Basin scale GPP change driving factor analysis method and system

Abstract

The invention discloses a watershed scale GPP change driving factor analysis method and system, and belongs to the technical field of ecological system carbon sink monitoring, and the method comprises the steps: firstly obtaining a multi-source data set, and completing data preprocessing through space-time matching and vegetation type classification; extracting regional features such as static features and dynamic features, and carrying out space-time alignment; a random forest regression model is constructed based on the classification and regression tree, and training and verification are carried out; and finally, calculating the relative contribution degree of each driving factor to GPP change based on a random forest regression model, identifying a key driving factor, and carrying out multi-scale driving mechanism analysis. By adopting the drainage basin scale GPP change driving factor analysis method and system, the problems of insufficient nonlinear analysis, low data fusion efficiency, extensive regional analysis and the like in the prior art are solved, the GPP driving factor analysis precision is remarkably improved, and scientific basis and technical support are provided for drainage basin carbon sink management.

Description

Technical Field

[0001] The present invention relates to the technical field of ecosystem carbon sink monitoring, and in particular to a method and system for analyzing driving factors of GPP changes at a watershed scale. Background Art

[0002] Gross primary productivity (GPP) is a core indicator for measuring the carbon sequestration capacity of terrestrial ecosystems. Accurately analyzing its driving factors is crucial for maintaining regional ecological security. Against the backdrop of global climate change and intensified human activities, basin-scale GPP dynamics are influenced by a combination of climatic factors (such as temperature, precipitation, and radiation); surface characteristics (such as elevation, slope, and vegetation type); and human activities (such as land use and population density), exhibiting highly complex nonlinear relationships. Traditional research methods rely on single data sources or simple models, making them difficult to address the demands of multi-dimensional data integration and the analysis of complex mechanisms.

[0003] Currently, traditional methods include mechanism model-driven method, statistical model-driven method and single remote sensing data-driven method: Among them, the mechanism model-driven method is to build models based on the physiological process of photosynthesis, such as the light use efficiency model (VPM) and the biogeochemical model (BESS), and estimate GPP by inputting meteorological data and vegetation parameters. This type of method relies on clear physical mechanisms, but has high requirements on data accuracy and is prone to errors caused by mechanism simplification in complex terrain and vegetation heterogeneity areas; the statistical model-driven method uses traditional statistical methods such as linear regression and stepwise regression to analyze the correlation between GPP and a single or a few driving factors. This method is simple and intuitive, but cannot handle the interactions and nonlinear relationships of multidimensional variables, and has limited analytical capabilities; the single remote sensing data-driven method is based on a single remote sensing data source such as vegetation index (such as NDVI) or sunlight-induced chlorophyll fluorescence (SIF) to establish a GPP estimation model. Although it can capture vegetation dynamics, it ignores the impact of surface characteristics and human activities, and is highly dependent on the spatiotemporal resolution and quality of remote sensing data.

[0004] At the same time, traditional mechanism models and simple statistical models lack the ability to analyze nonlinear relationships, making it difficult to characterize the complex nonlinear relationship between GPP and driving factors, resulting in inaccurate identification of key factors; existing solutions lack the depth of multi-source data integration, and fail to fully integrate multi-source data such as climate, surface, and human activities. The data temporal and spatial matching accuracy is low, and it is unable to fully reflect the comprehensive driving mechanism at the watershed scale; the existing models handle vegetation heterogeneity roughly, and do not classify and model the physiological characteristics and canopy structure differences of different vegetation types. The universality of model parameters is poor, and the estimation accuracy is limited; the regional scale analysis of existing models lacks refinement, and the analysis of differentiated driving mechanisms at multiple scales such as ecological zones, vegetation types, and administrative units is insufficient, making it difficult to provide targeted policy recommendations; at the same time, the existing model verification system is imperfect, lacks multi-source data comparative verification and cross-validation, and the model generalization ability and reliability are difficult to guarantee. Summary of the Invention

[0005] The purpose of this invention is to propose a method and system for analyzing the driving factors of GPP changes at the watershed scale. By integrating multi-source data, constructing a vegetation type differentiation model, and introducing machine learning algorithms such as random forests, the present invention solves the problems of insufficient nonlinear relationship analysis, low data fusion efficiency, and extensive regional scale analysis in the existing technology, and realizes the accurate identification and quantification of the key driving factors of GPP changes, providing a scientific basis for watershed ecological protection and carbon sequestration capacity improvement.

