Quantification method of meteorological and human contributions to NPP change based on pixel-level parameterization model

Abstract

This invention discloses a method for quantifying meteorological and anthropogenic contributions to NPP changes based on a pixel-level parameterized model, comprising the following steps: Step S1, data collection and preprocessing; Step S2, preparation of core variable data; Step S3, progressive diagnosis of meteorological driving factors; Step S4, construction of a pixel-level quantitative attribution model for NPP changes; Step S5, integration of product verification and operational services. This invention uses forest age structure dynamics as a representative indicator of the long-term effects and time lags of anthropogenic influences such as ecological engineering, and incorporates it into the indirect impacts of anthropogenic activities. It constructs a two-dimensional attribution framework of climate change and anthropogenic activities with ΔNPP as the object, achieving quantitative analysis of anthropogenic contributions by accurately stripping away the contribution of climate change. After removing the impact of climate change, the interannual change in net primary productivity (ΔNPP) can be systematically analyzed and quantified to reflect the contribution of anthropogenic factors, providing technical support for ecosystem protection and restoration, and eco-meteorological services serving dual-carbon goals.

Description

Technical Field

[0001] This invention relates to a method for quantifying meteorological and anthropogenic contributions to NPP changes, and more particularly to a method for quantifying meteorological and anthropogenic contributions to NPP changes based on a pixel-level parameterized model, belonging to the field of ecological environment remote sensing monitoring and carbon cycle assessment technology. Background Technology

[0002] Currently, net primary productivity (NPP) assessments still have significant shortcomings in quantitatively separating the independent contributions of climate change and anthropogenic activities. Existing technologies mainly focus on the overall monitoring and estimation of NPP, and their core limitation lies in the lack of precise analytical capabilities for the driving factors of change, failing to effectively distinguish and independently quantify the respective contributions of climate change and anthropogenic activities to the interannual change in net primary productivity (ΔNPP). This deficiency directly leads to NPP products being unable to support refined effectiveness assessments of ecological policies, nor can they meet the application needs of quantitatively tracing the sources of influencing factors in real-world scenarios. This is mainly reflected in the following aspects: (1) Existing technologies have limitations: On the one hand, the parameters of statistical model methods are mostly set with regional uniformity, which makes it difficult to achieve accurate attribution at the pixel scale and lacks ecological mechanism; on the other hand, meteorological contribution separation methods based on process models have problems of high uncertainty and poor robustness, making it difficult to directly apply to existing global standardized products or aggregate products. In addition, such methods generally lack the ability to extend across indicators (such as vegetation cover FVC, vegetation index DNVI, etc.), resulting in poor comparability of results when different indicators are evaluated simultaneously.

[0003] (2) Bias in the analysis object: Existing studies mostly focus on separating the meteorological and anthropogenic contributions to the total NPP, while neglecting the attribution analysis of its changes, which makes it impossible to accurately capture the impact of dynamic processes such as climate change and ecological protection and restoration.

[0004] (3) Insufficient quantification depth: Most results are still at the level of relative estimation of contribution rate, and have failed to further achieve quantitative separation of climate change and human activities, which limits the practical application value of the results in decision support.

[0005] (4) Low technological maturity: The relevant methods are still mainly based on scientific research and exploration, lacking systematic ground verification and business application, and have not yet formed a stable and reliable technical process, resulting in a serious lack of actual support capacity in ecological policy formulation and refined governance.

[0006] Therefore, the core technical problem this invention aims to solve is that existing NPP assessment techniques cannot accurately and quantitatively separate the independent contributions of climate change factors and human activities to NPP changes at the pixel scale, specifically in the following ways: 1) The accuracy and spatial texture of meteorological grid data in sparse meteorological stations and complex terrain areas are insufficient, which affects the product quality and application service effect of subsequent quantitative attribution models. 2) The lack of efficient driving factor diagnostic technology makes it difficult to accurately determine the effective meteorological driving factors for each pixel; 3) Traditional models struggle to balance ecological mechanisms and applicability, failing to achieve accurate mapping of "one pixel, one model"; 4) The lack of reliable verification methods and mature business application solutions makes it difficult to transform technological achievements into practical applications. Summary of the Invention

[0007] To address the shortcomings of the aforementioned technologies, this invention provides a method for quantifying meteorological and anthropogenic contributions to NPP changes based on a pixel-level parametric model. This method solves the challenge of quantitatively separating the contributions of climate change and human activities from NPP products. It enables quantitative attribution of climate change factors and anthropogenic activities in high spatial resolution ΔNPP products and provides a complete operational solution integrating technology development, product verification, and application feedback.

[0008] To address the above technical problems, the technical solution adopted in this invention is: a method for quantifying meteorological and anthropogenic contributions to NPP changes based on a pixel-level parameterized model, comprising the following steps: Step S1, Data Collection and Preprocessing: Collect multi-source environmental data within the target area and standardize it; Step S2: Prepare core variable data: Calculate the standardized data year by year to generate a gridded dataset; Step S3, Progressive meteorological driving factor diagnosis: Select at least 25 years of core variable data based on step S2, and use a three-level progressive factor detection and diagnosis framework of full-region scale, sub-region scale, and pixel scale for diagnosis; Step S4: Construct a pixel-level quantitative attribution model for NPP changes: Using pixels as independent analysis units, based on the diagnostic results of step S3, construct a pixel-level quantitative attribution model based on the ecological limiting factor law to quantitatively decompose the contributions of climate change and human activities to the interannual change in net primary productivity of each pixel. Step S5, Integrating Product Validation and Business Service Technology Applications: By establishing and running a collaborative iterative process of model simulation-product validation-application feedback, the credibility and practicality of the quantitative attribution product for the annual change in net primary productivity in Step S4 are continuously improved.

[0009] Preferably, step S1 includes the following steps: Step S11: Collect vegetation productivity data: Obtain gridded product data of annual net primary productivity of vegetation covering the target area; The data has a spatial resolution better than 1 kilometer, a time span of no less than 25 years, and is a standardized dataset with reliable accuracy.

[0010] Step S12: Collect meteorological station data: Obtain annual meteorological observation data from multiple meteorological stations covering the target area; The data start and end years are matched with NPP data, and the data content includes key meteorological elements such as average temperature, maximum temperature, minimum temperature, precipitation, relative humidity, and sunshine duration. At the same time, the geographic coordinate information of the corresponding stations is obtained, which is used to generate a grid dataset of each meteorological element that is spatially matched with the NPP grid data.

[0011] Step S13: Collect digital elevation model data: Obtain grid point data of the digital elevation model covering the target area; Among them, the spatial resolution of the data is better than 1 kilometer, and high-precision global public datasets are given priority.

[0012] Step S14: Collect ecological and geographical zoning data and target area surface vector data: Obtain ecological and geographical zoning vector data covering the target area and target area surface vector data; it should be noted that the data should be able to comprehensively reflect the differences in the macro-ecological and geographical patterns of the target area.

