Method for estimating grassland biomass based on domestic high-resolution remote sensing satellite data

Abstract

The invention discloses a grassland biomass estimation method based on domestic high-resolution remote sensing satellite data, and belongs to the technical field of remote sensing data analysis. According to the method, the biomass sample data is acquired, a vegetation index raster data set of global coverage of an evaluation area is established by utilizing domestic high-resolution remote sensing satellite data, a biomass statistical model is established based on the biomass sample data and the vegetation index raster data set, the biomass of the evaluation area is further estimated, and the scheme is simple in calculation, high in data precision and wide in application range.

Description

Technical Field

[0001] The present invention belongs to the technical field of remote sensing data analysis, and in particular relates to a grassland biomass estimation method based on domestically produced high-resolution remote sensing satellite data. Background Art

[0002] The key to maintaining the stability of grassland ecosystems lies in the rational use of grasslands based on biomass. Biomass is an important parameter for evaluating grassland ecosystem functions. Accurately assessing biomass is of great significance to the evaluation of grassland ecosystem functions and the rational use of grassland resources. At present, the use of remote sensing technology to estimate biomass is a common method. Satellite remote sensing technology has the advantages of strong timeliness, low cost, wide coverage area, and high observation frequency. When using remote sensing satellite monitoring technology to evaluate biomass, it is generally necessary to combine the vegetation index of grassland vegetation. The vegetation index mainly uses linear and nonlinear combinations of different bands to form a characteristic index that can reflect the growth status of green vegetation. It is a common practice to use the vegetation index to fit the biomass of a certain number of sample points on the grassland ground to establish a regression model. After statistical significance analysis and accuracy verification, the best model is selected for application.

[0003] At present, many scholars at home and abroad use remote sensing technology to conduct inversion research on grassland biomass. Most of the existing studies are based on foreign open-source remote sensing data products or open-source images, such as MODIS vegetation index data products, sentinel images, etc. A few involve domestic images, mainly medium and low-resolution satellite images, such as the China Environmental Disaster Reduction Satellite. In recent years, with the launch of domestic high-resolution series satellites, the resolution and coverage frequency of domestic satellites have become increasingly higher, but there are currently few studies using domestic high-resolution satellite data to estimate grassland biomass. In addition, most of the existing studies are based on two or three preset models to carry out statistical analysis, and there is a lack of more models to participate in statistical comparative analysis in order to obtain a better model. Based on this, the present invention proposes a model optimization method for using domestic high-resolution remote sensing satellites to calculate vegetation indices and conduct grassland biomass inversion research. Summary of the Invention

[0004] To address the above issues, the present invention provides a method for estimating grassland biomass based on domestically produced high-resolution remote sensing satellite data. By utilizing domestically produced high-resolution remote sensing satellite data for grassland biomass estimation, the accuracy of grassland biomass calculation is improved. In a first aspect, the present application discloses a method for estimating grassland biomass based on domestically produced high-resolution remote sensing satellite data, comprising: Step S1: Obtain grassland distribution survey data and high-resolution remote sensing satellite data in the assessment area; Step S2: Collecting biomass samples using the grassland distribution survey data in the assessment area, wherein the biomass sample data includes test sample data and verification sample data; Step S3: Calculate and evaluate regional vegetation index using domestic high-resolution remote sensing satellite data; Step S4: constructing a statistical model using the biomass sample data and vegetation index data, and performing statistical tests on the statistical model to select candidate models; Step S5: Based on the multiple candidate models selected in step S4, the predicted value of the biomass of the sample point is calculated using the validation sample data in step S2. Combined with the observed value, the simulation accuracy of each model is tested by leave-one-out cross-validation to select the optimal model; Step S6: Calculate and evaluate the grassland biomass in the region using the optimal model and the regional vegetation index raster dataset.

[0005] In addition, step S1 also includes: Step S11, collecting the latest annual land change survey data from the natural resources department of the assessment area to extract grassland distribution data as the grassland biomass assessment scope and sample collection area; Step S12, collects and pre-processes domestic high-resolution remote sensing satellite data during the period when the grassland vegetation in the assessment area is lush, obtains the reflectivity values of each band in the domestic high-resolution satellite, performs mosaic processing, and obtains an image covering the entire assessment area.

