Remote Sensing-Based Estimation Method for Transferable Eucalyptus Carbon Storage with Small Samples
By combining airborne lidar and Landsat data, a small sample migable eucalyptus carbon storage estimation method is established, which solves the problems of low estimation accuracy and poor migrationability of eucalyptus carbon storage estimation, and achieves high-precision estimation of eucalyptus carbon storage in the subtropical region.
Patent Information
- Application Number
- CN202311160455.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-09-11
- Publication Date
- 2025-07-11
- Estimated Expiration
- 2043-09-11
AI Technical Summary
In the prior art, the eucalyptus carbon storage estimation model has low accuracy and lacks mobility. Especially in subtropical areas, airborne lidar data acquisition costs are high and it is difficult to apply on a large scale.
By combining airborne lidar data with Landsat-derived vegetation age, a small sample of migratory eucalyptus carbon reserve estimation method was established, and a stepwise regression and nonlinear regression model was used, combined with random location segmentation and Zou's test, forest age information was extracted to improve model accuracy and migration.
The accuracy of eucalyptus carbon storage estimation and the mobility of the model are improved, and are suitable for accurate carbon storage estimation in large areas.
Smart Images
Figure CN117216725B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of forest carbon stock estimation, and in particular to a method for estimating the carbon stock of eucalyptus with small samples and migratability based on remote sensing. Background Art
[0002] As of 2018, eucalyptus covered more than 20 million hectares globally, and the significantly increased eucalyptus area was considered a huge carbon sink. Compared with other planted forests, eucalyptus has a fast growth rate, a short felling cycle, strong adaptability, and a wide range of uses, and is an important economic forest tree. Timely updating of eucalyptus carbon stock data is crucial for better understanding and quantifying its impact on ecological and hydrological processes.
[0003] In the past three decades, different types of remote sensing data, such as optical, radar, and Lidar data, have been widely used in forest carbon stock estimation. However, the data saturation of optical and radar data and their sensitivity to terrain, atmospheric conditions, and vegetation phenology result in low model accuracy and non-migratability. Therefore, there is an urgent need to extract variables from remote sensing data that are not affected by external factors and are closely related to forest carbon stock. Forest height and age variables are two parameters that can meet the requirements.
[0004] Airborne LiDAR has the ability to capture forest height information and is considered the most promising technology in forest carbon stock estimation. However, accurate tree heights are usually obtained from small-footprint airborne LiDAR data, which are generally limited to small areas considering data acquisition costs and data volumes. Compared with the more costly and difficult field measurement data of sample plots, in recent years, airborne LiDAR data has often been used as a bridge connecting sample plots and coarse-resolution remote sensing data. The use of such intermediate data sets provides a large number of samples for model modeling and validation to make up for the lack of spatial sampling in ground surveys, thereby establishing a reliable carbon stock estimation model suitable for large-scale use.
[0005] In addition to canopy height, forest age is another variable closely related to carbon stock. Currently, the method of extracting forest age by detecting disturbances in long-term remote sensing images (Landsat) and identifying deforestation and reforestation has been widely used. Previous studies used the forest age in years as a predictive variable for the forest carbon stock model. However, this data cannot effectively capture the growth characteristics of fast-growing eucalyptus in its young forest stage, indicating the necessity of using a more accurate forest age variable to establish an eucalyptus carbon stock estimation model. The new method developed by Li et al. to extract the forest age in months by combining random positioning segmentation and Chow test provides the possibility for model implementation. Summary of the Invention
[0006] The object of the present invention is to provide a method for estimating the carbon storage of eucalyptus in subtropical regions by using a plot-airborne lidar-Landsat time series data modeling strategy, so as to solve the technical problems of low accuracy of current carbon storage estimation and poor scalability of models. It relates to a method for estimating the carbon storage of eucalyptus in subtropical regions by using a plot-airborne lidar-Landsat time series data modeling strategy. Specifically, it refers to a method for linking plots with Landsat-derived vegetation ages through airborne lidar data and estimating the carbon storage of eucalyptus within a regional scope, belonging to the technical field of forest carbon storage estimation.