[0013] Step S15, Vector data standardization processing: The ecological geographic zoning data and target area surface vector data obtained in step S14 are converted to a preset geographic projection coordinate system, and the target area surface vector is used to perform mask extraction on the ecological geographic zoning data to obtain a zoning vector layer, which will be used as the basis for sub-region division in the subsequent step S31. Step S16: Raster data standardization processing: The net primary productivity (NPP) annual grid data from step S11 and the digital elevation model (DEM) grid data from step S13 are uniformly resampled and transformed to the same preset geographic projection coordinate system as the vector data. Subsequently, the surface vector of the target area processed in step S15 is used for mask extraction to ensure that all data layers have completely consistent spatial range, pixel size, and pixel alignment. A set of long-sequence standardized NPP annual grid datasets is generated, which will serve as the basis for preparing the interannual variation data of NPP in the subsequent step S21. At the same time, a set of standardized digital elevation model (DEM) grid data is generated and used as a covariate for generating a high-precision meteorological element grid dataset in the subsequent step S22.

[0014] Step S17: Station Data Standardization Processing: Extract the unique station identifier, longitude, latitude, and annual values ​​of various meteorological elements from the meteorological station observation data obtained in Step S12 to construct a structured dataset. This data will serve as the raw input for generating a high-precision meteorological element gridded dataset in subsequent Step S22.

[0015] Preferably, step S2 includes the following steps: Step S21: Prepare data on the interannual variation of net primary productivity: Based on the gridded dataset of annual net primary productivity values ​​standardized in step S16, calculate the difference between it and the long-term average state year by year to generate a gridded dataset of interannual variation of net primary productivity. Step S22: Prepare high-precision meteorological grid data: For each target meteorological element to be interpolated, run the dynamic adaptive K-nearest neighbor meteorological interpolation method based on elevation collaboration. Use the meteorological station geographic coordinates standardized in step S17 and the grid data of the digital elevation model standardized in step S16 as features, and use its observation values ​​as targets. Introduce the K-nearest neighbor machine learning algorithm framework. Within the set spatial search range, construct the K-nearest neighbor regression model through a dynamic adaptive mechanism to generate a high-precision meteorological element grid dataset year by year.

[0016] Preferably, step S3 includes the following steps: At least 25 years of core variable data prepared in step S2 were selected, and a three-level progressive factor detection and diagnostic framework of full-region scale, sub-region scale, and pixel scale was adopted.

[0017] Step S31, Regional-scale detection of meteorological driving factors: Based on the partition vector obtained in step S15, the target area is divided into multiple sub-regions; within the overall target area and each sub-region, the spatial differentiation relationship between the grid data of each meteorological element generated in step S22 and the interannual variation data of net primary productivity in step S21 is detected using the geographic detector model, and its q value and statistical significance p value are obtained. Step S32, Initial screening of meteorological driving factors at the regional scale: In all the gridded meteorological element data generated in step S31, meteorological factors that show low explanatory power q<0.05 and are not significant p>0.1 in both the overall target area and all sub-regions are removed. The meteorological factors that pass the screening are recorded in the core meteorological factor candidate set. Step S33, Pixel-scale fine screening of meteorological driving factors: Based on the core meteorological factors of the target area screened in step S32, the correlation coefficient between the interannual variation of net primary productivity over many years and the meteorological factor series is calculated pixel by pixel. The effective degrees of freedom are estimated and the significance test of the correlation coefficient is completed based on the effective degrees of freedom (p<0.05). Factors that pass the test are identified as effective meteorological driving factors that are significantly related to the interannual variation of net primary productivity.

[0018] Preferably, step S4 includes the following steps: Step S41, Classification of Effective Meteorological Driving Factors: Based on the light-temperature-water biometeorological element framework, all effective meteorological driving factors for each pixel determined in step S33 are identified and classified one by one: Step S42, Single Pixel Restriction Factor Screening: Using the pixel as an independent analysis unit, based on the classification results of Step S41, sort all effective meteorological driving factors in each category according to the absolute value of their correlation coefficient with the interannual variation of net primary productivity |r|, and select the one with the largest |r| as the restriction factor for that category; Step S43, Classification of Ecological Restriction Types of Single Pixels: After completing the data processing in Steps S33 and S42, based on the combination of restriction factors selected in Step S42, each pixel is defined as one of three ecological restriction types. Step S44, Construction of single-pixel quantitative regression model: Taking the pixel as an independent analysis unit, based on the different ecological restriction types and their restriction factors divided in step S43, and using the data used in step S3, establish appropriate regression models respectively. Step S45: Single-pixel quantitative regression model simulation: Based on the pixel-level parameterized quantitative attribution model constructed in step S44, input the core variable data prepared in step S2 as needed, run the model, and quantitatively decompose the interannual variation of net primary productivity ΔNPP of each pixel into the climate impact ΔNPP. 气象 Human-induced impact △NPP 人为 The formula is as follows: ΔNPP=ΔNPP 气象 +ΔNPP 人为 .

[0019] Preferably, in step S41, effective meteorological driving factors are classified into the following three categories: Solar energy supply factors include factors characterizing solar energy resources, such as sunshine duration; Thermal condition factors include those characterizing thermal resources: average temperature, maximum temperature, and minimum temperature; Water supply factors include those that characterize water resources, such as precipitation.

[0020] Preferably, in step S43, the three types of ecological constraints are: in the classification in step S41, if only one type contains a constraint factor, the pixel is a single-factor dominant type; if two types contain constraint factors, the pixel is a dual-factor synergistic type; and if all three types contain constraint factors, the pixel is a three-factor driven type. In step S44, a univariate regression model is used for the single-factor dominant type, a binary regression model is used for the two-factor collaborative type, and a ternary regression model is used for the three-factor driven type. Finally, a pixel-level parameterized quantitative attribution model for the interannual change of net primary productivity is integrated and constructed.

[0021] Preferably, step S5 includes the following steps: Step S51, Analysis of Typical Areas of Human Contribution: Based on the human influence quantity △NPP generated in step S45 人为Interannual series data products, through data analysis, identify representative geographical areas that are significantly affected by human activities during the study period, namely, typical areas of human contribution. Step S52, Multi-source data collection and collaborative preprocessing in typical areas: For each typical area in step S51, multi-source high-resolution remote sensing image data are collected and standardized preprocessed to achieve effective interannual comparative monitoring of surface processes in the typical area during the study period. Step S53, Structured Collection and Organization of Ecological Policy Information: In sync with the remote sensing data collection in Step S52, for the same typical area, extensively collect and systematically organize and study the ecological protection and governance policy information corresponding to the period.

[0022] Information sources include, but are not limited to: publicly available government policy documents, planning documents, implementation plans, project reports, and relevant academic literature. The collected information is then organized and analyzed.

[0023] Step S54, Change Detection and Attribution Verification Based on High-Resolution Imagery: For each typical area, the annual average change of NPP caused by human activities calculated according to the specific implementation method in Step S51 is spatially coupled and compared with the land cover change results identified from the high-resolution imagery in Step S52 and the ecological policy information sorted out in Step S53 for verification.

[0024] By analyzing spatial location matching and trend consistency, the reasonableness of significant negative values ​​or anomalous positive values ​​appearing in the annual average change of NPP due to anthropogenic influence is examined, thereby enabling the assessment of the anthropogenic influence (ΔNPP) generated in step S45. 人为 Indirect verification of the product.

[0025] Step S55, Post-verification Feedback Model Optimization: The verification results of step S54 are used as feedback information to optimize the model construction method of step S4; based on the optimized model and parameters, recalculation and verification are performed, iterating until all human contribution typical areas pass the coupling comparison verification of step S54;

[0026] Step S56, Application Feedback and Model Optimization: The gridded data products of climate impact and anthropogenic impact of the annual net primary productivity interannual variation generated after model optimization in step S55 are used for demonstration applications in specific scenarios.

