CN113297904A - Alpine grassland biomass estimation method and system based on satellite driving model - Google Patents
Alpine grassland biomass estimation method and system based on satellite driving model Download PDFInfo
- Publication number
- CN113297904A CN113297904A CN202110412645.6A CN202110412645A CN113297904A CN 113297904 A CN113297904 A CN 113297904A CN 202110412645 A CN202110412645 A CN 202110412645A CN 113297904 A CN113297904 A CN 113297904A
- Authority
- CN
- China
- Prior art keywords
- index
- model
- biomass
- alpine
- data information
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 239000002028 Biomass Substances 0.000 title claims abstract description 105
- 238000000034 method Methods 0.000 title claims abstract description 84
- 230000007613 environmental effect Effects 0.000 claims abstract description 80
- 238000004458 analytical method Methods 0.000 claims abstract description 67
- 238000005457 optimization Methods 0.000 claims abstract description 64
- 238000012545 processing Methods 0.000 claims abstract description 55
- 238000007781 pre-processing Methods 0.000 claims abstract description 49
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 40
- 238000010219 correlation analysis Methods 0.000 claims abstract description 24
- 238000000605 extraction Methods 0.000 claims abstract description 21
- 241000968928 Carex brizoides Species 0.000 claims abstract description 10
- 238000012544 monitoring process Methods 0.000 claims description 40
- 230000008569 process Effects 0.000 claims description 23
- 239000002689 soil Substances 0.000 claims description 17
- 239000004927 clay Substances 0.000 claims description 16
- 230000003595 spectral effect Effects 0.000 claims description 14
- 238000007637 random forest analysis Methods 0.000 claims description 13
- 238000012706 support-vector machine Methods 0.000 claims description 13
- 238000006243 chemical reaction Methods 0.000 claims description 12
- 238000012417 linear regression Methods 0.000 claims description 9
- IJGRMHOSHXDMSA-UHFFFAOYSA-N Atomic nitrogen Chemical compound N#N IJGRMHOSHXDMSA-UHFFFAOYSA-N 0.000 claims description 8
- 230000002159 abnormal effect Effects 0.000 claims description 8
- 230000004927 fusion Effects 0.000 claims description 8
- 238000003062 neural network model Methods 0.000 claims description 8
- 239000004576 sand Substances 0.000 claims description 8
- OAICVXFJPJFONN-UHFFFAOYSA-N Phosphorus Chemical compound [P] OAICVXFJPJFONN-UHFFFAOYSA-N 0.000 claims description 6
- 238000012952 Resampling Methods 0.000 claims description 6
- 230000008030 elimination Effects 0.000 claims description 6
- 238000003379 elimination reaction Methods 0.000 claims description 6
- 238000001556 precipitation Methods 0.000 claims description 6
- ZLMJMSJWJFRBEC-UHFFFAOYSA-N Potassium Chemical compound [K] ZLMJMSJWJFRBEC-UHFFFAOYSA-N 0.000 claims description 4
- 229910052757 nitrogen Inorganic materials 0.000 claims description 4
- 239000005416 organic matter Substances 0.000 claims description 4
- 229910052698 phosphorus Inorganic materials 0.000 claims description 4
- 239000011574 phosphorus Substances 0.000 claims description 4
- 239000011591 potassium Substances 0.000 claims description 4
- 229910052700 potassium Inorganic materials 0.000 claims description 4
- 238000010276 construction Methods 0.000 claims description 2
- 230000008859 change Effects 0.000 abstract description 15
- 238000010586 diagram Methods 0.000 abstract description 6
- 238000009826 distribution Methods 0.000 abstract description 5
- 230000006870 function Effects 0.000 description 23
- 230000014509 gene expression Effects 0.000 description 8
- 238000011160 research Methods 0.000 description 8
- 238000012549 training Methods 0.000 description 8
- 238000013528 artificial neural network Methods 0.000 description 5
- OYPRJOBELJOOCE-UHFFFAOYSA-N Calcium Chemical compound [Ca] OYPRJOBELJOOCE-UHFFFAOYSA-N 0.000 description 4
- 238000013145 classification model Methods 0.000 description 4
- 238000012937 correction Methods 0.000 description 4
- 238000003066 decision tree Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 238000011156 evaluation Methods 0.000 description 4
- 238000010801 machine learning Methods 0.000 description 4
- 230000005855 radiation Effects 0.000 description 4
- 230000009466 transformation Effects 0.000 description 4
- 238000011161 development Methods 0.000 description 3
- 238000013507 mapping Methods 0.000 description 3
- 238000012935 Averaging Methods 0.000 description 2
- 241001465754 Metazoa Species 0.000 description 2
- 241000497386 Silveira Species 0.000 description 2
- 238000013459 approach Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000002790 cross-validation Methods 0.000 description 2
- 230000001419 dependent effect Effects 0.000 description 2
- 230000009977 dual effect Effects 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000005192 partition Methods 0.000 description 2
- 238000009304 pastoral farming Methods 0.000 description 2
- 238000010845 search algorithm Methods 0.000 description 2
- 238000007619 statistical method Methods 0.000 description 2
- 238000002945 steepest descent method Methods 0.000 description 2
- 238000003860 storage Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 238000012546 transfer Methods 0.000 description 2
- 244000025254 Cannabis sativa Species 0.000 description 1
- 241000283984 Rodentia Species 0.000 description 1
- 230000004888 barrier function Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000001035 drying Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 238000005303 weighing Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/13—Satellite images
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/20—Ensemble learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/188—Vegetation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Molecular Biology (AREA)
- Medical Informatics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Multimedia (AREA)
- Computational Linguistics (AREA)
- Chemical & Material Sciences (AREA)
- Food Science & Technology (AREA)
- Astronomy & Astrophysics (AREA)
- Remote Sensing (AREA)
- Medicinal Chemistry (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a method and a system for estimating biomass of alpine grassland based on a satellite-driven model, wherein the method comprises the following steps: obtaining data information; preprocessing the data information to obtain a preprocessing result; performing index extraction processing based on the preprocessing result to obtain a remote sensing index and an environmental index; constructing an index system by utilizing the remote sensing index, the environmental index and the alpine grassland biomass based on the XGBoost algorithm and the correlation analysis; performing optimization selection processing on the satellite driving model based on an index system to obtain an optimization model for optimization selection; and (4) carrying out space-time dynamic analysis processing on the biomass of the alpine grasses based on the optimized and selected optimization model to obtain the biomass analysis result of the alpine grasses. In the embodiment of the invention, the optimal satellite-driven model can be determined through model precision comparison, the space-time distribution diagram of the AG-AGB is drawn, and the space-time dynamic change of the AG-AGB is analyzed.
Description
Technical Field
The invention relates to the technical field of environmental monitoring and evaluation, in particular to a alpine grassland biomass estimation method and system based on a satellite-driven model.
Background
The high-cold grassland in the Qinghai plateau is an important animal husbandry production base and an ecological safety barrier in China, and is extremely sensitive to global climate change and human activities (Fayiah et al, 2019; Liu et al, 2016; Zhang et al, 2014). However, over the last decades alpine grassland has significantly degraded due to excessive grazing, rodent activity and climate change (Chen et al, 2014; Liu et al, 2020; Sun et al, 2019). Alpine grassland biomass (AG-AGB) is used as an important index for monitoring a grassland ecosystem, directly influences grazing and bearing capacity and is closely related to animal husbandry development and herdsman income (Gao et al, 2019; Kong et al, 2019; Zeng et al, 2019). Therefore, timely and accurate estimation of AG-AGB can provide scientific reference for grassland resource management and sustainable utilization (Gao et al, 2020; Liang et al, 2016; Zeng et al, 2019).
