WO2024098444A1 - 一种乡村生态系统碳储量预测方法 - Google Patents

一种乡村生态系统碳储量预测方法 Download PDF

Info

Publication number
WO2024098444A1
WO2024098444A1 PCT/CN2022/132094 CN2022132094W WO2024098444A1 WO 2024098444 A1 WO2024098444 A1 WO 2024098444A1 CN 2022132094 W CN2022132094 W CN 2022132094W WO 2024098444 A1 WO2024098444 A1 WO 2024098444A1
Authority
WO
WIPO (PCT)
Prior art keywords
land use
data
land
total
type
Prior art date
Application number
PCT/CN2022/132094
Other languages
English (en)
French (fr)
Inventor
徐宁
池麦
王姁
郑琳
段皓然
成玉宁
潘可欣
何雪馨
Original Assignee
东南大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 东南大学 filed Critical 东南大学
Publication of WO2024098444A1 publication Critical patent/WO2024098444A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00Administration; Management
    • G06Q10/04Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10Complex mathematical operations
    • G06F17/18Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/04Architecture, e.g. interconnection topology
    • G06N3/0409Adaptive resonance theory [ART] networks
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/08Learning methods
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/10Services
    • G06Q50/26Government or public services
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T11/002D [Two Dimensional] image generation
    • G06T11/20Drawing from basic elements, e.g. lines or circles
    • G06T11/206Drawing of charts or graphs
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F2219/00Indexing scheme relating to application aspects of data processing equipment or methods
    • G06F2219/10Environmental application, e.g. waste reduction, pollution control, compliance with environmental legislation
    • YGENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02PCLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/80Management or planning
    • Y02P90/84Greenhouse gas [GHG] management systems

Definitions

  • the invention belongs to the technical field of ecological environment, and specifically relates to a method for predicting carbon reserves in a rural ecosystem.
  • Forests, grasslands, peat bogs, and other terrestrial ecosystems store far more carbon than the atmosphere does.
  • ecosystems influence climate by keeping climate-changing CO2 out of the atmosphere.
  • Disturbances to ecosystems, such as fire, disease, or vegetation conversion can release large amounts of CO2, while other management changes, such as forest restoration or alternative agricultural practices, can store large amounts of CO2.
  • the way ecosystems are managed is therefore critical to regulating the effects of CO2-driven climate change.
  • SSPs socio-economic pathways
  • the quantitative assessment methods of ecosystem services are divided into direct assessment and indirect assessment.
  • Direct assessment includes subjective and objective assessment methods
  • indirect assessment includes final and intermediate material conversion methods and energy conversion methods.
  • the current indicator system and conversion method of energy conversion method are not perfect.
  • the intermediate material conversion method has uncertainty in the selection of spatial data and methods.
  • the final material conversion method and direct assessment method need to overcome the problem of using points to represent the whole and low spatial resolution.
  • the InVEST model (Integrated Valuation of Ecosystem Services and Trade-offs) is a widely used tool that usually adopts the production function method to quantify and evaluate ecosystem services. Compared with other more complex tools, the InVEST model has lower data requirements and can achieve dynamic and sustainable assessment of the value of terrestrial, freshwater and marine ecosystem services.
  • the InVEST carbon storage and sequestration model is a sub-model in the InVEST model. It summarizes the amount of biophysical carbon stored in four carbon pools (aboveground biomass, belowground biomass, soil, and dead organic matter) based on current or future land use/land cover (LULC) maps to estimate the current amount of carbon stored; however, the InVEST model has fewer land scenario simulations and cannot flexibly meet research requirements.
  • LULC land use/land cover
  • the FLUS model (Future Land Use Simulation) is a widely used model for simulating land use changes under the influence of human activities and nature, as well as future land use scenarios. It couples the "top-down" system dynamics model and the "bottom-up” cellular automaton (CA) model.
  • the FLUS model uses a neural network algorithm (ANN) to combine land use data with a variety of driving factors including human activities and natural effects to calculate the suitability probability of each land use type within the study scope, which is more suitable for complex and changeable land scenarios in rural areas.
  • ANN neural network algorithm
  • the FLUS model proposes an adaptive inertial competition mechanism based on roulette selection, which can effectively deal with the uncertainty and complexity of the mutual transformation of various land use types under the joint influence of natural effects and human activities, so that the FLUS model has a high simulation accuracy and can obtain results similar to the actual land use distribution.
  • the purpose of the present invention is to provide a method for predicting carbon reserves in rural ecosystems, which solves the technical problems mentioned in the background technology.