[0027] Preferably, step S51 includes the following steps: A reference year is set. For each pixel in the data product, the difference between the anthropogenic impact data of each year during the study period and the reference year is calculated. Then, the average annual change of NPP anthropogenic impact of the pixel relative to the reference year is obtained by averaging. Through spatial feature analysis, typical areas with significantly negative annual changes in NPP anthropogenic impact and typical areas with abnormally positive values ​​are identified, namely, typical areas of anthropogenic contribution.

[0028] Preferably, the application in step S56 includes: collecting operational data and application service feedback, conducting evaluations on attribution accuracy, operational stability, and user support effectiveness, and further adjusting and optimizing the model based on the evaluation results.

[0029] Compared with existing technologies, the present invention has the following significant advantages: 1. At the data input level, a simultaneous leap in the accuracy and spatial detail of basic meteorological data has been achieved: Through the dynamic adaptive K-nearest neighbor meteorological interpolation method based on elevation collaboration, high-precision meteorological grid data has been generated in complex terrain areas with sparse stations, accurately reproducing the vertical distribution pattern and spatial details of meteorological elements, eliminating the systematic bias of traditional coarse data from the source. This progress has built a solid and reliable data foundation for the entire technology chain.

[0030] 2. At the process analysis level, pixel-level accurate identification of meteorological driving factors is achieved: overcoming the microscopic distortion problem caused by factor homogenization in traditional methods, it can accurately identify the effective meteorological driving factors of NPP changes in each pixel under different regions and different ecosystem types, making the attribution analysis more consistent with the actual ecological situation at the pixel scale, and significantly enhancing the diagnostic accuracy and spatial resolution.

[0031] 3. At the core model level, the ecological mechanism of statistical quantitative attribution has been enhanced: a pixel-level quantitative attribution model that deeply integrates statistics and ecological mechanisms incorporates the ecological limiting factor law into the core of the model, achieving a fundamental leap from mathematical correlation to ecological mechanism-driven attribution logic. Based on the effective meteorological driving factors diagnosed for each pixel, the model dynamically matches different modeling strategies, significantly improving the quantitative accuracy of climate change and human activity contributions, and enhancing the model's mechanistic explanatory power and robustness. Compared with traditional process models and statistical models, this invention effectively overcomes the shortcomings of high uncertainty in process models and weak mechanistic rationality in statistical models, making the attribution results more scientific and credible.

[0032] 4. In terms of output results, the dimensions and decision support value of attribution analysis have been expanded: it has achieved a leap from analyzing the total amount of NPP to analyzing its changes, and can keenly capture and quantify the net impact of dynamic processes such as ecological engineering and extreme climate events. This allows the analysis conclusions to directly serve the effectiveness evaluation of ecological protection and restoration projects and the dynamic adjustment of policies, and the application value has been substantially deepened.

[0033] 5. At the results verification level, an effective product verification method was proposed, overcoming the bottleneck of verification without ground-based measured data: An innovative ΔNPP attribution product-high-resolution image coupling verification method was adopted, enabling effective assessment of the spatial rationality of attribution products without relying on ground-based measured data. This verification method provides crucial support for the credibility of attribution results, promoting the advancement of related technologies from scientific research to operational and practical applications.

[0034] 6. At the technology maturity level, an integrated technology process of technology research and development, product verification, and application feedback is constructed to enhance business service capabilities: This invention connects the entire process from technology research and development and product verification to business application services. This technology process has been empirically proven in real-world application scenarios, and its service effectiveness has been adopted at the government level, fully demonstrating the high maturity and practicality of the technology, and achieving precise alignment between technological achievements and actual business needs such as ecological policy assessment and public communication. Attached Figure Description

[0035] Figure 1 This is the overall flowchart of the present invention.

[0036] Figure 2 This is a distribution map of the average annual change in NPP caused by human activities from 2001 to 2024, based on 2000, according to an embodiment of the present invention.

[0037] Figure 3 This is a comparison of high-resolution remote sensing images of Crescent Lake in Mohe from 2001 and 2020, based on an embodiment of the present invention.

[0038] Figure 4 This is a comparison of high-resolution remote sensing images of the border area between Tahe and Huma in the Greater Khingan Mountains in 2000 and 2017, representing an embodiment of the present invention.

[0039] Figure 5 This is a comparison of high-resolution remote sensing images of the Yinhe Reservoir area in Gannan Prefecture in 2000 and 2021, based on an embodiment of the present invention.

[0040] Figure 6 This is a comparison image of high-resolution remote sensing images of the Harbin area from 2005 and 2021, based on an embodiment of the present invention.

[0041] Figure 7 The image shown is a news photo of an algal bloom that occurred in Xingkai Lake, Heilongjiang Province in 2023, as an embodiment of the present invention.

[0042] Figure 8 This is a map showing the annual average variation of NPP climate impact and NPP anthropogenic impact in the Three-North Shelterbelt Project area of ​​Heilongjiang Province from 2001 to 2024 (based on 2000).

[0043] Figure 9 This is an example diagram of the service application products of the 7th Heilongjiang Provincial Tourism Industry Development Conference in 2025, as an embodiment of the present invention. Detailed Implementation

[0044] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0045] Spatiotemporal driving force analysis of net primary productivity (NPP) based on a geographic detector model reveals that, unlike the spatial pattern of NPP itself, which is significantly constrained by traditional geographic factors such as elevation, soil type, and land cover type, these factors have a significantly reduced explanatory power for the spatiotemporal differentiation of interannual NPP variation (ΔNPP), lacking a direct and significant impact. The fundamental reason is that these geographic factors exhibit long-term stability and relative static characteristics under natural conditions, making it difficult to effectively explain the dynamic fluctuations of NPP at interannual scales. Further analysis, combining ecological mechanisms with existing research consensus, indicates that regional interannual NPP variation is mainly synergistically regulated by three core driving factors: climate change, forest age structure dynamics, and human activities.

[0046] This invention uses forest age structure dynamics as a representative indicator of the long-term effects and time lags of anthropogenic influences such as ecological engineering, and incorporates it into the indirect impacts of anthropogenic activities. A two-dimensional attribution framework of climate change and anthropogenic activities is constructed, using ΔNPP as the object. By precisely separating the contribution of climate change, a quantitative analysis of the contribution of anthropogenic activities is achieved. The consideration of anthropogenic activities not only covers direct anthropogenic impacts such as afforestation, urbanization, and reservoir construction, but also incorporates the indirect enhancement of vegetation carbon sequestration capacity caused by ecological engineering measures such as forest closure for natural regeneration, forest tending, and grazing bans. Based on this, after removing the impact of climate change, the interannual change in net primary productivity (ΔNPP) can be systematically analyzed and quantified to reflect the contribution of anthropogenic factors, providing technical support for ecosystem protection and restoration, and for eco-meteorological services serving dual-carbon goals.

[0047] Example

[0048] like Figure 1 The method for quantifying meteorological and anthropogenic contributions to NPP changes based on a pixel-level parameterized model, as shown, mainly includes the following steps: Step 1: Data Collection and Preprocessing

[0049] Step 1.1: Vegetation productivity data collection: Extract annual gridded data covering Heilongjiang Province, with a spatial resolution of 500 meters and a time span from 2000 to 2024, from the Global Land Surface Characteristic Parameter Net Primary Product (GLASSNPP).

[0050] Step 1.2, Meteorological station data collection: Obtain annual meteorological observation data from multiple national meteorological stations covering Heilongjiang Province and surrounding areas from the Tianqing platform of the National Meteorological Administration.

