Intelligent land engineering carbon sink regulation and control system and method based on dynamic management

Abstract

The invention discloses an intelligent land engineering carbon sink regulation and control system and method based on dynamic management, and belongs to the technical field of land engineering carbon sink regulation and control. The multi-source spatio-temporal data comprises vegetation indexes, soil parameters, meteorological data and carbon dioxide flux data; the data processing and analyzing module is used for performing fusion processing on the multi-source spatio-temporal data, inverting and calculating the carbon sink or carbon source state of the target area in real time based on a built-in carbon sink accounting model, and generating a carbon sink spatio-temporal distribution diagram; the intelligent decision-making module is used for diagnosing a carbon sink capacity limiting factor based on the carbon sink space-time distribution diagram and generating land engineering regulation and control strategies for different sub-regions; and the execution and feedback module is used for issuing the regulation and control strategy to an execution terminal or a management platform. The method improves the carbon sequestration capacity and regulation and control efficiency of land engineering.

Description

Technical Field

[0001] This application relates to an intelligent land engineering carbon sequestration control system and method based on dynamic management, belonging to the field of land engineering carbon sequestration control technology. Background Technology

[0002] As a core carrier for enhancing carbon sequestration in terrestrial ecosystems, land engineering faces increasingly urgent needs for carbon sequestration regulation. Currently, land engineering carbon sequestration management relies heavily on traditional technologies, often employing either satellite remote sensing for macro-monitoring or ground-based fixed-point sampling. While satellite data offers broad coverage, its limited spatial resolution makes it difficult to capture micro-level differences within sub-regions. Ground sampling, though highly accurate, suffers from narrow monitoring ranges and long data acquisition cycles, hindering the formation of effective spatiotemporal correlations among multi-source data. Furthermore, strategies are often developed based on human experience or single-factor analysis, failing to accurately diagnose the core limiting factors of carbon sequestration capacity or design differentiated solutions for different land types and terrain features in various sub-regions. This results in weak targeting of regulation measures and low efficiency in carbon sequestration enhancement. Consequently, traditional land engineering carbon sequestration regulation struggles to meet the demands of refined and intelligent management, restricting the efficient release of the carbon sequestration potential of land ecosystems. Summary of the Invention

[0003] According to one aspect of this application, a smart land engineering carbon sequestration control system based on dynamic management is provided. This system improves the carbon sequestration capacity and control efficiency of land engineering, and provides reliable technical support for long-term dynamic monitoring and optimized management of carbon sequestration.

[0004] A smart land engineering carbon sequestration regulation system based on dynamic management, characterized in that it includes: The sensing module is used by the integrated monitoring network to acquire multi-source spatiotemporal data of the target area, including vegetation index, soil parameters, meteorological data and carbon dioxide flux data. The data processing and analysis module is used to fuse the multi-source spatiotemporal data and, based on the built-in carbon sink accounting model, to invert and calculate the carbon sink or carbon source status of the target area in real time, and generate a carbon sink spatiotemporal distribution map. The intelligent decision-making module is used to diagnose the limiting factors of carbon sink capacity based on the spatiotemporal distribution map of carbon sinks, and generate land engineering regulation strategies for different sub-regions; the regulation strategies include at least one of tree species optimization, precision water and fertilizer, and ecological restoration engineering. The execution and feedback module is used to send the control strategy to the execution terminal or management platform and to receive the control data fed back by the sensing module.

[0005] Furthermore, the sensing module includes: Satellite remote sensing units are used to acquire large-scale, periodic remote sensing images and generate satellite remote sensing data. The UAV remote sensing unit is used to acquire lidar and / or spectral data of key areas and generate UAV remote sensing data. The ground-based Internet of Things (IoT) unit includes soil sensors, weather stations, and carbon dioxide flux towers deployed in the target area, generating ground-based IoT data.

[0006] Furthermore, the multi-source spatiotemporal data is fused, including: The multi-source spatiotemporal data is preprocessed and spatiotemporally aligned to unify it to the same spatiotemporal reference. The preprocessed multi-source spatiotemporal data is input into a multi-source data fusion model for data-level, feature-level, or decision-level fusion processing to generate multi-source environmental factor data for the target area. Based on the multi-source environmental factor data, monitoring, assessment, or decision-making information for the target area is output.

[0007] Furthermore, the preprocessed multi-source spatiotemporal data will be fused at the decision level, including: Based on satellite remote sensing data, first-level decision information is obtained, which includes crop growth zoning of the target area, identification of drought stress areas, or preliminary assessment of carbon sequestration potential. Based on UAV remote sensing data, specific areas in the first-level decision information are identified and verified to obtain second-level decision information, which includes precise location of pests and diseases, inversion of plant-level physiological state, or verification of soil surface anomalies. Based on ground-based IoT data, third-level decision information is generated, including real-time soil moisture alerts, microclimate monitoring, or carbon dioxide flux change verification. By integrating the decision information from the first, second, and third levels, a comprehensive decision instruction is generated; The comprehensive decision-making instruction is as follows: A variable operation prescription model is used to guide agricultural machinery and equipment to perform precision irrigation, variable fertilization or pesticide spraying operations in the target area. The generation of the variable operation prescription model depends at least on the fused vegetation index, soil parameters and meteorological data. Alternatively, a carbon flux accounting and assessment report, which is generated based on preliminary results of regional carbon flux inversion from satellite data and verified and calibrated using carbon dioxide flux data collected by drones and the Internet of Things.

[0008] Furthermore, based on the built-in carbon sink accounting model, the carbon sink or carbon source status of the target area is inverted and calculated in real time, generating a spatiotemporal distribution map of carbon sinks, including: The carbon sink accounting model is driven by spatialized data provided by satellites and drones as input to simulate the first simulated value of carbon flux in the target area. Obtain the measured values ​​of carbon dioxide flux from IoT sensors within the same spatiotemporal range; The difference between the simulated value of the first carbon flux and the measured value of the carbon dioxide flux is calculated. Based on the difference, the key biophysical parameters of the carbon sink accounting model are optimized and adjusted in reverse through the data assimilation algorithm. The key biophysical parameters include the maximum carboxylation rate, the maximum photosynthetic rate, or the dark respiration coefficient. The carbon sink accounting model was run again using the optimized parameters to generate a second simulated carbon flux value. The carbon sink, carbon source status and intensity information of each spatial grid unit in the target area obtained through data assimilation and inversion are associated with geospatial coordinates to obtain a spatiotemporal distribution map of carbon sinks. The second simulated carbon flux value serves as the net ecosystem exchange in the target area. The determination of net ecosystem exchange includes: if the simulated value of the second carbon flux is less than zero, the corresponding area is determined to be a carbon sink; if the simulated value of the second carbon flux is greater than zero, the corresponding area is determined to be a carbon source.

[0009] Furthermore, based on the aforementioned spatiotemporal distribution map of carbon sequestration, the limiting factors of carbon sequestration capacity are diagnosed, including: Based on the spatiotemporal distribution map of carbon sinks and the multi-source environmental factor data, a diagnostic dataset is constructed with environmental factors as features and carbon sink values ​​as labels. Establish a mapping relationship between environmental factors and carbon sink values; Calculate the importance score of each environmental factor among the multi-source environmental factors; Based on the importance score and the values ​​of each environmental factor, the limiting factors of carbon sequestration capacity within the target area and their spatial distribution are diagnosed.

[0010] Furthermore, the multi-source environmental factor data includes meteorological factors, soil factors, vegetation factors, and topographic factors; The meteorological factors include one or more of temperature, precipitation, and photosynthetically active radiation; The soil factors include one or more of soil moisture, soil nitrogen content, and soil pH. The vegetation factors include one or more of leaf area index and vegetation cover. The terrain factors include one or more of elevation and slope.

[0011] Furthermore, after calculating the importance score of each environmental factor among the multi-source environmental factors, the process also includes: For a specific grid cell within the target area, calculate the contribution of each environmental factor to the carbon sink value of that specific grid cell; Environmental factors with the largest negative contribution value are diagnosed as the primary limiting factors for the specific grid cell. Based on the statistical and spatial visualization of the primary constraint factors of all grid cells, a spatial distribution map of constraint factors for the target region is generated.

[0012] Furthermore, the implementation of the control strategy includes: When the limiting factor is water stress and the land type is farmland, the control strategy is precision irrigation or the construction of water collection facilities, wherein the irrigation amount is adjusted according to crop type, growth period and soil texture. When the limiting factor is nitrogen deficiency and the vegetation type is forest, the regulation strategy is to apply slow-release nitrogen fertilizer; When the limiting factor is soil acidification, the control strategy is to apply lime, wherein the amount of lime is calculated based on the soil pH and the target pH. Lime requirement (t / ha) = (target pH - current pH) × soil cation exchange capacity × adjustment coefficient; When the limiting factor is low vegetation cover and steep terrain, the control strategy is horizontal terrace land preparation and planting soil-stabilizing vegetation.