Currently, two methods are mainly used for monitoring AG-AGB, namely traditional ground monitoring and satellite monitoring. Traditional ground-based methods estimate biomass by mowing in the field, drying and weighing in the laboratory (Yang et al, 2018). This conventional approach can provide high accuracy AG-AGB information for sample positioning. However, it is time consuming and expensive and ignores the spatial distribution of AG-AGB (Li et al, 2016; Liang et al, 2016; Zeng et al, 2019).
The satellite-driven approach is mainly to estimate biomass using a model of relationships between spectra (e.g., vegetation index and spectral reflectance data) and environmental (geographical, terrain, soil, weather, etc.) indicators and AG-AGB (Liang et al, 2016). At present, a great deal of research using the method mainly focuses on determining evaluation indexes and establishing a relationship model. Many studies use a single vegetation index, such as normalized vegetation index (NDVI), Enhanced Vegetation Index (EVI), Soil Adjusted Vegetation Index (SAVI) to estimate AG-AGB. Although the method is simple to apply, the method lacks accurate estimation capability on AG-AGB due to the influence of environmental factors such as geography, terrain, soil, weather and the like. Therefore, some scholars consider a variety of indicators, including vegetation index and environmental indicators, together to estimate AG-AGB (Liang et al, 2016; Silveira et al, 2019). Since these sensitivity indicators change with changes in the regional environment, there is no uniform indicator to estimate AG-AGB. To improve the estimation accuracy of grass biomass in a specific area, it remains a challenge to determine key estimation indicators from a large number of spectral and environmental indicators.
Existing AG-AGB estimation models can be divided into two categories: linear models and non-linear models. The linear model establishes a linear mathematical relationship between the estimation index and the AG-AGB. For example, Multiple Linear Regression (MLR) is commonly used to estimate AG-AGB (Liang et al, 2016; Silveira et al, 2019). However, these relationships often fail because they lack generality. Thus, as machine learning algorithms have evolved, many scholars (Gao et al, 2020; Yang et al, 2018; Zeng et al, 2019) have applied Artificial Neural Networks (ANNs), Support Vector Machines (SVMs) and Random Forest (RF) algorithms to construct satellite-driven models to estimate AG-AGBs. Satellite-driven models constructed using machine learning algorithms can learn highly complex non-linear mappings and achieve higher estimation accuracy than linear models (Xu et al, 2020; Yang et al, 2018; Zeng et al, 2019). However, these machine learning algorithms also have disadvantages. For example, the estimation accuracy of the Back Propagation Neural Network (BPNN) algorithm depends on the number and quality of samples, and the convergence process may be slow or encounter local minima problems (Yang et al, 2018). The support vector machine algorithm is mainly affected by the kernel function and the penalty factor because its parameters use only the expert experience of reference and are limited in the accuracy of the estimation result (Zhu et al, 2020). The results of the RF algorithm have limited interpretability because the relationship between the predicted values and the responses cannot be checked separately for each tree in the forest (Chagas et al, 2016). Therefore, establishing an accurate satellite-driven model by using a machine learning algorithm has become a research hotspot of high-precision AG-AGB estimation.
Disclosure of Invention
The invention aims to overcome the defects of the prior art, and provides a alpine grassland biomass estimation method and system based on a satellite-driven model.
In order to solve the above technical problem, an embodiment of the present invention provides a method for estimating alpine grassland biomass based on a satellite-driven model, where the method includes:
acquiring data information, wherein the data information comprises remote sensing image information, alpine grassland monitoring data information and environment data information;
preprocessing the data information to obtain a preprocessing result;
performing index extraction processing based on the preprocessing result to obtain a remote sensing index and an environmental index;
constructing an index system by utilizing the remote sensing index, the environmental index and the alpine grassland biomass based on the XGBoost algorithm and the correlation analysis;
performing optimization selection processing on the satellite driving model based on an index system to obtain an optimization model for optimization selection;
and (4) carrying out space-time dynamic analysis processing on the biomass of the alpine grasses based on the optimized and selected optimization model to obtain the biomass analysis result of the alpine grasses.
Optionally, the preprocessing the data information to obtain a preprocessing result includes:
initializing the remote sensing image information to obtain an initialization result of the remote sensing image information;
initializing the alpine grassland monitoring data information to obtain an alpine grassland monitoring data information initialization result;
and initializing the environmental data information to obtain an environmental data information initialization result.
Optionally, the initializing the remote sensing image information to obtain an initialization result of the remote sensing image information includes:
converting MOD09A1 and MCD12Q1 data in the remote sensing image information from sinusoidal projection to Albers projection by using an MODIS reprojection tool, converting an HDF format into a Geo-Tiff format, and then performing image fusion and MCD12Q1 data reclassification to obtain an initialization result of the remote sensing image information;
initializing the alpine grassland monitoring data information to obtain an alpine grassland monitoring data information initialization result, wherein the initializing process comprises the following steps:
vectorization, projection conversion and abnormal value elimination processing are sequentially carried out on the alpine grassland monitoring data information to obtain an alpine grassland monitoring data information initialization result;
and initializing the environmental data information to obtain an environmental data information initialization result. The method comprises the following steps:
and sequentially carrying out spatial interpolation, projection conversion and grid resampling on the environmental data information to obtain an environmental data information initialization result.
Optionally, the index extraction processing based on the preprocessing result to obtain the remote sensing index and the environmental index includes:
extracting and processing the remote sensing index and the environmental index in sequence based on the preprocessing result to obtain the remote sensing index and the environmental index;
the environmental data information includes in-ground data, soil data, and meteorological data.
Optionally, the remote sensing index extraction processing includes obtaining 7 VIs and 5 spectral indexes from MOD09a1 data in the preprocessing result based on ENVI software, and extracting the maximum value of each visible light and spectral index;
the environmental index extraction processing comprises the steps of acquiring longitude, latitude, altitude, gradient and slope terrain indexes from a space shuttle radar terrain task digital elevation model product in the preprocessing result based on ArcGIS software; obtaining soil indexes of 0-30 cm sand content, 0-30 cm clay content, 30-60 cm sand content, 30-60 cm clay content, 0-30 cm sand-clay ratio, 30-60 cm sand-clay ratio, organic matter content, total nitrogen content, total phosphorus content, total potassium content and pH from soil data; acquiring meteorological indexes of annual average temperature, annual total precipitation, annual solar total radiant quantity, > 0-year accumulated temperature and humidity from meteorological data.
Optionally, the XGboost algorithm and correlation analysis-based index system is constructed by using the remote sensing index, the environmental index and alpine grassland biomass, and includes:
performing index analysis processing of alpine grassland biomass estimation on the remote sensing index and the environmental index based on the correlation analysis to obtain an initial estimation index of alpine grassland biomass estimation;
based on an XGboost algorithm, carrying out redundancy removal on the initial estimation index of the alpine grassland biomass estimation to obtain an estimation index of the alpine grassland biomass estimation;
and constructing an index system based on the estimation index, the environmental index and the alpine grassland biomass.
Optionally, the satellite driving model includes a multiple linear regression model, a BP neural network model, a support vector machine model, and a random forest model.
Optionally, the performing, on the basis of the index system, optimization selection processing on the satellite driving model to obtain an optimization model for optimization selection includes:
determining precision parameters, wherein the precision parameters comprise a decision coefficient, a relative root mean square error and a residual prediction deviation;
and referencing residual prediction deviations of all models in the satellite driving model, comparing the decision coefficient of all models when the residual prediction deviations of all models are in the same interval, and then referencing the relative root mean square error to obtain an optimized and selected optimization model.
Optionally, the optimization model based on optimization selection performs space-time dynamic analysis processing on the alpine grassland biomass to obtain an alpine grassland biomass analysis result, including:
performing Sen and Mann-Kendall trend analysis on the biomass of the alpine grassland based on the optimized and selected optimization model to obtain a first analysis result;
performing R/S analysis on the biomass of the alpine grassland based on the optimized and selected optimization model to obtain a second analysis result;
and performing superposition judgment based on the first analysis result and the second analysis result to obtain a biomass analysis result of the alpine grassland.