[0007] Specifically, the following technical solutions are adopted:
[0008] A remotely sensed small-sample transferable eucalyptus carbon storage estimation method, characterized in that: the plot is associated with Landsat-derived vegetation ages through airborne lidar data, and the carbon storage of eucalyptus is estimated within a regional scope.
[0009] Furthermore, it includes the following steps:
[0010] Step 1: Use field per-tree measurement to obtain the observed values of the carbon storage of the plot.
[0011] Step 2: Preprocess the Lidar data in the study area to obtain the canopy height model CHM.
[0012] Step 3: Taking the plot range as the vector boundary, based on the Lidar CHM data, extract the stand height variables corresponding to the plot.
[0013] Step 4: Taking the plot carbon storage as the dependent variable and the Lidar variables as the independent variables, use stepwise regression to select statistically significant variables and establish a multiple linear regression model for the carbon storage of eucalyptus.
[0014] Step 5: According to the eucalyptus carbon storage model in the Lidar coverage area established in Step 4, calculate the estimated value of the carbon storage in the Lidar coverage area.
[0015] Step 6: Overlay the carbon storage distribution map of the Lidar coverage area with the eucalyptus distribution map, and use the method of stratified random sampling to select training samples and validation samples.
[0016] Step 7: Use the random location segmentation method and Chow test to conduct dynamic change monitoring on the Landsat time series NDVI data in the study area, so as to obtain the eucalyptus forest age.
[0017] Step 8: According to the sample positions collected in Step 6, extract the values of the eucalyptus forest age at the corresponding positions; then perform a logarithmic transformation on the eucalyptus forest age, and use the transformed values as independent variables and the training samples in Step 6 as dependent variables to establish a non-linear regression model for eucalyptus carbon storage;
[0018] Step 9: According to the eucalyptus carbon storage model of the study area established in Step 8, calculate each pixel of the eucalyptus forest age image obtained in Step 7 to obtain the eucalyptus carbon storage estimation result map of the entire study area.
[0019] Further, in Step 3, the stand height variables extracted for the corresponding sample plots include: height percentile, mean, skewness, kurtosis; then use the Pearson correlation analysis method to remove the variables that have a high correlation with other variables and a small correlation with eucalyptus carbon storage.
[0020] Further, in Step 4, the specific regression model is:
[0021] y = b0 + b1x1 + b2x2 + … + b i x i (1)
[0022] where y represents the measured value of the carbon storage in the sample plot, x i represents the Lidar variable set screened in Step 3, and b i represents the coefficients in the linear model;
[0023] The specific model and model parameters are obtained through the stepwise regression algorithm in the software SPSS.
[0024] Further, in Step 5, according to the eucalyptus carbon storage model of the Lidar coverage area established in Step 4, determine the selected Lidar variables; use the Create Fishnet tool in the software ArcMap to create polygon vector data of the sample plot size with the Lidar data as the range; then extract the corresponding Lidar statistical variables within the polygon range according to the vector data; finally, substitute the functional relationship into all polygons in the software ArcMap to obtain the carbon storage estimation value of the Lidar coverage area.
[0025] Further, in Step 8, according to the sample positions collected in Step 6, extract the values of the eucalyptus forest age at the corresponding positions; then perform a logarithmic transformation on the eucalyptus forest age, and use the transformed values as independent variables and the training samples in Step 6 as dependent variables to establish a non-linear regression model for eucalyptus carbon storage. The specific non-linear regression model is:
[0026] y = c0 + c1ln(x) (2)
[0027] Among them, y is the carbon storage value of the training sample in step 6, x represents the age of the eucalyptus forest in step 7, and c0 and c1 represent the coefficients in the model;
[0028] Establish a unary functional relationship between carbon storage and forest age through the software SPSS to obtain c0 and c1.
[0029] Furthermore, in step 9, according to the eucalyptus carbon storage model established in step 8, after inputting the model equation in the software ENVI, each pixel of the eucalyptus forest age image obtained in step 7 is calculated to obtain the eucalyptus carbon storage estimation result map of the entire study area.