[0013] According to another aspect of this application, a smart land engineering carbon sequestration regulation method based on dynamic management is also provided, comprising: Acquire multi-source spatiotemporal data of the target region; The multi-source spatiotemporal data are fused and processed, and the carbon sink status of the target area is calculated in real time based on the carbon sink accounting model to generate a carbon sink spatiotemporal distribution map. Based on the aforementioned spatiotemporal distribution map of carbon sinks, artificial intelligence optimization algorithms are used to diagnose the limiting factors of carbon sink capacity and generate land engineering regulation strategies. The control strategy is executed, and the control model is iteratively optimized based on the feedback data after control.

[0014] The beneficial effects that this application can produce include: This application provides a smart land engineering carbon sink regulation system and method based on dynamic management. By acquiring multi-source spatiotemporal data through an integrated monitoring network of sensing modules, it overcomes the limitations of traditional single data sources, which suffer from limited coverage and insufficient accuracy. This provides a continuous and comprehensive spatiotemporal foundation for carbon sink accounting. Simultaneously, the data processing and analysis module significantly improves the accuracy of carbon sink / carbon source status assessment through multi-source data fusion and real-time inversion of the carbon sink accounting model. The generated spatiotemporal distribution map of carbon sinks clearly presents regional differences, providing an intuitive basis for subsequent decision-making. Furthermore, based on the precise diagnosis of limiting factors according to the spatiotemporal distribution of carbon sinks, differentiated regulation strategies such as tree species optimization and precise water and fertilizer management are formulated for different sub-regions, avoiding the problem of weak targeting in traditional solutions and significantly improving the scientific nature and adaptability of regulation measures. In addition, the constructed closed-loop mechanism can receive post-regulation data in real time and feed back into system optimization, realizing dynamic iteration of carbon sink management. This effectively improves the carbon sink capacity and regulation efficiency of land engineering, reduces resource waste, and provides reliable technical support for long-term dynamic monitoring and optimized management of carbon sinks, promoting the efficient release of the carbon sink potential of the land ecosystem. Attached Figure Description

[0015] Figure 1 This is a block diagram of a smart land engineering carbon sequestration control system based on dynamic management, according to one embodiment of this application. Figure 2 This is a block diagram illustrating the principle of the intelligent decision-making module in one embodiment of this application. Figure 3 This is a schematic diagram of the spatiotemporal distribution of carbon sinks in one embodiment of this application; Figure 4 This is a flowchart of a smart land engineering carbon sequestration regulation method based on dynamic management in one embodiment of this application. Detailed Implementation

[0016] The present application is described in detail below with reference to the embodiments, but the present application is not limited to these embodiments.

[0017] See Figure 1-3 A smart land engineering carbon sequestration regulation system based on dynamic management, characterized in that it includes: The sensing module is used by the integrated monitoring network to acquire multi-source spatiotemporal data of the target area, including vegetation index, soil parameters, meteorological data and carbon dioxide flux data. The data processing and analysis module is used to fuse the multi-source spatiotemporal data and, based on the built-in carbon sink accounting model, to invert and calculate the carbon sink or carbon source status of the target area in real time, and generate a spatiotemporal distribution map of carbon sink. The intelligent decision-making module is used to diagnose the limiting factors of carbon sink capacity based on the spatiotemporal distribution map of carbon sinks, and generate land engineering regulation strategies for different sub-regions; the regulation strategies include at least one of tree species optimization, precision water and fertilizer, and ecological restoration engineering. The execution and feedback module is used to send the control strategy to the execution terminal or management platform and to receive the control data fed back by the sensing module.

[0018] Specifically, the perception module, as the data acquisition layer, constructs a multi-source spatiotemporal data monitoring network covering four dimensions: vegetation, soil, meteorology, and CO2 flux. It deploys drones and ground-based IoT devices, such as soil temperature and humidity sensors, NDVI spectrometers, and weather stations, to achieve high-frequency, high-precision data acquisition. It integrates satellite remote sensing data, such as Landsat, Sentinel, and near-Earth remote sensing data, to fill spatiotemporal blind spots in ground monitoring. It unifies multi-source data formats and establishes a spatiotemporal index database to support efficient subsequent processing. The data processing and analysis module, as the model calculation layer, performs data fusion, carbon sink inversion, and spatiotemporal modeling. It can use Kalman filtering or deep learning to fuse heterogeneous data, eliminating noise and extracting key features. The carbon sink accounting model simulates vegetation photosynthesis, respiration, and soil organic carbon decomposition. Based on historical data, it trains a carbon sink prediction model to improve real-time performance. Combined with GIS technology, it visualizes carbon sink results as dynamic heat maps, supporting multi-scale analysis at the regional / sub-regional / plot levels. The intelligent decision-making module, as the strategy generation layer, is used for limiting factor diagnosis and regulation strategy optimization. The contribution of various factors, such as soil pH and precipitation, to carbon sequestration is quantified using SHAP values ​​or LIME algorithms. A critical value for carbon sequestration capacity is set, triggering an alert when a sub-region falls below this threshold. Furthermore, suitable high-carbon-sinking tree species are screened, and planting combinations are optimized using genetic algorithms. Digital twin technology is used to simulate the impact of different water and fertilizer schemes on carbon sequestration, recommending optimal ratios. For degraded areas, vegetation restoration pathways are generated, such as succession sequences from herbaceous to shrub to tree. An execution and feedback module serves as a closed-loop control layer for strategy implementation and effect evaluation. Control commands are pushed to execution terminals, such as smart irrigation systems and drone seeding equipment, via API interfaces or blockchain technology. Carbon sequestration data before and after control are compared to calculate the increase in sequestration, such as tCO2 / ha, and model parameters are iteratively optimized.

[0019] It is worth noting that the sensing module, through an integrated monitoring network, can acquire rich multi-source spatiotemporal data of the target area, including vegetation indices, soil parameters, meteorological data, and carbon dioxide flux data. This data covers multiple aspects such as vegetation, soil, climate, and carbon flux, providing comprehensive and accurate foundational information for subsequent analysis and decision-making. The monitoring network integrates various monitoring methods, such as ground sensor networks, airborne remote sensing platforms, and satellite remote sensing systems, working collaboratively. For example, ground sensors can collect soil parameter and carbon dioxide flux data in real time, while remote sensing platforms can acquire vegetation indices and large-scale meteorological data at a macroscopic level, achieving multi-level, three-dimensional data acquisition and ensuring dual optimization of data coverage and spatiotemporal resolution. The data processing and analysis module fuses multi-source spatiotemporal data, integrating data from different sources and of different types, eliminating discrepancies and contradictions between data, and forming a unified and accurate dataset. This helps improve data quality and usability, providing a more reliable basis for subsequent analysis. The built-in carbon sink accounting model is the core of this module. Based on fused data, it can invert and calculate the carbon sink or carbon source status of a target area in real time and generate a spatiotemporal distribution map of carbon sinks. For example, by combining vegetation indices, soil organic carbon content, and meteorological data, and using specific algorithm models, it can accurately estimate the carbon sink amount in different regions at different times, intuitively displaying the spatiotemporal distribution characteristics of carbon sinks. The intelligent decision-making module, based on the spatiotemporal distribution map of carbon sinks, can deeply analyze and diagnose the limiting factors of carbon sink capacity. For example, by comparing the carbon sink amount and various influencing factors in different sub-regions, it can identify which factors restrict the improvement of carbon sink capacity. For instance, some areas may have low carbon sink capacity due to insufficient soil fertility, unsuitable tree species, or poor water conditions. Based on the diagnosed limiting factors, it generates land engineering control strategies for different sub-regions. These strategies are targeted and operable, including various methods such as tree species optimization, precision water and fertilizer management, and ecological restoration projects. For example, in areas with low carbon sequestration capacity, it is advisable to optimize the tree species structure by selecting species more suitable for the local environment and with stronger carbon sequestration capacity; or to improve vegetation growth and carbon sequestration capacity through precise water and fertilizer management, and the rational application of fertilizers and irrigation. The execution and feedback module can promptly distribute the control strategies generated by the intelligent decision-making module to the execution terminal or management platform to ensure effective implementation. This may involve interfacing with the information systems of relevant agricultural, forestry, and other departments or enterprises to achieve precise communication and implementation of control strategies. It also receives post-control data from the sensing module, forming a closed-loop management system. Through analysis and evaluation of the post-control data, the effectiveness of the control strategy can be understood, and it can be determined whether the expected carbon sequestration improvement target has been achieved. If the effect is unsatisfactory, the control strategy can be adjusted in a timely manner, and implementation and monitoring can be repeated to continuously optimize the land engineering carbon sequestration control process.

