A method for retrieving urban vegetation biomass by combining GEDI and remote sensing imagery

Abstract

This invention discloses a method for retrieving urban vegetation biomass by combining GEDI and remote sensing imagery. The method includes: acquiring GEDI data and remote sensing imagery within the same time frame of the study area; extracting vegetation height within the study area based on L1B level data; constructing a coupling model between the parameters of each band of the remote sensing imagery and the vegetation height; using the coupling model to retrieve vegetation height in areas where the remote sensing imagery and GEDI spots do not overlap; constructing a vegetation height-biomass model based on information from the L2A and L4A level data; and using the vegetation height and the vegetation height-biomass model to obtain biomass data for areas not covered by laser points, thereby retrieving vegetation height in areas not covered by laser points. This invention innovates the method for extracting vegetation height within urban areas by combining remote sensing imagery and satellite laser altimetry data to retrieve vegetation height and biomass within the remote sensing imagery coverage area, improving the accuracy of vegetation height retrieval and consequently increasing the precision of vegetation biomass retrieval.

Description

Technical Field

[0001] This invention relates to the field of remote sensing technology applications, and in particular to a method for retrieving urban vegetation biomass by combining GEDI and remote sensing images. Background Technology

[0002] Global change has brought about a series of problems that have seriously affected the development of human society. Although scientists still have some debate about the causes of global warming, the increase in greenhouse gas emissions is undoubtedly one of the important reasons for global warming. Current climate change is mainly attributed to the large-scale use of fossil fuels and the large-scale emission of greenhouse gases. The excessive emission of carbon dioxide, a greenhouse gas, leading to an increase in atmospheric temperature, is considered the main cause of climate change, and the carbon balance in ecosystems has become a key focus for ecologists.

[0003] Trees and other vegetation serve as a vast carbon sink, significantly mitigating the impacts of climate change. Biomass is a crucial source of carbon storage data and a vital component of the carbon cycle. Accurate acquisition of vegetation biomass data is therefore extremely important in addressing climate change and carbon balance. Existing biomass retrieval methods primarily target large-area monoculture regions such as forests and grasslands. Accurately measuring aboveground biomass in forests and grasslands on a large spatial scale plays a vital role in understanding the role of forests in the global carbon cycle, regional ecological environment monitoring, and the development of effective smart emission reduction strategies.

[0004] However, as a crucial component of terrestrial ecosystems, urban and surrounding ecosystems also hold a significant position in global change research. Biomass is a vital data source for carbon storage in urban ecosystems and an important part of the carbon cycle. However, urban vegetation is relatively more complex than forest vegetation, unlike the more uniform surface cover characteristics of forests. Methods for biomass retrieval based on vegetation height in forest areas cannot be directly applied to the surface of urban and surrounding areas. Therefore, accurately obtaining the height of trees and other vegetation is crucial for aboveground biomass retrieval in urban and surrounding areas.

[0005] In existing technologies, biomass research generally adopts field measurement methods, which have good results in small areas, but have limitations in large-scale measurements, are time-consuming, labor-intensive, and costly. Summary of the Invention

[0006] Purpose of the invention: To address the above problems, the purpose of this invention is to provide a method for retrieving urban vegetation biomass by combining GEDI and remote sensing imagery.

[0007] Technical solution: The present invention provides a method for retrieving urban vegetation biomass by combining GEDI and remote sensing imagery, comprising the following steps:

[0008] Step 1: Acquire GEDI data and remote sensing images of the study area within the same time range; wherein the GEDI data includes L1B, L2A and L4A level data, and the remote sensing images include multispectral remote sensing image data and raster digital elevation data.

[0009] Step 2: Extract vegetation height within the study area based on L1B level data; wherein the L1B level data includes geographic location data and waveform data;

[0010] Step 3: In the overlapping area where both the GEDI spot and the remote sensing image are covered, construct a coupling model between the parameters of each band of the remote sensing image and the vegetation height described in Step 2.