[0051] The start and end years of the data (2000-2024) perfectly match the NPP data in step 1.1. The collected meteorological indicators include: average temperature, maximum temperature, minimum temperature, surface temperature, precipitation, relative humidity, snow cover, evaporation, sunshine duration, and radiation. Simultaneously, the precise geographic coordinates (latitude and longitude) of each meteorological station are acquired. This station data will serve as the raw input, used to generate a gridded dataset of each meteorological element that perfectly matches the spatial resolution of the NPP gridded data through methods such as spatial interpolation.

[0052] Step 1.3, Digital Elevation Model Data Collection: Obtain SRTMDEM 90M resolution elevation model (DEM) product data from the Chinese Academy of Sciences Geospatial Data Cloud Platform, spatially covering the Heilongjiang Province region, and use it as a co-variable for subsequent generation of meteorological element grid datasets that spatially match the NPP grid data.

[0053] Step 1.4: Collection of Ecological Geographic Zoning Data and Regional Vector Data of Heilongjiang Province: Vector data of "Ecological Geographic Zoning of China" was obtained from the Resource and Environmental Science Data Center of the Chinese Academy of Sciences, spatially covering the Heilongjiang Province. This data integrates natural geographical elements such as climate, topography, vegetation, and soil, and can comprehensively reflect the differences in the macro-ecological geographical pattern of the target area, possessing authority and scientific validity.

[0054] This embodiment will use this data as the objective basis for subsequent sub-regional division, for regional-scale screening of meteorological driving factors, in order to eliminate interference from differences in macro-ecological background and improve the accuracy of regional screening of meteorological driving factors. Simultaneously, standard administrative area vector data of Heilongjiang Province will be acquired.

[0055] Step 1.5: Vector Data Standardization Processing: The surface vector data of the "China Eco-Geographical Zoning" and the administrative region of Heilongjiang Province obtained in Step 1.4 are uniformly converted to the preset Albers equal-area projection coordinate system. This projection can effectively control the area deformation of the mid-latitude regions of China and is suitable for spatial analysis and statistics at the provincial scale. Subsequently, the surface vector of Heilongjiang Province is used to perform mask extraction on the vector data of China's ecological geographical zoning. The resulting zoning vector layer will serve as the basis for sub-region division in the subsequent regional-scale meteorological driving factor detection (Step 3.1).

[0056] Step 1.6, Raster Data Standardization: To ensure that vegetation net primary productivity (NPP) and topographic (DEM) data are spatially comparable and computable, they need to be strictly standardized.

[0057] Specifically, the NPP grid data obtained in step 1.1 and the DEM grid data obtained in step 1.3 are uniformly resampled to a 1-kilometer resolution according to the service requirements of the embodiment, and transformed to the same Albers equal-area projection coordinate system as the vector data in step 1.5. Subsequently, the surface vector of Heilongjiang Province processed in step 1.5 is used for mask extraction to ensure that all data layers have completely consistent spatial range, pixel size and pixel alignment; a set of long-sequence standardized net primary productivity (NPP) annual value grid dataset is generated, which will serve as the basis for preparing the interannual variation data of net primary productivity in the subsequent step 2.1; at the same time, a set of standardized digital elevation model (DEM) grid data is generated and used as a covariate for generating a high-precision meteorological element grid dataset in the subsequent step 2.2.

[0058] Step 1.7, Standardization of Site Data: To achieve spatial continuity of discrete site data, the original data needs to be standardized, extracted, and structured to meet the input requirements of subsequent spatial interpolation algorithms.

[0059] First, from the annual observation data of meteorological stations in Heilongjiang Province obtained in step 1.2, the unique identifier, latitude and longitude coordinates, and annual observation values ​​of various meteorological elements for each station are extracted. Then, this information is organized into a standardized structured Excel dataset. This dataset will serve as the directly readable standardized input required for subsequent spatial interpolation (step 2.2) to generate a high-precision gridded dataset of meteorological elements.

[0060] Step 2: Preparation of core variable data

[0061] Step 2.1, Preparation of Interannual Variation of Net Primary Productivity (ΔNPP): Based on the long-sequence standardized net primary productivity (NPP) annual grid dataset from Step 1.6, firstly, the arithmetic mean of NPP at each spatial grid point within the study period (2000-2024) is calculated. Specifically, for any grid point in space... Its NPP average The calculation formula is as follows: in, This represents the total number of years (25 years in this example). This indicates that the grid point is at the 1st position. The annual NPP values ​​are calculated by iterating through all grid points, ultimately generating a multi-year average NPP baseline grid dataset with the same spatial resolution and extent as the input data. This dataset represents the mean or background level of net primary productivity (NPP) of vegetation during the study period.

[0062] After obtaining the multi-year average baseline field of NPP, year-by-year, grid-by-grid difference calculations are performed to quantify the deviation of NPP from the mean state each year. For a given year... sum point Its net primary productivity interannual change Defined as the difference between the actual NPP value in a given year and the multi-year average NPP value: This calculation process is repeated over all years (2000-2024) and on all spatial grid points. The value can be positive or negative, representing whether the NPP of the grid point in year t is higher or lower than its long-term average.

[0063] Through the above steps, a complete gridded dataset of net primary productivity interannual change (ΔNPP) is finally generated. The dataset consists of one image per year from 2000 to 2024, and the spatial attributes (spatial resolution, projection, etc.) remain completely consistent with the original data.

[0064] Step 2.2, High-precision meteorological grid data preparation: To achieve direct spatial matching analysis between discrete meteorological station observation data and rasterized ΔNPP data, and addressing the interpolation problem caused by the sparseness of meteorological stations in complex terrain, this invention introduces elevation as a core constraint variable into the K-Nearest Neighbor (KNN) machine learning algorithm framework. It innovatively proposes a dynamic adaptive K-Nearest Neighbor meteorological interpolation method based on elevation collaboration, significantly improving the interpolation accuracy and spatial texture detail representation capability of meteorological elements in sparse station areas. The specific implementation process is as follows: First, based on the spatial extent and resolution of the standardized DEM grid data obtained in step 1.6, a regular blank grid is generated that completely overlaps with the spatial extent and resolution of the long-sequence standardized net primary productivity (NPP) annual value grid data. This grid is used as the output template for the final interpolation result. Then, the coordinates of each meteorological station obtained in step 1.7 are converted to a projection coordinate system consistent with the DEM to ensure the uniformity of spatial reference. Simultaneously, the elevation values ​​of the corresponding locations of each meteorological station are extracted based on the DEM data. The station's planar coordinates (X, Y) and elevation (Z) information are integrated to construct a complete three-dimensional spatial dataset.

[0065] To eliminate the influence of different dimensions on interpolation accuracy, the constructed 3D feature data (X, Y, Z) is Z-score standardized to obtain a standardized 3D feature space. Within this standardized feature space, for each grid point to be interpolated in the output template, its 3D Euclidean distance to all meteorological stations is calculated, and interpolation is performed using the inverse distance weighting method. During the interpolation process, a search radius of 100 kilometers (R=100km) is set to select valid meteorological stations within this range of the grid point to be interpolated, excluding invalid stations outside the range to avoid interfering with the interpolation results. Based on this, K nearest neighbor stations are dynamically and adaptively selected to participate in the calculation (preset K=5; if the number of stations within the search radius is less than 5, all stations within the range are used). After the interpolation calculation is completed, the interpolation results are masked using the regional surface vector data of Heilongjiang Province to retain valid data within the region. Subsequently, through outlier and extreme value screening, data cleaning, and other processing, invalid data is removed and abnormal information is corrected. Finally, the processed results are output as raster data in GeoTIFF format for subsequent related analysis.

[0066] Through the above steps, a complete high-precision meteorological element grid dataset is finally generated. The dataset covers the period from 2000 to 2024, with one image for each element per year, and the spatial attributes (spatial resolution, projection, etc.) remain completely consistent with the original data.

[0067] Step 3: Progressive diagnosis of meteorological driving factors

[0068] This invention constructs a three-layer progressive diagnostic logic for meteorological driving factors of net primary productivity interannual variation (ΔNPP), consisting of regional initial screening, sub-regional verification, and pixel fine screening. Through a multi-scale, progressively focused analysis path, it achieves accurate identification and verification of effective meteorological driving factors, significantly improving the reliability and robustness of diagnostic results.

[0069] Specifically, based on the core variable dataset (ΔNPP and meteorological element grid data) prepared in step 2 for 2000-2024, the method first performs preliminary screening of potential driving factors at the macro-regional scale (Heilongjiang Province as a whole) based on global statistical characteristics, excluding factors that show weak influence at the regional scale. Then, the preliminary screening results are statistically validated within each sub-region (eco-geographical sub-region) to confirm that these factors, which show weak influence at the regional scale, also do not have a dominant influence at the sub-regional scale, thus effectively eliminating preliminary screening errors caused by regional heterogeneity. Finally, refined discrimination and calculation are performed at the pixel scale to ultimately identify the effective meteorological driving factors that dominate the spatial differentiation pattern of ΔNPP. This method, through the above-mentioned hierarchical, progressive, validation, and refinement mechanism, lays a solid scientific foundation for the subsequent construction of a quantitative attribution model. The specific implementation process is as follows: Step 3.1, Regional-scale detection of meteorological driving factors: Based on the partition vector obtained in Step 1.5, Heilongjiang Province is divided into multiple sub-regions; then, in the whole Heilongjiang Province and in each sub-region, the spatial differentiation relationship between the grid data of each meteorological element generated in Step 2.2 and the ΔNPP data is detected using the geographic detector model, and its q value and statistical significance p value are obtained. Step 3.2, Regional-scale preliminary screening of meteorological driving factors: In all the gridded meteorological element data in Step 3.1, meteorological factors that show low explanatory power (q<0.05) and are not significant (p>0.1) in the whole Heilongjiang Province region and all sub-regions are removed. The meteorological factors that pass the screening are recorded in the core meteorological factor candidate set. In this embodiment, the meteorological factors after regional preliminary screening still include average temperature, maximum temperature, minimum temperature, precipitation, relative humidity, sunshine duration, and radiation.

[0070] Step 3.3, Pixel-scale driver factor screening: Based on the core meteorological factors of Heilongjiang Province screened in Step 3.2, the correlation coefficient between the multi-year ΔNPP sequence and each meteorological factor sequence is calculated pixel by pixel, and the effective degrees of freedom are estimated to correct for time series autocorrelation; a t-distribution significance test is performed based on the effective degrees of freedom (p<0.05); the factors that pass the significance test are identified as effective meteorological drivers that are significantly correlated with the interannual variation of net primary productivity at that pixel.

[0071] Step 4: Construction of a quantitative attribution model for pixel-level NPP changes

[0072] Step 4.1, Classification of Effective Meteorological Driving Factors: Based on the light-temperature-water biometeorological element framework, to improve the model's mechanistic nature and reduce the impact of multicollinearity, all effective meteorological driving factors for each pixel identified in Step 3.3 are individually identified and classified into the following three categories: Solar energy supply factors: meteorological factors include sunshine duration and radiation.

[0073] Thermal conditions: Meteorological factors include average temperature, maximum temperature, and minimum temperature.

[0074] Moisture supply factors: meteorological factors include precipitation and relative humidity.

[0075] Step 4.2, Single-pixel limiting factor screening: Taking the pixel as an independent analysis unit, based on the classification results of Step 4.1, all effective meteorological driving factors in each category are sorted according to the absolute value of their ΔNPP-meteorological factor correlation coefficient |r|, and the one with the largest |r| is selected as the limiting factor for that category.

[0076] Step 4.3, Single-Pixel Ecological Limitation Type Classification: After completing the data processing in Step 3.3 (correcting for autocorrelation effects) and Step 4.2 (enhancing mechanisms and reducing multicollinearity), the processing achieved a balance between preserving key information and suppressing redundant noise, ensuring that the meteorological factors used in subsequent analyses possess both statistical independence and ecological representativeness. Thus, a complete and controllable set of limiting factors was ultimately obtained, providing ideal input for accurately defining ecological limitation types.

[0077] Based on the combination of limiting factors selected in step 4.2, each pixel is defined as one of three ecological limiting types. In the classification in step 4.1, if only one type contains a limiting factor, the pixel is a single-factor dominant type; if two types contain limiting factors, the pixel is a two-factor synergistic type; and if all three types contain limiting factors, the pixel is a three-factor driven type. For example, if the limiting factors in a pixel include maximum temperature and precipitation, i.e., it is affected by both heat conditions and water supply factors, then the pixel is a two-factor synergistic type.

[0078] Step 4.4, Construction of Single-Pixel Quantitative Regression Model: Taking the pixel as an independent analysis unit, based on the different ecological limitation types and their limiting factors identified in Step 4.3, and using the data from Step 3, appropriate regression models are established respectively. This breaks away from the traditional one-size-fits-all modeling approach, achieving precise matching of one model per pixel. The specific modeling strategy in this embodiment is as follows: For single-factor dominant regions, a univariate linear regression model is adopted, with the limiting factor corresponding to the pixel as the independent variable and the change in NPP as the dependent variable, to quantify the independent contribution of a single limiting factor to ΔNPP. For regions with two-factor synergistic characteristics, a binary linear regression model is adopted, and two types of limiting factors are introduced as independent variables to quantify the synergistic contribution of the two factors to ΔNPP. For regions driven by three factors, a ternary linear regression model is established, which comprehensively incorporates three limiting factors: light energy, heat, and moisture, to quantify the contribution of multiple factors to ΔNPP.

[0079] Finally, by integrating all pixel models, a pixel-level parameterized quantitative attribution model for ΔNPP was constructed. This model incorporates the ecological limiting factor law into its core, achieving a deep integration of statistics and ecological mechanisms, and completing the essential leap from mathematical correlation to ecological mechanism-driven attribution logic.

[0080] Step 4.5, Single-pixel quantitative regression model simulation: Based on the pixel-level parameterized quantitative attribution model constructed in Step 4.4, the model is run by inputting the core variable data prepared in Step 2 (2000-2024 in this example, which can be increased over time), and the ΔNPP of each pixel is quantitatively decomposed into the climate impact (ΔNPP). 气象 ) and human impact (△NPP) 人为 These figures represent the contributions of climate change and human activities, respectively. The formulas are as follows: ΔNPP = ΔNPP 气象 +ΔNPP 人为 Step 5: Technical Application Solutions Integrating Product Validation and Business Services Step 5.1, Analysis of Typical Areas of Anthropogenic Contribution: Based on the anthropogenic impacts (△NPP) generated in Step 4.5 and stripped of climate change effects from 2000 to 2024. 人为 This step in the study of interannual series data products aims to identify representative geographical areas that are significantly affected by human activities during the study period through data analysis, i.e., typical areas of human contribution.