[0020] like Figure 3 As shown in the carbon sink spatiotemporal distribution map, carbon emissions among cities within the Wei River Basin exhibit significant spatial correlations. Regarding spatial correlation, the spatial correlation of carbon sinks among cities remains relatively stable, while the spatial correlation of carbon emissions fluctuates considerably. Specifically, high-value carbon emission clusters are mainly concentrated in economically developed areas such as Xi'an and Xianyang, while low-value areas are more dispersed, and this clustering characteristic shows significant changes over time. From the perspective of carbon balance regional distribution, counties in the western and northern parts of the Wei River Basin have good carbon balances and are the main carbon sink areas; while counties in the southeast are high-carbon emission areas. Given the significant differences in carbon budgets across different regions, in practical work, the zoning type of carbon balance should be comprehensively considered, taking into full account the actual conditions of each region. Simultaneously, county-level cooperation and a point-to-area approach should be adopted to restructure the production pattern, accelerate industrial upgrading, and jointly promote carbon reduction and emission reduction through regional collaboration. High-high clustering refers to a high-carbon sink area surrounded by other high-carbon sink areas, characterized by a concentrated distribution of carbon sinks. High-low clustering refers to a high carbon sink area surrounded by a low carbon sink area, exhibiting a marginal distribution of carbon sinks. Low-high clustering refers to a low carbon sink area surrounded by a high carbon sink area, exhibiting an isolated distribution of carbon sinks. Low-low clustering refers to a low carbon sink area surrounded by other low carbon sink areas, exhibiting a sparse distribution of carbon sinks.

[0021] The sensing module includes: Satellite remote sensing units are used to acquire large-scale, periodic remote sensing images and generate satellite remote sensing data. The UAV remote sensing unit is used to acquire lidar and / or spectral data of key areas and generate UAV remote sensing data. The ground-based Internet of Things (IoT) unit includes soil sensors, weather stations, and carbon dioxide flux towers deployed in the target area, generating ground-based IoT data.

[0022] Specifically, satellite remote sensing units provide large-scale, periodic macroscopic monitoring data to support carbon sink baseline assessment and long-term trend analysis. Optical satellites are used for vegetation index inversion. Radar satellites are used to monitor surface deformation, such as soil erosion and vegetation height. Hyperspectral satellites are used for vegetation type classification and biomass estimation. Simultaneously, the 6S model or FLAASH algorithm can be used to eliminate aerosol and water vapor interference. Cloud areas are automatically identified based on deep learning (such as U-Net) to fill in missing data. TIMESAT software is used to extract vegetation growing season parameters, such as SOS / EOS dates and growing cycle length. UAV remote sensing units acquire high-precision three-dimensional structure and spectral information of key areas to compensate for the spatiotemporal blind spots of satellite remote sensing. LiDAR is used to generate point cloud data to extract parameters such as vegetation height, canopy density, and terrain slope. Multispectral / hyperspectral cameras are used for refined vegetation classification and chlorophyll content inversion. The flight path employs a gridded or adaptive sampling strategy to ensure that the coverage density of key areas, such as carbon sink hotspots, is greater than 10 points / m². 2 Ground-based IoT units monitor micro-environmental factors in real time, providing key input parameters for carbon sequestration process models. Soil sensors measure soil temperature (-40℃~85℃), humidity (0%~100%Vol), pH (0~14), and electrical conductivity (0~20dS / m). Weather stations integrate sensors for wind speed, wind direction, precipitation, light intensity, and air temperature and humidity. CO2 flux towers continuously monitor CO2 flux, water vapor flux, and energy balance. Soil sensors are deployed in a 500m×500m grid, and weather stations and flux towers cover different land use types, such as forests, farmland, and wetlands.

[0023] It is worth noting that satellite remote sensing units provide a high-altitude perspective for acquiring fundamental data related to carbon sequestration in large-scale land engineering areas. This is due to their wide coverage, fixed observation cycle, and high degree of data standardization, enabling them to provide regional-level spatiotemporal background information for carbon sequestration regulation. Satellites equipped with multispectral, hyperspectral, or thermal infrared sensors capture remote sensing images of the target area at fixed intervals. After preprocessing such as radiometric calibration, atmospheric correction, and geometric fine correction, standardized satellite remote sensing data is generated. Acquired data include Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), and Leaf Area Index (LAI), which reflect vegetation cover and growth vitality within the region, forming the basis for estimating vegetation carbon storage. Image interpretation distinguishes different land types such as forest, grassland, and cultivated land, clarifying the boundaries of areas with significant differences in carbon sequestration function. Simultaneously, macro-environmental parameters such as land surface temperature (LST) can assist in analyzing the indirect impacts of meteorological conditions on vegetation photosynthesis (carbon absorption) and soil respiration (carbon release). This system is suitable for preliminary assessment of the spatial distribution of carbon sinks in large-scale target areas such as provincial and municipal levels, as well as monitoring long-term dynamic trends of carbon sink changes, providing macro-level guidance for subsequent UAV and ground-based monitoring to pinpoint key areas. The UAV remote sensing unit is a crucial link connecting macro-level satellite data and micro-level ground data, focusing on key areas, small-scale, and high-resolution data acquisition. This overcomes the limitations of satellite remote sensing, such as insufficient resolution in local areas and significant cloud obstruction, enabling precise characterization of key carbon sink areas. Based on areas with abnormal carbon sink capacity identified by the satellite remote sensing unit, such as sparse vegetation areas or suspected soil degradation areas, UAVs equipped with lidar or hyperspectral sensors conduct low-altitude (typically 100-500 meters) autonomous or semi-autonomous flights to acquire three-dimensional structural and spectral data of key areas, generating high-resolution UAV remote sensing data. By acquiring lidar data and generating digital elevation models (DEMs), digital surface models (DSMs), and point cloud data for the target area, it can accurately calculate three-dimensional structural parameters such as vegetation canopy height, diameter at breast height (DBH), and canopy closure, significantly improving the estimation accuracy of vegetation biomass (a core indicator of carbon storage) for trees and shrubs. Hyperspectral data, compared to satellite multispectral data, offers higher spectral resolution, enabling the identification of specific tree species (e.g., distinguishing between coniferous and broadleaf forests), diagnosing vegetation nutrient stress (e.g., nitrogen and phosphorus deficiencies), and retrieving the spatial heterogeneity of soil organic matter content, providing direct evidence for subsequent tree species optimization and precise water and fertilizer regulation strategies. For sub-regions with insufficient resolution in satellite imagery (e.g., a plantation or ecological restoration pilot area), it conducts quarterly or monthly refined monitoring to obtain microscopic spatial differences in vegetation structure and soil properties, providing high-resolution input data for carbon sink accounting models. Ground-based IoT units are the ground-based tentacles of the sensing modules. Through a sensor network that is deployed at fixed points, continuously monitors, and transmits data in real time, they acquire the most direct and dynamic key parameters in the carbon sequestration process. They are the core data source for real-time inversion of carbon sequestration status and feedback on regulation effects.Based on the soil type, vegetation distribution, and topographic features of the target area, three types of core monitoring equipment are scientifically deployed to form a ground-based Internet of Things (IoT) monitoring network, enabling automatic data collection, storage, and remote transmission: Soil sensors are buried at different soil depths (e.g., 10cm, 30cm, 50cm) to collect parameters such as soil temperature, soil moisture, soil electrical conductivity (reflecting soil salinity), and soil organic matter content in real time, directly affecting vegetation root growth and the stability of the soil carbon pool (organic carbon); Meteorological stations are typically deployed at a density of 1-5 square kilometers to collect meteorological data such as temperature, precipitation, wind speed, wind direction, and sunshine hours in the target area, which are key driving factors for vegetation photosynthesis (carbon absorption) and soil respiration (carbon release); Carbon dioxide flux towers, as the gold standard for direct monitoring of carbon sink / carbon source status, use eddy covariance technology to measure the carbon dioxide exchange flux between the atmosphere and the surface in real time (the amount of carbon absorbed or released per unit area per unit time), directly determining whether an area is a carbon sink (negative flux, absorbing carbon dioxide) or a carbon source (positive flux, releasing carbon dioxide). Acquiring minute-level or hourly ground-based IoT data offers advantages such as high temporal resolution, strong data continuity, and direct correlation with carbon sequestration processes. It can be combined with spatial data from satellites and drones to form a comprehensive data support system with high temporal and spatial resolution. This provides real-time dynamic parameters for carbon sequestration accounting models (e.g., real-time meteorological data to correct photosynthetic rates) and serves as a sensor for the effects of implemented control strategies. For example, after applying precision water and fertilizer, soil sensors monitor changes in soil nutrients, and carbon dioxide flux towers monitor the improvement in carbon sequestration flux, providing direct evidence for feedback adjustments.

[0024] The fusion processing of the multi-source spatiotemporal data includes: The multi-source spatiotemporal data is preprocessed and spatiotemporally aligned to unify it to the same spatiotemporal reference. The preprocessed multi-source spatiotemporal data is input into a multi-source data fusion model for data-level, feature-level, or decision-level fusion processing to generate multi-source environmental factor data for the target area. Based on the multi-source environmental factor data, monitoring, assessment, or decision-making information for the target area is output.

