attained by the structure particularly pointed out in the written description, claims, and drawings.

The technical solutions of the present invention will be further described in detail below through the accompanying drawings and embodiments.

Description of drawings

The accompanying drawings are used to provide a further understanding of the present invention, and constitute a part of the specification, and are used to explain the present invention together with the embodiments of the present invention, and do not constitute a limitation to the present invention. In the attached image:

1 is a flowchart of a remote sensing inversion method for chlorophyll a concentration in plateau lakes provided by the embodiment of the present invention;

FIG. 2 is a schematic overall flowchart of an inversion method provided by an embodiment of the present invention.

Detailed ways

Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided so that the present disclosure will be more thoroughly understood, and will fully convey the scope of the present disclosure to those skilled in the art.

An embodiment of the present invention provides a remote sensing inversion method for chlorophyll a concentration in a plateau lake. Referring to FIG. 1 , the method includes the following steps:

S1. Obtain the chlorophyll a concentration of the plateau lakes and the remote sensing reflectance data of the plateau lakes respectively; the chlorophyll a concentration of the plateau lakes is obtained from the ground measuring points of the plateau lakes; the remote sensing reflectance data of the plateau lakes are obtained from the remote sensing satellites;

S2. Perform atmospheric correction on the remote sensing reflectivity data of the plateau lakes to obtain the bottom atmospheric reflectivity data;

S3. According to the inherent optical properties of chlorophyll a and the reflectivity data of the bottom atmosphere, respectively construct the chlorophyll a inversion spectral index; the chlorophyll a inversion spectral index includes: single-band index, ratio index, normalized chlorophyll-a index and three-band index;

S4, according to the chlorophyll a inversion spectral index, establish a decision tree model, and determine the spectral index that has the greatest impact on the chlorophyll a concentration result, as the preferred spectral index;

S5, perform linear regression on the optimal spectral index and the chlorophyll a concentration of the plateau lake, establish a chlorophyll a inversion model, and obtain the correlation coefficient of the regression model;

S6. According to the chlorophyll a inversion model and the atmospheric bottom reflectance data, the remote sensing inversion of the chlorophyll a concentration of the plateau lake is realized.

The remote sensing inversion method of chlorophyll a concentration in plateau lakes provided in this embodiment is suitable for plateau lakes. In the case of difficult acquisition of plateau lake spectral data, no field measurement is required, and plateau lakes can be quickly obtained only through the comparison of characteristic bands. The best model for chlorophyll a inversion realizes large-scale and periodic observation of plateau lakes, saves material and financial resources, and is of great significance to the intelligent monitoring and management of plateau lakes.

The following is a detailed description of the remote sensing inversion method of the chlorophyll a concentration of the plateau lake, and can be referred to Fig. 2. Fig. 2 is a schematic diagram of the overall flow of the inversion method provided in this embodiment:

Step 1. Obtain the chlorophyll a concentration data of the plateau lakes at the ground measuring points and the synchronized Sentinel-2MSI remote sensing reflectance data:

1) For the obtained plateau lake water samples of the ground measuring points, extract the pigment with acetone; through centrifugation, according to the absorption of the visible spectrum by the chlorophyll extract, use a laboratory spectrophotometer to measure the absorbance at its specific wavelength;

2) According to the empirical formula provided by the existing experimental principle, extract the content of chlorophyll a in the plateau lake in the lake sample;

3) Download Sentinel-2MSI remote sensing reflectance data (satellite data) through the official website of the United States Geological Survey (USGS).

Step 2: Perform atmospheric correction on the Sentinel-2MSI remote sensing reflectivity data to obtain the Sentinel-2MSI bottom atmospheric reflectivity data after atmospheric correction.

Step 3. Extract the plateau lake area through the atmospheric bottom reflectance data of Sentinel-2MSI after atmospheric correction:

Since Sentinel-2MSI atmospheric bottom reflectance data includes "water body" and "non-water body parts" (buildings, land, etc.), it is necessary to extract the plateau lake "water body" area, that is, the plateau lake area.

