Forest ecosystem carbon sink future scene simulation method and system

Abstract

The invention provides a forest ecosystem carbon sink future scene simulation method and system, and belongs to the technical field of administrative management purpose information. Based on a high-precision curved surface simulation HASM method, processing the climate data and the soil data respectively, and extracting pixel values of the position of the forest sample plot in the climate raster data and the soil raster data into forest sample plot data to obtain a forest sample plot environment information data set; designing a plurality of future scenes of the forest ecosystem under different management measures; and simulating the carbon density of each future scene by using a Forest-DNDC model, predicting the carbon density of other areas except a forest sample plot by using a random forest model, and carrying out HASM simulation by taking a prediction result as an initial field and taking a simulation result output by the Forest-DNDC model as a control condition to obtain forest and soil carbon density raster data in the future scenes. According to the method, the simulation precision can be effectively improved, and a scientific basis is provided for formulating county scale forest management and climate alleviation policies in the future.

Description

Technical Field

[0001] The present application relates to the field of information technology applicable to administrative management purposes, and in particular to a method and system for simulating future scenarios of carbon sinks in forest ecosystems. Background Art

[0002] Forests not only play a key role in achieving carbon neutrality by fixing carbon dioxide in the atmosphere, but also have a profound impact on the global carbon balance and the long-term control of climate change.

[0003] Although a large number of studies have explored the potential of forest carbon sinks and their importance in mitigating climate change, how to obtain future scenario simulation data of forest ecosystem carbon sinks at the county scale to support the study of the evolution of forest carbon density at the subtropical county scale under different climate change scenarios, especially the systematic analysis of the long-term impacts of mild and extreme climate scenarios on forest carbon density, and to support the complexity analysis of the spatial and temporal dimensions of the effects of different management policies under different climate scenarios at the subtropical county scale, is an unresolved problem.

[0004] Therefore, it is necessary to provide an improved technical solution to address the above-mentioned deficiencies in the prior art. Summary of the invention

[0005] The purpose of this application is to provide a method and system for simulating future scenarios of forest ecosystem carbon sinks to alleviate the problems existing in the above-mentioned prior art.

[0006] In order to achieve the above objectives, this application provides the following technical solutions:

[0007] The present application provides a method for simulating future scenarios of forest ecosystem carbon sinks, which uses the Forest-DNDC model as a scenario simulation model and includes the following steps:

[0008] Step S1: Obtain climate data, soil data, forest plot data, and management measures in the study area;

[0009] Step S2: Based on the high-accuracy surface modeling (HASM) method, the climate data and soil data are processed respectively to obtain climate raster data and soil raster data of specified resolution, and the pixel values ​​of the forest plot location in the climate raster data and the soil raster data are extracted into the forest plot data to obtain the forest plot environmental information data set;

[0010] Step S3: Based on the actual management measures recorded in the historical forest management of the study area, multiple scenarios of the future of the forest ecosystem under different management measures are designed;

[0011] Step S4: Based on the forest plot environmental information data set, considering various future climate change situations, the Forest-DNDC model is used to simulate the carbon density of each future scenario, and the simulation results of the soil and forest carbon density of all forest plots under different climate changes and different scenarios are obtained year by year;

[0012] Step S5: Input the simulation results of soil and forest carbon density of forest plots into the pre-trained random forest model, and combine with the environmental information data set of forest plots to predict the carbon density of other areas outside the forest plots, and obtain the spatial distribution raster data of soil carbon density and forest carbon density covering the entire study area under future scenarios;

[0013] Step S6: Using the spatially distributed raster data as the initial field and the simulated results of forest plot soil and forest carbon density output by the Forest-DNDC model as control conditions, a HASM simulation of the forest ecosystem carbon sink under future scenarios is performed to obtain high-precision future scenario raster data of forest and soil with a specified resolution.

[0014] In a possible implementation, in step S2, the climate data is processed based on the HASM method, including the following steps:

[0015] The historical data of all meteorological stations in the climate data are regressed and simulated, and then the inverse distance weighted method is used for interpolation to obtain the daily-scale climate trend surface with a specified resolution;

[0016] The residual of each meteorological station is obtained by subtracting the pixel predicted value at the location of each meteorological station in the daily climate trend surface from the actual observation value of the station.

[0017] Several meteorological stations in and around the study area were selected, and the spatial range of the selected meteorological stations was used to crop the daily-scale climate trend surface;

[0018] The clipping results are used as the climate initial field of the HASM method, and the residuals of multiple selected meteorological stations are used as the climate control conditions of the HASM method. The HASM simulation is carried out to obtain the daily-scale residual surface with a specified resolution.

[0019] The daily climate trend surface and the residual surface are summed to obtain the climate grid data of the specified resolution.

[0020] In a possible implementation, in step S2, the soil data is processed based on the HASM method to obtain soil grid data of a specified resolution, including:

[0021] Calculate the organic carbon density at different soil depths at each soil sampling point;

[0022] The calculated soil organic carbon density was used as the dependent variable, and multiple soil properties and environmental factors of the soil sampling point data were used as independent variables. The random forest model was used to predict the spatial distribution of soil organic carbon at different soil depths.

[0023] Taking the spatial distribution of soil organic carbon at different soil depths as the initial field and the soil sampling point data as the control condition, HASM simulation of soil organic carbon was carried out to obtain the soil organic carbon grid surface with a specified resolution.

[0024] Ordinary linear regression method was used to obtain trend surface data of other soil parameters. The soil census data and soil sampling point data were used as control conditions to carry out HASM simulation on other soil parameters to obtain grid surfaces of other soil parameters with specified resolution.

[0025] In a possible implementation, in step S4, before using the Forest-DNDC model to simulate the carbon density changes of each future scenario, the method further includes: improving the management measures module of the Forest-DNDC model, specifically as follows:

[0026] In the management measure module of the Forest-DNDC model, a determination condition is added to each management measure, and the management measure time is set as a cycle. The determination condition is used to determine whether the management measure should be executed.

[0027] In a possible implementation, in step S4, before simulating the carbon density of each future scenario using the Forest-DNDC model, the method further includes: performing a high-precision inversion of the forest physiological phenological parameters of the base year, as follows:

[0028] For each forest plot in the forest plot environmental information dataset, the forest carbon density is calculated using the empirical formula based on the measured data of the forest plot.

[0029] Determine the inversion parameter dataset of the current forest plot;

[0030] Traverse the inversion parameter data sets in order and perform the following steps for each set of parameters:

[0031] The current group parameters are set as the physiological phenological parameters of the Forest-DNDC model, and simulation is performed based on the Forest-DNDC model to obtain the simulated value of forest carbon density;

[0032] Calculate the error between the simulated forest carbon density value and the calculated forest carbon density value. If the error does not meet the preset threshold requirement, set the next set of parameters as the physiological phenological parameters of the Forest-DNDC model, and simulate again with the Forest-DNDC model to obtain a new simulated forest carbon density value. Calculate the error between the new simulated forest carbon density value and the calculated forest carbon density value until the error meets the preset threshold requirement, and end the cycle.

[0033] The corresponding physiological phenological parameters at the end of the cycle are taken as the final physiological phenological parameters of the current forest plot.

[0034] In a possible implementation, determining the inversion parameter dataset of the current forest plot includes:

[0035] The dominant forest species and forest age of the current forest plot are input into the Forest-DNDC model to obtain the default physiological phenological parameters generated by the Forest-DNDC model based on the empirical model, and then the forest carbon density simulation value is generated based on the default physiological phenological parameters;

[0036] Compare the size relationship between the simulated forest carbon density value and the calculated forest carbon density value, record the current default physiological phenological parameter value as the upper or lower limit of each parameter value according to the size relationship, and change the forest age value according to the size relationship, and re-input the changed forest age into the Forest-DNDC model to obtain new default physiological phenological parameters, use the new default physiological phenological parameters for simulation, and obtain new forest carbon density simulation values;

[0037] Compare the new simulated forest carbon density value with the calculated forest carbon density value again. If the relationship between the new simulated forest carbon density value and the calculated forest carbon density value changes, record the new default physiological phenological parameters as the lower or upper limits of each parameter, and then obtain the value range of each physiological phenological parameter based on the upper and lower limits of each parameter; otherwise, continue to change the forest age and re-simulate until the relationship between the new simulated forest carbon density value and the calculated forest carbon density value changes, and stop the cycle;

[0038] According to the value range of each physiological phenological parameter, multiple possible values ​​of each physiological phenological parameter are determined using an equal step method starting from the lower limit, and all possible values ​​of all physiological phenological parameters are combined into an inversion parameter data set for the current forest plot.

[0039] In a possible implementation, in step S5, before inputting the simulation result into the pre-trained random forest model, the following steps are further included:

[0040] At fixed annual intervals, the simulation results of soil and forest carbon density of all forest plots under different climate change and different scenarios were selected, and the data were formatted to obtain the forest plot simulation data set;

[0041] The formatted forest plot simulation dataset was input into the pre-trained random forest model.

[0042] This embodiment provides a future scenario simulation system for forest ecosystem carbon sinks, which uses the Forest-DNDC model as a scenario simulation model, including:

[0043] Data acquisition module, used to obtain climate data, soil data, forest plot data, and management measures in the study area;

[0044] The data processing module is used to process the climate data and soil data based on the high-accuracy surface modeling (HASM) method to obtain the climate grid data and soil grid data of the specified resolution, and extract the pixel values ​​of the forest plot location in the climate grid data and the soil grid data into the forest plot data to obtain the forest plot environmental information data set;

[0045] The scenario setting module is used to design multiple scenarios for the future of forest ecosystems under different management measures based on the actual management measures recorded in the historical forest management of the study area;

[0046] The sample plot carbon density simulation module is used to simulate the carbon density of each future scenario using the Forest-DNDC model based on the forest sample plot environmental information data set and considering various future climate change situations, so as to obtain the simulation results of the soil and forest carbon density of all forest sample plots under different climate changes and different scenarios year by year;

[0047] The carbon density initial field prediction module is used to input the simulation results of soil and forest carbon density in forest plots into the pre-trained random forest model, and combine it with the forest plot environmental information dataset to predict the carbon density of other areas outside the forest plot, and obtain the spatial distribution raster data of soil carbon density and forest carbon density covering the entire study area under future scenarios;

[0048] The carbon density downscaling module is used to use the spatially distributed raster data as the initial field and the simulated results of soil and forest carbon density of forest plots output by the Forest-DNDC model as control conditions to perform HASM simulation of forest ecosystem carbon sinks under future scenarios, and obtain high-precision future scenario raster data of forests and soils with specified resolution.

[0049] The technical solution of the embodiment of the present application has the following beneficial effects:

[0050] First, because the Forest-DNDC model is specially optimized for forest ecosystems, it provides multi-process coupling simulation capabilities, flexible input and management settings, and high-precision micro-process descriptions. It can better simulate tree species characteristics, phenological changes, and forest physiological processes. Therefore, this model can be used for forest carbon sink simulation with higher accuracy.

[0051] Second, the HASM method can be used to interpolate climate and soil data more accurately, making the data more continuous and smooth in space. Compared with traditional interpolation methods, HASM is more suitable for processing data in complex terrain areas and can improve the spatial resolution and accuracy of data. By obtaining accurate climate, soil and forest plot data, the uncertainty of input data in the model simulation process can be significantly reduced.

[0052] Third, the pixel values ​​of the forest plot locations in the climate raster data and soil raster data are extracted into the forest plot data to achieve spatial alignment of the data. This can ensure the consistency of the forest plot data with the data of its environment, guarantee the applicability of the model under different geographical conditions, and improve the reliability of the model simulation results.

[0053] Fourth, scenario design based on real data and historical management experience can make the scenario design more in line with the actual situation. At the same time, by designing multiple management measure scenarios, the impact of each measure on forest carbon sequestration capacity can be quantified.

[0054] Fifth, the use of forest plot environmental information datasets (including soil, climate, tree species, etc.) enables the Forest-DNDC model to perform accurate simulations based on the conditions of specific plots, capture subtle differences between different plots under future climate change conditions, and improve the accuracy of simulation results.

[0055] Sixth, the simulation results of soil and forest carbon density in forest plots were input into the pre-trained random forest model, and combined with the forest plot environmental information dataset, the carbon density of areas outside the forest plots was predicted. The random forest model's ability to handle nonlinear and complex relationships was used to effectively capture the impact of environmental factors on carbon density. The carbon density simulation results based on limited plots can be extended to the entire study area, and then used as the initial field for simulation using the HASM method to generate future scenario raster data of forest and soil carbon density with a more refined spatial pattern.

[0056] In summary, the method provided in this application improves the spatial resolution and accuracy of input data and the spatial alignment of input data through the HASM method, providing reliable and high-quality input data for the Forest-DNDC model. On this basis, Forest-DNDC is used as a scenario simulation model, combined with multi-scenario design and different climate change conditions to simulate the future scenarios of each sample plot, and the random forest model and HASM method are used to expand the simulation results of Forest-DNDC from the sample plot location to the entire study area, and finally obtain high-precision future scenario raster data of forest and soil carbon density under different climate change conditions in the study area, providing scientific support and basis for decision-making departments to respond to climate change and formulate carbon management strategies. At the same time, this method provides technical support for revealing the comprehensive impact of management measures on forest and soil carbon density under different climate change scenarios (such as SSP2-4.5 and SSP5-8.5) at the typical subtropical county scale. It helps to combine the spatial heterogeneity of different altitudes, tree species structure and management strategies to comprehensively evaluate the dynamic changes in carbon density of forest ecosystems at the typical subtropical county scale under different climate scenarios, provide a scientific basis for future forest management and climate mitigation policies at the subtropical county scale, and help achieve the global carbon neutrality goal. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 A schematic flow chart of a method for simulating future scenarios of forest ecosystem carbon sinks provided according to some embodiments of the present application.

[0058] Figure 2 Schematic diagram of the spatial distribution of forest plots in the study area.

[0059] Figure 3 Schematic diagram of the spatial distribution of soil organic carbon at different soil depths.

[0060] Figure 4 Schematic diagram of the process of Forest-DNDC inverting forest physiological phenological parameters.

[0061] Figure 5 This is the spatiotemporal distribution map of forest carbon density under the natural development scenario under SSP2-4.5 climate conditions.

[0062] Figure 6 This is the spatiotemporal distribution map of soil carbon density under the natural development scenario under SSP2-4.5 climate conditions.

[0063] Figure 7 This is a graph showing the changing trend of the impact of different climate conditions on forest carbon density under natural development scenarios.

[0064] Figure 8 This is a graph showing the changing trend of the impact of different climate conditions on soil carbon density under natural development scenarios. DETAILED DESCRIPTION

[0065] For ease of understanding, some terms involved in this embodiment are explained below. These explanations are only exemplary and should not be construed as limitations on the technical solutions of this application.

[0066] Forest ecosystem carbon sink refers to the process by which forests absorb carbon dioxide (CO2) from the atmosphere through photosynthesis and fix it in plants and soil.

[0067] Forest carbon density refers to the sum of the carbon density of the above-ground parts of trees, the carbon density of the underground parts of trees, the carbon density of the understory vegetation layer and the carbon density of the litter layer.

[0068] Soil carbon density mainly includes soil organic carbon, which is carbon in organic form in the soil. It is mainly converted from plant residues, dead branches and leaves, and root decomposition products. These organic substances are decomposed by soil microorganisms to form stable organic carbon, which can be stored in the soil for a long time.

[0069] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0070] The embodiments of the present application are described below in conjunction with the accompanying drawings.

[0071] Embodiment 1:

[0072] This embodiment provides a method for simulating future scenarios of forest ecosystem carbon sinks. Figure 1 As shown, the method includes:

[0073] Step S1: Obtain climate data, soil data, forest plot data, and management measures for the study area.

[0074] In this embodiment, the study area can be any geographical area, for example, it can be a region of global, national, or provincial scales. Preferably, the study area of ​​this embodiment is a region of county scale. Regions of global, national, and provincial scales are usually able to obtain more comprehensive and standardized data, and their models often use larger spatial units (i.e., lower spatial resolutions, such as low resolutions of 5 km and below), which are suitable for rough estimates. Compared with the global, national, and provincial scales, the simulation of forest ecosystem carbon sinks at the county level lacks high-resolution spatial data (e.g., high resolution of 100 m and above), and the data lacks systematicity and consistency, resulting in insufficient simulation accuracy and inability to provide effective decision-making support and scientific research basis.

[0075] In this embodiment, the technical solution is described in detail by taking Z County, J Province, which is located in the subtropical zone, as an example.

[0076] In this embodiment, the climate data includes historical climate site data and future climate scenario data.

[0077] The historical climate station data can be derived from the historical observation data of each meteorological station in Province J. The information recorded by each station can include daily-scale temperature, daily-scale precipitation, station number, longitude and latitude coordinates of the location, and date of recording data. Exemplarily, the historical climate station data of Province J includes 173 meteorological stations and observation data from 1957 to 2019.

[0078] The future climate scenario data comes from the official website of CMIP6, which provides multiple climate scenarios for simulating future climate change. Among them, SSP1-2.6 is a low-emission path, which mainly considers the more optimistic global emission reduction efforts, SSP2-4.5 represents the "intermediate development route", also known as the "sustainable development but not extreme scenario" and SSP5-8.5 "fossil fuel-driven economic development scenario", which represents the high-emission scenario of "economy takes precedence over environmental protection". SSP1-2.6 is a low-emission path, and the scenario is more idealistic, while SSP2-4.5 and SSP5-8.5 cover moderate and extreme climate change scenarios. These two scenarios can provide more diverse results for research, and can simultaneously show milder and more severe climate change impacts. Therefore, the future climate scenario data of this embodiment includes two future climate scenarios, SSP2-4.5 and SSP5-8.5, which can cover realistic scenarios from moderate emission reduction to no emission reduction, and the simulation of future change trends is more pragmatic, reflecting a wider range of possibilities.

[0079] Soil data include soil sampling measured data and soil profile data. Among them, the soil sampling measured data comes from the soil sampling measured data of two actual forest sample plots in the study area in 2023 and 2024, including multiple measured sample points (such as 37 sample points). The soil measurement indicators in each measured sample point include soil profile thickness, pH value, electrical conductivity, water content, organic carbon, soil total nitrogen, soil bulk density and soil mechanical composition. The soil profile data comes from the second national soil census data, which also includes multiple soil slope sample points. The total soil profile data of Z County, J Province, includes 74 sample points.

[0080] The data of forest plots (hereinafter referred to as plots and sample plots) come from multiple fixed plot forest resource inventories in the history of the study area (from 1977 to 2019), and its attribute information is forest information, such as dominant tree species, tree age, etc. Specifically, all plots in the forest plot data are square, with an area of ​​0.08 hectares, marked by three positioning trees in the southwest corner. For each sample plot, the sample collection team needs to fill in two main tables (i.e., inventory tables). The first table is the "Environmental Factor Record Table", which records information including plot ID, longitude, latitude, plot type, attributes, land type, landform type, elevation, slope aspect, slope position, soil name, soil thickness, forest species, source, dominant tree species, crown density, age group, average age, etc. The second table is the "Per Tree Measure Table", which records the ID of the sample tree, the tree species, and the diameter at breast height (DBH). If there are special tree species in the study area in a specific sample plot, such as bamboo forest, the sample collection team will also record the number of full-size bamboo and the number of mixed bamboo. For example, the spatial distribution of forest plots in the study area is as follows: Figure 2 shown.

[0081] The management measures are derived from the actual records of forest management in the study area over the years, including felling methods, felling areas, planting types, and planting ratios for different tree species. Specifically, when statistically analyzing the dominant tree species in all forest plots in the study area in 2019, combined with the characteristics of the subtropical climate, taking Z County, J Province as an example, the key targets considered were fir, pine, and bamboo. Therefore, only these three tree species were considered when classifying tree species, and other tree species were uniformly classified as "other tree species". There are 61 plots with Chinese fir as the dominant tree species, 5 plots with pine, 62 plots with bamboo, and 117 plots with other forest species as the dominant tree species (specifically, there is 1 plot with Chinese chestnut as the dominant tree species, 1 plot with oak, 1 plot with paulownia, 1 plot with arbor, 2 plots with liquidambar formosana, 3 plots with Schima superba, 4 plots with Alnus pseudoalpinia, 7 plots with Castanopsis fargesii, 8 plots with Castanopsis chinensis, 51 plots with Alnus rapa, 13 plots with other soft broad-leaved trees, and 71 plots with other hard broad-leaved trees). There are corresponding planting and felling management measures for Chinese fir and other tree species, while there are only annual felling policies for bamboo and pine, but no tree planting is carried out. The actual recorded management measures in the county are: 10% of fir is cut down each year, and 9.3% is planted; 12% of pine is cut down each year, and no planting is done; 0.5% of bamboo is cut down each year, and no planting is done; and other tree species are cut down an average of 2.5% each year, and 2% is planted.

[0082] Step S2: Based on the HASM method, the climate data and soil data are processed respectively to obtain climate raster data and soil raster data of specified resolution, and the pixel values ​​of the forest sample plot in the climate raster data and the soil raster data are extracted into the forest sample plot data to obtain the forest sample plot environmental information dataset.

[0083] In this embodiment, the purpose of step S2 is to use the HASM method to perform high-precision processing on the input data of the model (climate data, soil data) to improve the quality of the input data, thereby improving the accuracy of the model simulation and achieving the best simulation results.

[0084] Among them, specifying the resolution means explicitly setting the spatial resolution of climate raster data and soil raster data, that is, specifying the level of detail of the data in space, which is usually expressed as the actual area covered by each pixel (or each grid cell). For example, in climate raster data, if the specified resolution is 1km, it means that each grid cell represents an area of ​​1km square. In the field of geographic information processing, the higher the spatial resolution (such as 1km×1km), the more details the data can reflect and the more subtle changes can be captured; conversely, a lower resolution (such as 25km×25km) may not be able to effectively reflect local characteristics and changes.

[0085] It should be noted that, in this embodiment, the spatial resolution of the future climate scenario data provided by CMIP6 is 25km, which cannot be applied to the simulation of forest ecosystem carbon sinks at the county scale. The meteorological stations and soil data are point data with uneven spatial distribution and limited quantity, which makes it difficult to meet the high-precision simulation requirements covering the entire domain. Therefore, this embodiment uses the HASM method to process the climate data and soil data respectively to obtain climate raster data and soil raster data of specified resolution (high resolution) to improve data quality.

[0086] Specifically, firstly, high-precision simulation of climate elements is carried out, including the following steps:

[0087] Step S211: Perform regression simulation on the historical data of all meteorological stations in the climate data, and then use the Inverse Distance Weighting (IDW) method to interpolate to obtain a daily-scale climate trend surface with a specified resolution.

[0088] It should be noted that, in this embodiment, the coverage area of ​​all meteorological stations may be larger than the coverage area of ​​the study area. For example, if the study area is a county, all meteorological stations may cover the entire province where the county is located.

[0089] By using historical data from meteorological stations for regression simulation, we can better capture the long-term trend of climate variables and reduce the impact of short-term fluctuations and local drastic changes on the results. The IDW method can generate a relatively accurate and continuous climate trend surface by considering the information of multiple stations around each station, which helps to eliminate the limitations of only considering data from some meteorological stations and provide more consistent spatial distribution information.

[0090] In this embodiment, the specified resolution may be, for example, a spatial resolution of 90 m.

[0091] In order to improve the calculation efficiency, we can first set an intermediate-scale spatial resolution, for example, set a spatial resolution of 1 km, use all stations (i.e., 173 meteorological stations) to regress and simulate the daily-scale temperature and precipitation data of Province J and its surrounding areas, and obtain the daily-scale temperature and precipitation data with a resolution of 1 km. Then use the HASM method to downscale the daily-scale temperature and precipitation data with the intermediate-scale spatial resolution to the specified resolution (high resolution).

[0092] Illustratively, the equations used to return daily-scale temperature and precipitation data are as follows:

[0093] pr=a pr ×X+b pr ×Y+D pr (1)

[0094] tas=a tas ×X+b tas ×Y+c tas ×DEM+D tas (2)

[0095] Among them, X represents longitude, Y represents latitude, pr represents the precipitation forecast value, and a pr represents the longitude coefficient in the precipitation regression equation, b pr represents the latitude coefficient in the precipitation regression equation, D pr represents the intercept in the precipitation regression equation; tas represents the temperature prediction value, a tas Indicates the longitude coefficient in the regression temperature equation, b tas represents the latitude coefficient in the temperature regression equation, c tas represents the coefficient of DEM in the temperature regression equation, D tas represents the intercept in the temperature regression equation.

[0096] Based on the above equation, the least squares regression is used to obtain the coefficients of the regression equation, and based on the regression equation, the daily-scale climate trend surface of the specified resolution is obtained by the inverse distance weighted method.

[0097] Step S212: Use the pixel prediction value at the location of each meteorological station in the daily climate trend surface to subtract the actual observation value of the station to obtain the residual of each meteorological station.

[0098] Step S213: Select multiple meteorological stations in the study area and its surrounding areas (for example, 38 meteorological stations around the study area can be selected), and the spatial range of the selected meteorological stations is used for cropping in the daily-scale climate trend surface. To improve efficiency, cropping can be performed in the daily-scale climate trend surface with an intermediate spatial resolution to ensure that the climate data used for analysis matches the actual conditions of the study area, while ensuring the focus on specific areas. That is, the cropping process allows researchers to focus on the climate characteristics of specific areas, avoid unnecessary data processing, and improve the efficiency of HASM calculations.

[0099] Step S214: Using the clipping result as the climate initial field of the HASM method, using the residuals of multiple selected meteorological stations as the climate control conditions of the HASM method, launching the HASM simulation, and obtaining a daily-scale residual surface of a specified resolution.

[0100] Step S215: summing the daily-scale climate trend surface and the residual surface to obtain climate grid data of a specified resolution.

[0101] Specifically, the climate grid data (daily-scale high-precision climate surface data) of a specified resolution can be calculated from the following equation:

[0102] Surface=Surf IDW +Surf HASM (3)

[0103] Among them, Sruface represents the climate grid data of the specified resolution, that is, the high-precision grid surface data of temperature or precipitation on a daily scale, Surf IDW represents the daily climate trend surface obtained by interpolation using the IDW method, that is, the trend surface of daily temperature or precipitation. HASM Represents the daily-scale residual surface of the specified resolution obtained by interpolation using the HASM method, that is, the residual surface of daily temperature or precipitation.

[0104] The above is the process of high-precision simulation of climate elements. This process combines historical data analysis, spatial interpolation and regional clipping, HASM method and surface summation ideas to form a complete set of data integration methods, making climate data closer to research objectives and actual needs.

[0105] Secondly, based on the HASM method, the soil data is processed to obtain soil raster data of specified resolution, including:

[0106] Step S221: Calculate the organic carbon density at different soil layer depths at each soil sampling point.

[0107] The organic carbon density at each soil depth was calculated using the following formula:

[0108] SOC stocks =SOC×BD×d×(1-G) (4)

[0109] Among them, SOC stocks is the soil organic carbon density, in kg / m 2 , SOC is the organic carbon content, the unit is g / kg, BD is the soil bulk density, the unit is g / cm 3 , d is the soil thickness in cm, and G is the volume ratio of gravel with a diameter greater than 2 mm (%). This formula is also called the Rodríguez-Murillo formula. It combines multiple key soil parameters to more accurately reflect the organic carbon storage in the soil and reduce the impact of single parameter fluctuations on the calculation of soil organic carbon density. Its structure is simple and easy to understand, and does not require complex models and a large amount of input data, making it easy to apply in actual operations.

[0110] Step S222: Using the calculated soil organic carbon density as the dependent variable and multiple soil properties and environmental factors of the soil sampling point data as independent variables, a random forest model is used to predict the spatial distribution of soil organic carbon at different soil depths.

[0111] Among them, different soil layer depths include soil layer depths of 0-5cm and 0-100cm.

[0112] The spatial distribution of soil organic carbon at 0-5cm and 0-100cm was predicted by random forest combined with plot information. The random forest used 74 soil organic carbon (SOC) sample points as dependent variables and a total of 23 independent variables were input, namely: mountain shadow, slope, aspect, slope length, valley depth, elevation, convergence index, clay content (0-5cm, 0-100cm), sand content (0-5cm, 0-100cm), pH (0-5cm, 0-100cm), soil type, soil thickness, land use type, forest type, canopy density, average forest age, plane curvature, contour curvature, multi-year average rainfall, and terrain moisture index. Its sources are shown in Table 1, which is as follows:

[0113] Table 1 Random forest input data sources

[0114]

[0115]

[0116] The above data are input into the random forest model, and the final prediction result is the spatial distribution of soil organic carbon at different soil depths with high precision. The spatial resolution is the specified resolution, i.e. 90m. The soil organic carbon grid data includes two different soil depths: 0-5cm and 0-100cm. Figure 3 shown.

[0117] In this embodiment, in order to address the problem that ordinary regression models are difficult to express the nonlinear relationship between soil organic carbon density and soil properties and environmental factors, random forests are used as prediction models, and predictions are performed by integrating multiple decision trees. The input data uses multiple soil properties (such as soil type, texture, pH value, organic matter content) and environmental factors (such as temperature, precipitation, topography, vegetation type) as independent variables, so that the model can comprehensively consider the interaction of these factors, thereby more accurately estimating the spatial distribution of soil organic carbon.

[0118] Furthermore, this application also verified the accuracy of the model through a ten-fold cross-validation method, and the results showed that: SOC of 0-5cm soil layer: MSE: 35.516, MAE: 4.468, R2: 0.561, RMSE: 5.960; total SOC of all soil layers: MSE: 9.518, MAE: 2.221, R2: 0.530, RMSE: 3.085, indicating that the accuracy of the prediction results of the random forest model can meet the needs of HASM simulation.

[0119] Step S223: Using the spatial distribution of soil organic carbon at different soil depths as the initial field and the soil sampling point data as the control condition, a HASM simulation is performed on the soil organic carbon to obtain a soil organic carbon grid surface of a specified resolution.

[0120] The spatial distribution of soil organic carbon at different soil depths is used as the initial field to ensure that the starting point of the simulation is the best estimate based on the existing data, thereby improving the accuracy of the model. The soil sampling point data is used as a control condition, so that the HASM simulation can be closely interpolated and extended around the observed data, reducing the generation of deviations in the simulation process. At the same time, HASM can handle the non-uniformity in spatial data well, so that the generated soil organic carbon grid surface more realistically reflects the distribution characteristics of soil carbon in different regions and at different depths.

[0121] Step S224: Use the common linear regression method to obtain trend surface data of other soil parameters, use the soil census data and soil sampling point data as control conditions, carry out HASM simulation on other soil parameters, and obtain other soil parameter grid surfaces of specified resolution.

[0122] Among them, the general linear regression method can provide trend surface data for other soil parameters. As an initial estimate, this trend surface can capture the overall change trend of soil parameters and provide a reasonable starting point for subsequent HASM simulation. Using the soil census data and soil sampling point data as control conditions can ensure that the simulation process is closely dependent on real observation data, thereby improving the accuracy of the simulation results and ensuring that other soil parameter grid surfaces also meet the needs of carbon sinks in the forest ecosystem of the study area.

[0123] Finally, the pixel values ​​of the forest plot locations in the climate raster data (including temperature and precipitation) and soil raster data (including soil organic carbon and other soil parameters) obtained in the above steps are extracted into the forest plot data, so that the environmental information of all forest plots is consistent in spatial position and at a higher spatial resolution, and a forest plot environmental information dataset is obtained.

[0124] Step S3: Based on the actual recorded management measures of historical forest management in the study area, multiple scenarios of the future of the forest ecosystem under different management measures are designed.

[0125] The purpose of designing multiple scenarios is to better simulate and depict the future scenarios of carbon density of forest ecosystems in the study area under different development paths driven by the coupling of climate change and human activities. Since the management measures of the actual records of historical forest management are derived from real data, scenario design based on real data and the use of historical management records of the study area can make the scenario design more in line with the actual situation, thereby enabling the Forest-DNDC model to more accurately predict the carbon sequestration capacity and change trend of forest ecosystems under different future management strategies.

[0126] Exemplarily, three scenarios can be designed: natural development scenario, economic development priority scenario and management optimization scenario (maximum species diversity scenario).

[0127] Among them, the natural development scenario refers to the future scenario of the change in carbon density of the forest ecosystem in the study area under the actual management measures at the current stage of the study area. Taking Z County, J Province as an example, in this scenario, the amount of bamboo felling is extremely small, and the average annual felling in the county does not exceed 0.5%, but the bamboo is widely distributed. After visiting and field analysis, there are several reasons for this phenomenon: 1) The local bread industry is developed, most people are engaged in bread-related industries, and there is a shortage of manpower for bamboo felling; 2) Inconvenient transportation, most bamboo forests are distributed in relatively steep and dangerous terrains, and the felling is difficult and the risk factor is high; 3) Low profit, the current bamboo felling cost is high, and the purchase price of bamboo has declined compared with previous years, resulting in a small profit margin. For the above reasons, although the county has a large amount of bamboo, the felling volume is small. In addition, bamboo is not planted because bamboo itself reproduces naturally at a faster rate and does not require labor costs to plant. In this scenario, the amount of pine felling is slightly higher, but no planting is carried out. The overall average planting / felling ratio of the plots where the dominant tree species were Chinese fir forests and other forest species was about 0.9.

[0128] Economic development priority scenario: refers to the scenario of future changes in the carbon density of forest ecosystems driven by the development strategy planning that focuses on economic development in the future of the study area. In the economic priority development scenario, based on the actual management measures of the forest ecosystem in the natural development scenario, the forest land with the dominant tree species of fir will increase the felling by 20% and increase the planting by 7% on average each year, the forest land with the dominant tree species of pine will reduce the felling by 58% on average each year, the forest land with the dominant tree species of bamboo will increase the felling by 100% on average each year, and the forest land with the dominant tree species of other forest species will increase the felling by 40% on average each year while the planting will remain unchanged. The felling of all forest species will increase by 48% on average each year, and the planting will increase by 1.74%. In this scenario, the planting and felling ratio of the sample plot is about 1:1.2. For the pine forest, the reduction in its felling is to enable the study area to retain more abundant ecological species as much as possible to ensure that it will not face the situation of being felled. Doubling the amount of bamboo harvested will not only bring higher economic benefits, but also prevent bamboo from over-reproduction and encroaching on the breeding areas of other tree species, leading to the reduction of other forest species. At present, the situation is set relatively conservatively. After visits and investigations, the local natural reproduction rate of bamboo exceeds 20% each year, so the amount of bamboo harvested in the future can be higher.

[0129] Management optimization scenario (maximum species diversity scenario): refers to the development management optimization scenario with ecological conservation and species diversity as the priority under the coupling of natural climate and human activity interference. In this scenario, the optimal management model of the forest ecosystem for biodiversity conservation will be designed with the protection and realization of maximum species diversity as the development goal, so that under the combined effects of future climate change and human activities, the forest ecosystem in the study area will develop in the direction of maximizing species diversity.

[0130] On the basis of these three scenarios, we analyze the impacts of two climate change scenarios, SSP2-4.5 and SSP5-8.5, on the changes in carbon density of forest ecosystems in the study area under the management policies of the same scenario.

[0131] Step S4: Based on the forest plot environmental information dataset, considering various future climate change situations, the Forest-DNDC model is used to simulate the carbon density of each future scenario, and the simulation results of the soil and forest carbon density of all forest plots under different climate changes and different scenarios are obtained.

[0132] Specifically, based on the forest plot environmental information dataset and considering various future climate change scenarios, the Forest-DNDC model was used to simulate the location of each forest plot. The model input parameters included: climate data (daily temperature and precipitation), soil data (soil type, forest type, soil layer thickness, number of soil layers, pH, surface soil organic carbon content (0-5cm, kg C / kg), total soil organic carbon content (kg C / hectare), stone content, bulk density, clay content, hydrological conductivity, soil porosity, field water holding capacity, crop wilting point, etc.), forest data (forest species, forest age, tree physiological phenological parameters, etc.), and management measures (planting, felling, burning, drainage, and fertilization).

[0133] In this step, the forest plot environmental information dataset is used as input data, so that the model can accurately simulate the conditions of the forest plots in the study area. Combined with different climate change scenarios (such as temperature and precipitation changes), the simulation can better reflect the dynamic changes of forest ecosystems under different future climate change paths and different management measures, which is helpful to predict the impact of climate change on forest carbon density. At the same time, the simulation results provide information on year-by-year changes in soil and forest carbon density, which can quantify the interannual changes in forest carbon sinks under future scenarios and evaluate the long-term effects of different forest management measures. A long-term simulation is performed on each plot separately (for example, 42 years from 2019 to 2060), and finally the year-by-year simulation results of all plots are obtained. By simulating all forest plots separately according to their different environmental factors, it is ensured that the environment of each plot can be accurately considered.

[0134] It should be noted that in the existing Forest-DNDC model, management measures can only be set once during the simulation process, and this management measure can only be implemented at a specific point in time. However, in the simulation of long-term series of future scenarios (such as the study of forest ecosystem evolution over decades or even hundreds of years), the dynamic changes of the ecosystem often require multiple management measures to be fully described and understood. The above-mentioned setting method of management measures in the existing Forest-DNDC model leads to the inability of the existing Forest-DNDC model to reflect long-term management strategies and the lack of simulation of multiple intervention responses, which cannot meet the needs of simulating the dynamic changes of the ecosystem.

[0135] In view of this, in step S4, before using the Forest-DNDC model to simulate the changes in carbon density of each future scenario, it also includes: a step of improving the management measures module of the Forest-DNDC model, specifically as follows: adding judgment conditions to each management measure in the management measures module of the Forest-DNDC model, and setting the management measure time as a cycle, and the judgment conditions are used to determine whether the management measure should be executed.

[0136] Since the Forest-DNDC model adopts a modular design structure, in this embodiment, the management module is improved by modular design to support the simulation of multiple management measures. Specifically, a judgment condition (logical condition) is added to determine whether the management measure should be executed. By adding the judgment condition, the model can check whether the conditions for executing the management measure are met at each time step (such as year, month or day). If the condition is met, the preset management measure is executed at that time point. The judgment condition can be set according to the time step or other environmental variables. For example, the judgment condition can be set to "fertilize when soil nutrients drop to a certain threshold."

[0137] Setting the timing of management measures to a fixed cycle, such as implementing management measures once a year, every five years, or every ten years, means that management operations are no longer one-off, but are repeated regularly throughout the simulation process, that is, the model makes conditional judgments in each management cycle to determine whether to implement a management measure.

[0138] It should also be noted that in traditional simulations, forest physiological phenology parameters are usually set by the Forest-DNDC model based on empirical models. Such settings have little deviation for large-scale simulations, but when applied to county-scale simulations, they cannot reflect the local forest ecological characteristics, resulting in insufficient simulation accuracy. In order to solve the problem of forest physiological phenology parameters, some existing technologies use the method of actually measuring physiological phenology parameters in all forest plots to ensure the accuracy of physiological phenology parameters. However, such methods are very difficult to engineer and very costly.

[0139] In view of this, in the present application, in step S4, before using the Forest-DNDC model to simulate the carbon density of each future scenario, it also includes: using the Forest-DNDC model to perform high-precision inversion of forest physiological phenological parameters in the base year (for example, 2019).

[0140] The high-precision inversion of forest physiological phenological parameters in the base year (e.g. 2019) using Forest-DNDC is based on the following assumptions: the growth parameters of local tree species are consistent with the growth parameters of the model default tree species, referring to Figure 4 , including the following steps:

[0141] For each forest plot in the forest plot environmental information dataset, the forest carbon density calculation value (i.e., actual carbon density) is calculated using the empirical formula based on the forest plot measured data (i.e., measured data in the forest plot inventory table);

[0142] Determine the inversion parameter data set of the current forest plot; the inversion parameter data set includes multiple sets of values ​​of the physiological phenological parameters of the current forest plot. These parameter combinations will be used as candidate parameter groups for simulation of the Forest-DNDC model in the study area, and the optimal parameters of the current forest plot will be found through traversal. For example, assuming that the forest physiological phenological parameters include parameters A, B, and C, then one set of parameters includes {A = 9; B = 15; C = 30}, and another set of parameters can be {A = 12; B = 18; C = 33}, and so on.

[0143] Traverse the inversion parameter data sets in order and perform the following steps for each set of parameters:

[0144] The current group parameters are set as the physiological phenological parameters of the Forest-DNDC model, and simulation is performed based on the Forest-DNDC model to obtain the simulated value of forest carbon density under the current group parameters;

[0145] Calculate the error between the simulated forest carbon density value and the calculated forest carbon density value. If the error does not meet the preset threshold requirement (for example, the threshold requirement is an error of less than 5%), perform parameter screening, that is, exclude the current group of parameters from the candidate parameter values, and set the next group of parameters as the physiological phenological parameters of the Forest-DNDC model, re-simulate with the Forest-DNDC model, obtain a new simulated forest carbon density value, and calculate the error between the new simulated forest carbon density value and the calculated forest carbon density value, until the error meets the preset threshold requirement (that is, the error is less than 5%), and end the cycle;

[0146] The corresponding physiological phenological parameters at the end of the cycle are taken as the final physiological phenological parameters of the current forest plot.

[0147] Furthermore, the inversion parameter dataset of the current forest plot is determined, including:

[0148] The dominant forest species and forest age of the current forest plot are input into the Forest-DNDC model to obtain the default physiological phenological parameters generated by the Forest-DNDC model based on the empirical model, and then the forest carbon density simulation value is generated based on the default physiological phenological parameters; the dominant forest species and forest age of the current forest plot can be extracted from the forest plot data;

[0149] Compare the magnitude relationship between the simulated forest carbon density value and the calculated forest carbon density value (e.g., greater than or less than), and record the current default physiological phenological parameter values ​​as the upper limit (when the simulated forest carbon density value is greater than the calculated forest carbon density value) or lower limit (when the simulated forest carbon density value is less than the calculated forest carbon density value) of each parameter according to the magnitude relationship, and change the forest age value according to the magnitude relationship (if the simulated forest carbon density value is greater than the calculated forest carbon density value, reduce the forest age; if the simulated forest carbon density value is less than the calculated forest carbon density value, increase the forest age), and re-input the changed forest age into the Forest-DNDC model to obtain new default physiological phenological parameters, use the new default physiological phenological parameters for simulation, and obtain new simulated forest carbon density values;

[0150] Compare the new simulated forest carbon density value with the calculated forest carbon density value again. If the relationship between the new simulated forest carbon density value and the calculated forest carbon density value changes (for example, the original greater than relationship changes to a less than relationship), the new default physiological phenological parameters are recorded as the lower or upper limits of each parameter, and then the value range of each physiological phenological parameter is obtained according to the upper and lower limits of each parameter; otherwise, continue to change the forest age and re-simulate until the relationship between the new simulated forest carbon density value and the calculated forest carbon density value changes, and stop the cycle;

[0151] Through the above steps, it can be ensured that the physiological phenological parameter of the forest carbon density simulation value output by the Forest-DNDC model that is closest to the forest carbon density calculation value is included in the value range.

[0152] According to the value range of each physiological phenological parameter, multiple possible values ​​of each physiological phenological parameter are determined using an equal step method starting from the lower limit, and all possible values ​​of all physiological phenological parameters are combined into an inversion parameter data set for the current forest plot.

[0153] After executing the above steps for each forest sample plot, a high-precision forest physiological phenological parameter dataset of all sample plots in the base year can be inverted.

[0154] In order to facilitate understanding of the above-mentioned forest physiological phenological parameter inversion process, an example is given below.

[0155] Assume that the total forest plot data consists of 245 plots, and the following inversion logic is performed for each plot:

[0156] 1. Calculate forest carbon density based on existing empirical equations, assuming that the actual calculated carbon density is 8T C / ha;

[0157] 2. Input the dominant tree species and forest age of the sample plot into the Forest-DNDC model. Assuming that the dominant tree species input is Chinese fir and the forest age is 15 years, the model will generate a set of default physiological phenological parameters based on model experience. Assuming that the default physiological phenological parameters are {A: 9, B: 15, C: 30}, the Forest-DNDC model is executed based on this set of parameters. Assuming that the model simulation result is 5T C / ha, compared with the actual calculated carbon density of 8T C / ha, the former is smaller than the latter. At this time, {A: 9, B: 15, C: 30} is recorded as the lower limit of parameters A, B, and C, and then the forest age is increased, for example, set to 20 years, and re-simulate. The Forest-DNDC model generates a new set of physiological phenological parameters. Assuming that the parameters are {A: 12, B: 18, C: 33}, the new forest carbon density simulation value is 10T C / ha, which is different from the actual calculated carbon density of 8T C / ha, the former is greater than the latter, that is, the size relationship between the two has changed, the cycle is stopped, and {A: 12, B: 18, C: 33} is recorded as the upper limit of parameters A, B, and C, and the value interval of parameter A is [9, 12], the value interval of parameter B is [15, 18], and the value interval of parameter C is [30, 33]. At the same time, the above steps can ensure that the parameter whose simulated forest carbon density is closest to the actual carbon density is included in the above range.

[0158] 3. Starting from the lower limit, use the equal step method to determine multiple possible values ​​of each physiological phenological parameter. For example, set the parameter to 4 equal parts and divide the parameter into four parts to obtain the following inversion parameter data set: parameter A is [9, 10, 11, 12], parameter B is [15, 16, 17, 18], and parameter C is [30, 31, 32, 33];

[0159] 4. Traverse the inversion parameter data set and use each set of parameters to simulate a forest carbon density simulation value. Assume that the results are as follows:

[0160] {A[9],B

[15] ,C

[30] }, the simulated forest carbon density is 5T C / ha;

[0161] {A

[10] ,B

[16] ,C

[31] }, the simulated forest carbon density is 6.5TC / ha;

[0162] {A

[11] ,B

[175] ,C

[32] }, the simulated forest carbon density is 9T C / ha;

[0163] {A

[12] ,B

[18] ,C

[33] }, the simulated forest carbon density is 10T C / ha;

[0164] Obviously, the simulated forest carbon density value of 9T C / ha obtained by the parameter group {A

[11] , B

[175] , C

[32] } is the closest to the actual calculated carbon density of 8T C / ha. Therefore, this set of parameters was selected as the physiological phenological parameters of the sample site in 2019.

[0165] This process consumes computer energy to fill the gap in the inability to conduct actual measurements of the physiological phenological parameters of all forest plots. Through this inversion process, it can ensure that each forest plot obtains a set of physiological phenological parameters that can better reflect the actual carbon density of the plot, further improving the simulation accuracy of the Forest-DNDC model.

[0166] Step S5: Input the simulation results of forest plot soil and forest carbon density into the pre-trained random forest model, and combine them with the forest plot environmental information dataset to predict the carbon density of other areas outside the forest plot, and obtain the spatial distribution raster data of soil carbon density and forest carbon density covering the entire study area under future scenarios.

[0167] Furthermore, in order to improve the efficiency of the random forest model and simplify the calculation results, before inputting the simulation results into the pre-trained random forest model, it also includes: selecting the simulation results of forest sample soil and forest carbon density of all forest sample plots under different climate changes and different scenarios at fixed annual intervals (for example, every 5 years from 2020 to 2060), and formatting the data to obtain a forest sample plot simulation data set; and inputting the formatted forest sample plot simulation data set into the pre-trained random forest model.

[0168] Among them, data formatting processing refers to organizing the simulation results of forest plot soil and forest carbon density under different climate change scenarios output by the Forest-DNDC model into the input data format required by the random forest model.

[0169] Assuming that the DEM, slope and aspect of the area will not change in the future, the simulation results of soil and forest carbon density of the forest plot are input into the pre-trained random forest model. The input independent variables are: future annual total precipitation, elevation, slope and aspect, and the dependent variables are soil carbon density and forest carbon density, respectively. The random forest model is used to predict other areas outside the forest plot, and the spatial distribution raster data of soil carbon density and forest carbon density covering the entire study area under future scenarios are obtained.

[0170] Step S6: Using the spatially distributed raster data as the initial field and the simulated results of forest plot soil and forest carbon density output by the Forest-DNDC model as control conditions, a HASM simulation of the forest ecosystem carbon sink under future scenarios is performed to obtain high-precision future scenario raster data of forest and soil with a specified resolution.

[0171] Specifically, the spatial distribution raster data of soil carbon density and forest carbon density covering the entire study area under future scenarios predicted by random forests are used as trend surface data in HASM. Combined with the predicted values ​​of soil and forest carbon density output by the Forest-DNDC model, the precision is refined. Finally, high-precision carbon density surfaces of forests and soils are obtained every 5 years starting from 2020, that is, high-precision future scenario raster data of forests and soils with specified resolution.

[0172] Furthermore, in order to make the results more realistic, the results of farmland areas, water areas or other places without forests are masked in the final forest carbon density results, while soil carbon density does not need to be masked.

[0173] Finally, the spatiotemporal changes are analyzed by drawing and analyzing the spatiotemporal distribution maps of forest and soil carbon density under different climate conditions and scenarios. Taking the natural development scenario under the SSP2-4.5 climate conditions as an example, the high-precision carbon density surfaces of forests and soils are graphically displayed to form the spatiotemporal distribution maps of forest carbon density and soil carbon density, such as Figure 5 , Figure 6 shown.

[0174] By drawing the changing trend diagrams of the impact of different scenarios and different climate conditions on forest and soil carbon density, the changes in forest and soil carbon density under different scenarios and different climate conditions are analyzed. Taking the natural development scenario under SSP2-4.5 climate conditions as an example, the changing trends of forest and soil carbon density under different scenarios and different climate conditions are as follows Figure 7 , Figure 8 shown.

[0175] Based on high-precision simulation data on the future carbon sink of forest ecosystems covering the entire county scale, the above analysis can provide data support for the study of the evolution of forest carbon density at the subtropical county scale under different climate change scenarios, especially for the systematic analysis of the long-term impact of mild and extreme climate scenarios on forest carbon density. At the same time, it provides a technical basis for the complexity analysis of the spatial and temporal dimensions of the effects of different management policies under different climate scenarios at the subtropical county scale.

[0176] In summary, the method provided in this embodiment can reveal the complex impacts of climate change and management policies on the forest and soil carbon density and carbon sinks in the study area (county scale), and provide effective technical support for designing forest management policies that adapt to different climate scenarios. At the same time, the method provided in this embodiment provides reference opinions for understanding the joint impacts of climate change and management policies on the carbon density of forest ecosystems in the study area, and can provide a certain scientific basis for global carbon management and climate change response strategies.

[0177] Embodiment 2:

[0178] This embodiment provides a future scenario simulation system for forest ecosystem carbon sinks, which uses the Forest-DNDC model as a scenario simulation model, including:

[0179] Data acquisition module, used to obtain climate data, soil data, forest plot data, and management measures in the study area;

[0180] The data processing module is used to process the climate data and soil data based on the HASM method, respectively, to obtain the climate raster data and soil raster data of the specified resolution, and to extract the pixel values ​​of the forest plot location in the climate raster data and the soil raster data into the forest plot data, to obtain the forest plot environmental information data set;

[0181] The scenario setting module is used to design multiple scenarios for the future of forest ecosystems under different management measures based on the actual management measures recorded in the historical forest management of the study area;

[0182] The sample plot carbon density simulation module is used to simulate the carbon density of each future scenario using the Forest-DNDC model based on the forest sample plot environmental information data set and considering various future climate change situations, so as to obtain the simulation results of the soil and forest carbon density of all forest sample plots under different climate changes and different scenarios year by year;

[0183] The carbon density initial field prediction module is used to input the simulation results of soil and forest carbon density in forest plots into the pre-trained random forest model, and combine it with the forest plot environmental information dataset to predict the carbon density of other areas outside the forest plot, and obtain the spatial distribution raster data of soil carbon density and forest carbon density covering the entire study area under future scenarios;

[0184] The carbon density downscaling module is used to use the spatially distributed raster data as the initial field and the simulated results of soil and forest carbon density of forest plots output by the Forest-DNDC model as control conditions to perform HASM simulation of forest ecosystem carbon sinks under future scenarios, and obtain high-precision future scenario raster data of forests and soils with specified resolution.

[0185] The future scenario simulation system for forest ecosystem carbon sinks provided in this embodiment can implement the steps and processes of the future scenario simulation method for forest ecosystem carbon sinks provided in any of the above embodiments, and achieve the same technical effects, which will not be described one by one here.

[0186] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for simulating future scenarios of carbon sinks in forest ecosystems, the method using the Forest-DNDC model as a scenario simulation model, comprising the following steps: Step S1: Obtain climate data, soil data, forest plot data, and management measures in the study area; Step S2: Based on the high-precision surface simulation HASM method, the climate data and soil data are processed respectively to obtain climate raster data and soil raster data of specified resolution, and the pixel values ​​of the location of the forest sample plot in the climate raster data and the soil raster data are extracted into the forest sample plot data to obtain the forest sample plot environmental information data set; Step S3: Based on the actual management measures recorded in the historical forest management of the study area, multiple scenarios of the future of the forest ecosystem under different management measures are designed; Step S4: Based on the forest plot environmental information data set, considering various future climate change situations, the Forest-DNDC model is used to simulate the carbon density of each future scenario, and the simulation results of the soil and forest carbon density of all forest plots under different climate changes and different scenarios are obtained year by year; Step S5: Input the simulation results of soil and forest carbon density of forest plots into the pre-trained random forest model, and combine with the environmental information data set of forest plots to predict the carbon density of other areas outside the forest plots, and obtain the spatial distribution raster data of soil carbon density and forest carbon density covering the entire study area under future scenarios; Step S6: Using the spatially distributed raster data as the initial field and the simulated results of forest plot soil and forest carbon density output by the Forest-DNDC model as control conditions, a HASM simulation of the forest ecosystem carbon sink under future scenarios is performed to obtain high-precision future scenario raster data of forest and soil with a specified resolution.

2. The method according to claim 1, characterized in that In step S2, the climate data is processed based on the HASM method, including the following steps: The historical data of all meteorological stations in the climate data are regressed and simulated, and then the inverse distance weighted method is used for interpolation to obtain the daily-scale climate trend surface with a specified resolution; The residual of each meteorological station is obtained by subtracting the pixel predicted value at the location of each meteorological station in the daily climate trend surface from the actual observation value of the station. Several meteorological stations in and around the study area were selected, and the spatial range of the selected meteorological stations was used to crop the daily-scale climate trend surface; The clipping results are used as the climate initial field of the HASM method, and the residuals of multiple selected meteorological stations are used as the climate control conditions of the HASM method. The HASM simulation is carried out to obtain the daily-scale residual surface with a specified resolution. The daily climate trend surface and the residual surface are summed to obtain the climate grid data of the specified resolution.

3. The method according to claim 1, characterized in that In step S2, the soil data is processed based on the HASM method to obtain soil grid data of a specified resolution, including: Calculate the organic carbon density at different soil depths at each soil sampling point; The calculated soil organic carbon density was used as the dependent variable, and multiple soil properties and environmental factors of the soil sampling point data were used as independent variables. The random forest model was used to predict the spatial distribution of soil organic carbon at different soil depths. Taking the spatial distribution of soil organic carbon at different soil depths as the initial field and the soil sampling point data as the control condition, HASM simulation of soil organic carbon was carried out to obtain the soil organic carbon grid surface with a specified resolution. Ordinary linear regression method was used to obtain trend surface data of other soil parameters. The soil census data and soil sampling point data were used as control conditions to carry out HASM simulation on other soil parameters to obtain grid surfaces of other soil parameters with specified resolution.

4. The method according to claim 1, characterized in that: In step S4, before using the Forest-DNDC model to simulate the carbon density changes of each future scenario, it also includes: a step of improving the management measure module of the Forest-DNDC model, which is as follows: In the management measure module of the Forest-DNDC model, a determination condition is added to each management measure, and the management measure time is set as a cycle. The determination condition is used to determine whether the management measure should be executed.

5. The method according to claim 1, characterized in that In step S4, before simulating the carbon density of each future scenario using the Forest-DNDC model, it also includes: a step of high-precision inversion of the forest physiological phenological parameters of the base year, as follows: For each forest plot in the forest plot environmental information dataset, the forest carbon density calculation value is calculated using the empirical formula based on the measured data of the forest plot; Determine the inversion parameter dataset of the current forest plot; Traverse the inversion parameter data sets in order and perform the following steps for each set of parameters: The current group parameters are set as the physiological phenological parameters of the Forest-DNDC model, and simulation is performed based on the Forest-DNDC model to obtain the simulated value of forest carbon density; Calculate the error between the simulated forest carbon density value and the calculated forest carbon density value. If the error does not meet the preset threshold requirement, set the next set of parameters as the physiological phenological parameters of the Forest-DNDC model, and simulate again with the Forest-DNDC model to obtain a new simulated forest carbon density value. Calculate the error between the new simulated forest carbon density value and the calculated forest carbon density value until the error meets the preset threshold requirement, and end the cycle. The corresponding physiological phenological parameters at the end of the cycle are taken as the final physiological phenological parameters of the current forest plot.

6. The method according to claim 5, characterized in that The inversion parameter data set for determining the current forest plot includes: The dominant forest species and forest age of the current forest plot are input into the Forest-DNDC model to obtain the default physiological phenological parameters generated by the Forest-DNDC model based on the empirical model, and then the forest carbon density simulation value is generated based on the default physiological phenological parameters; Compare the size relationship between the simulated forest carbon density value and the calculated forest carbon density value, record the current default physiological phenological parameter value as the upper or lower limit of each parameter value according to the size relationship, and change the forest age value according to the size relationship, and re-input the changed forest age into the Forest-DNDC model to obtain new default physiological phenological parameters, use the new default physiological phenological parameters for simulation, and obtain new forest carbon density simulation values; Compare the new simulated forest carbon density value with the calculated forest carbon density value again. If the relationship between the new simulated forest carbon density value and the calculated forest carbon density value changes, record the new default physiological phenological parameters as the lower or upper limits of each parameter, and then obtain the value range of each physiological phenological parameter based on the upper and lower limits of each parameter; otherwise, continue to change the forest age and re-simulate until the relationship between the new simulated forest carbon density value and the calculated forest carbon density value changes, and stop the cycle; According to the value range of each physiological phenological parameter, multiple possible values ​​of each physiological phenological parameter are determined using an equal step method starting from the lower limit, and all possible values ​​of all physiological phenological parameters are combined into an inversion parameter data set for the current forest plot.

7. The method according to claim 1, characterized in that In step S5, before the simulation results are input into the pre-trained random forest model, the following steps are also included: At fixed annual intervals, the simulation results of soil and forest carbon density of all forest plots under different climate change and different scenarios were selected, and the data were formatted to obtain the forest plot simulation data set; The formatted forest plot simulation dataset was input into the pre-trained random forest model.

8. A future scenario simulation system for forest ecosystem carbon sinks, the system adopts the Forest-DNDC model as a scenario simulation model, including: Data acquisition module, used to obtain climate data, soil data, forest plot data, and management measures in the study area; The data processing module is used to process the climate data and soil data based on the high-precision surface simulation HASM method, respectively, to obtain climate raster data and soil raster data of specified resolution, and extract the pixel values ​​of the forest plot location in the climate raster data and the soil raster data into the forest plot data to obtain the forest plot environmental information data set; The scenario setting module is used to design multiple scenarios for the future of forest ecosystems under different management measures based on the actual management measures recorded in the historical forest management of the study area; The sample plot carbon density simulation module is used to simulate the carbon density of each future scenario using the Forest-DNDC model based on the forest sample plot environmental information data set and considering various future climate change situations, so as to obtain the simulation results of the soil and forest carbon density of all forest sample plots under different climate changes and different scenarios year by year; The carbon density initial field prediction module is used to input the simulation results of soil and forest carbon density in forest plots into the pre-trained random forest model, and combine it with the forest plot environmental information dataset to predict the carbon density of other areas outside the forest plot, and obtain the spatial distribution raster data of soil carbon density and forest carbon density covering the entire study area under future scenarios; The carbon density downscaling module is used to use the spatially distributed raster data as the initial field and the simulated results of soil and forest carbon density of forest plots output by the Forest-DNDC model as control conditions to perform HASM simulation of forest ecosystem carbon sinks under future scenarios, and obtain high-precision future scenario raster data of forests and soils with specified resolution.

Images