[0025] Specifically, coordinate transformation is performed on ground sensors, and UAV point clouds are registered with satellite imagery using the ICP algorithm. Super-resolution reconstruction is performed on the satellite imagery to match the resolution of the UAV imagery. Satellite transit times are synchronized with UAV flight times to ensure consistent phenological conditions. For minute-level data from the terrestrial IoT, cubic spline interpolation is used to generate time-series data with the same frequency as the satellite / UAV. During data-level fusion, satellite multispectral bands, UAV hyperspectral bands, and terrestrial IoT data are concatenated into a high-dimensional feature tensor. The t-SNE algorithm is used to reduce the features to 3 dimensions, retaining over 90% of the information entropy for visualization and subsequent analysis. During feature-level fusion, a three-branch 3D-CNN network is constructed to extract deep representations of satellite spatiotemporal features, UAV geometric features, and ground temporal features, respectively, and features are fused using a spatial attention mechanism. During decision-level fusion, dynamic weights are assigned to the prediction results of random forest, XGBoost, and LSTM, and the confidence levels of different models are converted into basic probability assignments. The final decision, such as carbon sink classification, is output after fusion.

[0026] It is worth noting that preprocessing and spatiotemporal alignment of multi-source spatiotemporal data are fundamental to data fusion. This process eliminates spatiotemporal differences and quality noise from multi-source data, establishes a unified data benchmark, and transforms satellite remote sensing data (typically using the WGS84 coordinate system), UAV data (often using local coordinate systems), and terrestrial IoT data (latitude and longitude coordinates) to the target area's standard projected coordinate system, such as the UTM partitioned coordinate system. Resampling adjusts data of different resolutions to the same grid scale, typically using the highest resolution data as the benchmark. Ground control points are used to perform precise geometric correction on UAV imagery, eliminating spatial offsets caused by flight attitude errors. High-frequency data from terrestrial IoT is converted into time series matching satellite and UAV data through moving averages. For data collected at different times, time interpolation is used to generate datasets at the same time node, resolving the issue of asynchronous observation times. Abnormal data points caused by sensor malfunctions and severe weather are removed using methods such as the 3σ criterion and sliding window detection. Based on data characteristics and application requirements, the fusion process is divided into three levels to realize the transformation from raw data to decision information: Landsat-8 NDVI data is fused with Sentinel-2 NDVI data to generate a vegetation index sequence with both high spatial resolution and high temporal frequency; organic matter data from soil sensors at different depths are fused to construct a vertical distribution profile of the soil carbon pool; the three-dimensional structural features of vegetation extracted by UAV lidar are fused with the biochemical features of vegetation retrieved by hyperspectral inversion to construct a more accurate feature set for estimating vegetation biomass; meteorological data and soil parameter features are fused to establish a combination feature of carbon flux influencing factors; and the final carbon sink intensity level distribution map is generated by integrating macroscopic carbon sink distribution from satellite remote sensing, carbon sink capacity of key areas monitored by UAVs, and measured carbon exchange values ​​from ground flux towers, thus solving the decision bias problem caused by a single data source. The fused multi-source environmental factor data, such as vegetation growth status, soil carbon pool characteristics, and carbon flux dynamics, are ultimately transformed into three types of core information products: Monitoring information: such as daily-scale net primary productivity (NPP) of vegetation and spatiotemporal distribution maps of soil organic carbon content per 500-meter grid; automatic triggering of early warning information when carbon flux in a certain area suddenly increases, such as due to carbon release from forest fires or a sharp drop in vegetation indices, such as due to pest and disease outbreaks. Assessment information: calculating key indicators such as total carbon storage, carbon sink growth rate, and carbon sink potential of the target area based on the fused data; outputting confidence scores for different fusion results, such as areas with high consistency between satellite and ground data receiving higher scores, providing a reference for subsequent decision-making. Decision support information: identifying key influencing factors of carbon sink capacity, such as the main reason for limited carbon sink in a certain area being insufficient soil moisture; determining areas with the most significant response to control measures based on the fused data, such as areas where precision fertilization can significantly improve carbon sink capacity, providing direct input to the intelligent decision-making module.

[0027] The preprocessed multi-source spatiotemporal data is fused at the decision level, including: Based on satellite remote sensing data, first-level decision information is obtained, which includes crop growth zoning of the target area, identification of drought stress areas, or preliminary assessment of carbon sequestration potential. Based on UAV remote sensing data, specific areas in the first-level decision information are identified and verified to obtain second-level decision information, which includes precise location of pests and diseases, inversion of plant-level physiological state, or verification of soil surface anomalies. Based on ground-based IoT data, third-level decision information is generated, including real-time soil moisture alerts, microclimate monitoring, or carbon dioxide flux change verification. By integrating the decision information from the first, second, and third levels, a comprehensive decision instruction is generated; The comprehensive decision-making instruction is as follows: A variable operation prescription model is used to guide agricultural machinery and equipment to perform precision irrigation, variable fertilization or pesticide spraying operations in the target area. The generation of the variable operation prescription model depends at least on the fused vegetation index, soil parameters and meteorological data. Alternatively, a carbon flux accounting and assessment report, which is generated based on preliminary results of regional carbon flux inversion from satellite data and verified and calibrated using carbon dioxide flux data collected by drones and the Internet of Things.

[0028] Specifically, the first-level decision-making information is driven by satellite remote sensing, inputting multispectral imagery and time series data to classify crop growth stages, such as seedling stage, jointing stage, and grain-filling stage. Combined with topographic data and historical yield data, the target area is divided into high / medium / low-yield zones. The Temperature-Vegetation Drought Index (TVDI) is calculated as follows: TVDI = (Ts - Ts min ) / (Ts max -Ts minThe system employs a threshold segmentation method, combining soil moisture products. Based on historical drought event data, an LSTM model is trained to predict the current drought level. A 95% confidence interval for carbon sink potential is generated using satellite LAI, meteorological data, and land cover type. Second-level decision information is validated and refined using drones. A pest and disease patch detector is trained using multispectral drone imagery and hyperspectral point clouds to generate a digital surface model and calculate the precise coordinates of pest and disease patches. Chlorophyll fluorescence is retrieved from hyperspectral imagery. Canopy temperature is calculated based on thermal infrared imagery to obtain moisture content. The 3σ principle is used to identify physiologically abnormal plants, such as those under water stress or nutrient deficiency. Soil cracks in drone imagery are segmented using U-Net++. Erosion risk levels are classified using a logistic regression model, combining topographic slope and vegetation cover. Third-level decision information is monitored in real-time via the Internet of Things (IoT). A 5TE soil sensor is deployed to measure EC, temperature, humidity, and tensiometers, and to measure matrix potential. Triggering is initiated when soil volumetric water content falls below 60% of field capacity and when matrix potential falls below -10 kPa. Net ecosystem exchange was measured using an EC155 open-circuit gas analyzer. The carbon flux retrieved from the satellite was then calibrated using linear regression with the ground-based measurements.

[0029] Further, spatial registration is performed, aligning UAV imagery with satellite grids using the ICP algorithm. Terrestrial IoT data is interpolated to the satellite grid using the Thiessen polygon method. Time synchronization is performed, aligning satellite transit times with UAV flight times. IoT data is generated using a sliding window averaging method to produce time-series matched data. A weighted voting method is used to assign dynamic weights to the hierarchical decision results. Satellite decisions have a weight of 0.4 due to their wide coverage but low resolution; UAV decisions have a weight of 0.5 due to their high accuracy but limited coverage; and IoT decisions have a weight of 0.1 due to their strong real-time capabilities but sparse spatial distribution. The hierarchical decisions are then transformed into basic probability assignments. For example, the basic probability assignment for crop growth zones (high / medium / low yield) is generated using a support vector machine classifier. The final decision is output after fusion.

[0030] When the integrated decision-making command is used as a variable-based operational prescription model, it integrates the satellite-derived normalized vegetation index (NDVI) and the drone-derived vegetation index; the electrical conductivity and pH values ​​measured by the Internet of Things (IoT) and the soil organic matter retrieved from the satellite; and the 10-day cumulative precipitation forecast. Irrigation zones are delineated based on soil moisture sensor data. Variable irrigation amounts are generated by combining crop water requirements. Variable fertilizer application amounts are generated using the nutrient index method based on the spatial distribution of soil nutrients. Variable spray concentrations are generated by combining pest and disease detection results and crop growth stages. When the integrated decision-making command is used as a carbon flux accounting and assessment report, it integrates the net primary productivity (NEE) and soil respiration products retrieved from the satellite, the total primary productivity (GPP) retrieved from the drone's sun-induced chlorophyll fluorescence (SIF), and the regional carbon flux measured by the ground eddy correlation method. Regional carbon flux NEE = GPP - RS (unit: gC / m³). 2 / day).