For the water body information of plateau lakes, the green light band (center wavelength 560nm) and the short-wave infrared band (center wavelength 1610nm) in satellite data are used for calculation, and the improved normalized difference water body index (MNDWI) is used to extract the water body region of plateau lakes. . MNDWI calculation results greater than 0 are plateau lake water areas; less than or equal to 0 are vegetation, land and other non-water areas. Its formula is expressed as:

In the formula, MNDWI represents the improved normalized difference water body index, ρ _Green represents the reflectivity in the green light band, and ρ _SWIR represents the reflectivity in the short-wave infrared band.

Step 4. Based on the reflection characteristics of the plateau lake water body, the inherent optical characteristics of chlorophyll a, and the bottom reflectance data of the Sentinel-2MSI atmosphere, construct the chlorophyll a inversion spectral index (including: single-band index, ratio index, normalized chlorophyll a index and three band index), and the linear expression formed by the inversion spectral index of each chlorophyll a and the concentration of chlorophyll a:

1) Construct a single band index:

The position of the reflection peak or absorption valley of the chlorophyll a reflection spectrum is mainly related to the red absorption peak band of Sentinel-2MSI. The linear expression formed by the single-band index and the concentration of chlorophyll a is:

C _chl-a =A+B· _ρa (2)

In the formula, C _chl-a represents the concentration of chlorophyll a in plateau lakes; A and B are the first correlation coefficients of the regression model; ρ _a is the water reflectance in the selected a band (the selected red light absorption peak band).

2) Construct the ratio index:

The position of the reflection peak or absorption valley of the chlorophyll a reflection spectrum is mainly related to the red absorption peak band and the fluorescence peak band of Sentinel-2MSI, and the two bands are selected as the a and b bands (respectively, the red absorption peak band and the fluorescence peak band). band, the central wavelengths are 665 nm and 708 nm, respectively), and the difference between the absorption valley and the reflection peak of chlorophyll a is enlarged by comparing ρ _a and ρ _b to construct a ratio index. The linear expression of ratio index and chlorophyll a concentration is:

In the formula, C _chl-a represents the concentration of chlorophyll a in plateau lakes; C and D are the second correlation coefficients of the regression model; ρ _a and ρ _b are the selected a and b bands (the red light absorption peak band and the fluorescence peak band), respectively. band) of water reflectance.

3) Construct the normalized chlorophyll a index:

The position of the reflection peak or absorption valley of the chlorophyll a reflection spectrum is mainly related to the red absorption peak band and the fluorescence peak band of Sentinel-2MSI, and the two bands are selected as the a and b bands (respectively, the red absorption peak band and the fluorescence peak band). band, the central wavelengths are 665nm and 708nm respectively), and the contrast of the reflectivity of the two bands is further enhanced by nonlinear stretching (the difference between ρ _a and ρ _b is compared with the sum of ρ _a and ρ _b ), and a normalized A chlorophyll a index. The linear expression formed by the normalized chlorophyll a index and the chlorophyll a concentration is:

In the formula, C _chl-a is the concentration of chlorophyll a in plateau lakes; J and K are the third correlation coefficients of the regression model; ρ _a and ρ _b are the selected a and b bands (the red light absorption peak band and the fluorescence peak band), respectively. band) of water reflectance.

4) Construct a three-band index:

Based on the chlorophyll a bio-optical model, according to the model requirements, select the red light absorption peak band, its adjacent fluorescence peak band, and the absorption peak band of pure water under ideal conditions. Combined with the band settings of Sentinel-2MSI, select this band The three bands are denoted as a, b and c bands (respectively, the red light absorption peak band, the fluorescence peak band and the absorption peak band, and the central wavelengths are 665 nm, 708 nm and 730 nm, respectively), and a three-band index is constructed. The three-band index is related to chlorophyll a The linear expression formed by the concentration is:

C _chl-a =M+N·(ρ _a ^-1 -ρ _b ^-1 )·ρ _c (5)

In the formula, C _chl-a represents the concentration of chlorophyll a in plateau lakes; M and N are the fourth correlation coefficients of the regression model; ρ _a ^-1 and ρ _b ^-1 are the selected a and b bands (the selected red light absorption peak), respectively. _ρc is the inverse of the water reflectance in the selected c-band (the selected absorption peak band).