[0011] Step 4: Use the coupling model to invert the vegetation height in the non-overlapping areas of the remote sensing image and the GEDI spot;

[0012] Step 5: Construct a vegetation height biomass model based on the information in the L2A and L4A level data;

[0013] Step 6: Use the vegetation height from Step 4 and the vegetation height biomass model to obtain biomass data for areas not covered by laser points, thereby realizing the inversion of vegetation height in areas not covered by laser points.

[0014] Furthermore, step 2 specifically includes the following sub-steps:

[0015] Step 21: Perform data normalization processing on the waveform data;

[0016] Step 22: Apply Gaussian filtering to the normalized waveform data for noise suppression;

[0017] Step 23: Determine the start and end positions of the waveform using the front and back threshold method;

[0018] Step 24: Calculate the vegetation height, which is the vertical distance between the first valid echo in the standardized smoothed GEDI waveform and the ground echo. The expression is:

[0019] H Waveform_extent =H ground_return -H signal_start

[0020] Among them, H Waveform_extent H represents the waveform length, i.e., the vegetation height. ground_return H represents the vertical height of the first valid echo in the GEDI waveform. signal_start Indicates the vertical height of the ground echo;

[0021] Step 25: Use the spot coarse spot rejection standard to reject the vegetation height extracted in step 24. For spot with vegetation height greater than the threshold height H0, extract the waveform of the spot and use the sliding threshold method for inspection.

[0022] Furthermore, step 25 specifically includes:

[0023] When the vegetation height is greater than the threshold height H0, the subsequent threshold is continuously increased, and the vegetation height is repeatedly extracted until the subsequent threshold equals the previous threshold.

[0024] If, during the extraction process after adding the threshold, the vegetation height is less than H0, the vegetation height under that threshold is directly output as the new vegetation height.

[0025] If, during the process of increasing the threshold extraction, the vegetation height is still greater than H0 when the threshold is equal to the pre-threshold, the vegetation height under the original threshold is output.

[0026] Furthermore, step 3 specifically includes the following sub-steps:

[0027] Step 31: First, based on the corresponding quarterly time, the remote sensing images are divided into four quarters. Then, the remote sensing images are classified under supervision into four types: urban areas, grasslands, woodlands, and cultivated land.

[0028] Step 32: Extract parameters for each band of the multispectral remote sensing image, 8 texture features, and 7 vegetation index parameters, denoted as x1, x2, ..., x... n ;

[0029] Step 33: Use raster digital elevation to extract ground elevation data of the study area and perform slope analysis to extract slope parameters;

[0030] Step 34, let the vegetation height be denoted as h, then the coupled model is expressed as:

[0031] h = f(x1, x2, ..., x) n )

[0032] Step 35: The random forest algorithm in machine learning is used to establish a fitting model between the independent variable x and the dependent variable h, and the accuracy is verified using UAV image data from field measurement points.

[0033] Furthermore, step 4 specifically includes:

[0034] By coupling remote sensing image parameters of areas not covered by GEDI spots into the model, the vegetation height in areas not covered by GEDI spots within the study area can be obtained.

[0035] Furthermore, step 5 specifically includes:

[0036] Using multi-term function fitting and random forest algorithms from machine learning, a vegetation height-biomass model was established based on tree height information from GEDI L2A and biomass information from GEDI L4A, with vegetation height as the independent variable and aboveground biomass as the dependent variable. The optimal model was then selected.

[0037] Beneficial effects: Compared with the prior art, the significant advantages of this invention are:

[0038] 1. This invention innovates the method for extracting vegetation height in urban areas. By combining remote sensing imagery and satellite laser altimetry data, it inverts the vegetation height and biomass within the area covered by the remote sensing imagery, thereby improving the accuracy of vegetation height inversion and, consequently, the accuracy of vegetation biomass inversion.

[0039] 2. Utilizing the accuracy of GEDI spot height measurement and the comprehensive coverage of remote sensing imagery, biomass estimation can be achieved over a large area in and around non-forested cities.

[0040] 3. Use the sliding threshold method to check the waveform to reduce errors;

[0041] 4. By using the vegetation height and biomass data itself to build a model, avoiding a large number of field measurements, a method for large-scale biomass inversion is provided in urban areas with diverse vegetation, thereby enabling large-scale biomass mapping. Attached Figure Description

[0042] Figure 1 This is a flowchart of the method for retrieving urban vegetation biomass by combining GEDI and remote sensing imagery in the embodiments.