In addition, the embodiment of the invention also provides a alpine grassland biomass estimation system based on the satellite-driven model, and the system comprises:
an obtaining module: the system comprises a data acquisition module, a data processing module and a data processing module, wherein the data acquisition module is used for acquiring data information, and the data information comprises remote sensing image information, alpine grassland monitoring data information and environmental data information;
a preprocessing module: the data information is preprocessed to obtain a preprocessing result;
an index extraction module: the system is used for extracting and processing indexes based on the preprocessing result to obtain remote sensing indexes and environmental indexes;
an index system construction module: the system is used for constructing an index system by utilizing the remote sensing index, the environmental index and the alpine grassland biomass based on the XGBoost algorithm and the correlation analysis;
an optimization selection module: the optimization model is used for carrying out optimization selection processing on the satellite driving model based on an index system to obtain an optimization model for optimization selection;
a space-time dynamic analysis module: and the method is used for carrying out space-time dynamic analysis processing on the alpine grassland biomass based on the optimization model selected by optimization to obtain the alpine grassland biomass analysis result.
In the embodiment of the invention, an XGboost algorithm and a correlation analysis method are mainly used, a grassland biomass monitoring index system is constructed by researching the relation between indexes and biomass, then a plurality of linear/nonlinear inversion models are constructed according to the constructed index system, an optimal satellite driving model can be determined through model precision comparison, a space-time distribution diagram of AG-AGB is drawn, and the space-time dynamic change of the AG-AGB is analyzed.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is a schematic flow chart of a method for estimating biomass in alpine grasses based on a satellite-driven model according to an embodiment of the present invention;
FIG. 2 is a schematic structural diagram of a alpine grassland biomass estimation system based on a satellite-driven model according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Example one
Referring to fig. 1, fig. 1 is a schematic flow chart of a method for estimating alpine grassland biomass based on a satellite-driven model according to an embodiment of the present invention
As shown in fig. 1, a method for estimating alpine grassland biomass based on a satellite-driven model includes:
s11: acquiring data information, wherein the data information comprises remote sensing image information, alpine grassland monitoring data information and environment data information;
in the specific implementation process of the method, the area needing to be subjected to alpine grassland biomass estimation is determined firstly, and then data information of the area is obtained, wherein the data information comprises remote sensing image information, alpine grassland monitoring data information and environment data information.
Specifically, the areas determined to require alpine grassland biomass estimation in the present application are located in the Qinghai province in the northwest of China (31 ° 36 '-39 ° 19' for north latitude and 89 ° 35 '-103 ° 04' for east longitude), and the total area is about 69.67 × 104km 2. The average altitude of the research area is more than 3000 m, belonging to continental plateau climate. The annual average air temperature is-5.7 ℃ to 8.5 ℃, the annual total precipitation is 50mm to 450mm, and the annual total solar radiation is 690.8kJ to 753.6kJ (Deng et al, 2017). The area of the alpine grassland is about 41.19 multiplied by 104km2, and accounts for 59.13 percent of the total area of the Qinghai province. Major steppe types include alpine meadows, alpine steppes, and temperate steppes (Qinghai et al, 2012). The growing season of the grassland starts from the beginning of 5 months and ends in the last 9 months (Qinghai et al, 2012).
S12: preprocessing the data information to obtain a preprocessing result;
in a specific implementation process of the present invention, the preprocessing the data information to obtain a preprocessing result includes: initializing the remote sensing image information to obtain an initialization result of the remote sensing image information; initializing the alpine grassland monitoring data information to obtain an alpine grassland monitoring data information initialization result; and initializing the environmental data information to obtain an environmental data information initialization result.
Further, the initializing the remote sensing image information to obtain an initialized result of the remote sensing image information includes: converting MOD09A1 and MCD12Q1 data in the remote sensing image information from sinusoidal projection to Albers projection by using an MODIS reprojection tool, converting an HDF format into a Geo-Tiff format, and then performing image fusion and MCD12Q1 data reclassification to obtain an initialization result of the remote sensing image information; initializing the alpine grassland monitoring data information to obtain an alpine grassland monitoring data information initialization result, wherein the initializing process comprises the following steps: vectorization, projection conversion and abnormal value elimination processing are sequentially carried out on the alpine grassland monitoring data information to obtain an alpine grassland monitoring data information initialization result; and initializing the environmental data information to obtain an environmental data information initialization result. The method comprises the following steps: and sequentially carrying out spatial interpolation, projection conversion and grid resampling on the environmental data information to obtain an environmental data information initialization result.
Specifically, the remote sensing image data preprocessing mainly comprises radiation correction, atmospheric correction and image fusion. MODIS re-projection tool (MRT) is used to convert MOD09A1 and MCD12Q1 data from sinusoidal projection to Albers projection, and HDF format to Geo-Tiff format, and image fusion and MCD12Q1 data re-classification are performed. The method mainly comprises vectorization, projection conversion and abnormal value elimination. The abnormal value is removed by a method of average value +/-2 multiplied by standard deviation. The preprocessing of environment data such as geography, soil, weather and the like mainly comprises spatial interpolation, projection conversion and grid resampling. The meteorological data is interpolated mainly by using Anuspli software (written by the teaching of Hutchinson of national university of Australia). In addition, for the convenience of calculating AG-AGB, projections of all the data described above are converted into Albers projections, the format thereof is adjusted to Geo Tiff, and environment data of different spatial resolutions are resampled by the Nearest Neighbor Method (NNM), and raster data of a spatial resolution of 500m × 500m is generated.
S13: performing index extraction processing based on the preprocessing result to obtain a remote sensing index and an environmental index;
in the specific implementation process of the present invention, the performing the index extraction processing based on the preprocessing result to obtain the remote sensing index and the environmental index includes: and extracting and processing the remote sensing index and the environmental index in sequence based on the preprocessing result to obtain the remote sensing index and the environmental index.
Further, the remote sensing index extraction processing comprises the steps of obtaining 7 VIS and 5 spectral indexes from MOD09A1 data in the preprocessing result based on ENVI software, and extracting the maximum value of each visible light and spectral index; and the environmental index extraction processing comprises the step of acquiring terrain indexes of longitude, latitude, altitude, gradient and slope direction from a space shuttle radar terrain task digital elevation model product in the preprocessing result based on ArcGIS software.
Specifically, the remote sensing index extraction mainly uses ENVI software to obtain 7 VIS (NDVI, EVI, SAVI, MSAVVI, OSAVI, SATVI and RVI) and 5 spectral indexes (B7/B2, B2-B7, B7/B5, B5/B7 and (B5-B7)/(B5+ B7)) from MOD09A1 data, and extracts the maximum value of each visible light and spectral index (only 7 months and 8 months) in the research area.
The environmental indexes mainly comprise data of terrain, soil, meteorological phenomena and the like. And (3) acquiring terrain indexes such as longitude, latitude, altitude, gradient, slope direction and the like from a space shuttle radar terrain task (SRTM) Digital Elevation Model (DEM) product by using ArcGIS software. Soil indexes such as sand content of 0-30 cm, clay content of 0-30 cm, sand content of 30-60 cm, clay content of 30-60 cm, sand-clay ratio of 0-30 cm, sand-clay ratio of 30-60 cm, organic matter content, total nitrogen content, total phosphorus content, total potassium content and pH are obtained from soil data with spatial resolution of 1 km. Acquiring meteorological indexes such as annual average temperature, annual total precipitation, annual solar total radiant quantity, > 0-year accumulated temperature, humidity and the like from meteorological data.