[0006] Furthermore, step S2 includes: Step S21, using the latest annual land change survey data of the assessment area, extracting three grassland types coded as 0401-natural grassland, 0402-marsh grassland, and 0403-artificial grassland as the scope of sample collection and biomass assessment; Step S22: Based on the grassland phenological characteristics in the assessment area, a sample plot survey and biomass acquisition are carried out during the period when the grassland vegetation is growing vigorously, so as to obtain biomass sample data in the assessment area.

[0007] Furthermore, step S22 includes: Step S221: selecting grassland plots with flat terrain and uniform vegetation composition covering different grassland community types and utilization patterns within the assessment area as sample points, with the number of sample points being greater than 100; Step S222: Each sample plot is separated by a certain distance, and the sample plot area is 200m×200m; three sample plots with an area of 1m×1m are randomly selected in each sample plot, and their location information is recorded using a handheld global positioning system terminal; Step S223, using the harvesting method to obtain the living parts of the above-ground plants in the sample plot, and weighing the fresh weight; drying at a constant temperature of 65°C for 24 hours to a constant weight, and weighing the dry weight; calculating the average biomass weight of the three sample plots in each sample plot as the average weight per unit area of the sample plot, and the center point of the sample plot as the sample point location.

[0008] Furthermore, step S3 includes: The normalized difference vegetation index (NDVI), enhanced vegetation index (EVI), ratio vegetation index (RVI), normalized difference red edge vegetation index (NDRE), and soil adjusted vegetation index (SAVI) were calculated using the pre-processed reflectance values of each band from domestic high-resolution satellites to form a vegetation index raster dataset covering the entire assessment area. The specific calculation formula is as follows: in, 、 、 、 Represent the reflectance of the near-infrared, red, blue, and red-edge bands respectively; L represents the green vegetation density in the assessment area, and its value range is [-1,1].

[0009] Furthermore, step S4 includes: In step S41, a statistical model is constructed using the vegetation index x at the same sampling point as an independent variable and the collected biomass sample weight y as a dependent variable. The constructed models include linear models, logarithmic models, inverse models, quadratic models, cubic models, composite models, power models, S models, growth models, exponential distribution models, and logistic models. The specific formulas are: Linear Model: Logarithmic model: Inverse model: Quadratic term model: Cubic Model: Composite Model: Power Model: S Model: Growth Model: Exponential distribution model: Logistic Model: in, 、 、 、 All are statistical test parameters; is the upper limit parameter of the Logisti model; in the above model, x is the vegetation index of the same sampling point, and y is the weight of the collected biomass sample; Step S42: Statistical test analysis is performed using statistical software to calculate test parameters of multiple models, and models with overall model significance and parameter statistical significance that meet the significance level are selected as candidate models.

[0010] Furthermore, step S42 specifically includes: The overall significance of the model satisfies: the p-value of the F statistic is less than 0.05 or 0.01; the statistical significance of the parameter satisfies: the p-value of the t statistic is less than 0.05 or 0.01.

[0011] Furthermore, in step S5, the leave-one-out cross-validation method is used to test the simulation accuracy of each model, including: The model was used to calculate the biomass prediction value of the test sample, and the relative root mean square error (RRMSE) and relative error (RE) were used to evaluate the accuracy of the model. The calculation formulas of RRMSE and RE are as follows: in: To verify the sample biomass predictions; is the measured value of aboveground biomass of the verification sample; n is the number of verification samples; , is the average of all predicted values of the validation sample.

[0012] In a second aspect, the present application provides a computer device comprising a memory, a processor, and program instructions stored in the memory for execution by the processor, wherein the processor executes the program instructions to implement the steps of any one of the methods described in the first aspect.

[0013] In a third aspect, the present application provides a computer-readable storage medium storing a computer program, which implements the method described in any one of the first aspects when executed by a processor.

[0014] This application discloses a grassland biomass estimation method based on domestic high-resolution remote sensing satellite data, which belongs to the technical field of remote sensing data analysis. This application obtains biomass sample data of the assessment area based on survey data, and uses domestic high-resolution remote sensing satellite data to establish a vegetation index raster dataset covering the entire assessment area. Based on the biomass sample data and the vegetation index raster dataset, a biomass statistical model is established to estimate the biomass of the assessment area. This solution has simple calculations, high data accuracy, and a wide range of applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The following exemplary embodiments and descriptions with reference to the accompanying drawings are intended to explain the present invention and are not intended to limit the present invention.