Step 5. Random forest algorithm feature importance measurement:

1) Random forest refers to a classifier that uses multiple trees to train and predict samples. For each decision tree, select the corresponding out-of-bag data to calculate the out-of-bag data error, denoted as err ₁ . When establishing the decision tree model, the spectral index established by the combination of bands (the chlorophyll a inversion spectral index) is used as the input of the feature of the decision tree, and the simulated training is carried out through the measured chlorophyll a concentration, and the output spectrum with high correlation with the chlorophyll a concentration is output. index. Among them, out-of-bag data means that every time a decision tree is built, a data is obtained by repeated sampling for training the decision tree. At this time, about 1/3 of the data is not used and does not participate in the establishment of the decision tree. This part of the data can be used to evaluate the performance of the decision tree and calculate the prediction error rate of the model, which is called the out-of-bag data error.

2) Randomly change the value of out-of-bag data at a certain feature, that is, add random noise, change the value of out-of-bag data under different spectral indices, and calculate the error of out-of-bag data again, which is recorded as err ₂ . Among them, the spectral index includes: single-band index, ratio index, normalized chlorophyll a index and three-band index.

3) Add random noise. If the accuracy rate of out-of-bag data drops significantly, that is, err ₁ decreases and err ₂ increases, indicating that this feature has a high degree of importance, which in turn indicates that this feature (specifically refers to the spectral index) has a great influence on the results of chlorophyll a concentration. . Assuming that there are N decision trees in the forest, the feature importance is calculated separately, and the spectral index that has a great influence on the chlorophyll a concentration result is obtained. The general formula is as follows:

In the formula, V represents the feature importance, err ₁ represents the out-of-bag data error, err ₂ represents the out-of-bag data error after changing the value of a feature, and N represents the number of decision trees in the random forest.

4) Arrange the feature importance calculation results from high to low, that is, the arrangement of the effect of spectral index on chlorophyll a concentration, and obtain the spectral index with the highest feature importance.

Step 6. Inversion of chlorophyll a concentration in plateau lakes:

1) Obtain the spectral index with the highest importance of the random forest feature as the preferred spectral index. According to formula (2)(3)(4)(5), the preferred spectral index is used as the independent variable, and the linear regression is performed with the dependent variable chlorophyll a concentration to obtain The correlation coefficient of the regression model (such as A, B in the formula (2)), this regression model is the chlorophyll a inversion model. Among them, the above formulas (2) (3) (4) (5) are linear expressions of the regression model.

2) Apply the chlorophyll a inversion model to the atmospheric bottom reflectance data of Sentinel-2MSI after atmospheric correction, and use the optimal spectral index as an independent variable to obtain the correlation coefficient of the model after linear regression, and realize the remote sensing inversion of chlorophyll a concentration in plateau lakes. . For example, for the plateau lakes in Yunnan Province, taking Dianchi Lake as an example, the final linear expression formula of the regression model is:

In the above formula, C _chl-a represents the concentration of chlorophyll a in plateau lakes; ρ _a and ρ _b represent the water reflectance in the red light absorption peak band and the fluorescence peak band, respectively.

The remote sensing inversion method for chlorophyll a concentration in plateau lakes provided in this embodiment adopts remote sensing quantitative inversion. Compared with traditional monitoring methods, it can realize rapid, accurate, and large-area monitoring of plateau lakes, which is suitable for complex and diverse geological and landforms. Highland lake water quality monitoring provides a more convenient means. Moreover, the method is simple and practical. Based on feature selection, various inversion indices can be compared, and for a certain plateau lake, the most effective model for inversion of chlorophyll a concentration can be quickly selected (establish a chlorophyll a inversion model), and further. Improved chlorophyll a inversion accuracy. It is suitable for plateau lakes. When spectral data of plateau lakes is difficult to obtain, field measurements are not required. Only through the comparison of characteristic bands, the best model for chlorophyll a inversion of plateau lakes can be quickly obtained, and a large-scale plateau lake can be realized. , Periodic observation, saving material and financial resources, is of great significance to the intelligent monitoring and management of plateau lakes.