S14: constructing an index system by utilizing the remote sensing index, the environmental index and the alpine grassland biomass based on the XGBoost algorithm and the correlation analysis;
in the specific implementation process of the invention, the XGBoost algorithm and correlation analysis based index system is constructed by utilizing the remote sensing index, the environmental index and the alpine grassland biomass, and comprises the following steps: performing index analysis processing of alpine grassland biomass estimation on the remote sensing index and the environmental index based on the correlation analysis to obtain an initial estimation index of alpine grassland biomass estimation; based on an XGboost algorithm, carrying out redundancy removal on the initial estimation index of the alpine grassland biomass estimation to obtain an estimation index of the alpine grassland biomass estimation; and constructing an index system based on the estimation index, the environmental index and the alpine grassland biomass.
Specifically, an AG-AGB estimation index is selected by adopting a correlation analysis method; and then, on the basis of the correlation analysis result, introducing an extreme gradient boost (XGboost) algorithm to determine an AG-AGB estimation index. The XGBoost algorithm, as a method of determining all relevant features or metrics, provides an intrinsic measure of Feature Importance (FI). The XGBoost algorithm may rank the entire FI by averaging the FI in each tree, which is calculated based on the amount of information obtained after segmenting the tree using the features. In order to accurately obtain key indexes and reduce data redundancy among the indexes, the research adopts an XGboost algorithm in combination with a correlation analysis method, and selects significant and irrelevant indexes to estimate AG-AGB. The XGboost algorithm is briefly described as follows:
and establishing a classification model based on an AGB estimation index system. The performance of the indicators in the classification model is checked to obtain FIs and sorted in descending order. The formula of FI is:
where H (T) and H (T | F) are the entropy of parent and child nodes of a partition based on F features, piIs the score of each labeled sample on a node. And selecting a plurality of indexes with higher FI values to generate index subsets, and classifying again. This process is repeated until all the metrics are selected, the classification of all subsets is examined, and the best subset of metrics (i.e., the subset with the relatively higher area under the curve value and fewer metrics) is selected. And finally, constructing an index system according to the estimation index, the environmental index and the biomass of the alpine grassland.
S15: performing optimization selection processing on the satellite driving model based on an index system to obtain an optimization model for optimization selection;
in the specific implementation process of the invention, the satellite driving model comprises a multiple linear regression model, a BP neural network model, a support vector machine model and a random forest model.
Further, the optimizing and selecting the satellite driving model based on the index system to obtain an optimized and selected optimizing model includes: determining precision parameters, wherein the precision parameters comprise a decision coefficient, a relative root mean square error and a residual prediction deviation; and referencing residual prediction deviations of all models in the satellite driving model, comparing the decision coefficient of all models when the residual prediction deviations of all models are in the same interval, and then referencing the relative root mean square error to obtain an optimized and selected optimization model.
Specifically, the satellite driving model comprises a multiple linear regression model, a BP neural network model, a support vector machine model and a random forest model.
The multivariate linear regression model takes AG-AGB as a dependent variable (Y) and takes a corresponding AG-AGB estimation index as an independent variable (X);
in the formula, betaiIs the ith weight coefficient determined from expert experience.
The BP neural network model is a multi-layer forward neural network based on an error back propagation algorithm, and generally consists of an input layer, an output layer and one or more hidden layers. It can use the guiding learning method to learn and train, and can simulate the relation between any nonlinear input variable and output variable. The learning process mainly includes forward propagation of the input signal and backward propagation of the error. The training process includes continuously adjusting the connection weights using a steepest descent method and a back propagation algorithm until the output mean square error meets a desired criterion. The equations for signal propagation and Mean Square Error (MSE) are respectively expressed as follows:
oj＝fi(∑wjioi+θj)；
ok＝fj(∑wkjoj+θk)；
in the formula, thetajIs the hidden layer information; oiIs input layer information (AG-AGB estimation index); w is ajiIs the weight of the input layer to the hidden layer; thetajIs the threshold of the hidden layer; f. ofiIs the rainlm function of the input layer to the hidden layer; okIs output layer information (AG-AGB estimate); w is akiIs the weight of the hidden layer to the output layer; thetakIs the threshold of the output layer; f. ofjIs the purelin function of the hidden layer to the output layer; o is AG-AGB field measurement data; n is the number of samples.
Support vector machine models are proposed by cortex and Vapnik (1995) for solving practical problems (small samples, non-linearity, high dimensions and local minimum points) by defining a kernel function, applying a non-linear transformation to a training data set and mapping it to a high-dimensional feature space. The expression of the SVM model is as follows:
wherein f (x) is an AG-AGB estimate; w is aiIs a weight vector; b is the deviation.
Representing a non-linear transfer function. Cortex and Vapnik (1995) introduced the following convex optimization problem with an e-insensitive penalty function to obtain a solution to the following equation:
wherein | | w | | non-conducting phosphor2Represents the flatness of the m-dimensional space; ε is a parameter representing the maximum allowable error between the observed value and the model prediction; xiiAnd
is a relaxation variable, penalizes training errors using a loss function on the error margin ξ; c is a positive compromise parameter that determines the magnitude of the empirical error in the optimization problem. The above equation is typically solved in the dual form of lagrange multipliers and applying optimality conditions.
Wherein
Is a transformation of w. The SVM function is represented as:
wherein
Is a kernel function. A Radial Basis Function (RBF) is selected as the kernel function. And determining a penalty parameter C and a kernel function parameter gamma of the support vector machine by adopting a grid search algorithm of cross validation.
The random forest model is an integrated learning method proposed by Breiman and Adele (2001), and has strong adaptability to a data set. It can handle high dimensional features and is not prone to overfitting. In addition, the model training speed is relatively fast. The expressions of the RF model and their errors are respectively as follows:
in the formula fKIs the predicted value of each decision tree; k is the number of decision trees;
is an AG-AGB estimate of out-of-bag data (OOB) that consists of the unextracted samples of the original samples. y isiIs and
the corresponding AG-AGB measured value; n is the OOB sample number.
In this study, three precision parameters were selected to verify the effect of the model: coefficient of determination (R2), Relative Root Mean Square Error (RRMSE) and residual prediction bias (PRD). Wherein R is2Is used to measure the stability of the model. R2The larger the model, the more stable. RRMSE reflects the accuracy of the model. The smaller the RRMSE, the better the prediction capability. The performance of the model was evaluated using RPD. When RPD is less than 1.4, the model is not good; when the RPD is more than or equal to 1.4 and less than or equal to 2, the model has certain prediction capability; when RPD is more than 2, the model has stronger prediction capability. Judging the criterion of the model precision: firstly, referring to the RPD value, and preferentially comparing the model R when the RPD is in the same interval2The size is considered, the RRMSE size is considered, and the optimal model judgment principle is R2Relatively large, and relatively small RRMSE.
The accuracy evaluation index expression is as follows:
wherein, yiIs the measured value of the ith sample;
is the predicted value of the ith sample;
is the measured average of the samples; n is the number of samples.
S16: and (4) carrying out space-time dynamic analysis processing on the biomass of the alpine grasses based on the optimized and selected optimization model to obtain the biomass analysis result of the alpine grasses.
In the specific implementation process of the invention, the optimization model based on optimization selection carries out space-time dynamic analysis processing on the alpine grassy biomass to obtain the alpine grassy biomass analysis result, and the method comprises the following steps: performing Sen and Mann-Kendall trend analysis on the biomass of the alpine grassland based on the optimized and selected optimization model to obtain a first analysis result; performing R/S analysis on the biomass of the alpine grassland based on the optimized and selected optimization model to obtain a second analysis result; and performing superposition judgment based on the first analysis result and the second analysis result to obtain a biomass analysis result of the alpine grassland.