  • a method for predicting carbon storage in a rural ecosystem characterized by comprising the following steps:
  • S1 obtain the initial land use data and climate scenario data of the study area, establish a land use and climate scenario database, and the climate scenario data is a shared socio-economic path; divide the study area into equidistant grids, and select driving force factors in the climate scenario to form driving force data;
  • S4 construct the InVest model; and input the four major carbon pool data obtained in S2 and the land use simulation results obtained in S3 into the InVEST model to obtain the distribution of land carbon storage in the study area for the corresponding scenarios (SSP1, SSP2, SSP5).
  • the steps of constructing the CARA model include:
  • the beneficial effects of the present invention are as follows: it can better combine the current new emission scenario, more effectively integrate data with a higher degree of collectability, and make more accurate spatiotemporal evolution simulation and prediction of the land carbon storage distribution in the study area, thereby providing an effective reference for optimizing regional carbon emission control strategies.
  • Figure 1 is a flow chart of the rural ecosystem carbon storage prediction method under the SSPs scenario
  • Figure 2 is the initial land use classification map of City A in 2010
  • Figure 3 is a collection of driving factors for various land use changes under the SSP1 scenario
  • Figure 4 is the simulated land use distribution map of City A in 2020 under the SSP1 scenario
  • Figure 5 is the simulated land use distribution map of City A in 2020 under the SSP2 scenario
  • Figure 6 is the simulated land use distribution map of City A in 2020 under the SSP5 scenario
  • Figure 7 is the actual land use distribution map of City A in 2020 for comparison and verification of simulation results
  • Figure 8 is the numerical results of the four major carbon pools of various types of land in rural areas of province A in 2020 output by the CARA model
  • FIG9 is a graph showing the carbon storage assessment results of the rural areas of province A meeting the scenario targets in 2020 output by an embodiment of the present invention under the SSPs1 scenario;
  • FIG10 is a graph showing the carbon storage assessment results of the rural areas of province A meeting the scenario targets in 2020 output by an embodiment of the present invention under the SSPs2 scenario;
  • FIG. 11 is a graph showing the carbon storage assessment results for rural areas of province A in 2020 that meet scenario targets, output by an embodiment of the present invention under the SSPs5 scenario.
  • a rural ecosystem carbon storage prediction method includes the following steps:
  • S1 establish a database and preprocess the data, including:
  • the initial land use data is first obtained by obtaining the initial high-resolution land image of the study area and preprocessing it, then interpreting the preprocessed high-resolution image to obtain the classified image, and finally obtaining the initial land use data from the classified image;
  • the climate scenario data is the Shared Socioeconomic Pathways (SSPs).
  • NPP net primary productivity
  • NDVI normalized difference vegetation index
  • SSPs shared socioeconomic pathways
  • NPP x,t APAR x,t ⁇ ⁇ x,t (2)
  • APAR x,t is the photosynthetically active radiation absorbed by pixel x in month t;
  • ⁇ x,t is the actual light energy rate of pixel x in month t;
  • f1 x,t and f2 x,t represent the stress effects of low and high temperatures on light energy utilization (unitless)
  • W x,t is the water stress influence coefficient (unitless) reflecting the influence of water conditions
  • ⁇ max is the maximum light utilization under ideal conditions (unit: gC ⁇ MJ-1)
  • f1, f2 and W are constants
  • the photosynthetically active radiation absorbed by vegetation depends on the total solar radiation and the photosynthetically active radiation component (FPAR), and the model is as follows:
  • S x,t is the total solar radiation of pixel x in month t
  • FPAR x,t is the absorption ratio of the incident photosynthetically active radiation by the vegetation layer
  • the constant 0.5 represents the ratio of the solar effective radiation (wavelength 0.38-0.71 ⁇ m) that can be used by vegetation to the total solar radiation
  • FPAR shows a good linear relationship with NDVI and ratio vegetation index (SR);
  • SR min and SR max correspond to the 5% and 95% lower percentiles of NDVI of each vegetation type, respectively; SR x,t can be obtained from NDVI x,t :
  • NDVI xt is the NDVI of pixel x in month t. Since NDVI comes from the NDVIMOD13Q1 dataset, the present invention can directly obtain the monthly time series NDVI dataset of the required year, thereby calculating the NPP of each year. Then, a linear regression analysis is performed on the obtained net primary productivity (NPP) of the study area in each year to establish a spatiotemporal evolution model of the net primary productivity (NPP) of the land in the study area.
