A method for inversion of water content and dry matter content of mangroves based on fractional-order assisted spectral angle mapping

Abstract

This paper proposes a method for inverting mangrove water content and dry matter content based on fractional-order assisted spectral angle mapping. The method includes: acquiring mangrove functional trait parameters, leaf hyperspectral data, and multi-source satellite remote sensing image data; proposing a novel fractional-order assisted spectral angle analysis method; proposing satellite-ground spectral matching and inversion verification methods to construct, reduce, and optimize the inversion dataset; and proposing multiple inversion schemes that fuse optical, radar, and thermal infrared data to construct a multi-source hybrid dataset and a hybrid estimation model for mangrove canopy water content and dry matter content. The method can match the sensitive spectral domain obtained from ground-based spectra with satellite bands, thereby enabling widespread and accurate inversion of mangrove water content and dry matter content, enabling low-cost, accurate, and rapid monitoring of mangrove growth status and coastal environment.

Description

Technical Field

[0001] The present invention belongs to the field of wetland vegetation monitoring, and in particular relates to a method for inverting the water content and dry matter content of a typical mangrove community. Background Art

[0002] Mangrove wetlands are among the world's most biodiverse ecosystems and play a crucial role in achieving the United Nations Sustainable Development Goals (SDGs). Leaf equivalent water thickness (EWT) and leaf mass per unit area (LMA) are key functional traits and important indicators for monitoring mangrove water content and dry matter content. Quantitative inversion of mangrove EWT and LMA is crucial for monitoring mangrove growth, observing coastal environmental changes, and maintaining marine and terrestrial biodiversity.

[0003] In recent years, active and passive remote sensing technologies have flourished alongside the global trend toward big data, and have proven their effectiveness in the quantitative estimation of vegetation functional trait parameters. For example, patent application 202211220766.1 discloses a mangrove extraction method that takes into account both phenological and water-level temporal characteristics. The method involves screening multispectral remote sensing images; constructing raw NDVI and MNDWI time series; reconstructing the NDVI and MNDWI time series for each pixel using a harmonic model; constructing a mangrove forest index (PWTMI) that takes into account both phenological and water-level temporal characteristics and calculating the PWTMI image for the target year; determining a threshold range based on the PWTMI statistical results for the sample to extract the preliminary mangrove area; and removing misclassified pixels to obtain the final mangrove extraction result. For example, patent application 202311282634.6 discloses a mangrove ecological detection method based on remote sensing image recognition, which belongs to the field of image processing technology. In the present invention, remote sensing images of mangrove ecological areas are collected, and the remote sensing images are first grayscale processed and contours are extracted to obtain a contour map. The contour map is used to extract contour features, and the remote sensing image is used to extract color features. The color features and contour features are combined to identify the mangrove targets, thereby determining the area of ​​the mangroves and realizing the monitoring and detection of the mangrove ecology.

[0004] However, mangroves grow on intertidal mudflats, making sampling extremely difficult and dangerous. The spectral response mechanisms of mangroves for water content and dry matter mass remain unclear, making quantitative inversion using remote sensing satellites difficult. Furthermore, due to the inherent spectral heterogeneity of vegetation, the spectral response mechanisms of other vegetation cannot be simply transferred to the quantitative estimation of mangrove functional trait parameters.

[0005] Therefore, there is an urgent need for a method that captures the sensitive spectral domains of different functional trait parameters based on ground-based hyperspectral data and applies it to satellites to achieve large-scale, stable and high-precision monitoring of functional trait parameters of mangrove communities. Summary of the Invention

[0006] Based on this, the primary purpose of the present invention is to provide a method for inverting the water content and dry matter content of mangroves based on fractional-order auxiliary spectral angle mapping. This method can match the sensitive spectral domain obtained based on the ground spectrum with the satellite band, and thus can widely and accurately realize the inversion of the water content and dry matter content of mangroves, so as to realize the monitoring of the growth status of mangroves and coastal environment at low cost, accurately and quickly.

[0007] Another object of the present invention is to provide a method for inverting the water content and dry matter content of mangroves based on fractional-order assisted spectral angle mapping. This method couples fractional-order derivatives and vector-based improved spectral angle mapping methods, which can accurately capture subtle spectral differences between mangrove communities, and combines multi-source remote sensing data to achieve spectral matching, as well as combining thermal infrared and SAR data with integrated learning algorithms to achieve high-precision inversion of the water content and dry matter content of mangrove communities.