Specifically, a Sen and Mann-Kendall trend analysis method is adopted, and two key parameters (a slope beta and a statistic | Z)s|) analyzed the dynamic trend of AG-AGB in 2018 in 2005. β was obtained by Sen trend analysis and is expressed as:
in the formula, AGBiAnd AGBjAGB values for year i and year j, respectively. β is the slope; β >0 indicates that AGB increases from 2005 to 2018 and vice versa. Parameter | Z is calculated using the Mann-Kendall test equations|；
Where n is the time series length (n-14), i is 2005, 2006, …, 2018, and | Z is the length of the time series given a significance level αs|＞u1-α/2In the case of time series data, it indicates a significant change in the studied time series data at α level, and α is generally 0.05, when | ZsWhen the value is greater than 1.96, the confidence level alpha of the time sequence is less than 0.05, and the value is ZsI.ltoreq.1.96 means a confidence level alpha > 0.05.
The Hurst index is a useful statistical method for estimating the autocorrelation characteristics of a time series, and predicting the sustainability of future vegetation trends through R/S analysis calculation. The expression for the Hurst index is:
Ru＝max1≤t≤uX(t，u)-min1≤t≤uX(t，u)，u＝1，2，…，n；
in the formula, RuIs a range sequence; suIs a standard deviation sequence. When R/S ^ u-HHowever, the AGB time series shows the Hurst phenomenon, and the Hurst index is represented by an H value. The H value can be obtained using a least squares method based on the equation:
wherein H is the slope of the line; a is the intercept of the straight line. When H > 0.5, the change in AGB is sustainable; when H ≦ 0.5, the change in AGB is not sustainable.
To clearly judge whether the development trend of AG-AGB is good or bad, Sen and Mann-Kendall trend analysis (beta and Z) was useds) And R/S analysis (H) for judgment of superposition. The criteria are shown in the following table.
AG-AGB dynamic change trend judgment standard table
In the embodiment of the invention, an XGboost algorithm and a correlation analysis method are mainly used, a grassland biomass monitoring index system is constructed by researching the relation between indexes and biomass, then a plurality of linear/nonlinear inversion models are constructed according to the constructed index system, an optimal satellite driving model can be determined through model precision comparison, a space-time distribution diagram of AG-AGB is drawn, and the space-time dynamic change of the AG-AGB is analyzed.
Example two
Referring to fig. 2, fig. 2 is a schematic structural composition diagram of a alpine grassland biomass estimation system based on a satellite-driven model according to an embodiment of the present invention.
As shown in fig. 2, a alpine grassland biomass estimation system based on a satellite-driven model, the system includes:
the obtaining module 21: the system comprises a data acquisition module, a data processing module and a data processing module, wherein the data acquisition module is used for acquiring data information, and the data information comprises remote sensing image information, alpine grassland monitoring data information and environmental data information;
in the specific implementation process of the method, the area needing to be subjected to alpine grassland biomass estimation is determined firstly, and then data information of the area is obtained, wherein the data information comprises remote sensing image information, alpine grassland monitoring data information and environment data information.
Specifically, the areas determined to require alpine grassland biomass estimation in the present application are located in the Qinghai province in the northwest of China (31 ° 36 '-39 ° 19' for north latitude and 89 ° 35 '-103 ° 04' for east longitude), and the total area is about 69.67 × 104km 2. The average altitude of the research area is more than 3000 m, belonging to continental plateau climate. The annual average air temperature is-5.7 ℃ to 8.5 ℃, the annual total precipitation is 50mm to 450mm, and the annual total solar radiation is 690.8kJ to 753.6kJ (Deng et al, 2017). The area of the alpine grassland is about 41.19 multiplied by 104km2, and accounts for 59.13 percent of the total area of the Qinghai province. Major steppe types include alpine meadows, alpine steppes, and temperate steppes (Qinghai et al, 2012). The growing season of the grassland starts from the beginning of 5 months and ends in the last 9 months (Qinghai et al, 2012).
The preprocessing module 22: the data information is preprocessed to obtain a preprocessing result;
in a specific implementation process of the present invention, the preprocessing the data information to obtain a preprocessing result includes: initializing the remote sensing image information to obtain an initialization result of the remote sensing image information; initializing the alpine grassland monitoring data information to obtain an alpine grassland monitoring data information initialization result; and initializing the environmental data information to obtain an environmental data information initialization result.
Further, the initializing the remote sensing image information to obtain an initialized result of the remote sensing image information includes: converting MOD09A1 and MCD12Q1 data in the remote sensing image information from sinusoidal projection to Albers projection by using an MODIS reprojection tool, converting an HDF format into a Geo-Tiff format, and then performing image fusion and MCD12Q1 data reclassification to obtain an initialization result of the remote sensing image information; initializing the alpine grassland monitoring data information to obtain an alpine grassland monitoring data information initialization result, wherein the initializing process comprises the following steps: vectorization, projection conversion and abnormal value elimination processing are sequentially carried out on the alpine grassland monitoring data information to obtain an alpine grassland monitoring data information initialization result; and initializing the environmental data information to obtain an environmental data information initialization result. The method comprises the following steps: and sequentially carrying out spatial interpolation, projection conversion and grid resampling on the environmental data information to obtain an environmental data information initialization result.
Specifically, the remote sensing image data preprocessing mainly comprises radiation correction, atmospheric correction and image fusion. MODIS re-projection tool (MRT) is used to convert MOD09A1 and MCD12Q1 data from sinusoidal projection to Albers projection, and HDF format to Geo-Tiff format, and image fusion and MCD12Q1 data re-classification are performed. The method mainly comprises vectorization, projection conversion and abnormal value elimination. The abnormal value is removed by a method of average value +/-2 multiplied by standard deviation. The preprocessing of environment data such as geography, soil, weather and the like mainly comprises spatial interpolation, projection conversion and grid resampling. The meteorological data is interpolated mainly by using Anuspli software (written by the teaching of Hutchinson of national university of Australia). In addition, for the convenience of calculating AG-AGB, projections of all the data described above are converted into Albers projections, the format thereof is adjusted to Geo Tiff, and environment data of different spatial resolutions are resampled by the Nearest Neighbor Method (NNM), and raster data of a spatial resolution of 500m × 500m is generated.
The index extraction module 23: the system is used for extracting and processing indexes based on the preprocessing result to obtain remote sensing indexes and environmental indexes;
in the specific implementation process of the present invention, the performing the index extraction processing based on the preprocessing result to obtain the remote sensing index and the environmental index includes: and extracting and processing the remote sensing index and the environmental index in sequence based on the preprocessing result to obtain the remote sensing index and the environmental index.
Further, the remote sensing index extraction processing comprises the steps of obtaining 7 VIS and 5 spectral indexes from MOD09A1 data in the preprocessing result based on ENVI software, and extracting the maximum value of each visible light and spectral index; and the environmental index extraction processing comprises the step of acquiring terrain indexes of longitude, latitude, altitude, gradient and slope direction from a space shuttle radar terrain task digital elevation model product in the preprocessing result based on ArcGIS software.
Specifically, the remote sensing index extraction mainly uses ENVI software to obtain 7 VIS (NDVI, EVI, SAVI, MSAVVI, OSAVI, SATVI and RVI) and 5 spectral indexes (B7/B2, B2-B7, B7/B5, B5/B7 and (B5-B7)/(B5+ B7)) from MOD09A1 data, and extracts the maximum value of each visible light and spectral index (only 7 months and 8 months) in the research area.
The environmental indexes mainly comprise data of terrain, soil, meteorological phenomena and the like. And (3) acquiring terrain indexes such as longitude, latitude, altitude, gradient, slope direction and the like from a space shuttle radar terrain task (SRTM) Digital Elevation Model (DEM) product by using ArcGIS software. Soil indexes such as sand content of 0-30 cm, clay content of 0-30 cm, sand content of 30-60 cm, clay content of 30-60 cm, sand-clay ratio of 0-30 cm, sand-clay ratio of 30-60 cm, organic matter content, total nitrogen content, total phosphorus content, total potassium content and pH are obtained from soil data with spatial resolution of 1 km. Acquiring meteorological indexes such as annual average temperature, annual total precipitation, annual solar total radiant quantity, > 0-year accumulated temperature, humidity and the like from meteorological data.