[0031] It is worth noting that the first-level decision information is based on the wide coverage of satellite remote sensing. It divides the basic carbon sink-related status of the target area at a macro scale. By analyzing the spatiotemporal changes of vegetation indices such as NDVI and EVI, the region is divided into three levels of growth: excellent, medium, and poor, to identify macro-spatial differences in carbon sink capacity. The vegetation water supply index (VSWI) is constructed by combining surface temperature and vegetation indices to delineate potential areas affected by drought and provide preliminary target areas for water regulation. Based on land use type, vegetation cover, and climate zone characteristics, a light energy utilization model is used to estimate the upper limit of regional carbon sink potential and select key areas with greater potential for carbon sink improvement. For abnormal or key areas identified in the first-level decision-making information, high-resolution data from drones is used for refined verification and supplementation. Specific bands of hyperspectral data (such as the red-edge band) are used to identify abnormal biochemical parameters caused by vegetation diseases and pests, refining satellite-identified areas of poor growth into specific disease and pest patches with positioning accuracy down to the individual tree level. Parameters such as height, crown width, and diameter at breast height of individual trees are extracted from lidar point cloud data and combined with chlorophyll content retrieved via hyperspectral inversion to accurately assess plant growth status and verify the accuracy of satellite growth zoning. For suspected soil degradation areas identified by satellite, features such as exposed ground and erosion gullies are identified using drone visible light imagery, and combined with soil organic matter content retrieved via hyperspectral inversion to confirm areas at risk of soil carbon pool loss. The third-level decision information uses fixed-point monitoring data as a benchmark to quantitatively verify and dynamically calibrate the decision information of the first two levels. Through measured data from soil sensors, it verifies the satellite drought stress identification results and triggers precise irrigation alerts for areas where the soil moisture content in the 0-50cm layer is below the threshold. Using hourly data from meteorological stations (temperature, humidity, and photosynthetically active radiation), it calibrates the macro-climate parameters retrieved by the satellite and improves the accuracy of the carbon sink model's response to the micro-environment. Through measured carbon exchange data from flux towers, it directly verifies the carbon sink intensity retrieved by the satellite and UAV and calculates the carbon flux correction coefficients for different regions. The three-level decision information is used to form the final decision instruction through a weighted fusion algorithm, specifically including two types of outputs. First, the variable-based operational prescription model serves as the instruction guiding precision agriculture operations. Its generation logic is as follows: merging satellite NDVI data (reflecting overall vegetation needs), leaf nitrogen content retrieved from UAV hyperspectral data (reflecting nutrient requirements), and soil moisture and organic matter data from ground sensors (reflecting basic soil conditions); based on the crop growth model, establishing a mapping relationship between vegetation index, soil moisture, and fertilizer application rate, such as prioritizing irrigation and reducing nitrogen fertilizer application in areas with NDVI < 0.3 and soil moisture < 15%; applying the decision rules to a 5m × 5m grid cell to generate an instruction containing irrigation amount (L / m²). 2 ), Fertilizer application rate (kg / m³) 2The system includes: 1) Spatial distribution maps of pesticide types and dosages, directly integrated with the variable operation system of intelligent agricultural machinery; 2) Carbon flux accounting and assessment report, a core document for evaluating the effectiveness of carbon sink regulation. Its generation logic is as follows: Monthly average carbon flux at a regional scale (1km×1km) is retrieved based on satellite data to obtain preliminary results of the total regional carbon sink; biomass increment calculated using UAV lidar data and measured data from ground flux towers are used to perform stratified calibration of the satellite results (e.g., different correction coefficients are used for forest land, grassland, and cultivated land); the output report includes the total carbon sink and error range, a spatial distribution map of carbon sink intensity, analysis of key influencing factors, and recommendations for regulation measures. The regulation recommendations must clearly define the carbon sink enhancement potential of different sub-regions. This hierarchical fusion mechanism achieves decision-making accuracy from macro to micro and problem diagnosis from phenomenon to cause. Satellite data ensures the spatial integrity of decision-making, avoiding the omission of large-scale carbon sequestration anomaly areas; UAV data solves the problem of unclear satellite images, enabling precise target area positioning; ground data solves the problem of inaccurate measurement in the former two, providing quantitative verification benchmarks; the final generated comprehensive decision-making instructions are both macro-level guiding, covering the entire target area, and micro-level operable, accurate to the operational parameters of specific plots, providing direct action basis for carbon sequestration regulation in intelligent land engineering.

[0032] Based on the built-in carbon sink accounting model, the carbon sink or carbon source status of the target area is inverted and calculated in real time, generating a spatiotemporal distribution map of carbon sinks, including: The carbon sink accounting model is driven by spatialized data provided by satellites and drones as input to simulate the first simulated value of carbon flux in the target area. Obtain the measured values ​​of carbon dioxide flux from IoT sensors within the same spatiotemporal range; The difference between the simulated value of the first carbon flux and the measured value of the carbon dioxide flux is calculated. Based on the difference, the key biophysical parameters of the carbon sink accounting model are optimized and adjusted in reverse through the data assimilation algorithm. The key biophysical parameters include the maximum carboxylation rate, the maximum photosynthetic rate, or the dark respiration coefficient. The carbon sink accounting model was run again using the optimized parameters to generate a second simulated carbon flux value. The carbon sink, carbon source status and intensity information of each spatial grid unit in the target area obtained through data assimilation and inversion are associated with geospatial coordinates to obtain a spatiotemporal distribution map of carbon sinks. The second simulated carbon flux value serves as the net ecosystem exchange in the target area. The determination of net ecosystem exchange includes: if the simulated value of the second carbon flux is less than zero, the corresponding area is determined to be a carbon sink; if the simulated value of the second carbon flux is greater than zero, the corresponding area is determined to be a carbon source.

[0033] Specifically, spatialized data is input into the built-in carbon sink accounting model to simulate the first carbon flux simulation value for the target area, in gC·m³. -2 ·d -1 By acquiring measured values ​​of carbon dioxide flux within the same spatiotemporal range using IoT sensors, the deviation between simulated and measured values ​​is calculated. Key biophysical parameters were obtained, showing that the maximum carboxylation rate affects photosynthetic efficiency. The maximum photosynthetic rate reflects the upper limit of vegetation productivity. The dark respiration coefficient regulates the intensity of nighttime respiration. The optimization objective was to minimize the root mean square error between simulated and measured values, ensuring that parameter adjustments are ecologically sound. The optimized parameters were used to generate a second simulated carbon flux value (NEE). sim2 As net ecosystem exchange volume Where NPP is net primary productivity, R het This is heterotrophic respiration. When the carbon sink (NEE) is less than 0, the ecosystem absorbs CO2, such as in forests and wetlands. When the carbon source (NEE) is greater than 0, the ecosystem releases CO2, such as in farmland and urban areas.

[0034] It is worth noting that the carbon sink accounting model uses spatialized data provided by satellites and UAVs as its core input to achieve a preliminary simulation of regional carbon flux: satellite remote sensing data, including NDVI / EVI, surface temperature, and solar radiation, is used to calculate the photosynthetically active radiation absorbed by vegetation; UAV remote sensing data provides high-resolution leaf area index, vegetation height, and canopy structure parameters to optimize the model's characterization of the three-dimensional structure of vegetation; meteorological data (temperature, precipitation, humidity), soil type spatial distribution maps, etc., are generated through spatial interpolation. Based on photosynthetic physiological mechanisms, the model converts the input data into total primary productivity of vegetation, and then subtracts ecosystem respiration, including autotrophic respiration of vegetation and heterotrophic respiration of soil, to obtain the first simulated carbon flux value, namely the preliminary net ecosystem exchange (NEE), which is presented spatially in the form of grid cells (e.g., 100m × 100m). Ground-based IoT measured data serve as the true benchmark to verify the model's simulation accuracy. Measured NEE values ​​within the same spatiotemporal range are obtained through eddy covariance flux towers, with a time resolution typically of 30 minutes, covering the complete carbon exchange process of the sunrise-sunset cycle. The simulated and measured values ​​of the first carbon flux at the location of the flux tower are spatiotemporally matched (e.g., compared to daily averages). The absolute error (|simulated value - measured value|) and relative error (|simulated value - measured value| / measured value) are calculated to identify time periods or regions with significant model simulation deviations. Simultaneously, key parameters are optimized: the maximum carboxylation rate affects the carbon fixation efficiency of vegetation photosynthesis, and adjustments towards the measured value are needed when the error is large; the maximum photosynthetic rate determines the upper limit of vegetation carbon absorption under specific environmental conditions; the dark respiration coefficient affects the calculation of nocturnal carbon release from vegetation and is related to temperature-sensitive parameters. The difference between the simulated and measured values ​​is used as the objective function, and the parameter combination that minimizes the error is found through iterative calculations, forming the optimized parameter set. For example, when the simulated carbon absorption is less than the measured value, i.e., the simulated carbon sink capacity is weak, it may be necessary to increase the maximum carboxylation rate parameter value. After parameter optimization, the simulated second carbon flux output by the model is the final net ecosystem exchange, and its error with the measured value can typically be reduced by 30%-50%, more accurately reflecting the regional carbon exchange status. The carbon sink / carbon source determination rule is as follows: if the simulated value of the second carbon flux is <0, it indicates that the amount of carbon dioxide absorbed by the ecosystem is greater than the amount released, and it is determined to be a carbon sink state; if the simulated value of the second carbon flux is >0, it indicates that the amount of carbon dioxide released by the ecosystem is greater than the amount absorbed, and it is determined to be a carbon source state; the larger the absolute value, the stronger the carbon sink capacity or the higher the carbon source intensity.