[0043] Figure 2 This is a schematic diagram of waveform processing in the embodiment. Detailed Implementation

[0044] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments.

[0045] like Figure 1 The diagram shown is a flowchart of a method for retrieving urban vegetation biomass by combining GEDI and remote sensing imagery, as described in one embodiment. The method includes the following steps:

[0046] Step 1: Acquire GEDI data and remote sensing images of the study area within the same time range; wherein the GEDI data includes L1B, L2A and L4A level data, and the remote sensing images include multispectral remote sensing image data and raster digital elevation data.

[0047] Specifically, GEDI (Global Ecosystem Dynamics Investigation) altimetry data of the study area at the same time was obtained from Earthdata Search (nasa.gov), including L1B, L2A, and L4A level data. L1B level data includes waveform and location information; L2A level data includes ground elevation, canopy top height, and relative elevation information; and L4A level data includes footprint-level aboveground biomass information. Remote sensing images of the same study area at the same time range in the same year were also obtained. These images, using Landsat 8 TM, are available from websites such as the European Space Agency and the USGS, and include multispectral remote sensing imagery and raster digital elevation data. GEDI spot data is accurate in altimetry measurement but cannot achieve full coverage, while remote sensing images offer comprehensive coverage. Optionally, UAV imagery data and quadrat biomass data from uniformly distributed field measurement points within the study area were obtained as reference controls.

[0048] Step 2: Extract vegetation height within the study area based on L1B level data; wherein the L1B level data includes geographic location data and waveform data.

[0049] Step 3: In the overlapping area where both the GEDI spot and the remote sensing image are covered, construct a coupling model between the parameters of each band of the remote sensing image and the vegetation height described in Step 2.

[0050] Specifically, for the GEDI spot and remote sensing image of the study area, there are overlapping and non-overlapping areas between them. In step 3 above, in the overlapping area, the parameters of each band of the remote sensing image are used as independent variables, and the vegetation height in the study area extracted from the L1B level data in step 2 is used as the dependent variable to establish a coupling model to construct the relationship between the remote sensing image and the vegetation height.

[0051] Step 4: Use the coupling model to invert the vegetation height in the non-overlapping areas of the remote sensing image and the GEDI spot.

[0052] Specifically, by substituting the parameters of each band of the remote sensing image of the study area not covered by the GEDI spot into the coupling model in step 3, the vegetation height in the area not covered by the GEDI spot within the study area can be obtained.

[0053] Step 5: Construct a vegetation height biomass model based on information from L2A and L4A level data.

[0054] Specifically, based on the information from L2A and L4A in GEDI, a model of vegetation height and biomass was established with vegetation height as the independent variable and aboveground biomass as the dependent variable, and the optimal model was selected as the vegetation height-biomass model.

[0055] Step 6: Use the vegetation height from Step 4 and the vegetation height biomass model to obtain biomass data for areas not covered by laser points, thereby realizing the inversion of vegetation height in areas not covered by laser points.

[0056] This implementation proposes a method for inverting urban vegetation biomass by combining GEDI and remote sensing imagery. By leveraging the accurate height data from satellite laser altimetry and the large-area coverage characteristics of remote sensing imagery, it enables the inversion of aboveground biomass over a large area of ​​the city and its surroundings, avoiding extensive field measurements and enabling large-scale aboveground biomass mapping.

[0057] In one embodiment, step 2 above specifically includes the following sub-steps:

[0058] Step 21: Perform data normalization on the waveform data in the L1B level data;

[0059] Step 22: Apply Gaussian filtering to the normalized waveform data for noise suppression;

[0060] Step 23: Use the front and back threshold method to determine the start and end positions of the waveform, and extract the vegetation height to calculate the corresponding waveform;

[0061] Step 24: Calculate the vegetation height based on the waveform corresponding to the vegetation height, which is the vertical distance between the first valid echo in the standardized smoothed GEDI waveform and the ground echo. The expression is:

[0062] H Waveform_extent =H ground_return -H signal_start

[0063] Among them, H Waveform_extent H represents the waveform length, i.e., the vegetation height. ground_return H represents the vertical height of the first valid echo in the GEDI waveform. signal_start This represents the vertical height of the ground echo. The first valid echo is defined as the rightmost intersection of the post-threshold and the normalized smooth waveform, and the ground echo is defined as the peak value of the Gaussian component on the leftmost part of the valid waveform.