[0006] To achieve the above objectives, the present invention proposes a method for analyzing the driving factors of GPP changes at the basin scale, which includes the following steps: Step S1: Acquire a multi-source dataset, including sunlight-induced chlorophyll fluorescence (SIF) remote sensing data, climate data, surface data, and human activity data, and complete data preprocessing through spatiotemporal matching and vegetation type classification; Step S2: extracting regional features based on the preprocessed multi-source data set, including static features and dynamic features, and performing spatiotemporal alignment of the data through spatiotemporal matching; Step S3: construct a random forest regression model based on the classification and regression tree, and generate multiple decision trees by randomly sampling the training data set and randomly selecting variable subsets; Step S4: Use stratified random sampling to divide the data set into training set and validation set and perform model training. Output the GPP change rate prediction results through a voting mechanism, and use 10-fold cross validation to evaluate the model performance. Step S5: Calculate the relative contribution of each driving factor to GPP changes based on the random forest regression model, identify key driving factors, and conduct multi-scale driving mechanism analysis.

[0007] Preferably, in step S1, the climate data includes climate driven datasets CMFD and GLADS data, the surface data includes digital elevation data and soil data, and the human activity data includes land cover or utilization data and population density data.

[0008] Preferably, in step S1, data preprocessing includes: The solar-induced chlorophyll fluorescence (SIF) remote sensing data, climate data, land surface data, and human activity data were unified into 0.05°×0.05° spatial resolution and 8-day temporal resolution through bilinear interpolation. The study area was divided into forest, grassland, farmland, and shrub vegetation types based on MODIS land cover data.

[0009] Preferably, in step S2, the static features are static attributes of the surface, including elevation, slope, aspect, soil type, and vegetation type; the dynamic features include the average annual temperature, annual precipitation, the interannual change rate of climate factors related to net radiation, and the change rate of land use type and population density.

[0010] Preferably, the spatiotemporal matching is performed by aligning various feature data in spatiotemporal order according to a grid unit of 0.05°×0.05° through a geographic information system (GIS).

[0011] Preferably, in step S4, the 10-fold cross validation evaluation model is used to determine the coefficient R 2 , root mean square error RMSE and mean absolute error MAE are evaluation indicators.

[0012] Preferably, in step S5, the multi-scale driving mechanism analysis performs multi-scale grouping modeling according to ecological zones, vegetation types, and administrative units.

[0013] The present invention also provides a basin-scale GPP change driving factor analysis system, comprising: Multi-source data preprocessing module: Integrates climate data, surface data, human activity data, and remote sensing data, and completes data preprocessing through spatiotemporal matching and vegetation type classification; Feature extraction module: extracts static surface features and dynamic features of climate and human activities, and constructs feature datasets; Model building module: Build a random forest regression model based on classification and regression trees, and predict the GPP change rate through training and verification; Driving factor analysis module: Calculates the importance score of driving factors based on the random forest regression model and conducts multi-scale heterogeneity analysis.

[0014] Therefore, the present invention proposes a method and system for analyzing the driving factors of GPP changes at the basin scale, which has the following beneficial effects: (1) By matching the spatiotemporal data of multiple sources and unifying the data resolution, the system effectively integrates SIF data, climate data, surface data, human activity data and other heterogeneous data, giving full play to the advantages of different types of data, improving the efficiency of data fusion, and overcoming the shortcomings of existing technologies in data spatiotemporal matching and heterogeneity processing; (2) Multi-scale group modeling based on ecological zones, vegetation types, administrative units, etc. can deeply reveal the heterogeneity of the driving mechanisms of GPP changes in different regions and vegetation types, changing the existing extensive regional scale analysis and providing precise guidance for the formulation of targeted ecological protection and carbon sequestration enhancement strategies in different regions; (3) Based on MODIS land cover data, vegetation types were divided in detail, and the specificity of different vegetation types was fully considered in the model construction, so that the model can better reflect the photosynthesis mechanism and canopy structure differences of different vegetation, reduce the uncertainty of model parameters, and enhance the adaptability of the model to complex ecosystems; (4) The training set and validation set were divided by stratified random sampling, and 10-fold cross-validation was used to evaluate the model performance. At the same time, multi-source data was used for comparative verification, which effectively improved the reliability and generalization ability of the model, ensured the credibility of the driving factor analysis results in different watersheds and application scenarios, and expanded the application scope of the technology; (5) The use of a vegetation type-differentiated random forest model can accurately capture the complex nonlinear relationship between GPP and multidimensional variables such as climate factors, surface characteristics, and human activities. Compared with traditional statistical models or simple mechanism models, it significantly improves the accuracy of analyzing the driving factors of GPP changes and provides a reliable basis for accurately assessing the carbon sequestration capacity of ecosystems.

