Methods for estimating grassland plant beta diversity based on airborne hyperspectral remote sensing technology

Abstract

This invention discloses a method for estimating grassland plant beta diversity based on airborne hyperspectral remote sensing technology. Several 1x1m quadrats are uniformly distributed within the survey area, their location information is recorded, and the beta diversity index is calculated. Hyperspectral remote sensing images of the survey area are acquired, with an image resolution of 0.25m, and each quadrat covers 16 pixels. The 16 pixels covered by each quadrat are sorted according to the normalized difference vegetation index. Multi-pixel spectral angles and multi-pixel spectral distances between quadrats are calculated. The image is resampled to the quadrat size. The single-pixel spectral distances between quadrats are calculated. A multiple regression model is constructed with multi-pixel spectral angles, multi-pixel spectral distances, and single-pixel spectral distances as independent variables and beta diversity as the dependent variable. Applying this model to the entire remote sensing imagery allows for the acquisition of vegetation beta diversity at any location within the survey area, making it suitable for estimating plant beta diversity over large areas in grassland regions.

Description

Technical Field

[0001] This invention relates to a method for estimating grassland plant beta diversity, specifically a method for estimating grassland plant beta diversity based on airborne hyperspectral remote sensing technology, belonging to the field of biodiversity remote sensing monitoring technology. Background Technology

[0002] Biodiversity is the dominant function of ecosystems and the foundation for functions such as food production, climate regulation, waste management, and soil and water conservation. In recent years, due to intensified human activities and global warming, species worldwide are disappearing at an unprecedented rate. Studies have shown that the current rate of species extinction in nature is 1000 times the normal rate, and this trend is accelerating. How to monitor biodiversity has become a global concern.

[0003] While traditional biodiversity surveys have made significant progress, these studies are primarily based on ground-based methods, focusing on the plot level. Plot-level survey results have provided accurate information for biodiversity assessment for a considerable period, but they cannot meet the needs of biodiversity monitoring at landscape, regional, or even global scales, let alone provide real-time monitoring results. Furthermore, ground-based surveys are time-consuming and labor-intensive, often requiring a large number of professionals. Remote sensing technology, which enables large-scale, rapid, and real-time monitoring, can effectively compensate for the time and labor costs of ground-based surveys. However, current research on using remote sensing technology for biodiversity monitoring is limited.

[0004] Biodiversity is divided into intra-community (quadrat) diversity and inter-community (quadrat) diversity, namely alpha diversity and beta diversity. Alpha diversity reflects the richness of species within a quadrat, while beta diversity reflects the variation in community composition between quadrats. Currently, there are many remote sensing methods for estimating alpha diversity, but methods for estimating beta diversity are lacking. Furthermore, due to the small size of herbaceous plants, estimating biodiversity in grassland areas presents greater challenges. However, a good method for estimating biodiversity in grassland areas still remains lacking. Summary of the Invention

[0005] To address the problems existing in the prior art, this invention provides a method for estimating grassland plant beta diversity based on airborne hyperspectral remote sensing technology. By utilizing remote sensing technology, only a small amount of ground survey is required to achieve rapid and accurate estimation of plant beta diversity.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for estimating grassland plant beta diversity based on airborne hyperspectral remote sensing technology, comprising the following steps:

[0007] Step 1: Evenly distribute several 1x1m square vegetation quadrats within the survey area, record the location information of the four corner points of the quadrats, and use GPS-RTK for precise positioning. Count the plant species and the number of each species in each quadrat, and calculate the beta diversity index.

[0008] Step 2: Acquire airborne hyperspectral remote sensing images of the survey area. Before aerial photography, adjust the image resolution to 0.25m to ensure that each sample plot can cover 16 pixels. Perform preprocessing such as radiometric calibration and atmospheric correction on the images to obtain surface reflectance data.

[0009] Step 3: Extract reflectance data of 16 pixels covered by each quadrat from the remote sensing image, calculate the normalized difference vegetation index of each pixel, and sort the 16 pixels according to the calculation results.