  • NPP net primary productivity
  • B is the long-term biomass of the plant (tonnes C/ha)
  • r is the intrinsic growth rate (year-1)
  • K is the carrying capacity (tonnes C/ha)
  • ⁇ i and kj represent the respective proportions of the different forms (leaves, large and small branches, stems, roots) in the total biomass and the turnover rate (year-1)
  • H is the harvest rate (tonnes ha-1 year-1).
  • the calculation steps of the carbon pool data of the land in the study area include:
  • mF, mR, and mS are the proportions of the carbon pools in plant leaves, roots, and stems, respectively;
  • x represents a type of land use, GAI x represents the total annual increment of biomass in type x; C above, x represents the aboveground biological carbon density in type x; and C below, x represents the underground plant root carbon density in type x.
  • the calculation steps to obtain the distribution probability of various land use types in the simulation area are:
  • the driving force data and initial land use data are randomly sampled using uniform sampling strategy or proportional sampling strategy.
  • uniform sampling mode the number of sampling points for each type of land use is the same; while in the random sampling mode, the number of sampling points for each type of land use is related to the proportion of each type of land use.
  • the sample formula after sampling is expressed as:
  • x n represents the variable of the nth driving force factor extracted at the first sampling point, and T is the transpose
  • the normalization calculation formula is:
  • max w and min w are the maximum and minimum values of the w-th driving force factor, respectively;
  • the parameter adaptive neural network algorithm can be expressed as follows:
  • ⁇ (n) is the learning rate of the nth iteration
  • E(n) and E(n-1) are the root mean square errors of the neural network outputs of two adjacent iterations
  • a, b, c are constants, and their value ranges are (1, 2), (0, 1), and (1, 1.1) respectively;
  • the parameter adaptive neural network algorithm includes an input layer, a hidden layer and an output layer. All driving force data are input into the trained neural network through the input layer. After the driving force input data is processed by the input layer, the hidden layer and the output layer in sequence, the distribution probability of each land use type in the simulation area is obtained.
  • w i,j is the one-to-one corresponding parameter between the input layer and the hidden layer, that is, the weight value between the two layers.
  • the optimizer in the neural adaptive network will train and calibrate the weight according to the value generated by the loss function;
  • xi (p, t) is the i-th variable related to the input neuron i on the grid cell p at time t;
  • connection between the input layer and the hidden layer is constructed using the sigmo activation function:
  • Each neuron in the output layer corresponds to a specific land use type.
  • the value of the lth neuron in the output layer will generate a value, which represents the probability of occurrence of the lth land use type in the grid cell. The higher the value, the greater the probability of occurrence of the target land use type in a specific grid cell.
  • the probability of occurrence of land use type k on grid cell p at time t is recorded as p(p, k, t):
  • wj ,k is the weight parameter between the hidden layer and the output layer, similar to wi,j ; the final p(p,k,t) is the distribution probability of various land use types in the simulation area calculated by the trained neural network.
  • a scanning window is constructed through the neighborhood function, and the number of various pixels in the scanning window is counted to measure the spatial mutual influence of various land use types.
  • the definition of the neighborhood function is as follows:
  • the FLUS model uses adaptive inertia to represent the inheritance of previous land use types and defines the inertia coefficient according to three cases:
  • Conversion cost is another factor that affects land use dynamics.
  • the conversion cost in FLUS is fixed.
  • the land use conversion cost from c to k is expressed as sc c ⁇ k , with a value range of (0,1);
  • the comprehensive probability formula for a particular land use type occupying a plot is calculated by combining the occurrence probability, neighborhood effect, inertia coefficient and conversion cost, that is, the global total probability synthesis formula of the roulette wheel is:
  • P p,k is the occurrence probability of land use type k on grid cell p
  • P p,k is the neighborhood effect of land use type k on grid cell p at iteration time t
  • sc c ⁇ k is the conversion cost of the original land use type c to the target type k.
  • the total distribution probability of various land use types on each pixel is used to form a roulette wheel, and the various land use types in the region compete on the pixel through a roulette wheel method, and the land use type that wins the competition occupies the pixel.
  • S34 go to S32, until all valid pixels of an image are iterated, the valid pixels are pixels whose pixel values in the land use data are not null values, and then return to S31 to refresh the initial impact and enter the next iteration; input the initial land use data and the total distribution probability synthesized in S31 and the restriction data constraining the land use change, set the number of iterations to stop after reaching the number of iterations, and the output result is the final simulation result of land use.