Index architecture building Module 24: the system is used for constructing an index system by utilizing the remote sensing index, the environmental index and the alpine grassland biomass based on the XGBoost algorithm and the correlation analysis;
in the specific implementation process of the invention, the XGBoost algorithm and correlation analysis based index system is constructed by utilizing the remote sensing index, the environmental index and the alpine grassland biomass, and comprises the following steps: performing index analysis processing of alpine grassland biomass estimation on the remote sensing index and the environmental index based on the correlation analysis to obtain an initial estimation index of alpine grassland biomass estimation; based on an XGboost algorithm, carrying out redundancy removal on the initial estimation index of the alpine grassland biomass estimation to obtain an estimation index of the alpine grassland biomass estimation; and constructing an index system based on the estimation index, the environmental index and the alpine grassland biomass.
Specifically, an AG-AGB estimation index is selected by adopting a correlation analysis method; and then, on the basis of the correlation analysis result, introducing an extreme gradient boost (XGboost) algorithm to determine an AG-AGB estimation index. The XGBoost algorithm, as a method of determining all relevant features or metrics, provides an intrinsic measure of Feature Importance (FI). The XGBoost algorithm may rank the entire FI by averaging the FI in each tree, which is calculated based on the amount of information obtained after segmenting the tree using the features. In order to accurately obtain key indexes and reduce data redundancy among the indexes, the research adopts an XGboost algorithm in combination with a correlation analysis method, and selects significant and irrelevant indexes to estimate AG-AGB. The XGboost algorithm is briefly described as follows:
and establishing a classification model based on an AGB estimation index system. The performance of the indicators in the classification model is checked to obtain FIs and sorted in descending order. The formula of FI is:
where H (T) and H (T | F) are the entropy of parent and child nodes of a partition based on F features, piIs the score of each labeled sample on a node. And selecting a plurality of indexes with higher FI values to generate index subsets, and classifying again. This process is repeated until all the metrics are selected, the classification of all subsets is examined, and the best subset of metrics (i.e., the subset with the relatively higher area under the curve value and fewer metrics) is selected. And finally, constructing an index system according to the estimation index, the environmental index and the biomass of the alpine grassland.
Optimization selection module 25: the optimization model is used for carrying out optimization selection processing on the satellite driving model based on an index system to obtain an optimization model for optimization selection;
in the specific implementation process of the invention, the satellite driving model comprises a multiple linear regression model, a BP neural network model, a support vector machine model and a random forest model.
Further, the optimizing and selecting the satellite driving model based on the index system to obtain an optimized and selected optimizing model includes: determining precision parameters, wherein the precision parameters comprise a decision coefficient, a relative root mean square error and a residual prediction deviation; and referencing residual prediction deviations of all models in the satellite driving model, comparing the decision coefficient of all models when the residual prediction deviations of all models are in the same interval, and then referencing the relative root mean square error to obtain an optimized and selected optimization model.
Specifically, the satellite driving model comprises a multiple linear regression model, a BP neural network model, a support vector machine model and a random forest model.
The multivariate linear regression model takes AG-AGB as a dependent variable (Y) and takes a corresponding AG-AGB estimation index as an independent variable (X);
in the formula, betaiIs the ith weight coefficient determined from expert experience.
The BP neural network model is a multi-layer forward neural network based on an error back propagation algorithm, and generally consists of an input layer, an output layer and one or more hidden layers. It can use the guiding learning method to learn and train, and can simulate the relation between any nonlinear input variable and output variable. The learning process mainly includes forward propagation of the input signal and backward propagation of the error. The training process includes continuously adjusting the connection weights using a steepest descent method and a back propagation algorithm until the output mean square error meets a desired criterion. The equations for signal propagation and Mean Square Error (MSE) are respectively expressed as follows:
oj＝fi(∑wjioi+θj)；
ok＝fj(∑wkjoj+θk)；
in the formula, thetajIs the hidden layer information; oiIs input layer information (AG-AGB estimation index); w is ajiIs the weight of the input layer to the hidden layer; thetajIs the threshold of the hidden layer; f. ofiIs the rainlm function of the input layer to the hidden layer; okIs output layer information (AG-AGB estimate); w is akjIs the weight of the hidden layer to the output layer; thetakIs the threshold of the output layer; f. ofjIs the purelin function of the hidden layer to the output layer; o is AG-AGB field measurement data; n is the number of samples.
Support vector machine models are proposed by cortex and Vapnik (1995) for solving practical problems (small samples, non-linearity, high dimensions and local minimum points) by defining a kernel function, applying a non-linear transformation to a training data set and mapping it to a high-dimensional feature space. The expression of the SVM model is as follows:
wherein f (x) is an AG-AGB estimate; w is aiIs a weight vector; b is the deviation.
Representing a non-linear transfer function. Cortex and Vapnik (1995) introduced the following convex optimization problem with an e-insensitive penalty function to obtain a solution to the following equation:
wherein | | w | | non-conducting phosphor2Represents the flatness of the m-dimensional space; ε is a parameter representing the maximum allowable error between the observed value and the model prediction; xiiAnd
is a relaxation variable, penalizes training errors using a loss function on the error margin ξ; c is a positive compromise parameter that determines the magnitude of the empirical error in the optimization problem. The above equation is typically solved in the dual form of lagrange multipliers and applying optimality conditions.
Wherein
Is a transformation of w. The SVM function is represented as:
wherein
Is a kernel function. A Radial Basis Function (RBF) is selected as the kernel function. And determining a penalty parameter C and a kernel function parameter gamma of the support vector machine by adopting a grid search algorithm of cross validation.
The random forest model is an integrated learning method proposed by Breiman and Adele (2001), and has strong adaptability to a data set. It can handle high dimensional features and is not prone to overfitting. In addition, the model training speed is relatively fast. The expressions of the RF model and their errors are respectively as follows:
in the formula fKIs the predicted value of each decision tree; k is the number of decision trees;
is an AG-AGB estimate of out-of-bag data (OOB) that consists of the unextracted samples of the original samples. y isiIs and
the corresponding AG-AGB measured value; n is the OOB sample number.
In this study, three precision parameters were selected to verify the effect of the model: determining the coefficient (R)2) Relative Root Mean Square Error (RRMSE) and residual prediction bias (PRD). Wherein R is2Is used to measure the stability of the model. R2The larger the model, the more stable. RRMSE reflects the accuracy of the model. The smaller the RRMSE, the better the prediction capability. The performance of the model was evaluated using RPD. When RPD is less than 1.4, the model is not good; when the RPD is more than or equal to 1.4 and less than or equal to 2, the model has certain prediction capability; when RPD is more than 2, the model has stronger prediction capability. Judging the criterion of the model precision: firstly, referring to the RPD value, and preferentially comparing the model R when the RPD is in the same interval2The size is considered, the RRMSE size is considered, and the optimal model judgment principle is R2Relatively large, and relatively small RRMSE.
The accuracy evaluation index expression is as follows:
wherein, yiIs the measured value of the ith sample;
is the predicted value of the ith sample;
is the measured average of the samples; n is the number of samples.
The spatio-temporal dynamic analysis module 26: and the method is used for carrying out space-time dynamic analysis processing on the alpine grassland biomass based on the optimization model selected by optimization to obtain the alpine grassland biomass analysis result.
In the specific implementation process of the invention, the optimization model based on optimization selection carries out space-time dynamic analysis processing on the alpine grassy biomass to obtain the alpine grassy biomass analysis result, and the method comprises the following steps: performing Sen and Mann-Kendall trend analysis on the biomass of the alpine grassland based on the optimized and selected optimization model to obtain a first analysis result; performing R/S analysis on the biomass of the alpine grassland based on the optimized and selected optimization model to obtain a second analysis result; and performing superposition judgment based on the first analysis result and the second analysis result to obtain a biomass analysis result of the alpine grassland.