[0030] Compared with the prior art, the carbon storage predicted by the present invention and its preferred solution using airborne lidar data is of great value for increasing the sample size of the regional-scale model, and this sample set can fully reflect the growth status of eucalyptus carbon storage in the study area; compared with using forest age data in years, the present invention uses forest age data in months to establish an eucalyptus carbon storage model to cope with the rapid changes in the eucalyptus in the young forest stage. The modeling strategy of the present invention gives full play to the advantages of airborne lidar and time-series remote sensing images in estimating carbon storage, effectively improving the accuracy of eucalyptus carbon storage estimation and solving the problem of model migration on the time scale. Description of the Drawings
[0031] The present invention will be further described in detail below in conjunction with the drawings and specific embodiments:
[0032] Figure 1 It is a flow chart of the method for estimating eucalyptus carbon storage in the embodiment of the present invention;
[0033] Figure 2 It is a scatter plot of the relationship between eucalyptus forest age and carbon storage in the embodiment of the present invention;
[0034] Figure 3 It is a map of the eucalyptus carbon storage estimation result in the embodiment of the present invention. Detailed Description of the Invention
[0035] To make the features and advantages of this patent more obvious and understandable, specific embodiments are given below and described in detail in conjunction with the drawings as follows:
[0036] The embodiment of the present invention is based on airborne Lidar and time-series Landsat series NDVI data to estimate the eucalyptus carbon storage in Yunxiao County, Fujian Province. Referring to Figure 1 , the specific steps of the embodiment of the present invention are as follows:
[0037] Step 1: Use on-site per-tree measurement to obtain the observed carbon storage value of the sample plot;
[0038] In this embodiment, according to the distribution of eucalyptus, plots were randomly selected, and diameter at breast height (DBH) of each tree in the plots was measured, and the tree species and DBH were recorded. The biomass of each tree was calculated according to the allometric growth equation of the main tree species in the study area (Du et al., 2014), and then the carbon storage was converted through the carbon coefficient corresponding to the tree species. Finally, the carbon storage of all individual trees in the plot was accumulated and converted into the carbon storage per hectare.
[0039] Step 2: Preprocess the Lidar data in the study area to obtain the Canopy Height Model (CHM).
[0040] Specifically, when implementing, the Lidar360 software package was used to preprocess the airborne lidar point cloud data, including denoising, filtering, and normalization. The inverse distance weighted interpolation algorithm was used for the ground points to generate the Digital Elevation Model (DEM), and the first echo interpolation was used to generate the Digital Surface Model (DSM), both with a spatial resolution of 1 m. Then, the Canopy Height Model (CHM) was obtained by raster difference operation on the DSM and DEM.
[0041] Step 3: Based on the Lidar CHM data, extract various stand height variables, including height percentiles, mean values, skewness, kurtosis, etc. Using the Pearson correlation analysis method, remove the variables that have a high correlation with other variables but a small correlation with the eucalyptus carbon storage.
[0042] In this embodiment, the height variables calculated based on the CHM are all common Lidar statistical variables, which will not be elaborated here. Regarding the screening of Lidar variables, Pearson correlation analysis was performed on all variables including the measured carbon storage observation values according to the samples, sorted from largest to smallest according to the correlation with carbon storage, and then the variables were taken out in turn to perform correlation analysis with other variables. Remove the variables with a correlation greater than 0.85 but a small correlation with carbon storage, and finally retain the candidate variable set.
[0043] Step 4: Using the plot carbon storage as the dependent variable and the Lidar variables as the independent variables, use stepwise regression to select the variables with statistical significance and establish a multiple linear regression model for the eucalyptus carbon storage. The specific regression model is:
[0044] y = b0 + b1x1 + b2x2 + … + b i x i (1)
[0045] where y represents the measured value of the plot carbon storage, x i represents the Lidar variable set screened in step 3, and b i represents the coefficient in the linear model.
[0046] The specific model and model parameters were obtained through the stepwise regression algorithm in the software SPSS.
[0047] In this embodiment, in the stepwise regression parameter setting, the p-values are set to 0.05 and 0.1 respectively to determine whether a variable should be included in the model or removed from the model.