It will be apparent to those skilled in the art that various modifications and variations can be made in the present invention without departing from the spirit and scope of the invention. Thus, provided that these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims (9)

1. a plateau lake chlorophyll a concentration remote sensing inversion method, is characterized in that, comprises the steps: S1. Obtain the chlorophyll a concentration of the plateau lake and the remote sensing reflectance data of the plateau lake respectively; the chlorophyll a concentration of the plateau lake is obtained from the ground measuring point of the plateau lake; S2, performing atmospheric correction on the remote sensing reflectivity data of the plateau lakes to obtain bottom atmospheric reflectivity data; S3. According to the intrinsic optical properties of chlorophyll a and the reflectivity data of the bottom layer of the atmosphere, respectively construct a chlorophyll a inversion spectral index; the chlorophyll a inversion spectral index includes: a single-band index, a ratio index, a normalized chlorophyll a index and three-band index; S4, inverting the spectral index according to the chlorophyll a, establishing a decision tree model, and determining the spectral index that has the greatest impact on the chlorophyll a concentration result, as the preferred spectral index; S5, perform linear regression on the preferred spectral index and the chlorophyll a concentration of the plateau lake, establish a chlorophyll a inversion model, and obtain the correlation coefficient of the regression model; S6. According to the chlorophyll a inversion model and the atmospheric bottom reflectance data, the remote sensing inversion of the chlorophyll a concentration of the plateau lake is realized. 2. a kind of plateau lake chlorophyll a concentration remote sensing inversion method as claimed in claim 1, is characterized in that, described step S2 also comprises: according to the green light wave band and short wave infrared wave band in described atmospheric bottom reflectivity data, The improved normalized differential water body index was used to extract the water body area of plateau lakes. 3. a kind of plateau lake chlorophyll a concentration remote sensing inversion method as claimed in claim 2, is characterized in that, extracts plateau lake water body area by following formula:

In the above formula, MNDWI represents the improved normalized difference water index; ρ _Green represents the reflectivity in the green light band in the bottom atmospheric reflectivity data; ρ _SWIR represents the reflectivity in the short-wave infrared band in the bottom atmospheric reflectivity data . 4. A kind of remote sensing inversion method of chlorophyll a concentration in plateau lakes as claimed in claim 1, it is characterized in that, in described step S3, according to described atmospheric bottom reflectivity data, select red light absorption peak waveband, construct single waveband index. 5. A kind of remote sensing inversion method of chlorophyll a concentration in plateau lakes as claimed in claim 1, it is characterized in that, in described step S3, according to described atmospheric bottom reflectivity data, select red light absorption peak band and fluorescence peak band , to construct the ratio index. 6. A kind of remote sensing inversion method of chlorophyll a concentration in plateau lakes as claimed in claim 1, it is characterized in that, in described step S3, according to described atmospheric bottom reflectivity data, select red light absorption peak band and fluorescence peak band ; Contrast the reflectance of the selected red light absorption peak band and fluorescence peak band by nonlinear stretching to construct a normalized chlorophyll a index. 7. A kind of remote sensing inversion method of chlorophyll a concentration in plateau lakes as claimed in claim 1, it is characterized in that, in described step S3, according to chlorophyll a intrinsic optical characteristic and described atmospheric bottom reflectivity data, select red light respectively The absorption peak band, the fluorescence peak band and the absorption peak band of pure water under ideal conditions are used to construct a three-band index; the selected fluorescence peak band is close to the selected red light absorption peak band. 8. a kind of plateau lake chlorophyll a concentration remote sensing inversion method as claimed in claim 1, is characterized in that, described step S4 comprises: Inverting the spectral index according to the chlorophyll a, establishing a decision tree model, selecting the corresponding out-of-bag data to calculate the out-of-bag data error, and generating the first error; adding random noise, respectively changing the value of the out-of-bag data corresponding to each of the chlorophyll-a inversion spectral indices, and calculating the out-of-bag data error again to generate a second error; According to the first error and the second error, the feature importance of each of the chlorophyll a inversion spectral indices is calculated respectively, and the spectral index corresponding to the highest feature importance is determined; The spectral index corresponding to the highest feature importance is taken as the spectral index that has the greatest influence on the chlorophyll a concentration result, as the preferred spectral index. 9. a kind of plateau lake chlorophyll a concentration remote sensing inversion method as claimed in claim 8, is characterized in that, calculates the characteristic importance of each described chlorophyll a inversion spectral index by following formula:

In the above formula, V represents feature importance; err ₁ represents the first error; err ₂ represents the second error; N represents the number of decision trees.

Images