[0064] Step 25: Use the spot coarse spot rejection standard to reject the vegetation height extracted in step 24. For spot with vegetation height greater than the threshold height H0, extract the waveform of the spot and use the sliding threshold method for inspection.

[0065] Optionally, the criteria for removing coarse light spots in step 25 above are a quality indicator not equal to 1 and a beam sensitivity less than 0.95. Vegetation height that does not meet the criteria for removing coarse light spots is removed from the data.

[0066] The pre- and post-thresholds in step 23 above include the pre-threshold for extracting canopy elevation and the post-threshold for extracting ground elevation, which are obtained using different constants, and their expressions are as follows:

[0067] h0 = h1 + x × h2

[0068] In the formula, h0 represents the front / back threshold, h1 represents the noise average value, h2 represents the noise standard deviation, and x is a constant. Different constant values ​​of x are used to obtain the front and back thresholds respectively.

[0069] Optionally, the expression for calculating the threshold H0 in step 25 above is:

[0070]

[0071] In the formula, h1 represents the average vegetation height, and h3 represents the standard deviation of vegetation height.

[0072] In one embodiment, step 25 specifically includes:

[0073] When the vegetation height is greater than the threshold height H0, the subsequent threshold is continuously increased, and the vegetation height is repeatedly extracted until the subsequent threshold equals the previous threshold.

[0074] If, during the extraction process after adding the threshold, the vegetation height is less than H0, the vegetation height under that threshold is directly output as the new vegetation height.

[0075] If, during the process of increasing the threshold extraction, the vegetation height is still greater than H0 when the subsequent threshold equals the previous threshold, then the vegetation height at the original threshold is output, as illustrated in the waveform diagram below. Figure 2 As shown.

[0076] In one embodiment, step 3 specifically includes the following sub-steps:

[0077] Step 31: First, based on the corresponding quarterly time, the remote sensing images are divided into four quarters. Then, the remote sensing images are classified under supervision into four types: urban areas, grasslands, woodlands, and cultivated land.

[0078] Step 32: Extract parameters for each band of the multispectral remote sensing image (including red band, green band, blue band, near-infrared band, and short-infrared band), 8 texture features (including mean, variance, contrast, etc.), and 7 vegetation index parameters (including NDVI, EVI, RVI, GNDVI, NBR, DVI, and ARVI), denoted as x1, x2, ..., xn ;

[0079] Step 33: Use raster digital elevation to extract ground elevation data of the study area and perform slope analysis to extract slope parameters;

[0080] Step 34, let the vegetation height be denoted as h, then the coupled model is expressed as:

[0081] h = f(x1, x2, ..., x) n )

[0082] Step 35: The random forest algorithm in machine learning is used to establish a fitting model between the independent variable x and the dependent variable h, and the accuracy is verified using UAV image data from field measurement points.

[0083] In one embodiment, step 4 specifically includes:

[0084] By coupling remote sensing image parameters of areas not covered by GEDI spots into the model, the vegetation height in areas not covered by GEDI spots within the study area can be obtained.

[0085] In one embodiment, step 5 specifically includes:

[0086] Using multi-term function fitting and random forest algorithms from machine learning, a model of vegetation height and biomass was established based on tree height information from GEDI L2A and biomass information from GEDI L4A, with vegetation height as the independent variable and aboveground biomass as the dependent variable. The coefficient of determination R0 was selected. 2 The model with the largest RMSE and the smallest vegetation height biomass is used as the vegetation height biomass model.