  • S4 construct the InVest model; input the four major carbon pool data (Cabove, Cbelow, Cdead, Csoil) calculated in S23 and the land use data simulated in S34 into the InVEST model to obtain the distribution of land carbon storage in the study area in the corresponding year; finally, provide an effective reference for further controlling carbon emissions in the study area.
  • major carbon pool data Cabove, Cbelow, Cdead, Csoil
  • the research object in the present invention is the rural area of province A; the data used in this research area are: the four major carbon pool data (Cabove, Cbelow, Cdead, Csoil), land use data and climate scenario data of the rural areas of province A in 2010, and a carbon storage database is established; among them, the rural carbon storage data of province A in 2010 is calculated by estimating the total carbon storage of City A in 2010 through the proportion of urban plant coverage in rural areas of province A, and the four major carbon pool data of rural areas of province A in 2010 are calculated according to the proportion of rural carbon storage data of province A.
  • the normalized difference vegetation index (NDVI) of province A is derived from the NDVI MOD13Q1 dataset.
  • the present invention can directly obtain the monthly time series NDVI dataset of the required year, thereby calculating the NPP data of province A from 2010 to 2019; finally, the data of the four major carbon pools in rural areas of province A in 2020 are obtained by superposition calculation; the data of the four major carbon pools in rural areas of province A in 2020 are used as verification data, and the source of the data of the four major carbon pools in rural areas of province A in 2020 is the same as that of the four major carbon pools in province A in 2010;
  • Land use data is interpreted from the initial high-resolution land image and is divided into six categories: cultivated land, forest land, grassland, water body, construction land and unused land; the land use change data interpreted from the initial high-resolution land image in 2020 is used as the verification data, and the land use change data of province A in 2010 has the same land use classification system as the data in 2020; based on the research experience of land use change simulation and the historical and existing data of rural areas in province A, a total of 11 driving factors of land use and land cover change are selected in this example.
  • the driving factors are DEM, slope, administrative division vector, precipitation, temperature, population density, distance to highway, distance to district and county center, distance to provincial road, distance to city center, distance to railway, and distance to county road;
  • the InVEST carbon storage and sequestration model estimates the current amount of carbon stored in the landscape and evaluates the amount of carbon sequestration over time;
  • the model first summarizes the amount of biophysical carbon stored in four carbon pools based on land use/land cover (LULC), and then estimates the change of carbon storage over time based on future LULC.
  • LULC land use/land cover
  • the climate scenario data uses shared socioeconomic pathways (SSPs) as the initial data for simulation, explores the changing trend of carbon storage in province A, and selects four shared socioeconomic scenarios: sustainable development pathway (SSP1), intermediate development pathway (SSP2), regional competition pathway (SSP3), and fossil fuel development pathway (SSP5).
  • SSP1 sustainable development pathway
  • SSP2 intermediate development pathway
  • SSP3 regional competition pathway
  • SSP5 fossil fuel development pathway
  • the main reason for selecting the rural areas of province A as the research object of this invention is that the social and economic development of province A has been rapid in recent years.
  • the accelerated urbanization process has increased the demand for rural construction land, which will occupy a large amount of land resources, resulting in a major change in the landscape pattern of rural areas in province A.
  • the prediction method includes the following steps:
  • Step 1 Refer to the "Rules for Compiling Statistical Zoning Codes and Urban-Rural Division Codes" (Guotongzi [2009] No. 91) issued by the National Bureau of Statistics, and according to the three-digit urban-rural classification code found on the official website of the National Bureau of Statistics, the areas with the first digit code of 1 are classified as towns, and the areas with the first digit code of 2 are classified as rural areas. In ArcGIS 10.7, the urban and rural areas of province A are divided;
  • Step 2 Using the SSPs scenario, from the five shared socio-economic scenarios of SSP1, SSP2, SSP3, SSP4 and SSP5, under the condition of considering climate policies, SSP1, SSP2 and SSP5 are selected as the application scenarios. Then, six driving factors affecting land use change, namely temperature, humidity, radiation, population, emissions and related policies, are selected from the selected climate scenario to form the driving force data.
  • a land use map ( Figure 2) and driving factor data ( Figure 3) with the same size as the standard raster image are generated.