[0016] Figure 1 This is a flow chart of the grassland biomass estimation method based on domestic high-resolution remote sensing satellite data in the present invention. DETAILED DESCRIPTION

[0017] The following will be combined with the drawings in the embodiments of this application to further clearly and completely describe the technical solutions in the embodiments of this application. It should be noted that the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making any creative work shall fall within the scope of protection of this application.

[0018] In order to make the purpose, technical solutions and advantages of this application more clear, the technical solutions of the present invention are clearly and completely described below with reference to the accompanying drawings and specific embodiments. It should be noted that those skilled in the art can make various changes and improvements without departing from the concept of the present invention, and these are all within the scope of protection of this application.

[0019] The embodiments of the present application are further described in detail below with reference to the accompanying drawings: like Figure 1 As shown, in the first embodiment of the present application, a grassland biomass estimation method based on domestic high-resolution remote sensing satellite data is disclosed, including Step S1: Obtain grassland distribution survey data and high-resolution remote sensing satellite data in the assessment area.

[0020] The most recent annual land change survey data from the natural resources department of the assessment area is collected to extract grassland distribution data, which serves as the scope of grassland biomass assessment and the area for sample collection.

[0021] High-resolution remote sensing satellite data from domestically produced satellites covering the assessment area was collected, primarily from Gaofen-1, Gaofen-2, and Ziyuan-1 satellites. The data primarily focused on July and August, when vegetation is lush and growing. Preprocessing, including radiometric calibration and orthorectification, was performed on the image processing platform to obtain reflectance values for each band from domestically produced high-resolution satellites. Mosaic processing was then performed to create an image covering the entire assessment area. This image was used to calculate vegetation indices.

[0022] Step S2: Collect biomass samples using the grassland distribution survey data in the assessment area, wherein the biomass sample data includes test sample data and verification sample data.

[0023] Using the most recent annual land change survey data for the assessment area, grassland land types coded 0401 (natural grassland), 0402 (swamp grassland), and 0403 (artificial grassland) were extracted for sample collection and biomass assessment. During the lush vegetation growth period, based on the phenological characteristics of the grasslands in the assessment area, plot surveys and biomass collection were conducted in July or August. Several grassland plots with relatively flat terrain and uniform vegetation composition were selected within the assessment area as sampling points, taking into account the coverage of different grassland community types and land use patterns. Each plot was spaced a certain distance apart, measuring 200 m x 200 m. Within each plot, three plots (1 m x 1 m) were randomly selected, and their locations were recorded using a handheld GPS device. Survey metrics included height, cover, and biomass. Live aboveground plant parts within the plots were harvested and weighed freshly. The plants were then dried at a constant temperature (65°C) for 24 hours to a constant weight, and the dry weight was then measured. The average biomass weight of the three plots in each plot was calculated as the average weight per unit area of the plot, and the center of the plot was used as the sample location. Depending on the size of the assessment area and the type of grassland community, the number of samples should be as large as possible, generally reaching more than 100 sample points.

[0024] Generally, based on the size of the assessment area and the number of samples, as many grassland biomass samples as possible should be collected. A subset of these samples should be selected for later model accuracy verification. Sampling can be tailored based on regional differences within the grassland, differences in vegetation growth, and other factors. Alternatively, samples can be numbered and validation data selected at regular intervals. Generally, when the sample size reaches hundreds, validation data should account for 10%.

[0025] Step S3: Calculate and evaluate regional vegetation index using domestic high-resolution remote sensing satellite data.

[0026] Using pre-processed reflectance values from various bands of domestically produced high-resolution satellite data, we calculate common vegetation indices to form a vegetation index grid dataset covering the entire assessment area. These include the Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), Ratio Vegetation Index (RVI), Normalized Difference Red Edge Vegetation Index (NDRE), and Soil Adjusted Vegetation Index (SAVI). The calculation formula is as follows: in: 、 、 、 L represents the reflectance of the near-infrared, red, blue, and red-edge bands, respectively. L varies from -1 to 1, depending on the density of green vegetation in the assessment area. In areas with high green vegetation, L = 0, in which case SAVI is identical to NDVI. Conversely, in areas with low green vegetation, L = 1. Typically, L is set to 0.5 to accommodate most land cover.