Specifically, a Sen and Mann-Kendall trend analysis method is adopted, and two key parameters (a slope beta and a statistic | Z)s|) analyzed the dynamic trend of AG-AGB in 2018 in 2005. β was obtained by Sen trend analysis and is expressed as:
in the formula, AGBiAnd AGBjAGB values in i and j years, respectively. β is the slope; β >0 indicates that AGB increases from 2005 to 2018 and vice versa. Parameter | Z is calculated using the Mann-Kendall test equations|；
Where n is the time series length (n-14), i is 2005, 2006, …, 2018, and | Z is the length of the time series given a significance level αs|＞u1-α/2In the case of time series data, it indicates a significant change in the studied time series data at α level, and α is generally 0.05, when | ZsWhen the value is greater than 1.96, the confidence level alpha of the time sequence is less than 0.05, and the value is ZsI.ltoreq.1.96 means a confidence level alpha > 0.05.
The Hurst index is a useful statistical method for estimating the autocorrelation characteristics of a time series, and predicting the sustainability of future vegetation trends through R/S analysis calculation. The expression for the Hurst index is:
Ru＝max1≤t≤uX(t，u)-min1≤t≤uX(t，u)，u＝1，2，…，n；
in the formula, RuIs a range sequence; suIs a standard deviation sequence. When R/S ^ u-HHowever, the AGB time series shows the Hurst phenomenon, and the Hurst index is represented by an H value. The H value can be obtained using a least squares method based on the equation:
wherein H is the slope of the line; a is the intercept of the straight line. When H > 0.5, the change in AGB is sustainable; when H ≦ 0.5, the change in AGB is not sustainable.
To clearly judge whether the development trend of AG-AGB is good or bad, Sen and Mann-Kendall trend analysis (beta and Z) was useds) And R/S analysis (H) for judgment of superposition. The criteria are shown in the following table.
AG-AGB dynamic change trend judgment standard table
In the embodiment of the invention, an XGboost algorithm and a correlation analysis method are mainly used, a grassland biomass monitoring index system is constructed by researching the relation between indexes and biomass, then a plurality of linear/nonlinear inversion models are constructed according to the constructed index system, an optimal satellite driving model can be determined through model precision comparison, a space-time distribution diagram of AG-AGB is drawn, and the space-time dynamic change of the AG-AGB is analyzed.
Those skilled in the art will appreciate that all or part of the steps in the methods of the above embodiments may be implemented by associated hardware instructed by a program, which may be stored in a computer-readable storage medium, and the storage medium may include: read Only Memory (ROM), Random Access Memory (RAM), magnetic or optical disks, and the like.
In addition, the satellite-driven-model-based alpine grassland biomass estimation method and system provided by the embodiment of the invention are described in detail, and a specific embodiment is adopted herein to explain the principle and the implementation of the invention, and the description of the embodiment is only used to help understanding the method and the core idea of the invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present invention.
Claims (10)
1. A alpine grassland biomass estimation method based on a satellite-driven model is characterized by comprising the following steps:
acquiring data information, wherein the data information comprises remote sensing image information, alpine grassland monitoring data information and environment data information;
preprocessing the data information to obtain a preprocessing result;
performing index extraction processing based on the preprocessing result to obtain a remote sensing index and an environmental index;
constructing an index system by utilizing the remote sensing index, the environmental index and the alpine grassland biomass based on the XGBoost algorithm and the correlation analysis;
performing optimization selection processing on the satellite driving model based on an index system to obtain an optimization model for optimization selection;
and (4) carrying out space-time dynamic analysis processing on the biomass of the alpine grasses based on the optimized and selected optimization model to obtain the biomass analysis result of the alpine grasses.
2. The method for estimating biomass in alpine grasses according to claim 1, wherein the preprocessing the data information to obtain a preprocessing result comprises:
initializing the remote sensing image information to obtain an initialization result of the remote sensing image information;
initializing the alpine grassland monitoring data information to obtain an alpine grassland monitoring data information initialization result;
and initializing the environmental data information to obtain an environmental data information initialization result.
3. The alpine grassland biomass estimation method according to claim 2, wherein the initializing the remote sensing image information to obtain an initialized result of the remote sensing image information comprises:
converting MOD09A1 and MCD12Q1 data in the remote sensing image information from sinusoidal projection to Albers projection by using an MODIS reprojection tool, converting an HDF format into a Geo-Tiff format, and then performing image fusion and MCD12Q1 data reclassification to obtain an initialization result of the remote sensing image information;
initializing the alpine grassland monitoring data information to obtain an alpine grassland monitoring data information initialization result, wherein the initializing process comprises the following steps:
vectorization, projection conversion and abnormal value elimination processing are sequentially carried out on the alpine grassland monitoring data information to obtain an alpine grassland monitoring data information initialization result;
and initializing the environmental data information to obtain an environmental data information initialization result. The method comprises the following steps:
and sequentially carrying out spatial interpolation, projection conversion and grid resampling on the environmental data information to obtain an environmental data information initialization result.
4. The alpine grassland biomass estimation method according to claim 1, wherein the performing index extraction processing based on the preprocessing result to obtain a remote sensing index and an environmental index includes:
extracting and processing the remote sensing index and the environmental index in sequence based on the preprocessing result to obtain the remote sensing index and the environmental index;
the environmental data information includes in-ground data, soil data, and meteorological data.
5. The alpine grassland biomass estimation method according to claim 4, wherein the remote sensing index extraction process includes obtaining 7 VIs and 5 spectral indexes from MOD09A1 data in the preprocessing result based on ENVI software, and extracting the maximum value of each visible light and spectral index;
the environmental index extraction processing comprises the steps of acquiring longitude, latitude, altitude, gradient and slope terrain indexes from a space shuttle radar terrain task digital elevation model product in the preprocessing result based on ArcGIS software; obtaining soil indexes of 0-30 cm sand content, 0-30 cm clay content, 30-60 cm sand content, 30-60 cm clay content, 0-30 cm sand-clay ratio, 30-60 cm sand-clay ratio, organic matter content, total nitrogen content, total phosphorus content, total potassium content and pH from soil data; acquiring meteorological indexes of annual average temperature, annual total precipitation, annual solar total radiant quantity, > 0-year accumulated temperature and humidity from meteorological data.
6. The alpine grassland biomass estimation method according to claim 1, wherein the XGBoost algorithm and correlation analysis based index system is constructed by using the remote sensing index, the environmental index and the alpine grassland biomass, and comprises the following steps:
performing index analysis processing of alpine grassland biomass estimation on the remote sensing index and the environmental index based on the correlation analysis to obtain an initial estimation index of alpine grassland biomass estimation;
based on an XGboost algorithm, carrying out redundancy removal on the initial estimation index of the alpine grassland biomass estimation to obtain an estimation index of the alpine grassland biomass estimation;
and constructing an index system based on the estimation index, the environmental index and the alpine grassland biomass.
7. The alpine grassland biomass estimation method of claim 1, wherein the satellite-driven models include a multivariate linear regression model, a BP neural network model, a support vector machine model, and a random forest model.
8. The alpine grassland biomass estimation method according to claim 7, wherein the optimizing and selecting the satellite-driven model based on the index system to obtain an optimized and selected optimized model comprises:
determining precision parameters, wherein the precision parameters comprise a decision coefficient, a relative root mean square error and a residual prediction deviation;
and referencing residual prediction deviations of all models in the satellite driving model, comparing the decision coefficient of all models when the residual prediction deviations of all models are in the same interval, and then referencing the relative root mean square error to obtain an optimized and selected optimization model.