[0015] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is a flow chart of the steps of a method for analyzing driving factors of GPP changes at a watershed scale according to the present invention; Figure 2 Schematic diagram of the importance scores of various variables estimated using a basin-scale GPP change driving factor analysis method of the present invention; wherein, Figure 2 (a) in the table is the overall importance score of each indicator. Figure 2 (b) is the importance score of each indicator in the GPP growth area. Figure 2 (c) in the figure is the importance score of each indicator in the GPP reduction area. DETAILED DESCRIPTION

[0017] The technical solution of the present invention is further described below with reference to the accompanying drawings and embodiments.

[0018] Unless otherwise defined, technical or scientific terms used in the present invention shall have the same meaning as commonly understood by one of ordinary skill in the art to which the present invention belongs.

[0019] Example 1 like Figure 1 As shown, the present invention provides a method for analyzing the driving factors of GPP changes at the basin scale, the steps are as follows: Step S1, acquisition and preprocessing of multi-source data; Acquire sunlight-induced chlorophyll fluorescence (SIF) remote sensing data with a spatial resolution of 0.05° and a temporal resolution of 8 days, climate data including temperature, precipitation, and net radiation, surface data including elevation, slope, and vegetation type, and human activity data including land use and population density. The solar-induced chlorophyll fluorescence (SIF) remote sensing data, climate data, land surface data, and human activity data were unified into 0.05°×0.05° spatial resolution and 8-day temporal resolution through bilinear interpolation. Based on the MODIS land cover data MCD12C1, the study area was divided into forest, grassland, farmland, and shrubland according to vegetation type, providing a basis for subsequent differentiated modeling.

[0020] Step S2: extract regional features, including static features and dynamic features, and perform spatiotemporal alignment of data through spatiotemporal matching; The extracted static features are static surface attributes, including elevation, slope, aspect, soil type, and vegetation type; the extracted dynamic features are the calculated interannual change rate of climate factors including annual average temperature, annual precipitation, and net radiation, as well as the change rate of land use type and population density; The various feature data were spatiotemporally aligned according to the grid unit of 0.05°×0.05° through the Geographic Information System (GIS) to ensure the consistency of the input data.

[0021] Step S3: Using an ensemble learning method, a random forest regression model is constructed based on the classification and regression tree CART, and multiple decision trees are generated by randomly sampling the training data set and randomly selecting variable subsets.

[0022] Step S4: Stratified random sampling is used to divide the training set into 70% and the validation set into 30% for model training. The GPP change rate prediction results are output through the voting mechanism. The model performance is evaluated by 10-fold cross validation to determine the coefficient R. 2 , root mean square error RMSE, and mean absolute error MAE are used as evaluation indicators.

[0023] Step S5: Calculate the relative contribution of each driving factor to the GPP change based on the feature importance algorithm of the random forest model (e.g. Figure 2As shown in the figure, key driving factors are identified, including population density, annual precipitation, and altitude; ecological regions are divided into the eastern monsoon region and the western arid region; vegetation types are divided into forest, grassland, farmland, and shrubland; different administrative units are divided, multi-scale driving mechanism analysis is conducted, and policy recommendations are proposed based on management measures analysis.

[0024] Example 2 The present invention also provides a basin-scale GPP change driving factor analysis system, comprising: Multi-source data preprocessing module: Integrates climate data, surface data, human activity data, and remote sensing data, and completes data preprocessing through spatiotemporal matching and vegetation type classification; Feature extraction module: extracts static surface features and dynamic features of climate and human activities, and constructs feature datasets; Model building module: Build a random forest regression model based on classification and regression trees, and predict the GPP change rate through training and verification; Driving factor analysis module: Calculates the importance score of driving factors based on the random forest regression model and conducts multi-scale heterogeneity analysis.