[0008] To achieve the above object, the technical solution of the present invention is:

[0009] A method for inverting water content and dry matter content of mangroves based on fractional-order assisted spectral angle mapping includes the following steps:

[0010] Step S101: Obtain optical remote sensing satellite images of the target mangrove area, calculate the equivalent water thickness (EWT) and leaf mass per unit area (LMA) of the mangrove samples as indicators of mangrove water content and dry matter content, and collect ground-based hyperspectral data (350-2500nm) of different types of mangrove communities.

[0011] Step S102: Constructing a fractional-order assisted spectral angle analysis method involves setting up two unique spectral vector forms to improve the spectral angle mapping method and simultaneously performing fractional-order derivative (FOD) processing on the in-situ hyperspectral data of the mangrove community. The spectral angles after in-situ and FOD processing are calculated, and the optimal order range for spectral analysis is determined. The Pearson correlation coefficient R and standard deviation of the spectra within this range with the EWT and LMA are calculated, and the optimal order is selected to capture the sensitive spectral domain.

[0012] Step S103: Construct a high-dimensional dataset based on the original satellite bands, perform dimensionality reduction and optimization, set a spectral matching index (SMI), verify the consistency between the mangrove-sensitive spectral domain obtained from the ground-based hyperspectral data and the sensitive bands of multiple corresponding optical satellites, and use these satellite bands to perform inversion verification.

[0013] Step S104: Based on the inversion verification using optical satellites, thermal infrared image data and radar image data are obtained, and the sensitive features of multi-source satellites are refined and combined into different inversion methods to construct a hybrid data set. Based on the ensemble learning algorithm, an inversion model for mangrove water content and dry matter content is constructed, and the determination coefficient and root mean square error are used for accuracy evaluation to obtain the optimal inversion scheme for mangrove water content and dry matter content.

[0014] Furthermore, in step S101, the leaf equivalent water thickness (EWT) and leaf mass per unit area (LMA) are measured as follows: the fresh weight (FW) of each mangrove sample is weighed and recorded using a 1 / 1000 electronic balance, and the leaf area (A) is measured. The leaves are then dried in an oven at 100 degrees for 72 hours, and the dry weight (DW) of the leaves is recorded. EWT and LMA are calculated according to the following formulas:

[0015]

[0016]

[0017] Furthermore, in step S102, the present invention sets two unique spectral vector forms θ(1) and θ(2) to improve the traditional spectral angle mapping method, and expresses the similarity between spectra through spectral angles; θ(1) reflects the numerical difference in reflectance of the two mangrove community spectra at wavelength w, and θ(2) characterizes the difference in spectral curve morphology of the two mangrove community spectra at wavelength w. The specific formula is as follows:

[0018]

[0019]

[0020] where θ i，j，w represents the spectral angle between mangrove community i and mangrove community j at wavelength i, represents the spatial vector of the spectral curve of mangrove i at wavelength w, h is the step size, rw represents the reflectance of the spectral curve at wavelength w. Since the interval of this hyperspectral data is 1 nm, h is set to 1.

[0021] Furthermore, FOD processing is performed on the in-situ hyperspectral data to provide a higher-dimensional amount of spectral information for the vector-based improved spectral angle mapping method. The FOD definition formula is as follows:

[0022]

[0023] Where α is the order of the derivative and α>0; a and b are the upper and lower limits of the difference, respectively; [(ba) / h] is the integer part of (ba) / h, and h is the step size:

[0024] Substituting the gamma function into the FOD definition formula and setting h to 1, we obtain the following formula:

[0025]

[0026] Where α = 0, 1, and 2 represent the original spectrum, the first-order derivative, and the second-order derivative, respectively. The present invention sets h to 0.25 and calculates the FOD spectrum transformation of the original spectrum from 0 to 2 orders to obtain 8 different trend transformation spectra for each mangrove community.

[0027] Furthermore, the Pearson correlation coefficient (R) and standard deviation (STD) of the in situ and FOD-processed hyperspectral data were calculated with the EWT and LMA measured in different mangrove groups. The principles for selecting the optimal fractional order for spectral analysis include:

[0028] ① Select the two orders with the highest R-mean between the mangrove community spectrum and EWT / LMA to ensure high overall correlation;

[0029] ② Select the one with higher STD among the two orders as the optimal fractional order for spectral analysis to ensure obvious correlation differences.

[0030] Furthermore, in step S103, the reflectivity of each band of the multi-source optical satellite is extracted, and the combination index is calculated to construct a high-dimensional data set. The calculation formula is as follows:

[0031] SI2=b i -b j ,

[0032] Where i, j, k, and l represent the different bands of the optical satellite used;

[0033] The Pearson correlation coefficient between the features in the high-dimensional data set and the measured EWT and LMA was calculated, and the correlation threshold was set to perform data dimensionality reduction. Maximum and minimum normalization, Box-Cox transformation, logarithmic transformation of target values, and outlier removal were performed.

[0034] Furthermore, a spectral matching index (SMI) is constructed to perform consistency testing on the sensitive spectral domain obtained by ground-based hyperspectral measurements and the sensitive bands obtained by optical satellites. The SMI definition formula is as follows:

[0035]

[0036] Where Y is the mangrove sensitive spectral domain W = {w1, w2, ..., w s} is a set of features composed of the corresponding bands yi in the satellite, and X is a feature set composed of the features Xi after dimensionality reduction of the optical satellite, where X i ={x j ,...,x k}, xi is the corresponding satellite band in each satellite feature;

[0037] The inversion and verification of mangrove EWT and LMA are performed using multiple optical satellite bands corresponding to the ground-based sensitive spectral domain. The formula for the verification model is as follows:

[0038] T=F(SI1(X′),SI2(X′),...,SI k (X′))

[0039]

[0040] Where T is the mangrove EWT or LMA estimation result, F is the estimation model, SI is the combined index calculated by the satellite band, and X′ is the screened satellite sensitivity feature.

[0041] Furthermore, in step S104, based on the high correlation features of the three optical satellites and with the GF-3SAR data and SDGSAT TIS data as auxiliary features, a hybrid data set is obtained as follows:

[0042] D (3×4) =D 基础 +D 辅助

[0043]

[0044] At the same time, an adaptive ensemble learning model was developed based on two traditional statistical models (Ridge, PLSR) and three machine learning models (K-NearestNeighbors, Random Forest and XGBoost) algorithms to quantitatively evaluate the inversion performance of the different inversion schemes for the water content and dry matter content of the mangroves.

[0045] The coefficient of determination (R2) and root mean square error (RMSE) were used to evaluate the accuracy of the inversion model for each mangrove community. The calculation formula of the evaluation index is as follows:

[0046]

[0047] Where n is the number of mangrove samples; yi and y are i are the measured and estimated results of functional trait parameters, respectively; is the average value of the measured results of functional trait parameters.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] The present invention can accurately capture the subtle spectral differences between mangrove communities by coupling fractional derivatives and a vector-based improved spectral angle mapping method, and accurately obtain their sensitive spectral domains for different functional trait parameters. It also combines multi-source remote sensing data to achieve spectral matching and combines thermal infrared and SAR data with an integrated learning algorithm to achieve high-precision inversion of the water content and dry matter content of mangrove communities.

[0050] At the same time, taking mangroves as an example, the feasibility of the fractional-order auxiliary spectral angle analysis method to reveal the spectral mechanism of important functional trait parameters was demonstrated, and the spectral matching technology was used to apply the sensitive spectral domain to satellites, breaking through the limitation of the existing optical inversion of vegetation functional trait parameters being limited to the visible light and near-infrared bands, and expanding it to the thermal infrared band, thereby realizing the inversion of mangrove water content and dry matter content in a broad and accurate manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a flow chart of the implementation of the present invention.

[0052] Figure 2 This is an example of the fractional-order assisted spectral angle analysis method implemented by the present invention.

[0053] Figure 3 Schematic diagram of hybrid data set construction implemented by the present invention.

[0054] Figure 4 It is a schematic diagram of the integrated learning model implemented by the present invention.

[0055] Figure 5 This is a comparison chart of the inversion results of mangrove water content and dry matter content under different schemes. DETAILED DESCRIPTION

[0056] In order to make the objectives, technical solutions, and advantages of the present invention more clearly understood, the technical methods of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, what is described is only a portion of the embodiments of the present application, not all of them. In addition, in the following description, descriptions of well-known structures and prior arts are omitted to avoid unnecessary confusion of the concepts of the present invention.

[0057] Figure 1 As shown in FIG, the method for inverting the water content and dry matter content of mangroves based on fractional-order auxiliary spectral angle mapping implemented by the present invention includes the following steps:

[0058] Step S101: Obtain optical remote sensing satellite images of the target mangrove area, calculate the equivalent water thickness (EWT) and leaf mass per unit area (LMA) of the mangrove samples as indicators of mangrove water content and dry matter content, and collect ground-based hyperspectral data (350-2500nm) of different types of mangrove communities.

[0059] The present invention acquires Sentinel-2A MSI imagery (458-2280nm), SDGSAT-1MII imagery (374-911nm), and OHS-3D CMOS imagery (443-940nm). Sentinel-2A MSI and SDGSAT-1MII imagery are L2A and L4B level data, requiring no additional processing. This paper uses OpenOHS software to perform radiometric calibration, atmospheric correction, and orthorectification on the OHS-3D CMOS, ultimately obtaining reflectance data for each optical image. The optical images described are not limited to the aforementioned satellite images.

[0060] The equivalent water thickness (EWT) and leaf mass per unit area (LMA) of leaves were measured in the laboratory. The fresh weight (FW) of each mangrove leaf was weighed and recorded using a 1 / 1000th electronic balance. The leaf area (A) was measured using a LI-3000C (LI-COR, USA). The leaves were then dried in an oven at 100°C for 72 hours. The dry weight (DW) of the leaves was recorded. EWT and LMA were calculated using the following formulas:

[0061]

[0062]

[0063] In situ hyperspectral data (350-2500nm) of mangrove communities were acquired using a portable American ASD spectrometer. Before performing leaf hyperspectral measurements, the ASD instrument was preheated for at least 20 minutes and the field of view was set to 25°. Calibration was performed using the calibration white plate at the bottom of the handheld leaf holder, and spectral radiometry measurements were converted to reflectance measurements. During the measurement process, a leaf from a sample was placed in the handheld leaf holder with its front facing the light source. After the spectral curve stabilized, 10 spectral curves were saved. This step was repeated seven times with each leaf, and the instrument was recalibrated after each sample was measured to ensure data accuracy and consistency. Ultimately, 70 spectral curves were collected for each mangrove sample, and the average was taken as the final spectral data for that leaf sample.

[0064] Step S102: Constructing a fractional-order assisted spectral angle analysis method involves setting up two unique spectral vector forms to improve the spectral angle mapping method and simultaneously performing fractional-order derivative (FOD) processing on the in-situ hyperspectral data of the mangrove community. The spectral angles after in-situ and FOD processing are calculated, and the optimal order range for spectral analysis is determined. The Pearson correlation coefficient R and standard deviation of the spectra within this range with the EWT and LMA are calculated, and the optimal order is selected to capture the sensitive spectral domain.

[0065] The present invention improves the traditional spectral angle mapping method by setting two unique spectral vector forms. The spectral angle shows the similarity between spectra. The smaller the angle, the higher the similarity between spectra. θ(1) reflects the numerical difference in reflectance of the two mangrove community spectra at wavelength w, and θ(2) characterizes the difference in spectral curve morphology of the two mangrove community spectra at wavelength w. The specific formula is as follows:

[0066]

[0067]

[0068] where θ i，j，w represents the spectral angle between mangrove community i and mangrove community j at wavelength i, represents the spatial vector of the spectral curve of mangrove i at wavelength w, h is the step size, rw represents the reflectance of the spectral curve at wavelength w. Since the interval of this hyperspectral data is 1 nm, h is set to 1.

[0069] In the present invention, FOD processing is performed on the in-situ hyperspectral image to provide a higher-dimensional spectral information quantity for the vector-based improved spectral angle mapping method. The FOD definition formula is as follows:

[0070]

[0071] Where α is the order of the derivative and α>0; a and b are the upper and lower limits of the difference, respectively; [(ba) / h] is the integer part of (ba) / h, and h is the step size. The gamma function is defined as follows:

[0072]

[0073] Substituting the gamma function into the FOD definition formula and setting h to 1, we can obtain the following formula:

[0074]

[0075] Where α = 0, 1, and 2 represent the original spectrum, the first-order derivative, and the second-order derivative, respectively. The present invention sets h to 0.25 and calculates the FOD spectrum transformation of the original spectrum from 0 to 2 orders to obtain 8 different trend transformation spectra for each mangrove community.

[0076] The Pearson correlation coefficient (R) and standard deviation (STD) of the in situ and FOD-processed hyperspectral data were calculated with the measured EWT and LMA of different mangrove communities. The principles for selecting the optimal fractional order for spectral analysis include:

[0077] ① Select the two orders with the highest R-mean between the mangrove community spectrum and EWT / LMA to ensure high overall correlation;

[0078] ② Select the one with higher STD among the two orders as the optimal fractional order for spectral analysis to ensure obvious correlation differences.

[0079] The present invention screens out a spectral band range above a correlation coefficient threshold as a mangrove-sensitive spectral domain.

[0080] Step S103: Construct a high-dimensional dataset based on the original satellite bands, perform dimensionality reduction and optimization, set a spectral matching index, verify the consistency between the mangrove-sensitive spectral domain obtained from the ground-based hyperspectral data and the sensitive bands of multiple corresponding optical satellites, and use these satellite bands to perform inversion verification.

[0081] The present invention extracts the reflectivity of each band of multi-source optical satellites, calculates the combination index, and constructs a high-dimensional data set. The calculation formula is as follows:

[0082] SI2=b i -b j ,

[0083] Where i, j, k, and 1 represent the different bands of the optical satellite used.

[0084] In addition, the Pearson correlation coefficients between the features in the high-dimensional dataset and the measured EWT and LMA were calculated, and the correlation threshold was set to perform data dimensionality reduction. Maximum and minimum normalization, Box-Cox transformation, logarithmic transformation of target values, and outlier removal were performed.

[0085] In this paper, a spectral matching index (SMI) is constructed to check the consistency between the sensitive spectral domain obtained by ground-based hyperspectral measurements and the sensitive bands obtained by optical satellites. Its value range is between 0 and 1. When the value is closer to 1, it indicates that the sensitive bands obtained by ground-based hyperspectral measurements and optical satellite reflectance are more consistent. The SMI definition formula is as follows:

[0086]

[0087] Where Y is the mangrove sensitive spectral domain W = {w1, w2, ..., w s} is a set of features composed of the corresponding bands yi in the satellite, and X is a feature set composed of the features Xi after dimensionality reduction of the optical satellite, where X i ={x j ,...,x k}, xi is the corresponding satellite band in each satellite feature.

[0088] The inversion and verification of mangrove EWT and LMA are performed using multiple optical satellite bands corresponding to the ground-based sensitive spectral domain. The formula for the verification model is as follows:

[0089] T=F(SI1(X′),SI2(X′),...,SI k (X′))

[0090]

[0091] Where T is the mangrove EWT or LMA estimation result, F is the estimation model, SI is the combined index calculated by the satellite band, and X′ is the screened satellite sensitivity feature.

[0092] Step S104: Based on the inversion verification using optical satellites, thermal infrared image data and radar image data are obtained, and the sensitive features of multi-source satellites are refined and combined into different inversion methods to construct a hybrid data set. Based on the ensemble learning algorithm, an inversion model for mangrove water content and dry matter content is constructed, and the determination coefficient and root mean square error are used for accuracy evaluation to obtain the optimal inversion scheme for mangrove water content and dry matter content.

[0093] The present invention is based on the highly correlated features of three optical satellites (SDGSAT-1MII, Sentinel-2AMSI, and OHS-03D CMOS), with GF-3SAR data and SDGSAT TIS data as auxiliary features. Finally, a hybrid dataset is obtained as follows:

[0094] D (3×4) =D 基础 +D 辅助

[0095]

[0096] An adaptive ensemble learning model was developed based on two traditional statistical models (Ridge, PLSR) and three machine learning models (K-NearestNeighbors, Random Forest and XGBoost) to quantitatively evaluate the inversion performance of the different inversion schemes for the water content and dry matter content of the mangroves.

[0097] The present invention uses the coefficient of determination (R2) and the root mean square error (RMSE) to evaluate the accuracy of the inversion model of each mangrove community. The calculation formula of the evaluation index is as follows:

[0098]

[0099] Where n is the number of mangrove samples; yi and y are i are the measured and estimated results of functional trait parameters, respectively; is the average value of the measured results of functional trait parameters.

[0100] In summary, the present invention has developed a method for inverting the water content and dry matter content of mangroves based on fractional-order assisted spectral angle mapping. This method couples fractional-order derivatives and vector-based improved spectral angle mapping methods, which can identify subtle differences between mangrove spectra and accurately capture the sensitive spectral domains of different mangrove functional trait parameters. It is applied to satellite inversion through spectral matching technology, and uses novel multiple inversion schemes that integrate optical, radar and thermal infrared data. The developed integrated learning algorithm is used to invert the water content and dry matter content of mangroves, breaking through the limitation that the inherent bands of multispectral sensors are limited to the visible light and near-infrared spectral bands, providing a basic scientific basis for monitoring the growth status of mangroves and environmental changes in coastal areas.

[0101] In addition, the fractional-order auxiliary spectral angle analysis method proposed in the present invention is not limited to the spectral analysis of mangrove communities, but can also be expanded to other spectral analysis fields such as vegetation and water bodies.

[0102] The specific embodiments described above are exemplary and non-limiting. Various solutions and all variations thereof that can be devised by those skilled in the art based on the disclosure of the present invention are encompassed by the present invention.

Claims

1. A method for inverting water content and dry matter content of mangroves based on fractional-order auxiliary spectral angle mapping, characterized in that: It includes the following steps: Step S101: Obtain optical remote sensing satellite images of the target mangrove area, calculate the equivalent water thickness (EWT) and leaf mass per unit area (LMA) of the mangrove samples as indicators of mangrove water content and dry matter content, and simultaneously collect ground-based hyperspectral data of different types of mangrove communities; Step S102: Constructing a fractional-order assisted spectral angle analysis method, including setting two unique spectral vector forms to improve the spectral angle mapping method, and simultaneously performing fractional-order derivative (FOD) processing on the in-situ hyperspectral data of the mangrove community; calculating the spectral angles after in-situ and FOD processing, determining the optimal order range of spectral analysis, calculating the Pearson correlation coefficient R and its standard deviation between the spectrum within this range and the EWT and LMA, and selecting the optimal order to capture the sensitive spectral domain; In step S102, two unique spectral vector forms θ(1) and θ(2) are set to improve the traditional spectral angle mapping method, and the similarity between spectra is expressed through spectral angles; θ(1) reflects the difference in reflectance values ​​of the two mangrove community spectra at wavelength w, and θ(2) represents the difference in spectral curve morphology of the two mangrove community spectra at wavelength w. The specific formula is as follows: where θ i,j,w represents the spectral angle between mangrove community i and mangrove community j at wavelength i, represents the spatial vector of the spectrum curve of mangrove i at wavelength w, h is the step size, rw represents the reflectance of the spectrum curve at wavelength w, the hyperspectral data interval is 1nm, and h is set to 1; Step S103: construct a high-dimensional dataset based on the original satellite bands, perform dimensionality reduction and optimization, set a spectral matching index (SMI), verify the consistency between the mangrove-sensitive spectral domain obtained from the ground-based hyperspectral data and the sensitive bands of multiple corresponding optical satellites, and use these satellite bands to perform inversion verification. Step S104: Based on the inversion verification using optical satellites, thermal infrared image data and radar image data are obtained, and the sensitive features of multi-source satellites are refined and combined into different inversion methods to construct a hybrid data set. Based on the ensemble learning algorithm, an inversion model for mangrove water content and dry matter content is constructed, and the determination coefficient and root mean square error are used for accuracy evaluation to obtain the optimal inversion scheme for mangrove water content and dry matter content.

2. The method for inverting water content and dry matter content of mangroves based on fractional-order auxiliary spectral angle mapping according to claim 1 is characterized in that: In step S101, the leaf equivalent water thickness (EWT) and leaf mass per unit area (LMA) were measured as follows: the fresh weight (FW) of each mangrove leaf was weighed and recorded using a 1 / 1000 electronic balance, and the leaf area (A) was measured. The leaves were then dried in an oven at 100 degrees for 72 hours, and the dry weight (DW) of the leaves was recorded. EWT and LMA were calculated according to the following formulas:

3. The method for inverting water content and dry matter content of mangroves based on fractional-order auxiliary spectral angle mapping according to claim 1 is characterized in that: In step S102, FOD processing is performed on the in-situ hyperspectral spectrum to provide a higher-dimensional spectral information quantity for the vector-based improved spectral angle mapping method. The FOD definition formula is as follows: Where α is the order of the derivative and α>0; a and b are the upper and lower limits of the difference, respectively; [(ba) / h] is the integer part of (ba) / h, and h is the step size; Substituting the gamma function into the FOD definition formula and setting h to 1, we obtain the following formula: Where α = 0, 1 and 2 represent the original spectrum, first-order derivative and second-order derivative, respectively; h is set to 0.25, and by calculating the FOD spectral transformation of the original spectrum from 0 to 2 orders, 8 different trend transformation spectra of each mangrove community are obtained.

4. The method for inverting water content and dry matter content of mangroves based on fractional-order assisted spectral angle mapping according to claim 3 is characterized in that: In step S102, the Pearson correlation coefficient (R) and its standard deviation (STD) are calculated between the hyperspectral data after in situ and FOD processing and the EWT and LMA measured in different mangrove groups. The selection principles of the optimal fractional order of spectral analysis include: ① Select the two orders with the highest R-mean between the mangrove community spectrum and EWT / LMA to ensure high overall correlation; ② Select the one with higher STD among the two orders as the optimal fractional order for spectral analysis to ensure obvious correlation differences.

5. The method for inverting water content and dry matter content of mangroves based on fractional-order auxiliary spectral angle mapping according to claim 1 is characterized in that: In step S103, the reflectivity of each band of the multi-source optical satellite is extracted, and the combination index is calculated to construct a high-dimensional data set. The calculation formula is as follows: Where i, j, k, and l represent the different bands of the optical satellite used; The Pearson correlation coefficient between the features in the high-dimensional data set and the measured EWT and LMA was calculated, and the correlation threshold was set to perform data dimensionality reduction. Maximum and minimum normalization, Box-Cox transformation, logarithmic transformation of target values, and outlier removal were performed.

6. The method for inverting water content and dry matter content of mangroves based on fractional-order auxiliary spectral angle mapping according to claim 5 is characterized in that: In step S103, a spectral matching index SMI is constructed to perform consistency check on the sensitive spectral domain obtained by ground-measured hyperspectral and the sensitive band obtained by optical satellite. The definition formula of SMI is as follows: Where Y is the mangrove sensitive spectral domain W = {w1,w2,...,w s } is a set of features composed of the corresponding bands yi in the satellite, and X is a feature set composed of the features Xi after dimensionality reduction of the optical satellite, where X i ={x j ,...,x k }, xi is the corresponding satellite band in each satellite feature; The inversion and verification of mangrove EWT and LMA are performed using multiple optical satellite bands corresponding to the ground-based sensitive spectral domain. The formula for the verification model is as follows: T=F(SI1(X'),SI2(X'),...,SI k (X ' )) Where T is the mangrove EWT or LMA estimation result, F is the estimation model, SI is the combined index calculated by satellite bands, and X' is the screened satellite sensitivity feature.

7. The method for inverting water content and dry matter content of mangroves based on fractional-order assisted spectral angle mapping according to claim 1 is characterized in that: In step S104, based on the high correlation features of the three optical satellites and with the GF-3SAR data and SDGSAT TIS data as auxiliary features, a hybrid data set is obtained as follows: D (3×4) =D 基础 +D 辅助 At the same time, an adaptive ensemble learning model was developed based on two traditional statistical models, Ridge and PLSR, and three machine learning model algorithms, K-Nearest Neighbors, Random Forest, and XGBoost, to quantitatively evaluate the inversion performance of the different inversion schemes for the water content and dry matter content of the mangroves. The coefficient of determination (R2) and root mean square error (RMSE) were used to evaluate the accuracy of the inversion model for each mangrove community. The calculation formula of the evaluation index is as follows: Where n is the number of mangrove samples; yi and y are i are the measured and estimated results of functional trait parameters, respectively; is the average value of the measured results of functional trait parameters.

Images