9. The alpine grassland biomass estimation method according to claim 1, wherein the optimization model based on optimization selection performs spatio-temporal dynamic analysis processing on the alpine grassland biomass to obtain the alpine grassland biomass analysis result, and the method comprises:
performing Sen and Mann-Kendall trend analysis on the biomass of the alpine grassland based on the optimized and selected optimization model to obtain a first analysis result;
performing R/S analysis on the biomass of the alpine grassland based on the optimized and selected optimization model to obtain a second analysis result;
and performing superposition judgment based on the first analysis result and the second analysis result to obtain a biomass analysis result of the alpine grassland.
10. A alpine grassland biomass estimation system based on a satellite-driven model, the system comprising:
an obtaining module: the system comprises a data acquisition module, a data processing module and a data processing module, wherein the data acquisition module is used for acquiring data information, and the data information comprises remote sensing image information, alpine grassland monitoring data information and environmental data information;
a preprocessing module: the data information is preprocessed to obtain a preprocessing result;
an index extraction module: the system is used for extracting and processing indexes based on the preprocessing result to obtain remote sensing indexes and environmental indexes;
an index system construction module: the system is used for constructing an index system by utilizing the remote sensing index, the environmental index and the alpine grassland biomass based on the XGBoost algorithm and the correlation analysis;
an optimization selection module: the optimization model is used for carrying out optimization selection processing on the satellite driving model based on an index system to obtain an optimization model for optimization selection;
a space-time dynamic analysis module: and the method is used for carrying out space-time dynamic analysis processing on the alpine grassland biomass based on the optimization model selected by optimization to obtain the alpine grassland biomass analysis result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110412645.6A CN113297904B (en) | 2021-04-16 | 2021-04-16 | Method and system for estimating alpine grassland biomass based on satellite driving model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110412645.6A CN113297904B (en) | 2021-04-16 | 2021-04-16 | Method and system for estimating alpine grassland biomass based on satellite driving model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113297904A true CN113297904A (en) | 2021-08-24 |
| CN113297904B CN113297904B (en) | 2023-11-21 |
Family
ID=77318822
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110412645.6A Active CN113297904B (en) | 2021-04-16 | 2021-04-16 | Method and system for estimating alpine grassland biomass based on satellite driving model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113297904B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115598164A (en) * | 2022-12-14 | 2023-01-13 | 南京师范大学（Cn） | Machine learning integrated soil heavy metal concentration detection method and system |
| CN117387581A (en) * | 2023-12-12 | 2024-01-12 | 北京麦麦趣耕科技有限公司 | Cinnamomum camphora biomass estimation method, apparatus and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103678914A (en) * | 2013-12-16 | 2014-03-26 | 中国科学院遥感与数字地球研究所 | Alpine grassland soil respiration estimation method based on satellite remote sensing data |
| CN104778451A (en) * | 2015-03-31 | 2015-07-15 | 中国科学院上海技术物理研究所 | Grassland biomass remote sensing inversion method considering grassland height factor |
| US20200141877A1 (en) * | 2018-11-06 | 2020-05-07 | Nanjing Agricultural University | Method for estimating aboveground biomass of rice based on multi-spectral images of unmanned aerial vehicle |
| US20200225075A1 (en) * | 2019-01-14 | 2020-07-16 | Wuhan University | Method and system for optical and microwave synergistic retrieval of aboveground biomass |
-
2021
- 2021-04-16 CN CN202110412645.6A patent/CN113297904B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103678914A (en) * | 2013-12-16 | 2014-03-26 | 中国科学院遥感与数字地球研究所 | Alpine grassland soil respiration estimation method based on satellite remote sensing data |
| CN104778451A (en) * | 2015-03-31 | 2015-07-15 | 中国科学院上海技术物理研究所 | Grassland biomass remote sensing inversion method considering grassland height factor |
| US20200141877A1 (en) * | 2018-11-06 | 2020-05-07 | Nanjing Agricultural University | Method for estimating aboveground biomass of rice based on multi-spectral images of unmanned aerial vehicle |
| US20200225075A1 (en) * | 2019-01-14 | 2020-07-16 | Wuhan University | Method and system for optical and microwave synergistic retrieval of aboveground biomass |
Non-Patent Citations (2)
| Title |
|---|
| 孟宝平;陈思宇;崔霞;冯琦胜;梁天刚;: "基于多源遥感数据的高寒草地生物量反演模型精度――以夏河县桑科草原试验区为例", 草业科学, no. 11 * |
| 葛静;孟宝平;杨淑霞;高金龙;殷建鹏;张仁平;冯琦胜;梁天刚;: "基于ADC和MODIS遥感数据的高寒草地地上生物量监测研究――以黄河源区为例", 草业学报, no. 07 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115598164A (en) * | 2022-12-14 | 2023-01-13 | 南京师范大学（Cn） | Machine learning integrated soil heavy metal concentration detection method and system |
| CN117387581A (en) * | 2023-12-12 | 2024-01-12 | 北京麦麦趣耕科技有限公司 | Cinnamomum camphora biomass estimation method, apparatus and storage medium |
| CN117387581B (en) * | 2023-12-12 | 2024-02-13 | 北京麦麦趣耕科技有限公司 | Cinnamomum camphora biomass estimation method, apparatus and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113297904B (en) | 2023-11-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zheng et al. | Improved estimation of rice aboveground biomass combining textural and spectral analysis of UAV imagery | |
| Muchoney et al. | Application of the MODIS global supervised classification model to vegetation and land cover mapping of Central America | |
| Meng et al. | Modeling alpine grassland above ground biomass based on remote sensing data and machine learning algorithm: A case study in east of the Tibetan Plateau, China | |
| CN115356249B (en) | Satellite polarization PM2.5 estimation method and system based on machine learning fusion model | |
| CN113297904B (en) | Method and system for estimating alpine grassland biomass based on satellite driving model | |
| Duadze | Land use and land cover study of the savannah ecosystem in the Upper West Region (Ghana) using remote sensing | |
| Yin et al. | Estimation of grassland height based on the random forest algorithm and remote sensing in the Tibetan Plateau | |
| CN113340836A (en) | Atmospheric temperature and humidity profile inversion method for high-latitude complex underlying surface | |
| Zhou et al. | Remote sensing of regional-scale maize lodging using multitemporal GF-1 images | |
| Shi et al. | High spatial-temporal resolution estimation of ground-based global navigation satellite system interferometric reflectometry (GNSS-IR) soil moisture using the genetic algorithm back propagation (GA-BP) neural network | |
| Lou et al. | An effective method for canopy chlorophyll content estimation of marsh vegetation based on multiscale remote sensing data | |
| CN113534083B (en) | SAR-based corn stubble mode identification method, device and medium | |
| Zhang et al. | Inversion models of aboveground grassland biomass in Xinjiang based on multisource data | |
| Zhu et al. | UAV flight height impacts on wheat biomass estimation via machine and deep learning | |
| CN117114147A (en) | Method and device for estimating forest vegetation carbon reserves based on radar and satellite remote sensing | |
| CN117075138A (en) | Remote sensing measurement and calculation method, system and medium for canopy height of 30-meter forest in area | |
| CN114782835B (en) | Crop lodging area proportion detection method and device | |
| CN114357885B (en) | Photosynthetic effective radiation scattering proportion prediction method fusing multisource data | |
| CN115496999A (en) | Method and device for estimating field straw yield | |
| CN114611699A (en) | Soil moisture downscaling method and device, electronic equipment and storage medium | |
| CN114047563A (en) | All-weather assimilation method of infrared hyperspectrum | |
| Tang et al. | UAV-based high spatial and temporal resolution monitoring and mapping of surface moisture status in a vineyard | |
| You et al. | Crop Mapping of Complex Agricultural Landscapes Based on Discriminant Space | |
| CN112380497A (en) | Straw coefficient estimation method and system for regional corn | |
| Yang et al. | Optimization of Hourly PM 2.5 Inversion Model Integrating Upper-Air Meteorological Elements |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |