Modeling method and system applied to ecological system carbon library monitoring under influence of hydropower engineering

Abstract

The invention provides a modeling method and system applied to ecological system carbon library monitoring under influence of water-power engineering, and the method comprises the steps: obtaining a drainage basin geographic information data set and a water-power engineering operation data set, and carrying out the water flow dynamic feature analysis of the drainage basin geographic information data set and the water-power engineering operation data set; the method comprises the following steps: obtaining a watershed water flow dynamic feature vector set, deducing soil physicochemical property offset and vegetation community structure variation according to the watershed water flow dynamic feature vector set, coupling the soil physicochemical property offset and the vegetation community structure variation to carry out carbon cycle process digital quantitative modeling, and generating a carbon reservoir reserve change sequence. And constructing an ecological system carbon library monitoring model under the influence of the hydropower engineering based on the carbon library reserve change sequence. According to the method, the accuracy of water-power engineering ecosystem carbon sink effect evaluation and prediction is effectively improved, and a more reliable basis is provided for related decisions.

Description

Technical Field

[0001] This invention relates to the field of data processing, and more specifically, to a modeling method and system for monitoring the carbon pool of an ecosystem under the influence of hydropower projects. Background Technology

[0002] Ecosystem carbon pool monitoring is a crucial process for assessing ecosystem carbon cycling processes and dynamic changes in carbon storage. In the context of hydropower project construction and operation, it is particularly significant for understanding the impact of engineering activities on the carbon balance of watershed ecosystems. Currently, monitoring of ecosystem carbon pools under the influence of hydropower projects involves sampling and observing carbon pool components such as soil, vegetation, and water bodies, or estimating based on empirical models. These methods often fail to systematically capture the comprehensive impact of complex hydrological processes triggered by hydropower project operation on various components of the ecosystem carbon pool. This results in an incomplete quantitative characterization of carbon pool dynamic changes, making it difficult to accurately reflect the true process of ecosystem carbon cycling under the influence of hydropower projects. Consequently, biases exist in the assessment and prediction of the ecosystem carbon sink effect of hydropower projects, affecting the accuracy and effectiveness of related decision-making. Summary of the Invention

[0003] This invention provides a modeling method and system for monitoring the carbon pool of an ecosystem under the influence of hydropower projects.

[0004] In a first aspect, embodiments of the present invention provide a modeling method for monitoring ecosystem carbon pools under the influence of hydropower projects. The method includes: acquiring a watershed geographic information dataset and a hydropower project operation dataset, wherein the watershed geographic information dataset represents topographic features, vegetation cover, and soil type, and the hydropower project operation dataset represents reservoir scheduling, dam discharge, and water level fluctuations; performing hydrodynamic feature analysis on the watershed geographic information dataset and the hydropower project operation dataset to obtain a watershed hydrodynamic feature vector set, wherein the watershed hydrodynamic feature vector set represents water flow velocity distribution, sediment transport, and water level fluctuations; and performing hydrodynamic feature analysis based on the watershed hydrodynamic features. The feature vector set is used to extrapolate soil physicochemical property shifts and vegetation community structure variability. The soil physicochemical property shifts characterize the changes in soil organic carbon content and bulk density, while the vegetation community structure variability characterizes the changes in vegetation biomass and type. The soil physicochemical property shifts and vegetation community structure variability are coupled to perform digital quantitative modeling of the carbon cycle process, generating a carbon pool storage change sequence. This sequence characterizes the storage changes in soil carbon pools, vegetation carbon pools, and water body carbon pools. Based on this carbon pool storage change sequence, an ecosystem carbon pool monitoring model under the influence of hydropower projects is constructed. This model is used to output carbon pool storage prediction curves under different hydrological scenarios.

[0005] Secondly, embodiments of the present invention provide a computer system, comprising: a memory storing a computer program; and a processor for loading the computer program to implement the modeling method described above for monitoring ecosystem carbon pools under the influence of hydropower projects.

[0006] This invention acquires watershed geographic information datasets and hydropower project operation datasets, performs hydrodynamic feature analysis on both to obtain a watershed hydrodynamic feature vector set, and infers soil physicochemical property shifts and vegetation community structure variability based on this feature vector set. Coupled with these shifts and variability, a digital quantitative model of the carbon cycle process is generated to produce a carbon pool storage change sequence, ultimately constructing an ecosystem carbon pool monitoring model under the influence of hydropower projects. This method uses hydrodynamic feature analysis as a key bridge to organically link hydropower project operation data with the ecosystem carbon cycle process, achieving a full-chain quantitative characterization from project operation data to carbon pool storage changes. It can effectively capture the dynamic impact of hydropower project operation on the ecosystem carbon pool. By comprehensively considering the synergistic changes of multiple carbon pools in soil, vegetation, and water, the generated carbon pool storage change sequence more comprehensively reflects the ecosystem carbon cycle process. The constructed monitoring model can output carbon pool storage prediction curves under different hydrological scenarios, providing strong technical support for the dynamic monitoring and management of ecosystem carbon pools under the influence of hydropower projects. It helps to more scientifically assess the impact of hydropower projects on ecosystem carbon cycling, effectively improves the accuracy of assessment and prediction of the ecosystem carbon sink effect of hydropower projects, and provides a more reliable basis for relevant decision-making. Attached Figure Description

[0007] Figure 1 This is a flowchart of a modeling method for monitoring the carbon pool of an ecosystem under the influence of hydropower projects, provided by an embodiment of the present invention.

[0008] Figure 2 This is a schematic diagram of the composition of a computer system provided in an embodiment of the present invention. Detailed Implementation

[0009] Please see Figure 1 This is a flowchart of a modeling method for monitoring ecosystem carbon pools under the influence of hydropower projects, provided by an embodiment of the present invention. The method can be executed by a computer system and may include the following steps: Step S100: Obtain the watershed geographic information dataset and the hydropower project operation dataset. The watershed geographic information dataset represents the topography, vegetation cover and soil type, while the hydropower project operation dataset represents the reservoir scheduling, dam discharge and water level fluctuation.

[0010] Watershed geographic information datasets describe the natural geographic features within a watershed. Topographic features include the undulating terrain such as mountains, valleys, and plains, as well as the distribution of rivers and lakes. Vegetation cover reflects the types, quantities, and distribution of vegetation within the watershed. Soil type describes the soil texture, pH, and other characteristics of different areas within the watershed. Hydropower project operation datasets record the operational status of hydropower projects. Reservoir scheduling includes information on reservoir storage and release times and flow rates. Dam discharge status involves the location, flow rate, and direction of dam discharge. Water level fluctuation records changes in reservoir water levels over time.

[0011] When acquiring watershed geographic information datasets, satellite remote sensing technology can be used to obtain elevation data of topography and vegetation cover imagery, and soil type conditions can be determined through field sampling and analysis. For hydropower project operation datasets, real-time data on reservoir scheduling, dam discharge, and water level fluctuations can be obtained from the hydropower project's monitoring system. Taking a specific watershed as an example, the topography and vegetation cover information of the watershed can be obtained through satellite remote sensing image interpretation. At the same time, soil samples are taken from different areas within the watershed to analyze its texture, pH, and other characteristics, thereby constructing a watershed geographic information dataset. The hydropower project operation data is exported from the hydropower project's automated monitoring system, including the reservoir's daily scheduling plan, the specific time and flow rate of dam discharge, and hourly water level changes.

[0012] Step S200: Perform hydrodynamic feature analysis on the watershed geographic information dataset and the hydropower project operation dataset to obtain the watershed hydrodynamic feature vector set. The watershed hydrodynamic feature vector set represents the distribution of water flow velocity, sediment transport, and water level fluctuation.

[0013] Hydrodynamics analysis is the process of analyzing and calculating the laws governing water flow within a watershed, aiming to extract flow-related features from watershed geographic information and hydropower project operation data. Water velocity distribution describes the magnitude and direction of water flow velocity at different locations within the watershed; sediment transport reflects the quantity, particle composition, and transport path of sediment carried by the water flow; and water level fluctuation reflects the range of water level changes over time in different water areas.

[0014] In one implementation, step S200 may specifically include the following steps S210 to S280: Step S210: Spatial discretization of the watershed geographic information dataset is performed according to topographic partitions to generate a hierarchical grid attribute matrix containing topographic elevation information, vegetation cover degree and soil texture identifier. The grid resolution of different topographic regions in the hierarchical grid attribute matrix is ​​different.

[0015] Spatial discretization divides a continuous watershed geographic information dataset into discrete grid cells for subsequent calculations and analysis. Topographic zoning divides the watershed into different regions based on its topographic features, such as mountains, plains, and valleys. A hierarchical grid attribute matrix is ​​a matrix used to store attribute information for each grid cell, where each cell contains topographic elevation information, vegetation cover, and soil texture information. Different grid resolutions exist for different topographic regions to better adapt to varying terrain complexities. For example, in mountainous areas with dramatic topographic changes, a higher grid resolution can more accurately describe terrain features, while in relatively flat plains, a lower grid resolution is sufficient.

[0016] When performing spatial discretization, a regular grid partitioning method can be used to divide the watershed into several square or rectangular grids of equal size. For each grid cell, its topographic elevation, vegetation cover, and soil texture identifier are obtained from the watershed geographic information dataset using an interpolation algorithm.

[0017] Step S220: Perform time series integration on the hydropower project operation dataset to generate reservoir scheduling time series data, dam discharge time series data, and water level fluctuation time series data with unified time stamps, and keep the time intervals of different types of time series data consistent.

[0018] Time series integration involves unifying the processing of data from different time scales and formats within a hydropower project operation dataset, ensuring they share the same time stamps and time intervals. Reservoir scheduling time series data records the reservoir's operation at different times, including the flow and timing of water storage and release; dam discharge time series data reflects the time, location, and flow rate of dam discharge; and water level fluctuation time series data records changes in reservoir water levels over time. Maintaining consistent time intervals across different types of time series data facilitates subsequent data analysis and calculations, ensuring accurate time correspondences between different data sets.

[0019] When integrating time series data, methods such as linear interpolation or spline interpolation can be used to unify data from different time scales to the same time interval.

[0020] Step S230: Spatial correlation calculation is performed between the data of different terrain regions in the hierarchical grid attribute matrix and the corresponding reservoir scheduling time series data to generate the initial boundary conditions of the flow field for different terrain regions. The initial boundary conditions of the flow field represent the spatial matching relationship between the terrain and the scheduling data.

[0021] Spatial correlation calculation involves mapping and calculating topographic region data in a hierarchical grid attribute matrix with reservoir scheduling time-series data to determine the initial state of the flow field in different topographic regions. The initial boundary conditions of the flow field describe information such as velocity, direction, and pressure of the water flow at the initial moment, reflecting the spatial matching relationship between the topography and reservoir scheduling data. For example, in mountainous terrain, reservoir release may lead to an increase in water flow velocity, while in plains terrain, the water flow velocity may be relatively slow. Through spatial correlation calculation, the initial boundary conditions of the flow field in different topographic regions can be accurately determined, providing accurate initial states for subsequent hydrodynamic calculations.

[0022] In one implementation, step S230 may specifically include the following steps S231 to S236: Step S231: Extract the terrain elevation information of the target terrain region from the layered grid attribute matrix, calculate the ratio of the elevation difference to the horizontal distance between adjacent grid cells, and generate the slope matrix of the terrain region. Each element of the slope matrix corresponds to the average slope of the grid cell.

[0023] Topographic elevation information is the altitude data of each grid cell in a hierarchical grid attribute matrix, reflecting the undulation of the terrain. The ratio of the elevation difference between adjacent grid cells to the horizontal distance is the slope, which describes the degree of inclination of the terrain. A slope matrix is ​​a matrix used to store the slope information of each grid cell, where each element corresponds to the average slope of a grid cell.

[0024] When extracting topographic elevation information for a target terrain region, the corresponding terrain region can be selected based on the terrain partition identifier in the hierarchical grid attribute matrix. Then, for each grid cell within this terrain region, its elevation difference and horizontal distance from adjacent grid cells are calculated. Taking a mountainous terrain region as an example, assuming the hierarchical grid attribute matrix of this region contains multiple grid cells, for one grid cell, the elevation difference is calculated by comparing its elevation values ​​with those of its four adjacent grid cells. Simultaneously, the horizontal distance is determined based on the size of the grid cell, and the slope of that grid cell is obtained by dividing the elevation difference by the horizontal distance. This process is repeated for all grid cells within the terrain region to generate a slope matrix.

[0025] Step S232: Perform seasonal pattern separation processing on the outflow data in the reservoir scheduling time series data, extract the outflow characteristics of different seasons, and generate seasonal flow characteristic values. The seasonal flow characteristic values ​​represent the flow differences in different seasons.

[0026] Seasonal pattern separation involves dividing and analyzing reservoir outflow data according to seasons within the reservoir's time-series scheduling data to extract flow characteristics for different seasons. Outflow data records the water release from the reservoir at different times and is affected by seasonal variations; for example, outflow may increase during the rainy season and decrease during the dry season. Seasonal flow characteristic values ​​are features used to describe the differences in outflow across different seasons and can include average flow, maximum flow, minimum flow, etc.

[0027] When separating seasonal patterns, time series analysis methods, such as seasonal decomposition algorithms, can be used. Taking the outflow data of a reservoir as an example, the annual data is divided into four time periods according to the seasons: spring, summer, autumn, and winter. For the data of each season, its statistical characteristics, such as average flow, maximum flow, and minimum flow, are calculated. By comparing these statistical characteristics of different seasons, seasonal flow characteristic values ​​are generated.

[0028] Step S233: Divide the slope matrix into different slope grade regions according to the slope condition. Different roughness coefficients are used for different slope grade regions. The roughness coefficient of the gentler slope region is higher than that of the steeper slope region.

[0029] Slope grading zones are categorized into different grade ranges based on slope values ​​in a slope matrix. For example, areas with a slope value less than 10% are classified as gentle slopes, areas with a slope value between 10% and 20% as medium slopes, and areas with a slope value greater than 20% as steep slopes. The roughness coefficient describes the resistance encountered by water flowing over different terrain surfaces and is related to factors such as terrain roughness and vegetation cover. The roughness coefficient is higher in gentler slopes than in steeper slopes because the water flow velocity is relatively slower in gentler slopes, resulting in a longer contact time with the terrain surface and greater resistance.

[0030] Step S234: Assign seasonal flow characteristic values ​​to the corresponding seasonal time windows, and calculate the initial flow allocation ratio of the terrain area by combining the roughness coefficient of different slope grades. The initial flow allocation ratio is negatively correlated with the slope condition.

[0031] A time window refers to a period of time divided according to the seasons, such as the spring time window from March to May, and the summer time window from June to August. The initial flow allocation ratio refers to the proportion of initial flow distributed across different terrain areas, and it is negatively correlated with slope; that is, areas with steeper slopes have a higher initial flow allocation ratio, while areas with gentler slopes have a lower initial flow allocation ratio. This is because areas with steeper slopes have faster water flow velocity and can accommodate more flow.

[0032] When allocating seasonal flow characteristic values ​​to corresponding seasonal time windows, the seasonal flow characteristic values ​​are correlated with the time windows according to the seasonal division. Then, combined with the roughness coefficient of different slope grades, hydraulic formulas are used to calculate the initial flow allocation ratio for that terrain area. For example, in spring, the seasonal flow characteristic values ​​of spring are allocated to the time window from March to May. For gentle slope areas, due to the higher roughness coefficient and greater flow resistance, the initial flow allocation ratio is relatively low; for steep slope areas, due to the lower roughness coefficient and less flow resistance, the initial flow allocation ratio is relatively high.

[0033] Step S235: Based on the initial flow distribution ratio and slope matrix, the initial velocity distribution of the terrain area is generated by spatial interpolation. In the initial velocity distribution, the velocity in the steeper slope area is higher than that in the gentler slope area.

[0034] Spatial interpolation is a method used to extrapolate data values ​​for unknown points based on known data values. It can convert discrete initial flow distribution ratios and slope matrix data into continuous initial velocity distributions. The initial velocity distribution describes the magnitude and direction of the initial velocity of water flow at different locations within a terrain area. The velocity is higher in areas with steeper slopes than in areas with gentler slopes because the water flow in steeper slopes experiences greater gravitational forces, resulting in a faster flow velocity.

[0035] When generating the initial velocity distribution, spatial interpolation methods such as Kriging interpolation or inverse distance weighted interpolation can be used. Taking Kriging interpolation as an example, the velocity values ​​of known points are first determined based on the initial flow distribution ratio and the slope matrix. Then, based on the data from these known points, the velocity values ​​of other unknown points within the terrain area are calculated using the Kriging interpolation algorithm. For areas with steeper slopes, the initial flow distribution ratio is higher, and the effect of gravity is greater, resulting in higher interpolated initial velocity values. Conversely, for areas with gentler slopes, the initial flow distribution ratio is lower, and the flow resistance is greater, resulting in lower interpolated initial velocity values.

[0036] Step S236: Extract vegetation cover data of the terrain region from the layered grid attribute matrix, adjust the initial velocity distribution by vegetation resistance, reduce the velocity in areas with higher vegetation cover, and generate the initial boundary conditions of the flow field for the terrain region. The initial boundary conditions of the flow field contain comprehensive information on velocity magnitude and direction.

[0037] Vegetation cover data refers to the proportion of vegetation cover in each grid cell of the hierarchical grid attribute matrix, reflecting the density of vegetation within the terrain area. Vegetation resistance adjustment is a process of correcting the initial velocity distribution by considering the obstruction effect of vegetation on water flow. In areas with higher vegetation cover, the water flow experiences greater resistance, resulting in lower velocity; conversely, in areas with lower vegetation cover, the water flow experiences less resistance, leading to relatively higher velocity. The initial boundary conditions of the flow field contain comprehensive information on the magnitude and direction of velocity, and are the state of the water flow at the initial moment, determined after considering factors such as topography, reservoir management, and vegetation.

[0038] When adjusting vegetation resistance, empirical formulas or numerical simulation methods can be used. Taking empirical formulas as an example, assume there exists a vegetation resistance coefficient that is proportional to the vegetation cover. For each grid cell, the vegetation resistance coefficient is determined based on its vegetation cover data. Then, the velocity value in the initial velocity distribution is multiplied by (1 - vegetation resistance coefficient) to obtain the adjusted velocity value. For areas with high vegetation cover, the vegetation resistance coefficient is larger, and the adjusted velocity value will decrease; for areas with low vegetation cover, the vegetation resistance coefficient is smaller, and the decrease in the adjusted velocity value is smaller.

[0039] Step S240: Combining the discharge location information in the discharge time series data of the dam body, the initial boundary conditions of the flow field in different terrain regions are adjusted for impact effects to generate a corrected boundary field that includes the influence of the discharge direction.

[0040] Impact effect adjustment is the process of modifying the initial boundary conditions of the flow field in different terrain regions, taking into account the impact of dam discharge on the water flow. Discharge location information refers to the specific location of the dam discharge recorded in the dam discharge time series data, which affects the direction and velocity of the water flow. The corrected boundary field is the boundary field obtained by adjusting the initial boundary conditions after considering the influence of the discharge direction, and it more accurately reflects the initial state of the water flow under actual conditions.

[0041] When adjusting for the impact effect, numerical simulation or physical model experiments can be used. Taking numerical simulation as an example, based on the discharge location information in the dam discharge time series data, an impact term related to the discharge direction and flow rate is added to the initial boundary conditions of the flow field. For terrain areas close to the discharge location, the impact term has a greater impact, causing significant changes in the velocity and direction of the water flow; for terrain areas far from the discharge location, the impact term has a smaller impact.

[0042] Step S250: Calculate the magnitude and direction of the water flow velocity in different grid cells based on the corrected boundary field, and generate water flow velocity distribution characteristics by analyzing the spatial distribution law of the velocity magnitude. The water flow velocity distribution characteristics characterize the spatial division of different flow regime regions.

[0043] After obtaining the corrected boundary field, the magnitude and direction of the water flow velocity in different grid cells are calculated using hydrodynamic equations. These equations describe the motion of water flow under different conditions, taking into account factors such as inertia, gravity, pressure, and viscosity. By solving these equations, the magnitude and direction of the water flow velocity in each grid cell can be obtained.

[0044] The calculated water flow velocities are analyzed for spatial distribution patterns to generate water flow velocity distribution characteristics. Cluster analysis or threshold partitioning methods can be used to determine the spatial division of different flow regimes. For example, areas with higher flow velocities can be classified as rapid flow regions, and areas with lower flow velocities as slow flow regions. This spatial division allows the water flow velocity distribution characteristics to clearly reflect the distribution of different flow regime regions.

[0045] Step S260: Calculate the water level change rate for different time periods based on the water level fluctuation time series data, and generate water level amplitude characteristics by combining the topographic elevation information of the grid cells. The water level amplitude characteristics represent the amplitude differences in different water areas.

[0046] Time-series data on water level fluctuations records the changes in reservoir water levels over time. By analyzing this data, the rate of change of water level in different time periods can be calculated. The rate of change of water level reflects the amount of change in water level per unit time, and can be positive (water level rises) or negative (water level falls). Combining the topographic elevation information of the grid cells and considering the elevation differences in different topographic regions, water level amplitude characteristics are generated. For water areas with lower elevations, water level changes may lead to larger changes in the inundated area, thus the water level amplitude characteristics are relatively large; for water areas with higher elevations, water level changes have a smaller impact, and the water level amplitude characteristics are relatively small.

[0047] Step S270: The characteristics of water flow velocity distribution and water level fluctuation are coupled by separating the sediment movement state calculation to generate sediment transport characteristics containing particle composition information.

[0048] Separate calculation of sediment movement states involves classifying the movement state of sediment into different types, such as suspended state and burial state, and performing calculations separately for each. The characteristics of water flow velocity distribution and water level fluctuation have a significant impact on sediment movement; faster water flow can carry more sediment, while larger water level fluctuations may lead to sediment erosion and deposition.

[0049] By coupling the characteristics of water flow velocity distribution and water level fluctuation, their combined impact on sediment transport is considered. The particle composition of the sediment is taken into account during the calculation, as sediment particles of different sizes exhibit different movement patterns and transport capacities in water. The sediment transport characteristics generated in this way accurately reflect the transport of sediment in water flow, including sediment quantity, particle composition, and transport paths.

[0050] Step S280: Integrate the characteristics of water flow velocity distribution, sediment transport, and water level variation to construct a set of watershed hydrodynamic feature vectors with spatiotemporal identifiers. The spatiotemporal identifiers represent the dual correlation between grid location and time marker.

[0051] By integrating the characteristics of water flow velocity distribution, sediment transport, and water level fluctuation, a vector set containing multiple hydrodynamic features is constructed. This vector set not only includes information such as water flow velocity, sediment transport, and water level fluctuation, but also has spatiotemporal identification.

[0052] Spatiotemporal identification refers to the association of each vector element with a defined grid location and time marker, reflecting the spatial and temporal variations of hydrodynamic characteristics. For example, for a given grid location and time point, the elements in the vector set can represent information such as the flow velocity, sediment transport, and water level fluctuation at that location and time.

[0053] Step S300: Based on the watershed hydrodynamic characteristic vector set, infer the soil physicochemical property offset and vegetation community structure variability. The soil physicochemical property offset represents the changes in soil organic carbon content and bulk density, while the vegetation community structure variability represents the changes in vegetation biomass and type.

[0054] In one implementation, step S300 may specifically include the following steps S310 to S380: Step S310: Establish a set of association rules between watershed hydrodynamic characteristics and soil property changes through historical data correlation analysis. The set of association rules represents the correspondence between water flow velocity distribution and soil erosion intensity, sediment transport and soil deposition thickness, and water level fluctuation and soil moisture content.

[0055] Historical data correlation analysis is the process of analyzing and mining data on the hydrodynamic characteristics and soil property changes of a watershed over a past period. By collecting and organizing this historical data, statistical analysis methods, such as correlation analysis and regression analysis, are used to identify the correlations between hydrodynamic characteristics and soil property changes. The set of correlation rules is a set of rules used to describe these correlations, including the correspondence between water flow velocity distribution and soil erosion intensity, sediment transport and soil deposition thickness, and water level fluctuations and soil moisture content. For example, areas with faster water flow velocity may have greater soil erosion intensity; areas with greater sediment transport may have thicker soil deposition thickness; and areas with larger water level fluctuations may have larger changes in soil moisture content.

[0056] Step S320: Perform multi-rule matching processing on the watershed hydrodynamic feature vector set based on the association rule set to generate the soil erosion and deposition intensity index and soil moisture fluctuation index for each grid cell.

[0057] After obtaining the set of association rules, the watershed hydrodynamic feature vector set is matched with these rules. For each grid cell, based on its hydrodynamic characteristics, such as water velocity distribution, sediment transport, and water level fluctuation, the corresponding rule is searched in the association rule set to calculate the soil erosion and deposition intensity index and the soil moisture fluctuation index for that grid cell. The soil erosion and deposition intensity index reflects the degree of erosion or deposition on the soil in that grid cell and is closely related to water velocity distribution and sediment transport. The soil moisture fluctuation index describes the changes in soil moisture content in that grid cell and is related to water level fluctuation.

[0058] Step S330: Calculate the organic carbon loss caused by erosion and the organic carbon accumulation caused by deposition based on the soil erosion and deposition intensity index. Combine the soil carbon mineralization process to generate a record of changes in soil organic carbon content. The record of changes in soil organic carbon content represents the annual and cumulative changes in organic carbon.

[0059] In one implementation, step S330 may specifically include the following steps S331 to S337: Step S331: Extract the erosion depth and deposition thickness from the soil erosion and deposition intensity index. Multiply the erosion depth by the initial soil bulk density and the initial organic carbon content to obtain the organic carbon loss caused by erosion. The initial soil bulk density and the initial organic carbon content are obtained from the watershed geographic information dataset.

[0060] The soil erosion-deposition intensity index contains information on erosion depth and deposition thickness. These two parameters can be extracted by analyzing the index. The initial soil bulk density and initial organic carbon content are the soil bulk density and organic carbon content recorded in the watershed geographic information dataset under their initial conditions. Multiplying the erosion depth by the initial soil bulk density and initial organic carbon content yields the amount of organic carbon lost due to erosion.

[0061] Step S332: Multiply the sediment thickness by the bulk density and organic carbon content of the sediment to obtain the amount of organic carbon accumulation caused by sedimentation. The bulk density and organic carbon content of the sediment are derived from the particle composition information in the sediment transport characteristics.

[0062] Sediment thickness reflects the thickness of soil deposition. The bulk density and organic carbon content of sediment can be derived from the particle composition information in sediment transport characteristics. Sediment particles of different sizes have different bulk densities and organic carbon contents. By analyzing and calculating the particle composition information of sediment, the average bulk density and organic carbon content of sediment can be obtained.

[0063] Multiplying the sediment thickness by the bulk density and organic carbon content of the sediment yields the amount of organic carbon accumulation caused by sedimentation. This is because sediment thickness represents the thickness of the newly added soil, bulk density represents the mass of sediment per unit volume, and organic carbon content represents the amount of organic carbon per unit mass of sediment. This calculation method allows for an accurate assessment of the impact of sedimentation on soil organic carbon content.

[0064] Step S333: Extract soil temperature change data and soil moisture change data from the watershed hydrodynamic feature vector set. Soil temperature change data and soil moisture change data characterize the dynamic changes of the soil environment.

[0065] The watershed hydrodynamic eigenvector set contains information related to the soil environment, with soil temperature and moisture changes reflecting dynamic changes in the soil environment. Soil temperature and moisture have a significant impact on soil carbon mineralization processes, affecting soil microbial activity and the rate of organic carbon decomposition. By extracting these data from the watershed hydrodynamic eigenvector set, changes in the soil environment can be monitored in real time.

[0066] Step S334: Calculate the soil carbon mineralization rate based on soil temperature change data and soil moisture change data. The soil carbon mineralization rate is affected by the combined effects of temperature and moisture changes.

[0067] Soil carbon mineralization rate refers to the rate at which organic carbon in the soil decomposes into carbon dioxide under the action of microorganisms, and is influenced by a combination of soil temperature and moisture changes. Generally, increased temperature and suitable moisture promote the activity of soil microorganisms, thereby increasing the soil carbon mineralization rate; conversely, excessively low or high temperatures, as well as insufficient or excessive moisture, inhibit the activity of soil microorganisms, reducing the soil carbon mineralization rate. Empirical formulas or models can be used to calculate the soil carbon mineralization rate, and these formulas or models typically consider the influence of soil temperature and moisture. For example, some models provide corresponding carbon mineralization rate coefficients based on different combinations of soil temperature and moisture. By substituting soil temperature and moisture variation data into these formulas or models, the soil carbon mineralization rate at different times can be calculated.

[0068] Step S335: Multiply the soil carbon mineralization rate by the initial organic carbon content and the time interval to obtain the soil carbon mineralization amount. The time interval is consistent with the time resolution of the soil organic carbon content change record.

[0069] After obtaining the soil carbon mineralization rate, multiplying it by the initial organic carbon content and the time interval yields the amount of soil carbon mineralization. The initial organic carbon content is the organic carbon content of the soil in its initial state, and the time interval is the temporal resolution of the record of changes in soil organic carbon content; for example, if the record shows annual changes, the time interval is one year. This calculation method allows for the accurate calculation of the amount of organic carbon decomposed in the soil over a given period.

[0070] Step S336: Calculate the initial organic carbon content, subtract the organic carbon loss due to erosion, add the organic carbon accumulation due to deposition, and then subtract the soil carbon mineralization to obtain the annual change value of organic carbon content.

[0071] Taking into account the loss of organic carbon due to erosion, the accumulation of organic carbon due to deposition, and the decomposition of organic carbon during soil carbon mineralization, the annual change in organic carbon content can be obtained through the above calculation steps.

[0072] Step S337: The annual change values ​​over several consecutive years are summed to obtain the cumulative change value. The annual change values ​​and the cumulative change values ​​are integrated to generate a record of soil organic carbon content change. The record of soil organic carbon content change includes grid location and time stamp information.

[0073] The annual changes in soil organic carbon content over several consecutive years are summed to obtain the cumulative change value. The cumulative change value reflects the overall change in soil organic carbon over a longer period of time.

[0074] By integrating annual and cumulative change values, a record of soil organic carbon content changes is generated. This record not only includes annual and cumulative changes in organic carbon, but also grid location and time stamp information, enabling accurate location of soil organic carbon content changes in each grid cell at different times.

[0075] Step S340: Calculate the soil porosity change based on the soil moisture fluctuation index, and generate a soil bulk density change record by combining the compaction effect information in the soil erosion and deposition intensity index. The soil bulk density change record represents the bulk density differences in different regions.

[0076] The soil moisture fluctuation index reflects changes in soil moisture content and is closely related to soil porosity. When soil moisture increases, air is expelled from the soil pores, reducing porosity; conversely, when soil moisture decreases, porosity may increase. By establishing a correlation model between the soil moisture fluctuation index and soil porosity, changes in soil porosity can be calculated based on the soil moisture fluctuation index.

[0077] The soil erosion and deposition intensity index includes compaction effect information, reflecting the degree of compaction experienced by the soil during erosion and deposition. By combining soil porosity changes and compaction effect information, the variation in soil bulk density is calculated. Different regions experience varying degrees of erosion, deposition, and compaction, resulting in differences in soil bulk density. Generating records of soil bulk density variations clearly reflects these regional differences.

[0078] Step S350: Calculate the annual inundation duration and average inundation depth of different elevation grid cells based on the water level fluctuation characteristics. The annual inundation duration is the ratio of the number of inundated days in a year to the total number of days in a year, and the average inundation depth is the average depth during the inundation period.

[0079] Water level fluctuation characteristics record the water level changes in different water areas. Based on this information, the annual inundation duration and average inundation depth of different elevation grid cells can be calculated. For each grid cell, the inundation period and inundation depth within a year are determined based on its topographic elevation and water level fluctuation characteristics.

[0080] Annual flood duration is the ratio of the number of flooded days in a year to the total number of days in a year, reflecting the proportion of time a grid cell is flooded during the year. Average flood depth is the average depth during the flood period, reflecting the average water depth of the grid cell during the flooding period.

[0081] Step S360: Compare the annual inundation duration and average inundation depth with the ecological adaptation range of vegetation types to generate a vegetation inundation stress index, which characterizes the degree to which vegetation is affected by inundation.

[0082] In one implementation, step S360 may specifically include the following steps S361 to S366: Step S361: Extract the ecological habit information of the main vegetation types from the watershed geographic information dataset, determine the suitable inundation duration range and suitable inundation depth range for each vegetation type, and construct a vegetation ecological adaptation matrix. The vegetation ecological adaptation matrix represents the ecological threshold of different vegetation types.

[0083] The watershed geographic information dataset contains ecological habit information of major vegetation types, including their environmental requirements and tolerance to inundation. By analyzing and organizing this information, the suitable inundation duration and depth ranges for each vegetation type are determined. This suitable range information is then organized into a matrix to construct a vegetation ecological adaptation matrix. Each row in this matrix corresponds to a vegetation type, and each column corresponds to an ecological threshold parameter, such as the lower and upper limits of the suitable inundation duration and depth ranges.

[0084] Step S362: Normalize the annual inundation duration, calculate the ratio of the actual annual inundation duration to the suitable inundation duration range, and generate the inundation frequency index. The magnitude of the inundation frequency index represents the suitability of the inundation duration.

[0085] Normalization is the process of converting annual inundation duration data into a relative value to facilitate comparison and calculation. The actual annual inundation duration is compared with the suitable inundation duration range, and their ratio is calculated to obtain the inundation frequency index.

[0086] If the flooding frequency index is close to 1, it means that the actual annual flooding duration is in line with the suitable flooding duration range, and the vegetation is less affected by flooding. If the flooding frequency index is greater than 1, it means that the actual annual flooding duration exceeds the suitable flooding duration range, and the vegetation may be subjected to greater flooding stress. If the flooding frequency index is less than 1, it means that the actual annual flooding duration is less than the suitable flooding duration range, and the vegetation may not be able to meet its growth needs.

[0087] Step S363: Normalize the average inundation depth using the same method to generate an inundation depth index. The calculation method of the inundation depth index is consistent with that of the inundation frequency index.

[0088] The average flood depth was normalized using the same method as the flooding frequency index. The actual average flood depth was compared with the suitable flood depth range, and their ratio was calculated to obtain the flood depth index. The magnitude of the flood depth index also characterizes the suitability of the average flood depth. By comparing the flood depth index with 1, it can be determined whether the vegetation is in a suitable growth environment in terms of flood depth.

[0089] Step S364: Extract the water flow velocity magnitude of the grid cells in the water flow velocity distribution characteristics, calculate the ratio of water flow velocity magnitude to vegetation erosion resistance, and generate the water flow stress index. The water flow stress index characterizes the erosion effect of water flow on vegetation.

[0090] The water flow velocity distribution characteristics include information on the magnitude of the water flow velocity in each grid cell. Vegetation erosion resistance refers to the ability of vegetation to resist water erosion, and is related to factors such as root structure and stem strength. By comparing the magnitude of the water flow velocity in each grid cell with the vegetation's erosion resistance and calculating their ratio, the water flow stress index is obtained. The water flow stress index reflects the degree of erosion impact of water flow on vegetation. A larger water flow stress index indicates a faster water flow velocity and a greater erosion impact on the vegetation; conversely, a smaller water flow stress index indicates a slower water flow velocity and a lesser erosion impact on the vegetation.

[0091] Step S365: Based on the sensitivity differences of vegetation types, the inundation frequency index, inundation depth index, and water flow stress index are weighted and integrated to generate the vegetation inundation stress index.

[0092] Different vegetation types exhibit varying sensitivities to inundation and water erosion. Therefore, a weighted composite analysis of the inundation frequency index, inundation depth index, and water stress index is necessary based on these sensitivity differences. Each index is assigned a different weight, determined by the vegetation type's sensitivity. For example, for vegetation types sensitive to inundation, the inundation frequency and inundation depth indices can have relatively larger weights; conversely, for vegetation types sensitive to water erosion, the water stress index can have a relatively larger weight. The three indices are then combined into a single vegetation inundation stress index through a weighted summation.

[0093] Step S366: Perform numerical range control processing on the vegetation submersion stress index, adjust the values ​​that exceed the reasonable range to the boundary value, and generate the final vegetation submersion stress index.

[0094] To ensure the rationality and effectiveness of the vegetation submersion stress index, a numerical range control process is implemented. A reasonable numerical range is set, such as between 0 and 1. If the vegetation submersion stress index exceeds this range, it is adjusted to the boundary value. Through this process, the final vegetation submersion stress index is generated, which accurately reflects the combined impact of vegetation submersion and water erosion.

[0095] Step S370: Calculate the probability of vegetation community structure type change by using the vegetation inundation stress index, and generate vegetation biomass change record and vegetation type change record by combining the correlation between vegetation biomass and stress index.

[0096] The vegetation submersion stress index reflects the combined impact of submersion and water erosion on vegetation, affecting its growth and survival. By establishing a correlation model between the vegetation submersion stress index and the probability of vegetation community structure type change, the likelihood of vegetation community structure type change can be calculated. Simultaneously, considering the correlation between vegetation biomass and the stress index, changes in vegetation biomass are calculated based on the vegetation submersion stress index. Records of vegetation biomass change and vegetation type change are generated, reflecting the changes in vegetation biomass and the change in vegetation type at different times, respectively, providing important evidence for assessing the variability of vegetation community structure.

[0097] Step S380: Integrate soil organic carbon content change records and soil bulk density change records to generate soil physicochemical property offsets, merge vegetation biomass change records and vegetation type succession records to obtain vegetation community structure variability. Both soil physicochemical property offsets and vegetation community structure variability include grid location and time stamp information.

[0098] By integrating records of changes in soil organic carbon content and soil bulk density, a soil physicochemical property offset is generated. This offset reflects the changes in soil organic carbon content and bulk density relative to their initial state, comprehensively considering the influence of factors such as soil erosion, deposition, and carbon mineralization.

[0099] By merging vegetation biomass change records and vegetation type succession records, vegetation community structure variability is obtained. Vegetation community structure variability reflects the changes in vegetation biomass and type over different time periods, comprehensively considering the impacts of factors such as inundation and water erosion on vegetation. Both soil physicochemical property offsets and vegetation community structure variability include grid location and time stamp information, enabling them to accurately locate the soil and vegetation changes in each grid cell at different times.

[0100] Step S400: Couple soil physicochemical property offset and vegetation community structure variability to perform digital quantitative modeling of carbon cycle process, generate carbon pool storage change sequence, and the carbon pool storage change sequence characterizes the storage change status of soil carbon pool, vegetation carbon pool and water body carbon pool.

[0101] In one implementation, step S400 may specifically include the following steps S410 to S460: Step S410: Extract key carbon exchange pathways from soil physicochemical property offsets and vegetation community structure variability through carbon cycle pathway identification. Key carbon exchange pathways characterize the carbon transfer process between different pools.

[0102] Carbon cycle pathway identification is the process of analyzing and identifying the transfer processes of carbon between soil, vegetation, and water bodies. Information related to carbon exchange is extracted from soil physicochemical property shifts and vegetation community structure variability to identify key carbon exchange pathways.

[0103] Key carbon exchange pathways include vegetation photosynthetic carbon fixation, litter decomposition input, and soil respiration release. These pathways describe the transfer of carbon between different carbon pools. For example, vegetation fixes atmospheric carbon dioxide into organic carbon through photosynthesis and stores it in the vegetation carbon pool; after litter decomposition, organic carbon enters the soil carbon pool; and soil microorganisms decompose organic carbon into carbon dioxide and release it into the atmosphere through respiration.

[0104] Step S420: Calculate the flux for each key carbon exchange pathway. The flux of the vegetation photosynthetic carbon fixation pathway is calculated by correlating the vegetation biomass change records with the photosynthetic efficiency information. The flux of the litter decomposition input pathway is calculated by correlating the vegetation biomass change records with the litter ratio information.

[0105] In one implementation, step S420 may specifically include the following steps S421 to S427: Step S421: Extract the annual increment of aboveground biomass and the annual increment of belowground biomass from the vegetation biomass change record in the vegetation community structure variability, and calculate the annual increment of total biomass, which is the sum of the annual increments of aboveground and belowground biomass.

[0106] The vegetation biomass change record contains the changes in aboveground and belowground biomass over different time periods. The annual increase in aboveground and belowground biomass, i.e., the increase in aboveground and belowground biomass throughout the year, is extracted from this record.

[0107] The annual increase in aboveground biomass and the annual increase in belowground biomass are added together to obtain the annual increase in total biomass. The annual increase in total biomass reflects the overall increase in vegetation biomass throughout the year, providing important basic data for subsequent calculations of photosynthetic carbon sequestration pathway fluxes.

[0108] Step S422: Retrieve the corresponding carbon content ratio parameter from the preset parameter library according to the vegetation type, and multiply the annual increase in total biomass by the carbon content ratio parameter to obtain the annual increase in total carbon.

[0109] The preset parameter library stores carbon content ratio parameters for different vegetation types, reflecting the carbon content per unit of biomass. The corresponding carbon content ratio parameter is retrieved from the preset parameter library based on the vegetation type. The annual increase in total biomass is multiplied by the carbon content ratio parameter to obtain the annual increase in total carbon. The annual increase in total carbon represents the amount of carbon accumulated by vegetation through growth in a year and is an important intermediate parameter for calculating the flux of the vegetation photosynthetic carbon fixation pathway.

[0110] Step S423: Retrieve the corresponding photosynthetic efficiency parameter from the preset parameter library according to the vegetation type. The photosynthetic efficiency parameter represents the ability of vegetation to utilize light energy.

[0111] The preset parameter library also stores photosynthetic efficiency parameters for different vegetation types. These parameters describe the vegetation's ability to fix carbon dioxide into organic carbon using light energy. The corresponding photosynthetic efficiency parameter is retrieved from the preset parameter library based on the vegetation type. The photosynthetic efficiency parameter is used to calculate the potential photosynthetic carbon sequestration of vegetation, reflecting its carbon sequestration capacity under certain light conditions.

[0112] Step S424: Extract the annual total solar radiation data from the watershed geographic information dataset, and multiply the annual total solar radiation data by the photosynthetic efficiency parameter to obtain the potential photosynthetic carbon sequestration.

[0113] The watershed geographic information dataset includes annual total solar radiation data, which reflects the solar radiation energy received by the watershed throughout the year. Multiplying the annual total solar radiation data by a photosynthetic efficiency parameter yields the potential photosynthetic carbon sequestration. Potential photosynthetic carbon sequestration represents the maximum amount of carbon that vegetation can fix using solar radiation energy under ideal conditions.

[0114] Step S425: Calculate the change in carbon dioxide dissolved amount based on the reservoir surface area change records in the hydropower project operation dataset, deduce the local change in atmospheric carbon dioxide concentration through the change in carbon dioxide dissolved amount, and generate a carbon dioxide concentration correction coefficient.

[0115] The reservoir surface area change records in the hydropower project operation dataset reflect the changes in reservoir surface area over time. Changes in reservoir surface area affect the amount of carbon dioxide dissolved in water, because a larger surface area means a greater contact area between carbon dioxide and water, potentially increasing dissolution. Based on these reservoir surface area change records, changes in carbon dioxide dissolution are calculated. Through analysis and calculation of these changes, local variations in atmospheric carbon dioxide concentration are derived. A carbon dioxide concentration correction coefficient is generated based on these local variations. This coefficient is used to correct for potential photosynthetic carbon sequestration to obtain the actual photosynthetic carbon sequestration.

[0116] Step S426: Multiply the potential photosynthetic carbon sequestration by the carbon dioxide concentration correction factor to obtain the actual photosynthetic carbon sequestration. Extract the vegetation respiration consumption from the vegetation biomass change record. The vegetation respiration consumption is the product of the annual increase in total carbon and the respiration consumption ratio. The respiration consumption ratio varies among different vegetation types.

[0117] The potential photosynthetic carbon sequestration is multiplied by a carbon dioxide concentration correction factor to obtain the actual photosynthetic carbon sequestration. The actual photosynthetic carbon sequestration considers the impact of local changes in atmospheric carbon dioxide concentration on vegetation photosynthetic carbon sequestration, and is closer to the actual carbon sequestration capacity of vegetation. Vegetation respiration consumption is extracted from vegetation biomass change records. Vegetation respiration consumption is the amount of carbon consumed by vegetation during its growth to maintain life activities, and is related to the annual increase in total carbon and the proportion of respiration consumption. The proportion of respiration consumption varies among different vegetation types. Based on the vegetation type, the corresponding respiration consumption proportion is retrieved from a preset parameter library, and the annual increase in total carbon is multiplied by the respiration consumption proportion to obtain the vegetation respiration consumption.

[0118] Step S427: Calculate the actual photosynthetic carbon sequestration amount minus the vegetation respiration consumption to obtain the vegetation photosynthetic carbon sequestration pathway flux. The vegetation photosynthetic carbon sequestration pathway flux characterizes the net carbon sequestration capacity of vegetation.

[0119] Subtracting vegetation respiration consumption from actual photosynthetic carbon sequestration yields the vegetation photosynthetic carbon sequestration pathway flux. This flux reflects the actual net carbon sequestration capacity of vegetation after accounting for respiration consumption. It is an important indicator for assessing vegetation's contribution to the carbon cycle.

[0120] In one implementation, step S420, the process of obtaining the input path flux for litter decomposition may include the following steps S428~S4213: Step S428: Determine the dominant vegetation type of the current grid cell from the vegetation type change record in the vegetation community structure variability. The dominant vegetation type is the vegetation type with the largest proportion in the grid cell.

[0121] The vegetation type succession record in the vegetation community structure variability records the changes in vegetation type over different time periods. From this record, the dominant vegetation type of the current grid cell, i.e., the vegetation type with the largest proportion, is determined.

[0122] Step S429: Retrieve the corresponding litter ratio parameter from the preset parameter library according to the dominant vegetation type. The litter ratio parameters of different types of vegetation are different. The ratio parameter of deciduous vegetation is higher than that of evergreen vegetation.

[0123] The preset parameter library stores litter proportion parameters for different vegetation types, reflecting the proportion of litter produced by each vegetation type each year. The corresponding litter proportion parameter is retrieved from the preset parameter library based on the dominant vegetation type. Generally, deciduous vegetation types shed a large amount of leaves in autumn, resulting in a relatively high litter proportion parameter; evergreen vegetation types have a relatively low litter proportion parameter.

[0124] Step S4210: Extract the aboveground biomass stock from the vegetation biomass change record, multiply the aboveground biomass stock by the litter ratio parameter to obtain the litter yield, which represents the total amount of litter in the year.

[0125] The vegetation biomass change record includes the current aboveground biomass, which is the actual amount of aboveground biomass at the current moment. Multiplying the current aboveground biomass by the litter proportion parameter yields the litter yield. Litter yield represents the total amount of litter produced by vegetation in a year and is an important parameter for calculating the input path flux of litter decomposition.

[0126] Step S4211: The litter production is divided into different litter components, and the carbon content of the different components varies.

[0127] Litter production includes various forms of litter, such as leaves, branches, and bark. Litter production is analyzed and divided into different litter components.

[0128] The carbon content varies among different components; for example, leaves may have a relatively low carbon content, while branches may have a relatively high carbon content. By classifying and analyzing litter components, the input pathway flux for litter decomposition can be calculated more accurately.

[0129] Step S4212: Extract soil temperature change records and soil moisture change records from soil physicochemical property offsets, calculate decomposition environmental parameters, and characterize the combined effects of temperature and moisture on the decomposition process.

[0130] The soil physicochemical property offset includes records of soil temperature and moisture changes, reflecting how these parameters change over time. Soil temperature and moisture significantly impact litter decomposition, affecting soil microbial activity and the decomposition rate. Decomposition environmental parameters are generated through analysis and calculation of these records.

[0131] Step S4213: Multiply the litter yield of each component by the corresponding carbon content and decomposition environment parameters to obtain the component decomposition flux. Summarize the decomposition flux of each component to generate the litter decomposition input path flux. The litter decomposition input path flux characterizes the important carbon input status of the soil carbon pool.

[0132] For each litter component, its yield is multiplied by the corresponding carbon content and decomposition environmental parameters to obtain the component decomposition flux. The component decomposition flux represents the amount of carbon that component of litter decomposes and inputs into the soil carbon pool within a certain time period. The decomposition fluxes of all components are then summed to obtain the litter decomposition input pathway flux.

[0133] Step S430: Classify and summarize the fluxes of each path according to the carbon pool type, and generate the sum of input fluxes and the sum of output fluxes for different carbon pools. The sum of input fluxes is the cumulative result of the fluxes of all input paths, and the sum of output fluxes is the cumulative result of the fluxes of all output paths.

[0134] After calculating the fluxes for each key carbon exchange pathway, these fluxes are categorized and summarized according to carbon pool type. Carbon pool types include soil carbon pool, vegetation carbon pool, and water carbon pool.

[0135] For each carbon pool, the fluxes of all input paths are summed to obtain the total input flux; the fluxes of all output paths are summed to obtain the total output flux. The total input flux represents the amount of carbon that the carbon pool obtains from other carbon pools within a certain period of time, and the total output flux represents the amount of carbon that the carbon pool releases to other carbon pools within a certain period of time.

[0136] Step S440: Calculate the net flux of each carbon pool based on the difference between the sum of input flux and the sum of output flux. The net flux characterizes the increase or decrease of the carbon pool.

[0137] The net flux of each carbon pool is obtained by calculating the difference between the sum of the input flux and the sum of the output flux. A positive net flux indicates that the carbon storage of that pool increases over a certain period of time; a negative net flux indicates that the carbon storage of that pool decreases over a certain period of time.

[0138] Step S450: Calculate the change in carbon pool reserves based on net flux and initial carbon pool reserves. The initial carbon pool reserves are obtained from the watershed geographic information dataset. The change in reserves is the sum of the products of the initial reserves, net flux, and time interval.

[0139] The initial carbon pool stock refers to the carbon storage capacity of the carbon pool at the initial moment, which can be obtained from watershed geographic information datasets. The change in carbon pool stock is calculated based on net flux and the initial carbon pool stock. The formula for calculating the change in stock is, for example: Change in stock = Initial stock + Net flux × Time interval. The time interval is the time span used to calculate the change in stock, such as one year.

[0140] Step S460: Perform spatiotemporal scale transformation on the carbon pool storage change, converting the short-term change at the grid scale into the long-term change at the watershed scale, and generating a carbon pool storage change sequence containing different carbon pool types. Each data point in the carbon pool storage change sequence contains a time stamp and a carbon pool type identifier.

[0141] Carbon pool changes are short-term variations calculated at the grid scale. To better reflect the carbon pool changes across the entire watershed, a spatiotemporal scale transformation is required. Spatial interpolation and time series analysis are used to convert the short-term grid-scale changes into long-term watershed-scale changes. This generates carbon pool change sequences containing different carbon pool types, with each data point in the sequence including a time stamp and a carbon pool type identifier.

[0142] Step S500: Construct an ecosystem carbon pool monitoring model under the influence of hydropower projects based on the carbon pool storage change sequence. The ecosystem carbon pool monitoring model under the influence of hydropower projects is used to output carbon pool storage prediction curves under different hydrological scenarios.

[0143] In one implementation, step S500 may specifically include the following steps S510 to S580: Step S510: Perform multi-scale trend separation processing on the carbon pool change sequence to separate the long-term trend component, seasonal fluctuation component and random disturbance component. The long-term trend component is calculated by moving average, and the seasonal fluctuation component is extracted by periodic signal.

[0144] Multi-scale trend separation decomposes the carbon pool change sequence into components at different scales, including long-term trend components, seasonal fluctuation components, and random disturbance components. The long-term trend component reflects the overall trend of carbon pool changes over a longer period, reflecting the direction of change in the ecosystem's carbon pool under the long-term impact of hydropower projects, such as whether the carbon pool is gradually increasing or decreasing. The seasonal fluctuation component reflects the periodic pattern of carbon pool changes with the seasons, which is related to seasonal changes in the natural environment (such as temperature and precipitation) and the seasonal scheduling of hydropower projects. The random disturbance component represents short-term fluctuations in carbon pool contents caused by random factors, such as sudden natural disasters or short-term abnormal operation of hydropower projects. In multi-scale trend separation, the moving average method is used for the long-term trend component. The moving average is a commonly used time series smoothing method that smooths the data by calculating the average value of data within a certain time window, thereby highlighting the long-term trend of the data.

[0145] For seasonal fluctuation components, a periodic signal extraction method is used. This can be achieved using Fourier transform or seasonal decomposition algorithms, among others.

[0146] Step S520: Use the long-term trend component as the model prediction target, and the seasonal fluctuation component and the key features in the watershed hydrodynamic feature vector set as model input variables. The key features characterize the main variation patterns of water flow, sediment and water level.

[0147] Long-term trend components are chosen as the model's prediction targets because they reflect the overall changing trend of carbon pool storage. Seasonal fluctuation components and key features from the watershed hydrodynamic eigenvector set are selected as model input variables. Seasonal fluctuation components contain information about the seasonal variation of carbon pool storage, helping the model capture seasonal patterns of carbon pool changes. Key features from the watershed hydrodynamic eigenvector set, such as flow velocity distribution, sediment transport, and water level fluctuations, are closely related to carbon cycle processes. Flow velocity affects carbon transport and diffusion, sediment transport may carry carbon and influence its deposition and redistribution, and water level fluctuations alter the inundation extent and ecological environment of the watershed, thus affecting carbon pool storage. By using these key features as input variables, the model can comprehensively consider the impact of multiple factors on carbon pool storage, improving prediction accuracy.

[0148] Step S530: Evaluate the feature importance of the model input variables, calculate the contribution ratio of each input variable to the prediction target through the feature influence degree analysis method, and select variables with contribution ratios higher than the ratio threshold to form an optimized input feature set.

[0149] Feature influence analysis methods, such as the feature importance calculation method in random forests, are employed. Random forests are ensemble learning algorithms that make predictions by constructing multiple decision trees and combining their results. In random forests, the average reduction in impurity for each input variable across all decision trees can be calculated; this reduction reflects the variable's contribution to the model's prediction. A threshold is set based on the calculated contribution proportion of each input variable. Variables with contribution proportions higher than this threshold are selected and grouped into an optimized input feature set. This approach reduces model complexity, avoids interference from too many irrelevant or redundant variables, and improves the model's generalization ability and prediction accuracy.

[0150] Step S540: Divide the optimized input feature set and long-term trend components into a training dataset and a validation dataset according to a preset ratio.

[0151] To train and evaluate an ecosystem carbon pool monitoring model under the impact of hydropower projects, the optimized input feature set and corresponding long-term trend component data need to be divided into training and validation datasets. The preset ratio is determined based on experience or the specific problem, for example, a ratio of 70:30 or 80:20, meaning 70% or 80% of the data is used for training the model, and 30% or 20% is used to validate the model's performance.

[0152] The purpose of dividing the dataset into training and validation datasets is to evaluate the model's generalization ability. The training dataset is used for parameter learning, where the model adjusts its parameters by learning from the training data to fit the relationship between the input variables and the prediction target. The validation dataset is used to evaluate the model's performance on unseen data. By comparing the model's predictions on the validation dataset with the actual values, it is determined whether the model is overfitting or underfitting, allowing for model adjustment and optimization.

[0153] Step S550: Adjust the parameters of the initial network structure using the training dataset. The initial network structure includes an input layer, multiple hidden layers, and an output layer. The number of neurons in the input layer is consistent with the number of variables in the optimized input feature set, and the number of neurons in the output layer corresponds to different carbon library types.

[0154] In one implementation, step S550 may specifically include the following steps S551 to S557: Step S551: Set the number of neurons in the input layer to the number of variables in the optimized input feature set, the number of neurons in the first hidden layer to a multiple of the number of neurons in the input layer, the number of neurons in the second hidden layer to a certain proportion of the first hidden layer, the number of neurons in the third hidden layer to a certain proportion of the second hidden layer, and the number of neurons in the output layer to the number of carbon library types.

[0155] When determining the specific parameters of the initial network structure, the number of neurons in the input layer is first set to be equal to the number of variables in the optimized input feature set to ensure that the input data can be accurately fed into the network. For the first hidden layer, the number of neurons is set to a multiple of the number of neurons in the input layer, for example, twice the number of neurons in the input layer. This increases the network's expressive power, enabling it to learn more complex features.

[0156] The number of neurons in the second hidden layer is set to a certain proportion of the number of neurons in the first hidden layer, for example, 0.8 times the number of neurons in the first hidden layer. As the number of network layers increases, appropriately reducing the number of neurons can avoid network overfitting and reduce computational cost. The number of neurons in the third hidden layer is also set to a certain proportion of the number of neurons in the second hidden layer, such as 0.8 times. The number of neurons in the output layer is set to the number of carbon library types, with each neuron corresponding to a prediction output for one carbon library type. This allows for direct acquisition of long-term trend component predictions for different carbon library types.

[0157] Step S552: Add data standardization processing between the input layer and the first hidden layer to perform numerical transformation on the input features so that the statistical properties of the transformed features remain consistent.

[0158] Data standardization aims to eliminate differences in units and numerical ranges among different input features, ensuring that all input features possess similar statistical properties. A data standardization layer is added between the input layer and the first hidden layer. Standardization methods can include Z-score standardization and Min-Max standardization. Taking Z-score standardization as an example, for each input feature, its mean and standard deviation are calculated. Then, the mean is subtracted from each feature value, and the result is divided by the standard deviation to obtain the standardized feature value. After this process, the mean of all input features is 0, and the standard deviation is 1, enabling the network to more effectively learn the relationship between the input features and the prediction target, thus improving the model's convergence speed and stability.

[0159] Step S553: ​​Use non-linear activation functions in the first and second hidden layers, and add random neuron deactivation processing between the second and third hidden layers to prevent the model from overfitting.

[0160] Non-linear activation functions, such as ReLU (Rectified Linear Unit), are used in the first and second hidden layers. The purpose of non-linear activation functions is to introduce non-linearity, enabling the neural network to learn complex non-linear relationships in the input data. Without non-linear activation functions, a multi-layer neural network would be equivalent to a linear model, significantly limiting its expressive power. Random neuron deactivation (Dropout) is added between the second and third hidden layers.

[0161] Step S554: A linear activation function is used between the third hidden layer and the output layer, and each neuron in the output layer corresponds to a predicted value of a carbon library type.

[0162] A linear activation function, f(x) = x, is used between the third hidden layer and the output layer, directly taking the input value as the output value. This is because the output layer aims to obtain the actual predicted values ​​of the long-term trend components of the carbon pool type, without requiring nonlinear transformations. Each neuron in the output layer corresponds to a predicted value for a carbon pool type, and the linear activation function can directly output the prediction result, facilitating subsequent analysis and application.

[0163] Step S555: Initialize the network weight parameters. Different initialization methods are used for the weight parameters of different layers. The bias parameters are initialized to preset values.

[0164] The initialization of network weight parameters is crucial for the training performance and convergence speed of the model. Different initialization methods are used for the weight parameters of different layers. For example, the weight parameters from the input layer to the first hidden layer use the He initialization method. He initialization is an initialization method designed for the ReLU activation function, which ensures that the variance of the input signal remains constant as it passes through the network layers, thereby avoiding the vanishing or exploding gradient problems. For the bias parameters, they are initialized to a set value, or they can be initialized to 0. This ensures that the network output is not affected by the bias at the beginning of training, allowing the model to start learning from a relatively neutral state.

[0165] Step S556: Set the loss function as a measure of the difference between the predicted value and the actual value. The optimizer adopts an adaptive learning rate adjustment method, and the learning rate decays according to a preset rule during training.

[0166] Set the loss function as a measure of the difference between the predicted and actual values. For example, the mean squared error (MSE) loss function measures the average squared error between the predicted and actual values, which can effectively reflect the prediction accuracy of the model.

[0167] The optimizer employs adaptive learning rate adjustment methods, such as Adagrad, Adadelta, or Adam. Taking the Adam optimizer as an example, it combines the advantages of Adagrad and RMSProp, adaptively adjusting the learning rate for each parameter. During training, the learning rate decays according to a preset rule, such as an exponential decay strategy, gradually decreasing as the number of training iterations increases. This allows the model to converge quickly in the early stages of training and enables more fine-tuning of parameters in the later stages, improving model performance.

[0168] Step S557: Input the training dataset into the network in batches, obtain the predicted value through forward computation, calculate the error through the loss function, adjust the weight parameters through backpropagation, and iterate the above process until the early stopping condition is met or the maximum number of iterations is reached.

[0169] The training dataset is fed into the network in batches for training. Batch training refers to dividing the training dataset into several small batches, and feeding one small batch of data into the network for forward computation at a time. Forward computation refers to the process of input data passing through the input layer, various hidden layers, and finally reaching the output layer to obtain the predicted value.

[0170] The error between the predicted and actual values ​​is calculated using a loss function, and then the gradient of the error with respect to the weight parameters is calculated using the backpropagation algorithm. Backpropagation is a gradient calculation method based on the chain rule, which can efficiently calculate the gradient of the error with respect to all weight parameters.

[0171] Based on the calculated gradient, the optimizer updates the weight parameters, continuously reducing the error. This process is iterated—that is, continuously performing forward computation, error calculation, backpropagation, and parameter updates—until the early stopping condition is met or the maximum number of iterations is reached.

[0172] Early stopping is a strategy implemented to prevent overfitting. For example, a validation set is used to evaluate the model's performance after each training iteration. If the error on the validation set does not significantly improve over several consecutive iterations, training is stopped, and the current model parameters are saved. The maximum number of iterations is a pre-defined upper limit on the number of training iterations to ensure that the training process does not continue indefinitely.

[0173] Step S560: During parameter tuning, an early stopping strategy is adopted. When the prediction error of the validation dataset fails to improve after multiple iterations, training is stopped, and the current network parameters are saved as the optimal parameters.

[0174] Early stopping is an effective method to prevent model overfitting. During model training, in addition to the training dataset, a validation dataset is used to evaluate the model's generalization ability. After each training iteration, the validation dataset is input into the model for prediction, and the prediction error is calculated.

[0175] If the prediction error on the validation dataset does not improve significantly over multiple iterations (e.g., no decrease in validation error over 10 consecutive iterations), the model is considered to be overfitting, and training is stopped. After stopping training, the current network parameters are saved as optimal parameters. The model corresponding to these optimal parameters exhibits good generalization performance on the validation dataset and can make accurate predictions on unseen data.

[0176] Step S570: Based on the validation dataset, perform performance testing on the network structure corresponding to the optimal parameters, calculate the prediction error index, and determine that the model performance is qualified when the prediction error index meets the preset requirements.

[0177] After obtaining the optimal parameters, the performance of the network structure corresponding to the optimal parameters is evaluated. A validation dataset is used, input into the network model with the optimal parameters, and prediction error metrics are calculated. Prediction error metrics can include, for example, mean squared error (MSE), mean absolute error (MAE), and coefficient of determination (R²). 2 )wait.

[0178] Mean squared error measures the average squared error between the predicted and actual values, mean absolute error measures the average absolute deviation between the predicted and actual values, and the coefficient of determination indicates how well the model fits the data; the closer its value is to 1, the better the model fits.

[0179] The preset requirements are error metric thresholds set based on specific application scenarios and needs. For example, a mean squared error (MSE) threshold of 0.1 is set. When the calculated MSE is less than this threshold, the model's prediction error is considered to be within an acceptable range, and the model's performance is deemed satisfactory. If the prediction error metric does not meet the preset requirements, the model structure or parameters need to be readjusted, and retraining and validation performed.

[0180] Step S580: Integrate the optimal parameters, the statistical parameters of the optimized input feature set, and the multi-scale trend separation parameters to construct an ecosystem carbon pool monitoring model under the influence of hydropower projects. The model input is the optimized input feature set, and the output is the predicted value of the long-term trend components of different carbon pool types.

[0181] By integrating the optimal parameters, statistical parameters for optimizing the input feature set, and multi-scale trend separation parameters, a final monitoring model for ecosystem carbon pools under the influence of hydropower projects is constructed. The optimal parameters are those obtained during training using an early stopping strategy that maximize the model's performance on the validation dataset.

[0182] The statistical parameters for optimizing the input feature set include standardized parameters such as mean and standard deviation. These parameters are used to standardize the input features, ensuring that the statistical characteristics of the input data are consistent with those used during training. Multi-scale trend separation parameters include the time window size of the moving average and the period of seasonal decomposition. These parameters are used to separate the multi-scale trends of carbon pool changes. The model input is the optimized input feature set, i.e., the input variables after feature importance assessment and filtering. The model output is the predicted long-term trend components for different carbon pool types. By inputting the current optimized input feature set, the model can directly output the predicted long-term trend components for different carbon pool types over a future period, providing decision support for hydropower project planning and ecosystem carbon pool management.

[0183] As one implementation, after step S500, the method provided in this embodiment of the invention may further include the following steps S600~S1300: Step S600: Collect multi-scenario hydrological datasets for historical periods within the basin. The multi-scenario hydrological datasets include precipitation data, evaporation data, and hydropower project operation and control schemes under different hydrological conditions, representing the basin conditions under different hydrological conditions.

[0184] Collecting multi-scenario hydrological datasets from historical periods within the watershed aims to provide a comprehensive understanding of the watershed's condition under varying hydrological conditions. These datasets cover diverse hydrological scenarios, including wet and dry seasons, as well as different precipitation patterns and evaporation intensities.

[0185] Precipitation data records the amount of precipitation at different times and locations within the watershed, and is an important factor affecting the watershed's water resources and ecosystem. Evaporation data reflects the evaporation loss of water within the watershed, and together with precipitation data, it affects the water balance of the watershed. Hydropower project operation and control schemes record the scheduling strategies of hydropower projects under different hydrological conditions, such as the timing and flow of reservoir storage and release.

[0186] Step S700: Spatial refinement processing is performed on the precipitation data of each scenario set to convert low-resolution precipitation data into a high-resolution precipitation field that matches the watershed grid. The spatial refinement processing adopts the terrain-assisted interpolation method.

[0187] Because raw precipitation data may have low spatial resolution and cannot accurately reflect the differences in precipitation across different regions within a watershed, spatial refinement of precipitation data for each scenario set is necessary. Topographic-assisted interpolation is an interpolation method that incorporates topographic information, considering the influence of topography on precipitation distribution. Topography affects airflow and precipitation formation; for example, in mountainous areas, windward slopes typically receive more precipitation. In interpolation, the topographic elevation data of the watershed is first acquired, and then the watershed is divided into different regions based on topographic features. For each region, an interpolation algorithm, such as Kriging interpolation or inverse distance weighted interpolation, is used, combined with topographic information, to interpolate the low-resolution precipitation data, resulting in a high-resolution precipitation field.

[0188] Step S800: Input the high-resolution precipitation field, evaporation data and hydropower project operation and control scheme of each scenario set into the ecosystem carbon pool monitoring model under the influence of hydropower projects, and generate the carbon pool storage prediction curve for the corresponding scenario. The prediction curve represents the change trajectory of different carbon pool storage over time.

[0189] In one implementation, step S800 may specifically include the following steps S810~S870: Step S810: Extract monthly average precipitation data from the high-resolution precipitation field, and calculate the monthly average effective precipitation by combining it with evaporation data. The monthly average effective precipitation is the difference between the monthly average precipitation data and the monthly average evaporation data.

[0190] Monthly average precipitation data is extracted from high-resolution precipitation fields to obtain the average precipitation amount for each month. Monthly average precipitation data reflects the overall precipitation situation for that month. Combined with evaporation data, monthly average effective precipitation is calculated. Evaporation data records the water loss through evaporation each month. Monthly average effective precipitation is the difference between monthly average precipitation data and monthly average evaporation data, representing the actual amount of precipitation that can impact the watershed ecosystem and carbon cycle after accounting for evaporation losses.

[0191] Step S820: Analyze the reservoir scheduling rules in the hydropower project operation and control scheme, extract water level control parameters, and generate water level control curves. The water level control curves represent the control water level status in different months.

[0192] The operation and control scheme of hydropower projects includes reservoir scheduling rules, which specify the reservoir's water storage and release strategies under different time and hydrological conditions. The reservoir scheduling rules are analyzed to extract water level control parameters, such as target water levels and water level variation ranges for different months. Based on the extracted water level control parameters, a water level control curve is generated. The water level control curve, with the month on the horizontal axis and water level on the vertical axis, shows the controlled water level situation for different months. For example, during the rainy season, the reservoir water level may be raised to store water, while during the dry season, the water level may be lowered to meet downstream water demand.

[0193] Step S830: Input the monthly average effective precipitation and water level control curve into the hydrological driving module of the monitoring model of ecosystem carbon pool under the influence of hydropower projects, and generate the hydrological feature vector input to the model. The hydrological feature vector represents the hydrological factors affecting the carbon cycle.

[0194] For example, the hydrological-driven module spatially allocates the monthly average effective precipitation based on the topographic information of the watershed. For instance, in mountainous areas, due to the large topographic relief, the distribution of precipitation may be affected by factors such as slope and aspect. The module uses a topographic-assisted interpolation method, combined with digital elevation model (DEM) data, to accurately allocate the monthly average effective precipitation to each grid cell of the watershed.

[0195] The water level control curve represents the controlled water level status in different months. The hydrological-driven module analyzes the changing trends of the water level control curve, including the rate of rise and fall of water level and the water level difference between different months. Simultaneously, combining the water system distribution information of the basin, the water level control curve is transformed into water level change information for each grid unit. For example, water level changes may differ in different river sections; based on the river's topology and hydrodynamic principles, the water level change for each grid unit is calculated. Next, considering the spatial distribution of monthly average effective precipitation and water level change information, the water balance of each grid unit is calculated. Water balance refers to the difference between the input and output water volume in a given area over a certain period, which equals the change in water storage within that area. During the calculation, factors such as precipitation, evaporation, surface runoff, and groundwater runoff are considered. For example, using a soil moisture model, soil infiltration and evapotranspiration are calculated to determine the magnitude of surface and groundwater runoff. Finally, the water balance calculation results and water level change information are integrated to generate the hydrological feature vector input to the model. This vector contains information in multiple dimensions, such as water volume changes, water level change rates, and surface runoff intensity for each grid cell. This information comprehensively characterizes the hydrological factors that affect the carbon cycle.

[0196] Step S840: Standardize the hydrological feature vectors, converting each feature value into a numerical range that meets the model input requirements. The conversion process is based on the statistical parameters used in the model construction process.

[0197] Before inputting hydrological feature vectors into the model, they need to be standardized. The purpose of standardization is to convert the feature values ​​in the hydrological feature vectors into numerical ranges that meet the model's input requirements, ensuring that the model can effectively process these input data.

[0198] The transformation process is based on statistical parameters from the model building process, such as the mean and standard deviation calculated during the standardization of input features in the model training phase. The same standardization method used during model training, such as Z-score standardization, is applied to each feature value in the hydrological feature vector. This ensures that the input hydrological feature vector has the same statistical properties as the input data during model training, improving the model's prediction accuracy.

[0199] Step S850: Input the standardized hydrological feature vector into the prediction module of the ecosystem carbon pool monitoring model under the influence of hydropower projects, and obtain the predicted values ​​of long-term trend components for different carbon pool types through forward calculation.

[0200] When the standardized hydrological feature vector is input into the prediction module, it first enters the input layer. The number of neurons in the input layer is consistent with the dimension of the standardized hydrological feature vector, ensuring that each feature is accurately passed into the network. After receiving the data, the input layer passes it to multiple hidden layers. The hidden layers use non-linear activation functions, such as ReLU (Rectified LinearUnit). Each neuron in the hidden layer performs a weighted summation of the input information and performs a non-linear transformation through the activation function, gradually extracting higher-level feature representations. As the data is passed between the hidden layers, the network continuously abstracts and integrates the features, uncovering the potential correlation between the hydrological feature vector and the long-term trend of carbon pool reserves. Finally, the data is passed to the output layer. The number of neurons in the output layer is the same as the number of carbon pool types, with each neuron corresponding to a prediction output for one carbon pool type. The output layer uses a linear activation function to directly output the predicted values ​​of the long-term trend components of different carbon pool types.

[0201] The entire forward computation process is performed according to the network architecture and the learned weight parameters. The weight parameters are continuously adjusted during the model training phase through the backpropagation algorithm and optimizer to minimize the model's prediction error.

[0202] Step S860: Perform multi-scale trend synthesis processing on the long-term trend component prediction values, and superimpose them with the seasonal fluctuation component and random disturbance component to generate a complete carbon pool storage prediction sequence.

[0203] After obtaining the predicted long-term trend components for different carbon pool types, multi-scale trend synthesis is required. During model construction, multi-scale trend separation was performed on the carbon pool storage change sequence to obtain the long-term trend components, seasonal fluctuation components, and random perturbation components.

[0204] By overlaying the long-term trend component forecast with the seasonal fluctuation component and the random disturbance component, a complete carbon pool storage forecast sequence is generated. The seasonal fluctuation component reflects the periodic pattern of carbon pool storage changes with the seasons, while the random disturbance component reflects short-term fluctuations caused by some accidental factors.

[0205] Step S870: Arrange the carbon pool reserve prediction sequence in chronological order to generate a carbon pool reserve prediction curve containing an uncertainty range. The uncertainty range is calculated through the prediction error distribution during model training. The horizontal and vertical axes of the curve represent time and carbon pool reserves, respectively.

[0206] The carbon pool reserve prediction sequence is arranged chronologically, with time on the horizontal axis and carbon pool reserve on the vertical axis to plot a carbon pool reserve prediction curve. Simultaneously, considering the inherent uncertainty in model predictions, the uncertainty range is calculated using the prediction error distribution during model training. During model training, the prediction error for each iteration is recorded, and statistical analysis of these error data yields the error distribution characteristics, such as mean and standard deviation. Based on these error distribution characteristics, the uncertainty range is calculated; for example, a range within one standard deviation above and below the predicted value is considered the uncertainty range. When plotting the carbon pool reserve prediction curve, this uncertainty range is also plotted on the curve, typically represented by a shaded area. The resulting carbon pool reserve prediction curve not only displays the predicted carbon pool reserve value but also includes the predicted uncertainty range.

[0207] Step S900: Obtain measured carbon pool data for different periods within the watershed. The measured carbon pool data includes soil sampling data, vegetation survey data, and water monitoring data. The point data is converted into area data through spatial interpolation.

[0208] Obtaining measured carbon pool data from different periods within the watershed is crucial for verifying the accuracy of the model's predictions. This measured carbon pool data includes soil sampling data, vegetation survey data, and water monitoring data. Soil sampling data is obtained by collecting soil samples at various locations within the watershed and analyzing parameters such as organic carbon content and bulk density. Vegetation survey data is obtained through field surveys of vegetation within the watershed, recording information such as vegetation species and biomass. Water monitoring data is obtained by monitoring water bodies within the watershed and measuring parameters such as carbon content and dissolved oxygen. Since the measured data are mostly obtained from discrete sampling points, spatial interpolation is needed to convert the point data into surface data for comparison with the surface data predicted by the model. Spatial interpolation methods include, for example, Kriging interpolation and inverse distance weighted interpolation. Through spatial interpolation, measured surface data of the carbon pool across the entire watershed can be obtained, providing an accurate data foundation for subsequent model validation.

[0209] Step S1000: Perform time series matching processing on the carbon pool reserve prediction curve and the corresponding measured carbon pool surface data to generate a comparison data group with the same time stamp. The comparison data group includes the prediction results and the measured results.

[0210] To accurately compare carbon pool prediction curves and measured carbon pool surface data, time series matching is required. Since prediction curves and measured data may have different time resolutions and time ranges, they need to be aligned to have the same time stamp.

[0211] The carbon pool reserve prediction curve and the measured carbon pool surface data are arranged in chronological order, and data from the same time point are selected to form a comparative data group. For example, if the time resolution of the prediction curve is monthly and the time resolution of the measured data is quarterly, the measured data is averaged quarterly to convert it into monthly data, and then matched with the prediction curve.

[0212] Step S1100: Calculate the model prediction error by comparing the data sets. The prediction error includes absolute difference, relative difference, and trend consistency difference. The trend consistency difference is calculated by the proportion of the consistency between the change direction of the predicted curve and the measured curve.

[0213] The accuracy of the model is assessed by calculating the prediction error of the data sets. The prediction error includes absolute difference, relative difference, and trend consistency difference.

[0214] Absolute difference refers to the absolute value of the difference between the predicted result and the measured result, reflecting the actual deviation between the predicted and actual values. Relative difference refers to the ratio of the absolute difference to the measured value, reflecting the proportion of the prediction error relative to the actual value.

[0215] Trend consistency error is calculated by the proportion of the consistency between the predicted and measured curves in their direction of change. It involves statistically analyzing the percentage of time points at which the predicted and measured curves show the same direction of change (upward or downward). A higher percentage indicates a more accurate prediction of the carbon pool's changing trend. By comprehensively considering these three prediction error indicators, the model's performance can be fully evaluated.

[0216] Step S1200: When the relative difference in the prediction error exceeds the preset limit, new feature variables are added from the watershed hydrodynamic feature vector set, and the monitoring model of ecosystem carbon pool under the influence of hydropower projects is retrained until the relative difference is lower than the preset limit.

[0217] A preset limit is set as a threshold for relative difference. When the calculated relative difference exceeds this threshold, it indicates that the model's predictive accuracy does not meet the requirements and the model needs to be adjusted. New feature variables are added from the watershed hydrodynamic feature vector set. These feature variables may contain previously underutilized information and can further improve the model's explanatory power for changes in carbon pool storage. For example, seasonal variations in water flow velocity or particle size distribution characteristics of sediment transport can be added.

[0218] The monitoring model for ecosystem carbon pool under the impact of hydropower projects was retrained. The dataset, supplemented with new feature variables, was trained following the previous training procedure, including data standardization and model parameter adjustment. The training process was iterated until the relative differences fell below a preset limit, at which point the model's predictive accuracy was considered to have reached an acceptable level.

[0219] Step S1300: Integrate the carbon pool storage prediction curves and corresponding prediction errors of each scenario set to generate a multi-scenario prediction report of ecosystem carbon pool under the influence of hydropower projects. This report includes the changing trend and uncertainty range of carbon pool storage under different scenarios.

[0220] The carbon pool prediction curves and corresponding prediction errors from each scenario set are integrated to generate a multi-scenario prediction report on ecosystem carbon pools under the impact of hydropower projects. This report summarizes and presents the entire modeling and prediction process, providing important decision-making basis for hydropower project planning and ecosystem carbon pool management.

[0221] The report includes trends in carbon pool storage under different scenarios. The carbon pool storage prediction curves provide a clear visual representation of how carbon pool storage changes over time under varying hydrological conditions (such as wet and dry seasons). The report also illustrates the range of prediction uncertainty, allowing decision-makers to understand the reliability and risk level of the model predictions through prediction error analysis. Furthermore, the report allows for comparison and analysis of results under different scenarios, such as comparing the differences in carbon pool storage changes under different hydropower project operation strategies, providing a reference for formulating optimal hydropower project operation plans.

[0222] Please see Figure 2 , Figure 2 This is a schematic diagram of a computer system provided in an embodiment of the present invention. The computer system includes at least a processor 101, a communication interface 102, and a memory 103. The processor 101, communication interface 102, and memory 103 can be connected via a bus or other means. The processor 101 (or Central Processing Unit, CPU) is the computing and control core of the computer system, capable of parsing various instructions and processing various data within the computer system. The communication interface 102 may optionally include a standard wired interface or a wireless interface (such as Wi-Fi, mobile communication interface, etc.), and can be used to send and receive data under the control of the processor 101; the communication interface 102 can also be used for data transmission and interaction within the computer system. The memory 103 is a storage device in the computer system used to store programs and data. It is understood that the memory 103 here can include the computer system's built-in memory, or it can include extended memory supported by the computer system. The memory 103 provides storage space, which stores the computer system's operating system; this invention does not limit this storage space.

[0223] In one embodiment, the processor 101 executes the modeling method for monitoring ecosystem carbon pools under the influence of hydropower projects provided above in the embodiments of the present invention by running a computer program in the memory 103.

Claims

1. A modeling method for monitoring ecosystem carbon pools under the influence of hydropower projects, characterized in that, The method includes: Obtain a watershed geographic information dataset and a hydropower project operation dataset. The watershed geographic information dataset represents the topography, vegetation cover, and soil type. The hydropower project operation dataset represents the reservoir scheduling, dam discharge, and water level fluctuation. The watershed geographic information dataset and the hydropower project operation dataset are analyzed for hydrodynamic features to obtain a watershed hydrodynamic feature vector set, which represents the distribution of water flow velocity, sediment transport and water level fluctuation. Based on the watershed hydrodynamic characteristic vector set, the soil physicochemical property offset and vegetation community structure variability are deduced. The soil physicochemical property offset represents the changes in soil organic carbon content and bulk density, and the vegetation community structure variability represents the changes in vegetation biomass and type. The carbon cycle process is digitally quantified and modeled by coupling the soil physicochemical property offset and the vegetation community structure variability, generating a carbon pool storage change sequence, which characterizes the storage changes of soil carbon pool, vegetation carbon pool and water body carbon pool. Based on the carbon pool storage change sequence, an ecosystem carbon pool monitoring model under the influence of hydropower projects is constructed. The ecosystem carbon pool monitoring model under the influence of hydropower projects is used to output carbon pool storage prediction curves under different hydrological scenarios.

2. The method according to claim 1, characterized in that, The process of analyzing the watershed geographic information dataset and the hydropower project operation dataset to obtain a watershed hydrodynamic feature vector set includes: The watershed geographic information dataset is spatially discretized according to topographic partitions to generate a hierarchical grid attribute matrix containing topographic elevation information, vegetation cover degree and soil texture identification. The grid resolution of different topographic regions in the hierarchical grid attribute matrix is ​​different. The hydropower project operation dataset is integrated into time series data to generate reservoir scheduling time series data, dam discharge time series data and water level fluctuation time series data with unified time stamps, and the time intervals of different types of time series data are kept consistent. Spatial correlation calculation is performed between the data of different terrain regions in the hierarchical grid attribute matrix and the corresponding reservoir scheduling time series data to generate the initial boundary conditions of the flow field for different terrain regions. The initial boundary conditions of the flow field characterize the spatial matching relationship between the terrain and the scheduling data. By combining the discharge location information in the discharge time series data of the dam body, the initial boundary conditions of the flow field in different terrain regions are adjusted for impact effects to generate a corrected boundary field that includes the influence of discharge direction; Based on the corrected boundary field, the magnitude and direction of the water flow velocity in different grid cells are calculated. The spatial distribution pattern of the velocity magnitude is analyzed to generate the water flow velocity distribution characteristics, which characterize the spatial division of different flow regime regions. The water level change rate at different time periods is calculated based on the water level fluctuation time series data, and the water level amplitude characteristics are generated by combining the topographic elevation information of the grid cells. The water level amplitude characteristics represent the amplitude differences in different water areas. By separating and calculating the sediment movement state, the water flow velocity distribution characteristics and the water level fluctuation characteristics are coupled to generate sediment transport characteristics that include particle composition information. By integrating the water flow velocity distribution characteristics, sediment transport characteristics, and water level fluctuation characteristics, a set of watershed hydrodynamic feature vectors with spatiotemporal identifiers is constructed. The spatiotemporal identifiers represent the dual correlation between grid location and time marker.

3. The method according to claim 2, characterized in that, The step of spatially correlating the data of different terrain regions in the hierarchical grid attribute matrix with the corresponding reservoir scheduling time series data to generate initial boundary conditions for the flow field in different terrain regions includes: Extract the topographic elevation information of the target terrain region from the hierarchical grid attribute matrix, calculate the ratio of the elevation difference to the horizontal distance between adjacent grid cells, and generate the slope matrix of the terrain region. Each element of the slope matrix corresponds to the average slope of the grid cell. The outflow data in the reservoir scheduling time series data is subjected to seasonal pattern separation processing to extract the outflow characteristics of different seasons and generate seasonal flow characteristic values, which represent the flow differences in different seasons. The slope matrix is ​​divided into different slope grade regions according to the slope condition, and different roughness coefficients are used for different slope grade regions. The seasonal flow characteristic value is assigned to the time window of the corresponding season, and the initial flow allocation ratio of the terrain area is calculated by combining the roughness coefficient of different slope grades. The initial flow allocation ratio is negatively correlated with the slope condition. Based on the initial flow distribution ratio and slope matrix, the initial velocity distribution of the terrain area is generated using a spatial interpolation method. The vegetation cover data of the terrain region is extracted from the layered grid attribute matrix. The vegetation resistance is adjusted to adjust the initial velocity distribution to generate the initial boundary conditions of the flow field in the terrain region. The initial boundary conditions of the flow field include comprehensive information on velocity magnitude and direction.

4. The method according to claim 1, characterized in that, The deduction of soil physicochemical property shifts and vegetation community structure variability based on the watershed hydrodynamic feature vector set includes: By using historical data correlation analysis, a set of correlation rules is established between the characteristics of watershed hydrodynamics and changes in soil properties. The set of correlation rules represents the correspondence between water flow velocity distribution and soil erosion intensity, sediment transport and soil deposition thickness, and water level fluctuation and soil moisture content. Based on the set of association rules, the watershed hydrodynamic feature vector set is subjected to multi-rule matching processing to generate the soil erosion and deposition intensity index and soil moisture fluctuation index for each grid cell; Based on the soil erosion and deposition intensity index, the loss of organic carbon caused by erosion and the accumulation of organic carbon caused by deposition are calculated. Combined with the soil carbon mineralization process, a record of changes in soil organic carbon content is generated. The record of changes in soil organic carbon content characterizes the annual and cumulative changes in organic carbon. Soil porosity changes are calculated based on the soil moisture fluctuation index, and soil bulk density change records are generated by combining the compaction effect information in the soil erosion and deposition intensity index. These soil bulk density change records characterize the bulk density differences in different regions. Based on the water level fluctuation characteristics, the annual inundation duration and average inundation depth of different elevation grid cells are calculated. The annual inundation duration is the ratio of the number of inundated days in a year to the number of days in a year, and the average inundation depth is the average depth during the inundation period. The annual inundation duration and average inundation depth are compared with the ecological adaptation range of vegetation types to generate a vegetation inundation stress index, which characterizes the degree to which vegetation is affected by inundation. The vegetation submersion stress index is used to calculate the probability of type change in vegetation community structure. The vegetation biomass change record and vegetation type change record are generated by combining the correlation between vegetation biomass and stress index. The soil organic carbon content change records and the soil bulk density change records are integrated to generate soil physicochemical property offset. The vegetation biomass change records and the vegetation type succession records are merged to obtain vegetation community structure variability. Both the soil physicochemical property offset and the vegetation community structure variability include grid location and time stamp information.

5. The method according to claim 4, characterized in that, The calculation of organic carbon loss due to erosion and organic carbon accumulation due to deposition based on the soil erosion-deposition intensity index, combined with the soil carbon mineralization process, generates a record of changes in soil organic carbon content, including: The erosion depth and deposition thickness are extracted from the soil erosion and deposition intensity index. The erosion depth is multiplied by the initial soil bulk density and the initial organic carbon content to obtain the organic carbon loss caused by erosion. The initial soil bulk density and the initial organic carbon content are obtained from the watershed geographic information dataset. The amount of organic carbon accumulation caused by deposition is obtained by multiplying the deposition thickness by the bulk density and organic carbon content of the sediment. The bulk density and organic carbon content of the sediment are derived from the particle composition information in the sediment transport characteristics. Soil temperature change data and soil moisture change data are extracted from the watershed hydrodynamic feature vector set, and the soil temperature change data and soil moisture change data represent the dynamic changes of the soil environment. The soil carbon mineralization rate is calculated based on the soil temperature change data and soil moisture change data, and the soil carbon mineralization rate is affected by the combined effects of temperature and moisture changes. The soil carbon mineralization rate is multiplied by the initial organic carbon content and the time interval to obtain the soil carbon mineralization amount, wherein the time interval is consistent with the time resolution of the soil organic carbon content change record. The annual change in organic carbon content is obtained by subtracting the amount of organic carbon lost due to erosion from the initial organic carbon content, adding the amount of organic carbon accumulated due to deposition, and then subtracting the amount of soil carbon mineralization. The annual change values ​​over several consecutive years are summed to obtain the cumulative change value. The annual change values ​​and the cumulative change values ​​are then integrated to generate a record of soil organic carbon content changes. The record of soil organic carbon content changes includes grid location and time stamp information.

6. The method according to claim 4, characterized in that, The process of comparing the annual inundation duration and average inundation depth with the ecological adaptation range of vegetation types to generate a vegetation inundation stress index includes: Ecological habit information of major vegetation types is extracted from watershed geographic information datasets to determine the suitable inundation duration range and suitable inundation depth range for each vegetation type. A vegetation ecological adaptation matrix is ​​constructed, which represents the ecological threshold of different vegetation types. The annual inundation duration is normalized, and the ratio of the actual annual inundation duration to the suitable inundation duration range is calculated to generate an inundation frequency index. The magnitude of the inundation frequency index represents the suitability of the inundation duration. The average flood depth is normalized using the same method to generate a flood depth index, and the calculation method of the flood depth index is consistent with that of the flood frequency index. Extract the water flow velocity magnitude of the grid cells in the water flow velocity distribution characteristics, calculate the ratio of water flow velocity magnitude to vegetation erosion resistance, and generate a water flow stress index, which characterizes the erosion effect of water flow on vegetation. Based on the sensitivity differences of vegetation types, the inundation frequency index, inundation depth index, and water flow stress index are weighted and integrated to generate the vegetation inundation stress index. The vegetation submersion stress index is subjected to numerical range control processing, and the values ​​that exceed the reasonable range are adjusted to the boundary value to generate the final vegetation submersion stress index.

7. The method according to claim 1, characterized in that, The carbon cycle process is digitally quantified by coupling the soil physicochemical property offsets and the vegetation community structure variability, generating a carbon pool storage change sequence, including: Key carbon exchange pathways are extracted from the soil physicochemical property offsets and the vegetation community structure variability by carbon cycle pathway identification. These key carbon exchange pathways characterize the carbon transfer process between different pools. Flux was calculated for each key carbon exchange pathway. The flux of the vegetation photosynthetic carbon fixation pathway was calculated by correlating vegetation biomass change records with photosynthetic efficiency information, and the flux of the litter decomposition input pathway was calculated by correlating vegetation biomass change records with litter ratio information. The fluxes of each path are categorized and summarized according to the carbon pool type, generating the sum of input fluxes and the sum of output fluxes for different carbon pools. The sum of input fluxes is the cumulative result of the fluxes of all input paths, and the sum of output fluxes is the cumulative result of the fluxes of all output paths. The net flux of each carbon pool is calculated based on the difference between the sum of input flux and the sum of output flux, and the net flux represents the increase or decrease of the carbon pool. The change in carbon pool storage is calculated based on the net flux and the initial carbon pool storage, wherein the initial carbon pool storage is obtained from the watershed geographic information dataset, and the change in storage is the sum of the products of the initial storage, the net flux, and the time interval. The changes in carbon pool storage are processed by spatiotemporal scale transformation, converting short-term changes at the grid scale into long-term changes at the watershed scale, generating carbon pool storage change sequences containing different carbon pool types. Each data point in the carbon pool storage change sequence includes a time stamp and a carbon pool type identifier.

8. The method according to claim 7, characterized in that, The flux calculation for each key carbon exchange pathway, wherein the flux of vegetation photosynthetic carbon sequestration pathway is obtained by correlating vegetation biomass change records with photosynthetic efficiency information, includes: Extract the annual increments of aboveground biomass and belowground biomass from the vegetation biomass change records in the vegetation community structure variability, and calculate the annual increment of total biomass, which is the sum of the annual increments of aboveground and belowground biomass. Based on the vegetation type, the corresponding carbon content ratio parameter is retrieved from the preset parameter library, and the annual increase in total biomass is multiplied by the carbon content ratio parameter to obtain the annual increase in total carbon. The corresponding photosynthetic efficiency parameter is retrieved from the preset parameter library according to the vegetation type. The photosynthetic efficiency parameter characterizes the vegetation’s ability to utilize light energy. The annual total solar radiation data is extracted from the watershed geographic information dataset, and the potential photosynthetic carbon sequestration is obtained by multiplying the annual total solar radiation data by the photosynthetic efficiency parameter. The changes in carbon dioxide dissolved amount are calculated based on the reservoir surface area change records in the hydropower project operation dataset. The local changes in atmospheric carbon dioxide concentration are deduced from the changes in carbon dioxide dissolved amount, and a carbon dioxide concentration correction coefficient is generated. The actual photosynthetic carbon sequestration is obtained by multiplying the potential photosynthetic carbon sequestration by the carbon dioxide concentration correction factor. The vegetation respiration consumption is extracted from the vegetation biomass change record. The vegetation respiration consumption is the product of the annual increase in total carbon and the respiration consumption ratio. The respiration consumption ratio varies among different vegetation types. The vegetation photosynthetic carbon fixation pathway flux is obtained by subtracting the vegetation respiration consumption from the actual photosynthetic carbon fixation amount. The vegetation photosynthetic carbon fixation pathway flux characterizes the net carbon fixation capacity of vegetation.

9. The method according to claim 8, characterized in that, The process of obtaining the input path flux for litter decomposition includes: The dominant vegetation type of the current grid cell is determined from the vegetation type change record in the vegetation community structure variability. The dominant vegetation type is the vegetation type with the largest proportion in the grid cell. The corresponding litter ratio parameter is retrieved from the preset parameter library according to the dominant vegetation type. The litter ratio parameters of different types of vegetation are different, with the ratio parameter of deciduous vegetation being higher than that of evergreen vegetation. The aboveground biomass stock is extracted from the vegetation biomass change record, and the aboveground biomass stock is multiplied by the litter ratio parameter to obtain the litter yield, which represents the total amount of litter in the year. The litter production was divided into different litter components, and the carbon content of the different components varied. Soil temperature change records and soil moisture change records are extracted from the soil physicochemical property offsets, and decomposition environmental parameters are calculated. These decomposition environmental parameters characterize the combined effects of temperature and moisture on the decomposition process. The litter yield of each component is multiplied by the corresponding carbon content and decomposition environment parameters to obtain the component decomposition flux. The decomposition flux of each component is then summed to generate the litter decomposition input path flux.

10. A computer system, characterized in that, include: A memory, wherein a computer program is stored; A processor is used to load the computer program to implement the modeling method for monitoring ecosystem carbon pools under the influence of hydropower projects as described in any one of claims 1-9.

Images