[0027] Step S4: construct a statistical model using the biomass sample data and vegetation index data, and perform statistical tests on the statistical model to select alternative models.

[0028] A statistical model is constructed using the vegetation index x calculated at the same sampling point as the independent variable and the biomass sample weight y as the dependent variable. y can be either dry weight or fresh weight. Common candidate models include linear models, logarithmic models, inverse models, quadratic models, cubic models, composite models, power models, S models, growth models, exponential distribution models, and logistic models. The formulas are: Linear Model: Logarithmic model: Inverse model: Quadratic term model: Cubic Model: Composite Model: Power Model: S Model: Growth Model: Exponential distribution model: Logistic Model: in, 、 、 、 All are statistical test parameters; is the upper limit parameter of the Logisti model; in the above model, x is the vegetation index of the same sampling point, and y is the weight of the collected biomass sample; Step S5: After selecting multiple candidate models in step S4, use the validation sample data in step S2 to calculate the predicted value of the sample point biomass. Combined with the observed value, the simulation accuracy of each model is tested using the leave-one-out cross-validation method to select the optimal model.

[0029] Statistical tests of candidate models include overall model significance and parameter significance. These parameters can be directly derived when analyzing using common statistical software. Regarding model significance, the first step is to use the F-test to test overall model significance. If the p-value of the F-statistic is less than the significance level (e.g., 0.05 or 0.01), the model as a whole is significant. The second step is to use the t-test for each variable to test the parameter significance of each independent variable. If the p-value of the t-statistic is less than the significance level (e.g., 0.05 or 0.01), the independent variable has a significant impact on the dependent variable. The third step is to use the R² / Adjusted R² method to assess the explanatory power of the model; a larger value indicates a better fit. The fourth step is to ensure that the regression assumptions hold true through residual analysis and VIF. Only when both overall model significance and parameter significance are met can the model be considered as a candidate model and participate in the next step of precision verification.

[0030] After selecting multiple candidate models, the validation sample data retained by S2 was used to calculate the predicted biomass values at the sample points. Combined with the observed values, the simulation accuracy of each prediction model was tested using the leave-one-out cross-validation method. The relative root mean square error (RRMSE) and relative error (RE) were mainly used to evaluate the accuracy of the model. RRMSE is used to judge the quality of the regression model and measures the mean deviation between the model prediction value and the mean value. The value range is 0-100%. The smaller the value, the higher the model prediction accuracy. The relative error ranges from 0-100%. The smaller the value, the smaller the error and the higher the model accuracy. The calculation formulas for RRMSE and RE are as follows: in: To verify the sample biomass predictions; is the measured value of aboveground biomass of the verification sample; n is the number of verification samples; , is the average of all predicted values of the validation sample.

[0031] Step S6: Calculate and evaluate the grassland biomass in the region using the optimal model and the regional vegetation index raster dataset.

[0032] This application discloses a grassland biomass estimation method based on domestic high-resolution remote sensing satellite data, which belongs to the technical field of remote sensing data analysis. This application obtains biomass sample data of the assessment area based on survey data, and uses domestic high-resolution remote sensing satellite data to establish a vegetation index raster dataset covering the entire assessment area. Based on the biomass sample data and the vegetation index raster dataset, a biomass statistical model is established to estimate the biomass of the assessment area. This solution has simple calculations, high data accuracy, and a wide range of applications.

Claims

1. A grassland biomass estimation method based on domestic high-resolution remote sensing satellite data, characterized in that: include: Step S1: Obtain grassland distribution survey data and high-resolution remote sensing satellite data in the assessment area; Step S2: collecting biomass samples using the grassland distribution survey data in the assessment area, wherein the biomass sample data includes test sample data and verification sample data; Step S3: Calculate and evaluate regional vegetation index using domestic high-resolution remote sensing satellite data; Step S4: constructing a statistical model using the biomass sample data and vegetation index data, and performing statistical tests on the statistical model to select candidate models; Step S5: Based on the multiple candidate models selected in step S4, the predicted value of the biomass of the sample point is calculated using the validation sample data in step S2. Combined with the observed value, the simulation accuracy of each model is tested by leave-one-out cross-validation to select the optimal model; Step S6: Calculate and evaluate the grassland biomass in the region using the optimal model and the regional vegetation index raster dataset.