[0081] Specifically, the process includes the following: 1. Benchmark setting and difference calculation

[0082] In 2000 (the starting year of the study period), Heilongjiang Province was officially approved as a national pilot province for ecological construction. This embodiment sets 2000 as the reference year. For each pixel in the data product Calculate the amount of human impact year by year. Corresponding value to the base year The difference: in, Indicates a specific year within the research period ( ).

[0083] 2. Calculation of the average change: for each pixel Based on the above difference sequence, calculate its average change over the entire study period: In the formula, =24, which represents the total number of years in the study period excluding the base year. This yields the average anthropogenic impact data of the interannual change in net primary productivity (ΔNPP) from 2001 to 2024, with 2000 as the base year. This is referred to as the average annual anthropogenic impact of NPP.

[0084] 3. Typical region identification: Through analysis Based on the spatial distribution characteristics, five typical regions were identified, such as... Figure 2 As shown, the four typical areas with significantly negative NPP values ​​are: ① the area with negative anthropogenic impact at Crescent Lake in Mohe, ② the area with negative anthropogenic impact near Yinhe Reservoir in Gannan, ③ the area with negative anthropogenic impact at the junction of Tarim and Huma, and ④ the area with negative anthropogenic impact near Harbin. The annual average change in NPP caused by anthropogenic impact in these areas is shown in the figure. Generally below -50 gC / m². However, in the area of ​​abnormally positive anthropogenic influence within Xingkai Lake (section ⑤), unlike other water bodies, the annual average change in NPP due to anthropogenic influence is... There are regions where the concentration is significantly greater than 0 gC / m².

[0085] Step 5.2, Multi-source data collection and collaborative preprocessing in typical areas: For each typical area in Step 5.1, multi-source high-resolution remote sensing image data are collected, and the collected images are standardized and preprocessed to achieve effective interannual comparative monitoring of surface processes in the typical area during the study period.

[0086] Step 5.3, Structured Collection and Organization of Ecological Policy Information: Simultaneously with the remote sensing data collection in Step 5.2, for the same typical area, extensively collect and systematically organize ecological protection and governance policy information corresponding to the research period. Information sources include, but are not limited to: publicly available government policy documents, planning texts, implementation plans, project reports, and relevant academic literature. The collected information will then be organized and analyzed.

[0087] Step 5.4: Change Detection and Attribution Verification Based on High-Resolution Imagery: The ΔNPP attribution product-high-resolution imagery coupling verification method is used. For each typical area, the annual average change in NPP anthropogenic impact calculated according to the specific implementation method in Step 5.1 is spatially coupled and compared with the land cover change results identified from high-resolution imagery in Step 5.2 and the ecological policy information compiled in Step 5.3. Attribution verification is achieved through spatial location matching and trend consistency analysis. The analysis shows that these outlier areas correspond to clear, high-intensity human activity disturbances (such as urbanization, reservoir construction, etc.) or indirect human activity impacts (such as forest fires, eutrophication), thus strongly supporting the reliability of the results. Specifically: 1. In the Crescent Lake area of ​​Mohe County, Heilongjiang Province, high-resolution remote sensing imagery clearly shows that compared with 2000 ( Figure 3 Compared to (left), 2020 (left) Figure 3(Right) Urban construction land shows a significant expansion, mainly manifested in a large-scale outward spread from the original urban area in the center. This urbanization process has led to a large area of ​​vegetation cover being replaced by urban land, thus providing direct and clear evidence of anthropogenic activity (urban expansion) driving the significant negative spatial pattern of the annual average change in NPP in this region.

[0088] 2. In the Tahe and Huma areas of the Greater Khingan Mountains region of Heilongjiang Province, a comparison of high-resolution remote sensing images shows that compared with 2000 (… Figure 4 Compared to (left), 2017 (left) Figure 4 (Right) The remote sensing image shows a significant left-right color difference. Combined with field investigations and ecological process analysis, this difference primarily stems from the burn marks left by the "5.17" Fulahan fire in 2003. This manifests as significant spatial heterogeneity between different forest ages and vegetation types and the surrounding native vegetation during the vegetation restoration process in the burned area. The fire caused widespread and severe damage to vegetation in the area within a short period, significantly altering the land cover pattern. The spatial pattern of the fire-affected area and the area with significantly negative annual NPP (Annual Percentage Change) values ​​provides direct surface observation evidence for the reliability of the ΔNPP quantitative attribution product.

[0089] 3. In the Yinhe Reservoir area of ​​Gannan County, Heilongjiang Province, high-resolution remote sensing imagery clearly reveals significant changes in land cover between 2000 and 2021. Image comparison shows that in 2021 (… Figure 5 (Right) Compared to 2000 ( Figure 5 (Left) The natural vegetation cover around the reservoir in this region has clearly shrunk and become fragmented, with some original vegetation areas having been converted into urban land; at the same time, the area of ​​the reservoir's water body has also shown a significant expansion. This land cover change process, characterized by vegetation reduction, artificial surface expansion, and reservoir water storage area expansion, is highly consistent with the spatial distribution characteristics of the significant negative annual change in NPP (Nuclear Power Product) in the region during the same period. This provides intuitive and reliable remote sensing observation evidence for the conclusion that human activities are the main driving force behind the decline in ecosystem productivity in this region.

[0090] 4. In Harbin City and surrounding areas of Heilongjiang Province, high-resolution remote sensing image comparisons revealed two significant anthropogenic-driven land surface change characteristics: (1) Compared with 2005 ( Figure 6 Compared to (left), 2021 (left) Figure 6 (Right) The width of the Songhua River in the Harbin section has increased significantly and the water area has expanded significantly. Based on the on-site investigation and analysis, this change is mainly attributed to the flooding effect of the Dadingshan Reservoir (construction started in 2004 and water was impounded in 2007). (2) The comparison of images from the same period also shows that the built-up area of ​​Harbin has expanded significantly from the original urban area to the surrounding areas, and the urban surface area has increased significantly.

[0091] The aforementioned urban expansion and reservoir inundation together led to large-scale loss of vegetation cover, causing a systematic decline in vegetation productivity. The spatial distribution of this loss closely matches the significant negative anomaly areas in the annual average change of NPP due to anthropogenic factors. This consistency provides direct observational evidence for the reliability of the ΔNPP quantitative attribution product.

[0092] 5. In some areas of Lake Xingkai, the annual average change in NPP (Neuro-Protected Protein) due to anthropogenic factors shows an abnormally positive value. Analysis of literature and other data suggests this phenomenon is mainly attributed to eutrophication and periodic algal blooms caused by human activities such as watershed pollution. Studies (Li Liangfang, 2024) indicate that since 2000, the eutrophication trend of Lake Xingkai has been increasing, with a significant trend of ecological degradation and frequent algal blooms. Figure 7 As shown, during algal blooms, the explosive proliferation of phytoplankton (mainly algae) in the lake water leads to abnormally high levels of signals such as chlorophyll a concentration, which in turn significantly increases the NPP in the region. This process clearly reveals the specific cause of the abnormally positive annual variation in NPP due to anthropogenic influence: abnormal vegetation growth indirectly caused by human activities such as eutrophication, rather than the natural recovery or improvement of terrestrial vegetation. The areas with abnormally positive annual variation in NPP due to anthropogenic influence are highly correlated with algal bloom events in time and space. This correspondence provides intuitive observational evidence for the reliability of the ΔNPP quantitative attribution product.