[0035] Based on the aforementioned spatiotemporal distribution map of carbon sequestration, the limiting factors of carbon sequestration capacity are diagnosed, including: Based on the spatiotemporal distribution map of carbon sinks and the multi-source environmental factor data, a diagnostic dataset is constructed with environmental factors as features and carbon sink values ​​as labels. Establish a mapping relationship between environmental factors and carbon sink values; Calculate the importance score of each environmental factor among the multi-source environmental factors; Based on the importance score and the values ​​of each environmental factor, the limiting factors of carbon sequestration capacity within the target area and their spatial distribution are diagnosed.

[0036] Specifically, all factors are resampled to match the spatiotemporal distribution of carbon sinks. Figure 1 A consistent grid scale is used. If the carbon sink value is interannual data, then environmental factors should be multi-year averages or key quarterly averages. Feature and label pairing is performed: feature X is a multi-source environmental factor matrix (n grid cells × m factors). Label Y is the corresponding grid's carbon sink value (NEE). sim2 Negative values ​​indicate carbon sequestration intensity. The determination of limiting factors requires that they rank within the top 30% of importance and that their actual values ​​exceed the ecologically suitable range. For example, soil pH < 5.5 (acidity inhibits microbial activity) or annual precipitation < 400 mm (drought limits photosynthesis). Major limiting factors are those affecting more than 20% of the area, such as soil organic matter deficiency. Minor limiting factors are locally significant but have a small impact range, such as frequent forest fires.

[0037] The quality of the diagnostic dataset directly determines the accuracy of limiting factor identification. Its construction requires precise matching between environmental factors and carbon sink values. The basic sample unit is a grid cell (e.g., 50m × 50m) representing the spatiotemporal distribution of carbon sinks. Each cell contains two parts of information: feature variables and label variables. Feature variables are multi-source environmental factor data, including vegetation characteristics (LAI, vegetation type, chlorophyll content), soil parameters (organic matter content, pH, nitrogen, phosphorus, potassium content, water content), meteorological data (annual precipitation, accumulated temperature, sunshine hours), and topographic factors (slope, aspect, altitude). The label variable is the carbon sink value of that grid cell (i.e., the absolute value of the second carbon flux simulation, reflecting the intensity of carbon sink capacity). Data standardization (e.g., Z-score standardization) is performed to eliminate dimensional differences. K-means clustering or spatial stratified sampling methods are used to ensure the representativeness of the samples in terms of carbon sink value gradient and spatial distribution. Multiple imputation is used to handle missing values, avoiding diagnostic bias caused by missing key environmental factor data. This study quantifies the impact of environmental factors on carbon sequestration capacity using machine learning and statistical analysis. It captures the complex nonlinear relationship between environmental factors and carbon sequestration values, such as how soil nitrogen content promotes carbon sequestration within a certain range while inhibiting it when excessive. It identifies environmental factors with significant linear effects on carbon sequestration values, such as the positive correlation between annual precipitation and carbon sequestration. It analyzes the interactions between environmental factors, such as the synergistic effect of soil moisture and temperature on carbon sequestration. The study outputs a feature importance score, reflecting the factor's contribution to reducing the prediction error of carbon sequestration values ​​across all decision trees; a higher score indicates a greater impact on carbon sequestration capacity. By fixing other factors and individually changing the value of a single factor, the rate of change in carbon sequestration values ​​is calculated; a larger rate of change indicates higher sensitivity and importance of the factor. The study interprets the contribution of each factor to carbon sequestration values, using the mean absolute value of the SHAP score as the importance score. All environmental factor importance scores are normalized to 0-100 points, forming an intuitive importance ranking. Typically, the top 5-8 factors are selected as key influencing factors. By combining importance scores and actual factor values, key limiting factors and their spatial distribution that restrict carbon sequestration capacity are identified: For key factors with high importance scores, threshold and effect analysis are used to determine whether they are limiting factors. If a factor has an importance score ≥60, it is considered a core influencing factor, and if the actual value of the factor in a certain region is lower than its critical threshold for promoting carbon sequestration, such as soil organic carbon content <15g / kg and annual precipitation <400mm, then it is determined to be a carbon sequestration limiting factor in that region. If multiple limiting factors exist in the same region, they are ranked according to importance scores to identify 1-2 dominant limiting factors. For example, arid and semi-arid regions may be limited by both water and soil nutrients, and the water factor with a higher score is the dominant limiting factor.

[0038] When drawing spatial distribution maps, different colors are used to mark the dominant limiting factors in each region. For the same limiting factor, the limiting intensity is divided according to the difference between its actual value and the critical threshold. For example, a soil nitrogen content difference of more than 30% from the critical value is considered a severe limitation. The limiting factor distribution map is overlaid with the carbon sink spatiotemporal distribution map to verify the spatial consistency between the limiting factors and the low carbon sink area. The higher the consistency, the more reliable the diagnostic results.

[0039] Therefore, by identifying limiting factors with clear spatial orientation and targeted regulation, the intelligent decision-making module can prioritize the generation of precision irrigation strategies for water-limited areas; formulate variable fertilization plans for soil nutrient-limited areas; and propose tree species optimization and stand transformation suggestions for vegetation structure-limited areas, such as monoculture and unreasonable stand density. This realizes the carbon sequestration enhancement path of identifying where, what is limited, and how to regulate it, greatly improving the efficiency and targeting of land engineering regulation.

[0040] The multi-source environmental factor data includes meteorological factors, soil factors, vegetation factors, and topographic factors; The meteorological factors include one or more of temperature, precipitation, and photosynthetically active radiation; The soil factors include one or more of soil moisture, soil nitrogen content, and soil pH. The vegetation factors include one or more of leaf area index and vegetation cover. The terrain factors include one or more of elevation and slope.

[0041] Specifically, meteorological factors are fundamental drivers of the carbon exchange rate in ecosystems, directly influencing the intensity of photosynthesis and respiration in vegetation. Within a certain range (e.g., 5-25℃), rising temperatures accelerate photosynthetic enzyme activity, increasing the carbon absorption rate. However, temperatures exceeding the optimum, such as above 30℃ for most temperate plants, inhibit photosynthesis, exhibiting an exponential positive correlation. For every 10℃ increase in temperature, soil and vegetation respiration rates can increase by 1-2 times, directly impacting ecosystem carbon release. Precipitation indirectly regulates vegetation growth by affecting soil moisture. In arid and semi-arid regions, annual precipitation <400mm is the primary limiting factor for carbon sequestration capacity. Extreme precipitation can lead to decreased soil aeration, inhibiting root respiration and microbial activity, resulting in a short-term reduction in carbon release. Photosynthetically active radiation (PADR) is the energy source for vegetation photosynthesis, directly determining the upper limit of total primary productivity. In areas with high vegetation cover (e.g., forests), it is a key driver of carbon sequestration capacity. Significant spatial heterogeneity exists in mountainous areas due to the significant influence of cloud cover and topographic shading.

[0042] Soil factors are the material basis and environmental carrier of carbon sinks. Soil is the largest carbon pool in terrestrial ecosystems, and its physicochemical properties directly affect the processes of carbon fixation, storage, and release. Soil moisture affects the efficiency of root water absorption and nutrient uptake; the optimal moisture range (60-80% of field capacity) promotes vegetation growth and carbon absorption. Excessive moisture leads to anaerobic environments, promoting methane (a greenhouse gas) emissions; excessive moisture inhibits microbial activity and slows organic matter decomposition. Soil nitrogen content is a key element in chlorophyll and enzyme synthesis in vegetation. In nitrogen-limited areas (such as most temperate forests), increasing nitrogen content can significantly enhance photosynthetic rates and biomass accumulation. Excessive nitrogen input may lead to soil acidification and increased greenhouse gas (such as N2O) emissions, producing negative effects. Soil pH affects the structure and activity of soil microbial communities. Neutral to slightly acidic soils (pH 6.0-7.5) are generally conducive to organic matter decomposition and nutrient release. Strongly acidic (pH < 5.0) or strongly alkaline (pH > 8.5) soils inhibit root growth and carbon fixation efficiency.

[0043] Vegetation factors directly reflect carbon sequestration capacity, and vegetation structure parameters directly reflect carbon sequestration potential and status. Leaf area index (LAI), which is half the total leaf area per unit area, is positively correlated with photosynthetic rate within a certain range and is a key parameter for estimating total primary productivity. Excessively high LAI (e.g., LAI > 6) may lead to insufficient light penetration within the canopy, resulting in a self-shading effect and reducing light energy utilization efficiency. Vegetation cover, the vertical projection of the vegetation canopy onto the ground, directly determines the size of the regional carbon sequestration area and is a fundamental indicator for distinguishing between carbon sequestration and carbon source areas. It is related to factors such as soil erosion and surface temperature; high cover can reduce soil carbon loss.

[0044] Topographic factors indirectly shape the spatial differentiation of carbon sinks. Topography, by altering the spatial pattern of hydrothermal conditions and soil distribution, indirectly influences regional differences in carbon sink capacity. With increasing altitude, temperatures decrease and radiation increases, creating different climate zones, which directly affect vegetation type distribution (e.g., from tropical rainforest to alpine meadow), leading to vertical differentiation of carbon sink capacity. High-altitude areas typically have shallower soils and limited carbon storage capacity, but the low temperatures can slow down organic matter decomposition. Slope affects soil thickness and water retention capacity; steep slopes are prone to soil erosion, leading to the loss of soil carbon pools and limited vegetation growth. Gentle slopes typically have fertile soils and good water conditions, which are conducive to the formation of high carbon sink areas.

[0045] Meanwhile, these environmental factors do not act in isolation, but rather influence carbon sink capacity through complex synergistic effects. For example, in low-altitude, gentle slope areas, ample rainfall (meteorological factor) and fertile soil (soil factor) may support high LAI vegetation (vegetation factor), forming a strong carbon sink zone; while in high-altitude, steep slope areas, low temperature (meteorological factor), thin soil (soil factor), and low vegetation cover (vegetation factor) work together to form a weak carbon sink or carbon source zone. This multi-factor synergistic effect is the core reason for the heterogeneity of carbon sink distribution in time and space, and also the scientific basis for intelligent decision-making modules to formulate differentiated regulation strategies.

[0046] The implementation of the control strategy includes: When the limiting factor is water stress and the land type is farmland, the control strategy is precision irrigation or the construction of water collection facilities, wherein the irrigation amount is adjusted according to crop type, growth period and soil texture. When the limiting factor is nitrogen deficiency and the vegetation type is forest, the regulation strategy is to apply slow-release nitrogen fertilizer; When the limiting factor is soil acidification, the control strategy is to apply lime, wherein the amount of lime is calculated based on the soil pH and the target pH. Lime requirement (t / ha) = (target pH - current pH) × soil cation exchange capacity × adjustment coefficient; When the limiting factor is low vegetation cover and steep terrain, the control strategy is horizontal terrace land preparation and planting soil-stabilizing vegetation.

[0047] Specifically, when water is diagnosed as the main limiting factor for farmland carbon sequestration, irrigation quotas are calculated based on the water requirements of crop type (e.g., wheat, corn, rice), the water sensitivity of growth stage (e.g., jointing stage, grain-filling stage), and soil texture (sandy soil has low water holding capacity and requires frequent, small-volume irrigation, while clay soil has high water holding capacity and requires frequent, large-volume irrigation). Irrigation is triggered when the soil moisture content in the 0-60cm layer, monitored in real-time by soil sensors, falls below 60% of field capacity, enabling precise application of water and fertilizer through an intelligent irrigation system. For areas experiencing seasonal drought or uneven rainfall, small water collection facilities such as fish-scale pits and drainage ditches are constructed at the edges of farmland to improve rainwater utilization efficiency. The ratio of catchment area to irrigated area is no less than 1:5, ensuring that a moderate rainfall (10-20mm) can meet the needs of 1-2 irrigations. When forest carbon sinks are limited by nitrogen, a slow-release nitrogen fertilizer application scheme should be adopted, prioritizing slow-release nitrogen fertilizers such as coated urea and urea-formaldehyde. The nitrogen release cycle should be matched with the tree growing season to reduce nitrogen leaching loss. In young forests, ring trench application should be used, digging a ring trench 20-30 cm deep 50-100 cm around the tree trunk, applying the fertilizer evenly, and then covering with soil. In mature forests, broadcasting combined with shallow tillage should be used, spreading the fertilizer evenly in the forest and then tilling to a depth of 5-10 cm to promote root absorption. The application rate should be determined based on the leaf nitrogen content obtained from UAV hyperspectral inversion and the measured soil nitrogen content. Generally, the annual nitrogen application rate should not exceed 50 kg / ha for coniferous forests and 80 kg / ha for broadleaf forests. To address carbon sequestration limitations caused by low soil pH (typically pH < 5.5), lime application is used for adjustment. The lime requirement (t / ha) is calculated as: (target pH - current pH) × soil cation exchange capacity × adjustment coefficient. The target pH is typically set at 6.5-7.0 for farmland and 6.0-6.5 for forest land. The adjustment coefficient is 0.8-1.0 for sandy soil, 1.0-1.2 for loamy soil, and 1.2-1.5 for clay soil. For example, if the current pH is 5.0, the target pH is 6.5, the cation exchange capacity is 15 cmol / kg, and the adjustment coefficient for clay soil is 1.3, then the lime requirement is (6.5 - 5.0) × 15 × 1.3 = 29.25 t / ha. For farmland, apply the lime evenly during tillage, then rotary till to a depth of 15-20 cm to ensure thorough mixing with the soil. For forest land, apply the lime in holes or strips, avoiding direct contact with tree roots, and lightly till the topsoil after application. To address the low vegetation cover in steep slopes (>25°), a combined engineering and biological approach was adopted. Horizontal terraces, 1-1.5m wide and 0.5-0.8m deep, were excavated along contour lines, with a spacing of 2-3m between terraces. Topsoil was retained within each terrace, and 20-30cm high earthen embankments were constructed on the outer sides to retain water. This reduces soil erosion, improves soil water and fertilizer retention capacity, and creates basic conditions for vegetation restoration. Priority was given to drought-tolerant, barren-tolerant, and well-developed native species, such as alfalfa and *Alternanthera philoxeroides* (legumes) (for nitrogen fixation and soil improvement), paired with shrubs such as sea buckthorn and *Caragana korshinskii*.Planting methods include hole sowing or seedling planting within horizontal rows, with a plant spacing of 1×1m and a sowing rate of 15-20kg / ha to ensure that the vegetation coverage reaches more than 60% in the same year.

[0048] In areas with multiple limiting factors, combined measures should be adopted. For example, in farmland with dual limitations of water and nitrogen, precision irrigation and slow-release nitrogen fertilizer application can be implemented simultaneously. After all measures are implemented, monthly monitoring should be carried out through ground-based Internet of Things and drone remote sensing to evaluate the control effect and provide a basis for subsequent strategy optimization. Eco-friendly measures should be given priority, such as replacing some chemical fertilizers with organic fertilizers and giving priority to local species, to avoid secondary disturbance to the ecosystem caused by control measures.

[0049] See Figure 4 A smart land engineering carbon sequestration regulation method based on dynamic management includes: Acquire multi-source spatiotemporal data of the target region; The multi-source spatiotemporal data are fused and processed, and the carbon sink status of the target area is calculated in real time based on the carbon sink accounting model to generate a carbon sink spatiotemporal distribution map. Based on the aforementioned spatiotemporal distribution map of carbon sinks, artificial intelligence optimization algorithms are used to diagnose the limiting factors of carbon sink capacity and generate land engineering regulation strategies. The control strategy is executed, and the control model is iteratively optimized based on the feedback data after control.

[0050] Specifically, satellite remote sensing data acquires large-scale data on vegetation indices (NDVI / EVI), land use types, and surface temperature, reflecting macro-scale vegetation growth and environmental conditions. Unmanned aerial vehicle (UAV) remote sensing data collects high-resolution imagery, lidar point cloud data, and hyperspectral data of key areas, obtaining three-dimensional vegetation structure, tree species composition, and soil surface characteristics. Ground-based IoT data, through soil sensors, weather stations, and carbon flux towers, monitors soil temperature and humidity, nitrogen content, pH value, meteorological parameters, and carbon dioxide exchange flux in real time. These data possess multi-scale and multi-dimensional characteristics, with temporal resolutions ranging from minutes to days / weeks, providing comprehensive input for subsequent analysis. Systematic processing of multi-source data accurately inverts carbon sink status, performing data cleaning, spatiotemporal alignment, and format standardization to unify coordinate systems and time scales. Through data-level, feature-level, and decision-level fusion, a spatiotemporally continuous multi-source environmental factor dataset is generated, including meteorological, soil, vegetation, and topographic factors. The fused spatialized data is input into the carbon sink accounting model to obtain the first simulated carbon flux value. Combined with ground-measured carbon dioxide flux data, model parameters (such as maximum photosynthetic rate and dark respiration coefficient) are optimized using algorithms such as Kalman filtering. Based on the optimized second simulated carbon flux value, a spatial distribution map containing the state and intensity of carbon sinks / carbon sources in each grid cell is generated. Based on the spatiotemporal distribution characteristics of carbon sinks, a sample database is established using environmental factors as features and carbon sink values ​​as labels. Algorithms such as random forest and gradient boosting tree are used to establish the mapping relationship between environmental factors and carbon sink values. Through feature importance scoring and sensitivity analysis, key limiting factors of carbon sink capacity (such as water stress and nitrogen deficiency) and their spatial distribution are identified. Based on the diagnostic results, and considering land type, vegetation type, and topographic features, differentiated regulation strategies are generated. In farmland areas experiencing water stress, precision irrigation plans are developed, adjusting irrigation amounts according to crop type and soil texture. In forest areas lacking nitrogen, slow-release nitrogen fertilizer application plans are designed, determining appropriate fertilization methods and dosages. In soil acidification areas, lime requirements are calculated, and scientific application plans are developed. In steep slopes with low cover, leveling projects and soil-stabilizing vegetation planting schemes are planned. Furthermore, a closed-loop feedback mechanism continuously improves the regulation effect, translating the regulation strategies into actionable instructions and distributing them to execution terminals to guide precise operations by agricultural machinery or ecological engineering. Multi-source data after regulation is collected through a sensing network, including vegetation growth status, soil parameter changes, and carbon flux dynamics. Feedback data is input into the system to evaluate the implementation effect of the regulation strategies. Artificial intelligence algorithms are used to optimize the parameters of the carbon sink accounting model and the regulation decision-making model, improving the accuracy and relevance of subsequent decisions. Through continuous data updates and model iterations, refined management of land engineering carbon sinks throughout their entire lifecycle is achieved, continuously improving the efficiency and scientific rigor of carbon sink regulation.

[0051] The above description is merely a few embodiments of this application and is not intended to limit this application in any way. Although this application discloses preferred embodiments as described above, it is not intended to limit this application. Any changes or modifications made by those skilled in the art without departing from the scope of the technical solution of this application using the disclosed technical content are equivalent to equivalent implementation cases and fall within the scope of the technical solution.

Claims

1. A smart land engineering carbon sequestration regulation system based on dynamic management, characterized in that, include: The sensing module is used by the integrated monitoring network to acquire multi-source spatiotemporal data of the target area, including vegetation index, soil parameters, meteorological data and carbon dioxide flux data. The data processing and analysis module is used to fuse the multi-source spatiotemporal data and, based on the built-in carbon sink accounting model, to invert and calculate the carbon sink or carbon source status of the target area in real time, and generate a spatiotemporal distribution map of carbon sink. The intelligent decision-making module is used to diagnose the limiting factors of carbon sink capacity based on the spatiotemporal distribution map of carbon sinks, and generate land engineering regulation strategies for different sub-regions; the regulation strategies include at least one of tree species optimization, precision water and fertilizer, and ecological restoration engineering. The execution and feedback module is used to send the control strategy to the execution terminal or management platform and to receive the control data fed back by the sensing module.

2. The intelligent land engineering carbon sequestration regulation system based on dynamic management according to claim 1, characterized in that, The sensing module includes: Satellite remote sensing units are used to acquire large-scale, periodic remote sensing images and generate satellite remote sensing data. The UAV remote sensing unit is used to acquire lidar and / or spectral data of key areas and generate UAV remote sensing data. The ground-based Internet of Things (IoT) unit includes soil sensors, weather stations, and carbon dioxide flux towers deployed in the target area, generating ground-based IoT data.

3. The intelligent land engineering carbon sequestration regulation system based on dynamic management according to claim 2, characterized in that, The fusion processing of the multi-source spatiotemporal data includes: The multi-source spatiotemporal data is preprocessed and spatiotemporally aligned to unify it to the same spatiotemporal reference. The preprocessed multi-source spatiotemporal data is input into a multi-source data fusion model for data-level, feature-level, or decision-level fusion processing to generate multi-source environmental factor data for the target area. Based on the multi-source environmental factor data, monitoring, assessment, or decision-making information for the target area is output.

4. The intelligent land engineering carbon sequestration regulation system based on dynamic management according to claim 3, characterized in that, The preprocessed multi-source spatiotemporal data is fused at the decision level, including: Based on satellite remote sensing data, first-level decision information is obtained, which includes crop growth zoning of the target area, identification of drought stress areas, or preliminary assessment of carbon sequestration potential. Based on UAV remote sensing data, specific areas in the first-level decision information are identified and verified to obtain second-level decision information, which includes precise location of pests and diseases, inversion of plant-level physiological state, or verification of soil surface anomalies. Based on ground-based IoT data, third-level decision information is generated, including real-time soil moisture alerts, microclimate monitoring, or carbon dioxide flux change verification. By integrating the decision information from the first, second, and third levels, a comprehensive decision instruction is generated; The comprehensive decision-making instruction is as follows: A variable operation prescription model is used to guide agricultural machinery and equipment to perform precision irrigation, variable fertilization or pesticide spraying operations in the target area. The generation of the variable operation prescription model depends at least on the fused vegetation index, soil parameters and meteorological data. Alternatively, a carbon flux accounting and assessment report, which is generated based on preliminary results of regional carbon flux inversion from satellite data and verified and calibrated using carbon dioxide flux data collected by drones and the Internet of Things.

5. The intelligent land engineering carbon sequestration regulation system based on dynamic management according to claim 1, characterized in that, Based on the built-in carbon sink accounting model, the carbon sink or carbon source status of the target area is inverted and calculated in real time, generating a spatiotemporal distribution map of carbon sinks, including: The carbon sink accounting model is driven by spatialized data provided by satellites and drones as input to simulate the first simulated value of carbon flux in the target area. Obtain the measured values ​​of carbon dioxide flux from IoT sensors within the same spatiotemporal range; The difference between the simulated value of the first carbon flux and the measured value of the carbon dioxide flux is calculated. Based on the difference, the key biophysical parameters of the carbon sink accounting model are optimized and adjusted in reverse through the data assimilation algorithm. The key biophysical parameters include the maximum carboxylation rate, the maximum photosynthetic rate, or the dark respiration coefficient. The carbon sink accounting model was run again using the optimized parameters to generate a second simulated carbon flux value. The carbon sink, carbon source status and intensity information of each spatial grid unit in the target area obtained through data assimilation and inversion are associated with geospatial coordinates to obtain a spatiotemporal distribution map of carbon sinks. The second simulated carbon flux value serves as the net ecosystem exchange in the target area. The determination of net ecosystem exchange includes: if the simulated value of the second carbon flux is less than zero, the corresponding area is determined to be a carbon sink; if the simulated value of the second carbon flux is greater than zero, the corresponding area is determined to be a carbon source.

6. The intelligent land engineering carbon sequestration regulation system based on dynamic management according to claim 3, characterized in that, Based on the aforementioned spatiotemporal distribution map of carbon sequestration, the limiting factors of carbon sequestration capacity are diagnosed, including: Based on the spatiotemporal distribution map of carbon sinks and the multi-source environmental factor data, a diagnostic dataset is constructed with environmental factors as features and carbon sink values ​​as labels. Establish a mapping relationship between environmental factors and carbon sink values; Calculate the importance score of each environmental factor among the multi-source environmental factors; Based on the importance score and the values ​​of each environmental factor, the limiting factors of carbon sequestration capacity within the target area and their spatial distribution are diagnosed.

7. The intelligent land engineering carbon sequestration regulation system based on dynamic management according to claim 6, characterized in that, The multi-source environmental factor data includes meteorological factors, soil factors, vegetation factors, and topographic factors; The meteorological factors include one or more of temperature, precipitation, and photosynthetically active radiation; The soil factors include one or more of soil moisture, soil nitrogen content, and soil pH. The vegetation factors include one or more of leaf area index and vegetation cover. The terrain factors include one or more of elevation and slope.

8. The intelligent land engineering carbon sequestration regulation system based on dynamic management according to claim 6, characterized in that, After calculating the importance score of each environmental factor among the multi-source environmental factors, the process also includes: For a specific grid cell within the target area, calculate the contribution of each environmental factor to the carbon sink value of that specific grid cell; Environmental factors with the largest negative contribution value are diagnosed as the primary limiting factors for the specific grid cell. Based on the statistical and spatial visualization of the primary constraint factors of all grid cells, a spatial distribution map of constraint factors for the target region is generated.

9. The intelligent land engineering carbon sequestration regulation system based on dynamic management according to claim 1, characterized in that, The implementation of the control strategy includes: When the limiting factor is water stress and the land type is farmland, the control strategy is precision irrigation or the construction of water collection facilities, wherein the irrigation amount is adjusted according to crop type, growth period and soil texture. When the limiting factor is nitrogen deficiency and the vegetation type is forest, the regulation strategy is to apply slow-release nitrogen fertilizer; When the limiting factor is soil acidification, the control strategy is to apply lime, wherein the amount of lime is calculated based on the soil pH and the target pH. Wherein, lime requirement (t / ha) = (target pH - current pH) × soil cation exchange capacity × adjustment coefficient; When the limiting factor is low vegetation cover and steep terrain, the control strategy is horizontal terrace land preparation and planting soil-stabilizing vegetation.

10. A method for regulating carbon sequestration in intelligent land engineering based on dynamic management, characterized in that, include: Acquire multi-source spatiotemporal data of the target region; The multi-source spatiotemporal data are fused and processed, and the carbon sink status of the target area is calculated in real time based on the carbon sink accounting model to generate a carbon sink spatiotemporal distribution map. Based on the aforementioned spatiotemporal distribution map of carbon sinks, artificial intelligence optimization algorithms are used to diagnose the limiting factors of carbon sink capacity and generate land engineering regulation strategies. The control strategy is executed, and the control model is iteratively optimized based on the feedback data after control.

Images