[0048] Step 5: According to the established eucalyptus carbon storage model in the Lidar coverage area, determine the selected Lidar variables. Using the Create Fishnet tool of the software ArcMap, with the Lidar data as the range, create polygon vector data of the plot size. Then, according to the vector data, extract the corresponding Lidar statistical variables within the polygon range. Finally, substitute the functional relationship into all polygons in the software ArcMap to obtain the estimated value of the carbon storage in the Lidar coverage area.
[0049] In this embodiment, the size of the fishnet polygon is set to the plot size of 20*20m, so the spatial resolution of the eucalyptus carbon storage distribution map in the Lidar coverage area is 20m.
[0050] Step 6: Overlay the carbon storage distribution map of the Lidar coverage area with the eucalyptus distribution map, and use the stratified random sampling method to select training samples and validation samples.
[0051] In this embodiment, the eucalyptus distribution map was obtained by Li et al. (2022) using the random forest classification algorithm based on time series Landsat NDVI and topographic features. The producer accuracy of this eucalyptus distribution map is 88.2%, and the user accuracy is 89.6%. After overlaying the carbon storage distribution map of the Lidar coverage area with the eucalyptus distribution map, sample points are randomly selected by stratification according to the carbon storage size, and try to make the sample points fall in different forest stands. After determining the samples, randomly select 60% of the samples as the training set for the subsequent model, and 40% of the samples as the test set.
[0052] Step 7: Use the random location segmentation method and Chow test to monitor the dynamic changes of Landsat time series NDVI data in the study area, so as to obtain the eucalyptus forest age.
[0053] In this embodiment, the algorithm for extracting the eucalyptus forest age comes from Li et al. (2022), and uses the Landsat time series NDVI data from August 1986 to January 2021. The main steps include using random location segmentation to detect NDVI time series change points, determining candidate change points, removing false change points, using random forest to classify the segmented segments, and identifying eucalyptus reforestation, rotation cycle and age. The product accuracy of the eucalyptus forest age in the embodiment is R 2 = 0.91, RMSE = 13.3 months, and the spatial resolution is 30m.
[0054] Step 8: According to the sample positions collected in Step 6, extract the values of the eucalyptus forest ages at the corresponding positions. Then, perform a logarithmic transformation on the eucalyptus forest ages. Using this variable as the independent variable and the training samples in Step 6 as the dependent variable, establish a non-linear regression model for the eucalyptus carbon storage. The specific non-linear regression model is:
[0055] y = c0 + c1ln(x) (2)
[0056] Where y is the carbon storage value of the training samples in Step 6, x represents the eucalyptus forest age in Step 7, and c0 and c1 represent the coefficients in the model.
[0057] Establish a unary functional relationship between carbon storage and forest age through the software SPSS to obtain c0 and c1.
[0058] In this embodiment, considering that the relationship between eucalyptus forest age and carbon storage is non-linear, convert the forest age into the logarithmic form with base e, and then obtain the regression model between forest age and carbon storage: y = -70.31 + 29.57ln(x). Figure 2 The model fitting situation shown in
[0059] Step 9: According to the established eucalyptus carbon storage model for the study area in Step 8, input the model equation into the software ENVI and calculate each pixel of the eucalyptus forest age image obtained in Step 7 to obtain the eucalyptus carbon storage estimation result map for the entire study area.
[0060] In this embodiment, since the spatial resolution of the eucalyptus forest age distribution map obtained in Step 7 is 30m, after inputting it into the carbon storage model for estimation, the eucalyptus carbon storage estimation result map for the study area is obtained ( Figure 3 ), with a spatial resolution of 30m.
[0061] Those skilled in the art should understand that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0062] This application is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, as well as the combination of flows and / or blocks in the flowchart and / or block diagram. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate a means for implementing the functions specified in the Figure 1 one or more flows and / or blocks Figure 1 one or more blocks.
[0063] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including an instruction means that implements the functions specified in the Figure 1 one or more flows and / or blocks Figure 1 one or more blocks.
[0064] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, so that the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in the Figure 1 one or more flows and / or blocks Figure 1 one or more blocks.
[0065] As described above, it is only the preferred embodiment of the present invention, and it is not a limitation of the present invention in other forms. Any person skilled in the art may use the disclosed technical content to make changes or modifications into equivalent embodiments with equivalent changes. However, any simple modification, equivalent change, and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution content of the present invention still fall within the protection scope of the technical solution of the present invention.
[0066] This patent is not limited to the above best mode. Anyone inspired by this patent can obtain various other forms of small-sample transferable eucalyptus carbon storage estimation methods based on remote sensing. All equivalent changes and modifications made according to the scope of the patent application of the present invention shall fall within the scope covered by this patent.
Claims
1. A small-sample transferable eucalyptus carbon storage estimation method based on remote sensing, characterized in that: Associate the sample plots with the Landsat-derived vegetation ages using airborne lidar data, and estimate the carbon storage of eucalyptus within the regional scope; Including the following steps: Step 1: Use field per-tree measurement to obtain the observed values of carbon storage in the sample plots; Step 2: Preprocess the Lidar data in the study area to obtain the canopy height model CHM; Step 3: Take the sample plot range as the vector boundary, and based on the Lidar CHM data, extract the stand height variables corresponding to the sample plots; Step 4: Use stepwise regression to select statistically significant variables with the carbon storage of the sample plots as the dependent variable and the Lidar variables as the independent variables, and establish a multiple linear regression model for the carbon storage of eucalyptus; Step 5: According to the eucalyptus carbon storage model established in Step 4 for the Lidar coverage area, calculate the estimated value of carbon storage in the Lidar coverage area; Step 6: Overlay the carbon storage distribution map of the Lidar coverage area with the eucalyptus distribution map, and use the stratified random sampling method to select training samples and validation samples; Step 7: Use the random location segmentation method and the Chow test to monitor the dynamic changes of the Landsat time series NDVI data in the study area, so as to obtain the eucalyptus forest age; Step 8: According to the sample positions collected in Step 6, extract the values of the eucalyptus forest age at the corresponding positions; then logarithmically transform the eucalyptus forest age, and use the transformed values as the independent variables and the training samples in Step 6 as the dependent variables to establish a non-linear regression model for the carbon storage of eucalyptus; Step 9: According to the eucalyptus carbon storage model established in Step 8 for the study area, calculate each pixel of the eucalyptus forest age image obtained in Step 7 to obtain the estimated result map of the eucalyptus carbon storage in the entire study area; In Step 4, the specific regression model is: y = b0 + b1x1 + b2x2 + … + b i x i (1) where y represents the measured value of plot carbon storage, and x i represents the Lidar variable set screened in step 3, and b i represents the coefficient in the linear model; Obtain the specific model and model parameters through the stepwise regression algorithm in the software SPSS; In Step 8, according to the sample positions collected in Step 6, extract the values of the eucalyptus forest age at the corresponding positions; then logarithmically transform the eucalyptus forest age, and use the transformed values as the independent variables and the training samples in Step 6 as the dependent variables to establish a non-linear regression model for the carbon storage of eucalyptus. The specific non-linear regression model: y = c0 + c1ln(x) (2) Where y is the carbon storage value of the training samples in Step 6, x represents the eucalyptus forest age in Step 7, and c0 and c1 represent the coefficients in the model; Establish a unary functional relationship between carbon storage and forest age through the software SPSS to obtain c0 and c1.
2. The remote sensing-based small-sample transferable eucalyptus carbon storage estimation method according to claim 1, characterized in that: In Step 5, according to the eucalyptus carbon storage model established in Step 4 for the Lidar coverage area, determine the selected Lidar variables; use the Create Fishnet tool in the software ArcMap to create polygon vector data of the sample plot size with the Lidar data as the range; then according to the vector data, extract the corresponding Lidar statistical variables within the polygon range; finally, substitute the functional relationship into all polygons in the software ArcMap to obtain the estimated value of carbon storage in the Lidar coverage area.
3. The method for estimating the carbon storage of eucalyptus based on remote sensing with small samples and transferability according to claim 1, characterized in that: In step 9, according to the eucalyptus carbon storage model of the study area established in step 8, after inputting the model equation in the software ENVI, each pixel of the eucalyptus forest age image obtained in step 7 is calculated to obtain the eucalyptus carbon storage estimation result map of the entire study area.