[0025] It is worth noting that the contents not elaborated in detail in the present invention are all prior art and are well known to those skilled in the art.

[0026] Therefore, the present invention provides a method and system for analyzing the drivers of GPP change at the watershed scale. By integrating multi-source heterogeneous data and applying machine learning algorithms such as random forests, this method addresses existing issues such as insufficient analysis of nonlinear relationships, inefficient data fusion, and extensive regional-scale analysis. This method accurately identifies and quantifies the key drivers of GPP change, enabling accurate analysis of GPP changes. This provides a scientific basis and strong technical support for watershed ecological protection, carbon sequestration enhancement, and related management decisions, helping to maintain regional ecological security goals.

[0027] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit the same. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solutions of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for analyzing the driving factors of GPP changes at the basin scale, characterized by: The following steps are involved: Step S1: Acquire a multi-source dataset, including sunlight-induced chlorophyll fluorescence (SIF) remote sensing data, climate data, surface data, and human activity data, and complete data preprocessing through spatiotemporal matching and vegetation type classification; Step S2: extracting regional features based on the preprocessed multi-source data set, including static features and dynamic features, and performing spatiotemporal alignment of the data through spatiotemporal matching; Step S3: construct a random forest regression model based on the classification and regression tree, and generate multiple decision trees by randomly sampling the training data set and randomly selecting variable subsets; Step S4: Use stratified random sampling to divide the data set into training set and validation set and perform model training. Output the GPP change rate prediction results through a voting mechanism, and use 10-fold cross validation to evaluate the model performance. Step S5: Calculate the relative contribution of each driving factor to GPP changes based on the random forest regression model, identify key driving factors, and conduct multi-scale driving mechanism analysis.

2. The method for analyzing driving factors of GPP changes at the basin scale according to claim 1 is characterized in that: In step S1, the climate data includes climate driven datasets (CMFD) and GLADS data, the surface data includes digital elevation data and soil data, and the human activity data includes land cover or utilization data and population density data.

3. The method for analyzing driving factors of GPP changes at the basin scale according to claim 1 is characterized in that: In step S1, data preprocessing includes: The solar-induced chlorophyll fluorescence (SIF) remote sensing data, climate data, land surface data, and human activity data were unified into 0.05°×0.05° spatial resolution and 8-day temporal resolution through bilinear interpolation. The study area was divided into forest, grassland, farmland, and shrub vegetation types based on MODIS land cover data.

4. The method for analyzing driving factors of GPP changes at the basin scale according to claim 1 is characterized in that: In step S2, the static features are static attributes of the surface, including elevation, slope, aspect, soil type, and vegetation type; the dynamic features include the interannual change rate of climate factors related to average annual temperature, annual precipitation, net radiation, and the change rate of land use type and population density.

5. The method for analyzing driving factors of GPP changes at the basin scale according to claim 4 is characterized in that: The spatial and temporal matching was carried out by aligning various feature data in a grid unit of 0.05°×0.05° through the Geographic Information System (GIS).

6. The method for analyzing driving factors of GPP changes at the basin scale according to claim 1 is characterized in that: In step S4, the model is evaluated by 10-fold cross validation to determine the coefficient R 2 , root mean square error RMSE and mean absolute error MAE are evaluation indicators.

7. The method for analyzing driving factors of GPP changes at the basin scale according to claim 1 is characterized in that: In step S5, the multi-scale driving mechanism analysis is performed by multi-scale grouping modeling according to ecological zones, vegetation types, and administrative units.

8. A basin-scale GPP change driving factor analysis system, characterized by: include: Multi-source data preprocessing module: Integrates climate data, surface data, human activity data, and remote sensing data, and completes data preprocessing through spatiotemporal matching and vegetation type classification; Feature extraction module: extracts static surface features and dynamic features of climate and human activities, and constructs feature datasets; Model building module: Build a random forest regression model based on classification and regression trees, and predict the GPP change rate through training and verification; Driving factor analysis module: Calculates the importance score of driving factors based on the random forest regression model and performs multi-scale heterogeneity analysis.

Images