[0010] Step 4: Calculate the multi-pixel spectral angle and multi-pixel spectral distance between sample plots based on the sorted reflectance data, and normalize the results to between 0 and 1.

[0011] Step 5: Resample the image to the size of the quadrat, i.e., 1m spatial resolution, and extract the reflectance data of the single pixel covered by each quadrat.

[0012] Step 6: After performing a first-order derivative transformation on the extracted reflectance data, calculate the spectral distance of a single pixel between sample plots and normalize the result to between 0 and 1.

[0013] Step 7: Construct a multiple regression model with the normalized multi-pixel spectral angle, multi-pixel spectral distance, and single-pixel spectral distance as independent variables and the beta diversity index as the dependent variable; apply this model to the entire remote sensing image to obtain the vegetation beta diversity index at any location within the entire survey area.

[0014] Furthermore, considering the accuracy requirements of the modeling, the number of vegetation quadrats in step one is T, and T≥9; when calculating the beta diversity index of paired quadrats, a total of [number] quadrats should be calculated. Data sets; quadrats are preferably evenly distributed within the survey area, and the coordinates of the four corner points of the quadrats are recorded after the survey is completed.

[0015] Furthermore, the formula for calculating the multi-pixel spectral angle between sample plots in step four is as follows:

[0016]

[0017] In equation (1), S is the number of bands in the hyperspectral remote sensing image, and N is the number of pixels covered within the quadrat. Let be the reflectance of the j-th pixel in sample plot X after sorting, in band i. Let be the reflectance of the j-th pixel in quadrat Y after sorting, in band i. Quadrats X and Y are sorted in the same way, either from largest to smallest or smallest to largest. Equation (1) calculates the angle between the reflectances of the two quadrats in the S-dimensional feature space. The larger the angle, the larger the value of Spectral_angle_Mutul, indicating that the species composition between the quadrats is less similar, representing higher plant beta diversity in the quadrat pair. Therefore, Spectral_angle_Mutul is positively correlated with the beta diversity index.

[0018] Furthermore, the formula for calculating the multi-pixel spectral distance between sample plots in step four is as follows:

[0019]

[0020] In equation (2), S is the number of bands in the hyperspectral remote sensing image, and N is the number of pixels covered within the quadrat. Let be the reflectance of the j-th pixel in sample plot X after sorting, in band i. Let be the reflectance of the j-th pixel in quadrat Y after sorting in band i. Quadrats X and Y are sorted in the same way, either from largest to smallest or smallest to largest. Equation (2) calculates the distance between the reflectances of the two quadrats in the S-dimensional feature space. The larger the distance, the larger the value of Spectral_distance_Mutul, indicating that the species composition between the quadrats is less similar, which means that the plant beta diversity of the quadrat pair is higher. Therefore, Spectral_distance_Mutul is positively correlated with the beta diversity index.

[0021] Furthermore, in step five, the spatial resolution of the image needs to be reduced so that the reduced spatial resolution is consistent with the side length of the square quadrat, i.e., the pixel size equals the quadrat size.

[0022] Furthermore, the formula for calculating the spectral distance of a single pixel between sample plots in step six is ​​as follows:

[0023]

[0024] In equation (3), S represents the number of bands in the hyperspectral remote sensing image. Let X be the value of the reflectance of the pixel corresponding to sample plot X in band i after performing a first-order derivative transformation. The value of the reflectance of the pixel corresponding to sample Y in band i is the result of the first derivative transformation. Equation (3) calculates the distance between the reflectances of the two sample plots in the S-dimensional feature space. The larger the distance, the less similar the species composition between the sample plots. The larger the value of Spectral_distance_Single, the higher the plant beta diversity of the sample plot pair. Therefore, Spectral_distance_Single is positively correlated with the beta diversity index.