2. The method according to claim 1, characterized in that The step S1 includes: Step S11, collecting the latest annual land change survey data from the natural resources department of the assessment area to extract grassland distribution data as the grassland biomass assessment scope and sample collection area; Step S12, collects and pre-processes domestic high-resolution remote sensing satellite data during the period when the grassland vegetation in the assessment area is lush, obtains the reflectivity values of each band in the domestic high-resolution satellite, performs mosaic processing, and obtains an image covering the entire assessment area.

3. The method according to claim 2, characterized in that The step S2 includes: Step S21, using the latest annual land change survey data of the assessment area, extracting three grassland types coded as 0401-natural grassland, 0402-marsh grassland, and 0403-artificial grassland as the scope of sample collection and biomass assessment; Step S22: Based on the grassland phenological characteristics in the assessment area, a sample plot survey and biomass acquisition are carried out during the period when the grassland vegetation is growing vigorously, so as to obtain biomass sample data in the assessment area.

4. The method according to claim 3, characterized in that The step S22 includes: Step S221: selecting grassland plots with flat terrain and uniform vegetation composition covering different grassland community types and utilization patterns within the assessment area as sample points, with the number of sample points being greater than 100; Step S222: Each sample plot is separated by a certain distance, and the sample plot area is 200m×200m; three sample plots with an area of 1m×1m are randomly selected in each sample plot, and their location information is recorded using a handheld global positioning system terminal; Step S223, using the harvesting method to obtain the living parts of the above-ground plants in the sample plot, and weighing the fresh weight; drying at a constant temperature of 65°C for 24 hours to a constant weight, and weighing the dry weight; calculating the average biomass weight of the three sample plots in each sample plot as the average weight per unit area of the sample plot, and the center point of the sample plot as the sample point location.

5. The method according to claim 4, characterized in that The step S3 includes: The normalized difference vegetation index (NDVI), enhanced vegetation index (EVI), ratio vegetation index (RVI), normalized difference red edge vegetation index (NDRE), and soil adjusted vegetation index (SAVI) were calculated using the pre-processed reflectance values of each band from domestic high-resolution satellites to form a vegetation index raster dataset covering the entire assessment area. The specific calculation formula is as follows: in, 、 、 、 Represent the reflectance of the near-infrared, red, blue, and red-edge bands respectively; L represents the green vegetation density in the assessment area, and its value range is [-1,1].

6. The method according to claim 5, characterized in that The step S4 includes: In step S41, a statistical model is constructed using the vegetation index x at the same sampling point as an independent variable and the collected biomass sample weight y as a dependent variable. The constructed models include linear models, logarithmic models, inverse models, quadratic models, cubic models, composite models, power models, S models, growth models, exponential distribution models, and logistic models. The specific formulas are: Linear Model: Logarithmic model: Inverse model: Quadratic term model: Cubic Model: Composite Model: Power Model: S Model: Growth Model: Exponential distribution model: Logistic Model: in, 、 、 、 All are statistical test parameters; is the upper limit parameter of the Logisti model; in the above model, x is the vegetation index of the same sampling point, and y is the weight of the collected biomass sample; Step S42: Statistical test analysis is performed using statistical software to calculate test parameters of multiple models, and models with overall model significance and parameter statistical significance that meet the significance level are selected as candidate models.

7. The method according to claim 6, characterized in that In step S42, the overall significance of the model and the statistical significance of the parameters satisfying the significance level include: The overall significance of the model satisfies: the p-value of the F statistic is less than 0.05 or 0.01; the statistical significance of the parameter satisfies: the p-value of the t statistic is less than 0.05 or 0.

01.

8. The method according to claim 7, characterized in that The step S5 specifically includes: The model was used to calculate the biomass prediction value of the test sample, and the relative root mean square error (RRMSE) and relative error (RE) were used to evaluate the accuracy of the model. The calculation formulas of RRMSE and RE are as follows: in: To verify the sample biomass predictions; is the measured value of aboveground biomass of the verification sample; n is the number of verification samples; , is the average of all predicted values of the validation sample.

9. A computer device, characterized in that: The method comprises a memory, a processor and program instructions stored in the memory and executable by the processor, wherein the processor executes the program instructions to implement the steps of the method according to any one of claims 1 to 8.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images