[0093] Step 5.5, Post-Verification Feedback Model Optimization: If the verification result in Step 5.4 fails, the relevant cases will be used as key feedback information to backtrack and optimize the model type parameters or algorithm structure in Step 4. The updated model is then recalculated and verified, and this iterative process continues until the anthropogenic influence results for all typical regions pass the attribution verification in Step 5.4. The final result presented in this embodiment is the result after iterative model optimization.

[0094] Step 5.6, Product Demonstration Application and Service Effectiveness: Based on the gridded data products of ΔNPP climate impact and anthropogenic impact generated after model optimization in Step 5.5, operational demonstration applications were carried out in typical areas such as the Three-North Shelterbelt Project area in Heilongjiang Province and Fuyuan City. Application results show that this achievement has achieved significant results in serving local ecological protection and sustainable development, with outstanding socio-economic benefits. At the operational integration level, this achievement has been deeply integrated into the Heilongjiang Provincial Ecological Meteorological Decision-Making Service System, expanding its service scope from the provincial level to multiple city (prefecture) and county-level units such as Fuyuan, Yichun, and Harbin, providing scientific and reliable data support for optimizing local ecological protection policies.

[0095] Demonstration application products in the Three-North Shelterbelt Project area of ​​Heilongjiang Province, such as Figure 8As shown, from 2001 to 2024, the net primary productivity (NPP) of vegetation in this region increased overall compared to 2000, with a total change of 762.5 TgC. In terms of driving factors, climate change and anthropogenic activities contributed 41.6% and 58.4% of the total change, respectively, indicating that anthropogenic factors played a dominant role in the NPP change in this region, and that ecological protection and restoration policies have been remarkably effective. Further spatial analysis shows (… Figure 8 In most parts of Heilongjiang Province, both the magnitude of climate impacts and anthropogenic impacts increased compared to 2000. Among them, the most significant increase in the average annual change in climate impacts was concentrated in the southwestern semi-arid region, with a relatively concentrated spatial distribution; while the increase in the average annual change in anthropogenic impacts was widespread, its spatial distribution was more dispersed and did not show obvious clustering characteristics.

[0096] Demonstration application service effectiveness such as Figure 9 As shown in the "Fuyuan City Vegetation Ecological Monitoring Report", the monitoring results indicate that the area where human factors have significantly enhanced the promotion of vegetation growth accounts for 69.3%, highlighting the effectiveness of ecological protection policies and projects.

[0097] In summary, the present invention has the following beneficial effects: 1. To address the meteorological interpolation distortion caused by sparse meteorological stations in complex terrain, this invention proposes a K-Nearest Neighbor (KNN) meteorological interpolation framework that integrates elevation constraints and a dynamic adaptive mechanism. This method introduces elevation as a key spatial variable, constructs a comprehensive similarity metric that integrates geographical distance and elevation gradient, and achieves gridded collaborative computation based on a digital elevation model (DEM). Simultaneously, a dynamic adaptive strategy intelligently adjusts the number of neighboring stations, thereby significantly improving the interpolation accuracy and spatial detail representation capability of meteorological elements in sparsely populated areas.

[0098] 2. This invention abandons the traditional globally uniform factor screening strategy and creates a progressive diagnostic mechanism from macro to micro. This mechanism, through the integration of methods such as geographic detectors, forms a progressively converging judgment logic that can adaptively diagnose and identify the unique key meteorological driving factors for each pixel's spatiotemporal dynamic characteristics, laying the factor foundation for accurate pixel-level attribution modeling.

[0099] 3. This invention breaks through the limitations of traditional statistical models by introducing the ecological limiting factor law, pioneering a pixel-level quantitative attribution model that deeply integrates with ecological mechanisms. Based on the effective meteorological driving factors identified through prior diagnosis, this model categorizes them according to the "light-temperature-water" biometeorological element framework. It dynamically matches differentiated modeling paths—"single-factor dominant," "dual-factor synergistic," or "triple-factor driven"—based on the number of limiting factors, achieving precise mapping of "one pixel, one model," ensuring the ecological mechanism of the attribution process and the interpretability of the results.

[0100] 4. To address the core contradiction of the disconnect between technological development and business application in the field of NPP quantitative attribution, this invention constructs a complete technical process integrating technology development, product verification, and typical business demonstration. The core is the first-time use of the "ΔNPP attribution product-high-resolution image coupling verification method," which, through spatial anchoring and multi-source data matching, achieves indirect validity verification under conditions lacking ground-based measured data. This drives a virtuous cycle of technological iteration, business empowerment, and value feedback, ensuring the practicality and viability of the technology. Finally, a successful government-level application case demonstrates the high maturity of this technical system and its direct support capability for ecological governance.

[0101] The above embodiments are not intended to limit the present invention, and the present invention is not limited to the examples given above. Any changes, modifications, additions or substitutions made by those skilled in the art within the scope of the technical solution of the present invention are also within the protection scope of the present invention.

Claims

1. A method for quantifying meteorological and anthropogenic contributions to NPP changes based on a pixel-level parameterized model, characterized by: Includes the following steps: Step S1, Data Collection and Preprocessing: Collect multi-source environmental data within the target area and standardize it; Step S2: Prepare core variable data: Calculate the standardized data year by year to generate a gridded dataset; Step S3, Progressive meteorological driving factor diagnosis: Select at least 25 years of core variable data based on step S2, and use a three-level progressive factor detection and diagnosis framework of full-region scale, sub-region scale, and pixel scale for diagnosis; Step S4: Construct a pixel-level quantitative attribution model for NPP changes: Using pixels as independent analysis units, based on the diagnostic results of step S3, construct a pixel-level quantitative attribution model based on the ecological limiting factor law to quantitatively decompose the contributions of climate change and human activities to the interannual change in net primary productivity of each pixel. Step S5: Integrating the application of technologies for product verification and business services.

2. The method for quantifying meteorological and anthropogenic contributions to NPP changes based on a pixel-level parameterized model according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: Collect vegetation productivity data: Obtain gridded product data of annual net primary productivity of vegetation covering the target area; Step S12: Collect meteorological station data: Obtain annual meteorological observation data from multiple meteorological stations covering the target area; Step S13: Collect digital elevation model data: Obtain grid point data of the digital elevation model covering the target area; Step S14: Collect ecological geographic zoning data and target area surface vector data: Obtain ecological geographic zoning vector data covering the target area and target area surface vector data; Step S15, Vector data standardization processing: The ecological geographic zoning data and target area surface vector data obtained in step S14 are converted to a preset geographic projection coordinate system, and the target area surface vector is used to perform mask extraction on the ecological geographic zoning data to obtain a zoning vector layer, which will be used as the basis for sub-region division in the subsequent step S31. Step S16, Raster data standardization processing: The net primary productivity annual value grid data from step S11 and the digital elevation model grid data from step S13 are uniformly resampled and transformed to the same preset geographic projection coordinate system as the vector data. Then, the target area surface vector processed in step S15 is used for mask extraction to ensure that all data layers have completely consistent spatial range, cell size and cell alignment. Step S17: Standardization of station data: Extract the unique station identifier, longitude, latitude and annual values ​​of various meteorological elements from the meteorological station observation data obtained in step S12, and construct a structured dataset.