[0025] Compared with existing technologies, this invention creatively proposes to characterize species composition differences by adjusting the spatial resolution of images and utilizing angular and distance differences between multiple pixel spectra, as well as distance differences between individual pixel spectra. Furthermore, it constructs a biodiversity estimation model using multi-pixel spectral angles, multi-pixel spectral distances, and individual pixel spectral distances. By first sorting the pixels and then calculating the multi-pixel spectral angles and distances, the influence of different species spatial distributions can be effectively eliminated. In addition, this invention scientifically specifies the number of ground samples and the image spatial resolution, thus ensuring high model accuracy. This invention requires only a small amount of ground surveys to achieve rapid and accurate estimation of plant beta diversity in large-scale grassland areas. It is of great significance for large-scale biodiversity monitoring and assessment in grassland areas. Attached Figure Description

[0026] Figure 1 This is a flowchart of the present invention;

[0027] Figure 2 The correlation coefficient and scatter plot between the Spectral_angle_Mutil index and the Bray-Curtis index in this embodiment of the invention;

[0028] Figure 3 The correlation coefficient and scatter plot between the Spectral_distance_Mutil index and the Bray-Curtis index in this embodiment of the invention;

[0029] Figure 4 This is a correlation coefficient and scatter plot between the Spectral_distance_Single index and the Bray-Curtis index in an embodiment of the present invention;

[0030] Figure 5 This is a correlation coefficient diagram between the actual value and the estimated value in an embodiment of the present invention. Detailed Implementation

[0031] The invention will now be further described with reference to the accompanying drawings.

[0032] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0033] The following examples are located in the Xilinhot Grassland of Inner Mongolia Autonomous Region in northern China. The grassland is situated in a semi-arid continental monsoon climate zone, and its dominant species include Leymus chinensis, Stipa capillata, Cleistogenes squarrosa, and Chloris virgata. The Xilinhot Grassland is rich in energy resources and has a long history of mineral mining, with several mining areas including Shengli and Datang. Furthermore, a large number of herders live on the grassland, engaging in grazing. Therefore, the grassland is subject to the dual disturbances of grazing and mining, resulting in a rapid decline in biodiversity. Due to its vast area, obtaining biodiversity data solely through ground surveys is impractical. This invention provides a technical solution: a method for estimating grassland plant beta diversity based on airborne hyperspectral remote sensing technology.

[0034] like Figure 1 As shown, the present invention provides a technical solution for estimating grassland plant beta diversity based on airborne hyperspectral remote sensing technology.

[0035] First, vegetation quadrats were established within the survey area. A total of 15 square vegetation quadrats (1x1m) were evenly distributed throughout the survey area. The number of species and the number of plants of each species within each quadrat were counted, and the beta diversity index was calculated using formula (4):

[0036]

[0037] In equation (4), θ represents the total number of species in quadrats X and Y. This represents the number of the u-th species in quadrat X. This represents the number of the μ-th species in quadrat Y. The value represents the smaller number of the u-th species in quadrats X and Y. A larger Bray-Curtis value indicates fewer shared species in quadrats X and Y, and thus greater beta diversity. A total of 105 sets of beta diversity data were calculated for the 15 quadrats. After the vegetation survey, GPS-RTK technology was used to locate the four corner points of the square quadrats with an accuracy of 2 cm.

[0038] Secondly, airborne hyperspectral remote sensing imagery of the survey area was acquired. In this embodiment, a DJI Wind4 drone equipped with an imaging spectrometer system was used to acquire hyperspectral images. The imaging spectrometer system mainly consists of three parts: an optical scanner (SPECIM FX10), a control system (DPU), and a GNSS / IMU integrated navigation system. Clear, cloudless weather was selected for data acquisition, with the flight time at noon. Before flight, parameters such as heading, altitude, and heading / lateral overlap were set, and whiteboards were placed on the ground for later whiteboard calibration. Image preprocessing mainly included geometric correction and radiometric calibration, whiteboard calibration, registration, and mosaicking. The final imagery has a spatial resolution of 0.25m, a spectral resolution of 5.5nm, and 112 bands. The coordinates of the four corner points of the quadrat were imported into the remote sensing imagery; upon inspection, each quadrat occupied at least 16 pixels.

[0039] Third, reflectance was extracted and sorted. Using ENVI software, reflectance data for the 16 pixels covered by each quadrat were extracted. The normalized difference vegetation index (NDVI) for each pixel was calculated, and the 16 pixels were sorted (from smallest to largest) based on the calculation results. Since beta diversity only considers differences in plant species and quantity, and not differences in the spatial distribution of plants, the pixels were sorted according to the magnitude of the NDVI, thus eliminating errors caused by differences in the spatial distribution of plants.

[0040] Fourth, calculate the multi-pixel spectral angle and multi-pixel spectral distance of the paired sample plots.

[0041] The formula for the multi-pixel spectral angle is as follows:

[0042]

[0043] In equation (1), S is the number of bands in the hyperspectral remote sensing image, which is 112 in this embodiment, and N is the number of pixels covered within the quadrat, which is 16 in this embodiment. Let be the reflectance of the j-th pixel in sample plot X after sorting, in band i. Let be the reflectance of the j-th pixel in quadrat Y after sorting, in band i. Equation (1) calculates the angle between the reflectances of the two quadrats in the S-dimensional feature space. The larger the angle, the larger the value of Spectral_angle_Mutil, representing higher species beta diversity of the quadrat pair. The calculation of Spectral_angle_Mutil is performed in R language. After normalizing the calculation results, a correlation analysis is performed between the spectral angle Spectral_angle_Mutil and the Bray-Curtis index. Figure 2As shown, the correlation coefficient between the two is 0.363, and it passed the 0.01 significance test, indicating that the Spectral_angle_Mutil proposed in this invention has a high correlation with the Bray-Curtis index.

[0044] The formula for multi-pixel spectral distance is as follows:

[0045]

[0046] In equation (2), S is the number of bands in the hyperspectral remote sensing image, which is 112 in this embodiment, and N is the number of pixels covered within the sample plot, which is 16 in this embodiment. Let be the reflectance of the j-th pixel in sample plot X after sorting, in band i. Let be the reflectance of the j-th pixel in quadrat Y after sorting, in band i. Equation (2) calculates the distance between the reflectances of two quadrats in the S-dimensional feature space. The larger the distance, the larger the value of Spectral_distance_Mutil, representing the higher the species beta diversity of the quadrat pair. The calculation of Spectral_distance_Mutil is performed in R language. After normalizing the calculation results, a correlation analysis is performed between Spectral_distance_Mutil and the Bray-Curtis index. Figure 3 As shown, the correlation coefficient between the two is 0.583, and it passed the 0.01 significance test, indicating that there is a high correlation between the Spectral_distance_Mutil proposed in this invention and the Bray-Curtis index.

[0047] Fifth, image resampling. In ENVI, the 0.25m spatial resolution image is upscaled to 1m spatial resolution, so that the sample plot size matches the pixel size. The reflectance data of the single pixel covered by each sample plot is extracted from the resampled hyperspectral image.

[0048] Sixth, calculate the spectral distance of a single pixel. In Origin, perform a first-order derivative transformation on the reflectance extracted in step five and calculate the spectral distance of a single pixel. The formula is:

[0049]

[0050] In equation (3), S represents the number of bands in the hyperspectral remote sensing image, which is 112 in this embodiment. Let X be the value of the reflectance of the pixel corresponding to sample plot X in band i after performing a first-order derivative transformation. The value is the first derivative of the reflectance of the pixel corresponding to sample plot Y in band i. Equation (3) calculates the distance between the reflectances of the two sample plots in the S-dimensional feature space. The larger the distance, the larger the value of Spectral_distance_Single, which represents the higher the plant beta diversity of the sample plot pair. The calculation of Spectral_distance_Single is completed in R language. After normalizing the calculation results, a correlation analysis is performed between Spectral_distance_Single and the Bray-Curtis index. Figure 4 As shown, the correlation coefficient between the two is 0.810, and it passed the 0.01 significance test, indicating that there is a high correlation between the Spectral_distance_Single proposed in this invention and the Bray-Curtis index.

[0051] Finally, the model was constructed. A multiple regression model was built in SPSS 23 using normalized multi-pixel spectral angles, multi-pixel spectral distances, and single-pixel spectral distances as independent variables, and the Bray-Curtis index as the dependent variable. The model is as follows:

[0052] Y=0.152X1+0.119X2+0.398X3+0.177 (5);

[0053] In equation (5), Y is the Bray-Curtis index, X1 is the multi-pixel spectral angle, X2 is the multi-pixel spectral distance, and X3 is the single-pixel spectral distance. The model's R... 2 The value was 0.759, and it passed the significance test at 0.001. The model was validated, and the results are as follows: Figure 5 As shown, the correlation coefficient between the estimated value and the actual value is 0.834, indicating that the method proposed in this invention has high reliability.

[0054] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0055] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any minor modifications, equivalent substitutions, and improvements made to the above embodiments based on the technical essence of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for estimating beta diversity of grassland plants based on airborne hyperspectral remote sensing technology, characterized in that, The method comprises the following steps: Step one, evenly arrange a plurality of 1x1m square vegetation quadrats in the investigation area, record the position information of the four corner points of the quadrats, count the plant species and the number of each species in each quadrat, and calculate the beta diversity index; Step two, obtain the airborne hyperspectral remote sensing image of the investigation area, the image spatial resolution is 0.25m, 16 pixels can be covered in each quadrat, and the surface reflectivity data is obtained by pre-processing the image; Step three, extract the reflectivity data of 16 pixels covered by each quadrat on the remote sensing image, calculate the normalized difference vegetation index of each pixel, and sort the 16 pixels according to the calculation results; Step four, calculate the multi-pixel spectral angle and multi-pixel spectral distance between the quadrats based on the sorted reflectivity data, and normalize the results; The formula for calculating the multi-pixel spectral angle between the quadrats in step four is as follows: where S is the number of bands of the hyperspectral remote sensing image, N is the number of pixels covered in the sample plot, is the reflectance of the jth pixel in the sample plot X in the ith band, is the reflectance of the jth pixel in the sample plot Y in the ith band. The formula for calculating the multi-pixel spectral distance between the quadrats in step four is as follows: where S is the number of bands of the hyperspectral remote sensing image, N is the number of pixels covered in the sample plot, is the reflectance of the jth pixel in the sample plot X in band i, is the reflectance of the jth pixel in the sample plot Y in band i; Step five, resample the image to the size of the quadrat, and extract the reflectivity data of a single pixel covered by each quadrat; Step six, after first derivative transformation of the extracted reflectivity data, calculate the single-pixel spectral distance between the quadrats, and normalize the results; The formula for calculating the single-pixel spectral distance between the quadrats in step six is as follows: Wherein, S is the band number of the hyperspectral remote sensing image, Xi is the value of the reflectance of the pixel corresponding to the sample X after the first derivative transformation of the band i, Xi is the value of the reflectance of the pixel corresponding to the sample X after the first derivative transformation of the band i, Step seven, take the normalized multi-pixel spectral angle, multi-pixel spectral distance and single-pixel spectral distance as independent variables, and the beta diversity index as dependent variable, to construct a multiple regression model; apply the model to the entire remote sensing image, and the beta diversity index of vegetation at any location in the entire investigation area can be obtained.

2. The method for estimating beta diversity of grassland plants based on airborne hyperspectral remote sensing technology according to claim 1, characterized in that, The number of vegetation quadrats in step one is T and T≥9.

3. The method for estimating beta diversity of grassland plants based on airborne hyperspectral remote sensing technology according to claim 1, characterized in that, In step five, the image spatial resolution needs to be reduced, and the reduced image spatial resolution is consistent with the side length of the square quadrat.

Images