[0087] This invention provides a method for retrieving urban vegetation biomass by combining GEDI and remote sensing imagery. It verifies the accuracy of vegetation height extraction from spaceborne lidar waveforms, and based on this, uses remote sensing imagery to retrieve vegetation height, establishing a model of vegetation height and biomass, ultimately achieving large-scale biomass mapping in urban areas. This invention avoids extensive field measurements and provides a method for large-scale biomass retrieval in areas with diverse urban vegetation, thus enabling large-scale biomass mapping.

Claims

1. A method for retrieving urban vegetation biomass by combining GEDI and remote sensing imagery, characterized in that, Includes the following steps: Step 1: Acquire GEDI data and remote sensing images of the study area within the same time range; wherein the GEDI data includes L1B, L2A and L4A level data, and the remote sensing images include multispectral remote sensing image data and raster digital elevation data. Step 2: Extract vegetation height within the study area based on L1B level data; wherein the L1B level data includes geographic location data and waveform data; Step 3: Within the overlapping area covered by both the GEDI spot and the remote sensing image, construct a coupling model between the parameters of each band of the remote sensing image and the vegetation height described in Step 2; specifically: Step 31: First, based on the corresponding quarterly time, the remote sensing images are divided into four quarters. Then, the remote sensing images are classified under supervision into four types: urban areas, grasslands, woodlands, and cultivated land. Step 32: Extract parameters for each band, eight texture features, and various vegetation index parameters from the multispectral remote sensing image, denoted as... ; Step 33: Use raster digital elevation to extract ground elevation data of the study area and perform slope analysis to extract slope parameters; Step 34, let the vegetation height be denoted as h, then the coupled model is expressed as: ； Step 35: The random forest algorithm in machine learning is used to establish a fitting model between the independent variable x and the dependent variable h, and the accuracy is verified using UAV image data from field measurement points. Step 4: Use the coupling model to invert the vegetation height in the non-overlapping areas of the remote sensing image and the GEDI spot; specifically: Substituting the remote sensing image parameters of the areas not covered by GEDI spots into the coupling model, the vegetation height in the areas of the study region not covered by GEDI spots is obtained; Step 5: Based on the information in the L2A and L4A level data, construct a vegetation height biomass model; specifically: Using multi-term function fitting and random forest algorithms from machine learning, a vegetation height-biomass model was established based on tree height information in GEDI L2A and biomass information in GEDI L4A, with vegetation height as the independent variable and aboveground biomass as the dependent variable. The optimal model was then selected. Step 6: Using the vegetation height and the vegetation height biomass model from Step 4, obtain the biomass data of the area not covered by the laser point, and realize the inversion of the vegetation height in the area not covered by the laser point.

2. The method for retrieving urban vegetation biomass by combining GEDI and remote sensing imagery according to claim 1, characterized in that, Step 2 specifically includes the following sub-steps: Step 21: Perform data normalization processing on the waveform data; Step 22: Apply Gaussian filtering to the normalized waveform data for noise suppression; Step 23: Determine the start and end positions of the waveform using the front and back threshold method; Step 24, calculate the vegetation height, which is the vertical distance between the first valid echo in the standardized smoothed GEDI waveform and the ground echo. The expression is: ； in, This indicates the waveform length, i.e., the vegetation height. This indicates the vertical height of the first valid echo in the GEDI waveform. Indicates the vertical height of the ground echo; Step 25: Use the spot coarse spot rejection standard to reject the vegetation height extracted in step 24. For spot with vegetation height greater than the threshold height H0, extract the waveform of the spot and use the sliding threshold method for inspection.

3. The method for retrieving urban vegetation biomass by combining GEDI and remote sensing imagery according to claim 2, characterized in that, Step 25 specifically includes: When the vegetation height is greater than the threshold height H0, the subsequent threshold is continuously increased, and the vegetation height is repeatedly extracted until the subsequent threshold equals the previous threshold. If, during the extraction process after adding the threshold, the vegetation height is less than H0, the vegetation height under that threshold is directly output as the new vegetation height. If, during the process of increasing the threshold extraction, the vegetation height is still greater than H0 when the threshold is equal to the pre-threshold, the vegetation height under the original threshold is output.

Images