3. The method for quantifying meteorological and anthropogenic contributions to NPP changes based on a pixel-level parameterized model according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: Prepare data on the interannual variation of net primary productivity: Based on the gridded dataset of annual net primary productivity values ​​standardized in step S16, calculate the difference between it and the long-term average state year by year to generate a gridded dataset of interannual variation of net primary productivity. Step S22: Prepare high-precision meteorological grid data: For each target meteorological element to be interpolated, run the dynamic adaptive K-nearest neighbor meteorological interpolation method based on elevation collaboration. Use the meteorological station geographic coordinates standardized in step S17 and the grid data of the digital elevation model standardized in step S16 as features, and use its observation values ​​as targets. Introduce the K-nearest neighbor machine learning algorithm framework. Within the set spatial search range, construct the K-nearest neighbor regression model through a dynamic adaptive mechanism to generate a high-precision meteorological element grid dataset year by year.

4. The method for quantifying meteorological and anthropogenic contributions to NPP changes based on a pixel-level parameterized model according to claim 1, characterized in that: Step S3 includes the following steps: Step S31, Regional-scale detection of meteorological driving factors: Based on the partition vector obtained in step S15, the target area is divided into multiple sub-regions; within the overall target area and each sub-region, the spatial differentiation relationship between the grid data of each meteorological element generated in step S22 and the interannual variation data of net primary productivity in step S21 is detected using the geographic detector model, and its q value and statistical significance p value are obtained. Step S32, Initial screening of meteorological driving factors at the regional scale: In all the gridded meteorological element data generated in step S31, meteorological factors that show low explanatory power q<0.05 and are not significant p>0.1 in both the overall target area and all sub-regions are removed. The meteorological factors that pass the screening are recorded in the core meteorological factor candidate set. Step S33, Pixel-scale fine screening of meteorological driving factors: Based on the core meteorological factors of the target area screened in step S32, the correlation coefficient between the interannual variation of net primary productivity over many years and the meteorological factor series is calculated pixel by pixel. The effective degrees of freedom are estimated and the significance test of the correlation coefficient is completed based on the effective degrees of freedom (p<0.05). Factors that pass the test are identified as effective meteorological driving factors that are significantly related to the interannual variation of net primary productivity.

5. The method for quantifying meteorological and anthropogenic contributions to NPP changes based on a pixel-level parameterized model according to claim 1, characterized in that: Step S4 includes the following steps: Step S41, Classification of Effective Meteorological Driving Factors: Based on the light-temperature-water biometeorological element framework, all effective meteorological driving factors for each pixel determined in step S33 are identified and classified one by one: Step S42, Single Pixel Restriction Factor Screening: Using the pixel as an independent analysis unit, based on the classification results of Step S41, sort all effective meteorological driving factors in each category according to the absolute value of their correlation coefficient with the interannual variation of net primary productivity |r|, and select the one with the largest |r| as the restriction factor for that category; Step S43, Classification of Ecological Restriction Types of Single Pixels: After completing the data processing in Steps S33 and S42, based on the combination of restriction factors selected in Step S42, each pixel is defined as one of three ecological restriction types. Step S44, Construction of single-pixel quantitative regression model: Taking the pixel as an independent analysis unit, based on the different ecological restriction types and their restriction factors divided in step S43, and using the data used in step S3, establish appropriate regression models respectively. Step S45: Single-pixel quantitative regression model simulation: Based on the pixel-level parameterized quantitative attribution model constructed in step S44, input the core variable data prepared in step S2 as needed, run the model, and quantitatively decompose the interannual variation of net primary productivity ΔNPP of each pixel into the climate impact ΔNPP. 气象 Human-induced impact △NPP 人为 The formula is as follows: ΔNPP=ΔNPP 气象 +ΔNPP 人为。 6. The method for quantifying meteorological and anthropogenic contributions to NPP changes based on a pixel-level parameterized model according to claim 5, characterized in that: In step S41, effective meteorological driving factors are classified into the following three categories: Solar energy supply factors include factors characterizing solar energy resources, such as sunshine duration; Thermal condition factors include those characterizing thermal resources: average temperature, maximum temperature, and minimum temperature; Water supply factors include those that characterize water resources, such as precipitation.

7. The method for quantifying meteorological and anthropogenic contributions to NPP changes based on a pixel-level parameterized model according to claim 5, characterized in that: In step S43, the three types of ecological constraints are as follows: In the classification in step S41, if only one type contains a constraint factor, the pixel is a single-factor dominant type; if two types contain constraint factors, the pixel is a dual-factor synergistic type; and if all three types contain constraint factors, the pixel is a three-factor driven type. In step S44, a univariate regression model is used for the single-factor dominant type, a binary regression model is used for the two-factor collaborative type, and a ternary regression model is used for the three-factor driven type. Finally, a pixel-level parameterized quantitative attribution model for the interannual change of net primary productivity is integrated and constructed.

8. The method for quantifying meteorological and anthropogenic contributions to NPP changes based on a pixel-level parameterized model according to claim 1, characterized in that: Step S5 includes the following steps: Step S51, Analysis of Typical Areas of Human Contribution: Based on the human influence quantity △NPP generated in step S45 人为 Interannual series data products, through data analysis, identify representative geographical areas that are significantly affected by human activities during the study period, namely, typical areas of human contribution. Step S52, Multi-source data collection and collaborative preprocessing in typical areas: For each typical area in step S51, multi-source high-resolution remote sensing image data are collected and standardized preprocessed to achieve effective interannual comparative monitoring of surface processes in the typical area during the study period. Step S53, Structured Collection and Organization of Ecological Policy Information: In sync with the remote sensing data collection in Step S52, for the same typical area, extensively collect and systematically organize and study the ecological protection and governance policy information corresponding to the period. Step S54, Change Detection and Attribution Verification Based on High-Resolution Imagery: For each typical area, the annual average change of NPP caused by human activities calculated according to the specific implementation method in Step S51 is spatially coupled and compared with the land cover change results identified from the high-resolution imagery in Step S52 and the ecological policy information sorted out in Step S53 for verification. Step S55, Post-verification Feedback Model Optimization: The verification results of step S54 are used as feedback information to optimize the model construction method of step S4; based on the optimized model and parameters, recalculation and verification are performed, iterating until all human contribution typical areas pass the coupling comparison verification of step S54; Step S56, Application Feedback and Model Optimization: The gridded data products of climate impact and anthropogenic impact of the annual net primary productivity interannual variation generated after model optimization in step S55 are used for demonstration applications in specific scenarios.

9. The method for quantifying meteorological and anthropogenic contributions to NPP changes based on a pixel-level parameterized model according to claim 8, characterized in that: Step S51 includes the following steps: A reference year is set. For each pixel in the data product, the difference between the anthropogenic impact data of each year during the study period and the reference year is calculated. Then, the average annual change of NPP anthropogenic impact of the pixel relative to the reference year is obtained by averaging. Through spatial feature analysis, typical areas with significantly negative annual changes in NPP anthropogenic impact and typical areas with abnormally positive values ​​are identified, namely, typical areas of anthropogenic contribution.

10. The method for quantifying meteorological and anthropogenic contributions to NPP changes based on a pixel-level parameterized model according to claim 8, characterized in that: The application in step S56 includes: collecting operational data and application service feedback, conducting evaluations on attribution accuracy, operational stability, and user support effectiveness, and then adjusting and optimizing the model again based on the evaluation results.

Images