  • a vegetation carbon pool ratio data set is established. The distance calculation formula from the spatial raster to the driving factor is calculated using the spatial Euclidean distance formula:
  • Step 3 Select 3 to 4 dominant plant species in a unit pixel as representative species, establish a relationship model between net primary productivity (NPP) and photosynthetically active radiation, and use relevant data under the SSPs scenario to calculate the net primary production (NPP) of the land pixel in the corresponding year:
  • NPP x,t APAR x,t ⁇ ⁇ x,t (2)
  • APAR x,t is the photosynthetically active radiation absorbed by pixel x in month t;
  • ⁇ x,t is the actual light energy rate of pixel x in month t;
  • f1 x,t and f2 x,t represent the stress effects of low and high temperatures on light energy utilization (unitless)
  • W x,t is the water stress influence coefficient (unitless) reflecting the influence of water conditions
  • ⁇ max is the maximum light utilization under ideal conditions (unit: gC ⁇ MJ-1)
  • f1, f2 and W are constants
  • the photosynthetically active radiation absorbed by vegetation depends on the total solar radiation and the photosynthetically active radiation component (FPAR), and the model is as follows:
  • S x,t is the total solar radiation of pixel x in month t
  • FPAR x,t is the absorption ratio of the incident photosynthetically active radiation by the vegetation layer
  • the constant 0.5 represents the ratio of the solar effective radiation (wavelength 0.38-0.71 ⁇ m) that can be used by vegetation to the total solar radiation
  • FPAR shows a good linear relationship with NDVI and ratio vegetation index (SR);
  • SR min and SR max correspond to the 5% and 95% lower percentiles of NDVI of each vegetation type, respectively; SR x,t can be obtained from NDVI x,t :
  • the NDVI data of province A in 2010 was put into ArcGIS tools for statistical calculation, and a confidence interval was determined by the cumulative percentage, which is generally 2%-90%.
  • the NDVI values at the cumulative percentages of 5% and 95% were taken as the minimum and maximum values, respectively, as shown in Table 1:
  • Step 4 Establish a calculation model for the total annual increment of biomass (GAI) and its relationship model with net primary productivity (NPP):
  • B is the long-term biomass of the plant (tons C/ha)
  • r is the intrinsic growth rate (year-1)
  • K is the carrying capacity (tons C/ha)
  • ⁇ i and kj represent the respective proportions of different forms (leaves, large and small branches, stems, roots) in the total biomass and turnover rates (year-1)
  • H is the harvest rate (tons ha-1 year-1);
  • mF, mR, and mS are the proportions of the carbon pools in plant leaves, roots, and stems, respectively;
  • x represents a type of land use, GAI x represents the total annual increment of biomass in type x; C above, x represents the aboveground biological carbon density in type x; and C below, x represents the underground plant root carbon density in type x.
  • Step 5 Perform linear regression analysis on the NPP data obtained in step 3, the total annual increment of biomass (GAI) calculated in step 4, and the rural land carbon storage data of City A in 2010 in step 1 to obtain the rural land carbon storage data of City A in 2020, and compare it with the measured rural land carbon storage data of province A in 2020; as shown in Table 8:
  • Step 6 Use uniform sampling strategy or proportional sampling strategy to randomly sample the initial land use data ( Figure 2) and driving force data ( Figure 3).
  • the uniform sampling mode the number of sampling points for each type of land use is the same; while in the random sampling mode, the number of sampling points for each type of land use is related to the proportion of each type of land use.
  • the sample formula after sampling is expressed as:
  • x n represents the variable of the nth driving force factor extracted at the first sampling point, and T is the transpose
  • the normalization calculation formula is:
  • max w and min w are the maximum and minimum values of the w-th driving force factor, respectively;
  • the parameter adaptive neural network algorithm can be expressed as follows:
  • ⁇ (n) is the learning rate of the nth iteration
  • E(n) and E(n-1) are the root mean square errors of the neural network outputs of two adjacent iterations
  • a, b, c are constants, and their value ranges are (1, 2), (0, 1), and (1, 1.1) respectively;
  • the parameter adaptive neural network algorithm includes an input layer, a hidden layer and an output layer. All driving force data are input into the trained neural network through the input layer. After the driving force input data is processed by the input layer, the hidden layer and the output layer in sequence, the distribution probability of each land use type in the simulation area is obtained.
  • w i,j is the one-to-one corresponding parameter between the input layer and the hidden layer, that is, the weight value between the two layers.
  • the optimizer in the neural adaptive network will train and calibrate the weight according to the value generated by the loss function;
  • xi (p, t) is the i-th variable related to the input neuron i on the grid cell p at time t;
  • connection between the input layer and the hidden layer is constructed using the sigmo activation function:
  • Each neuron in the output layer corresponds to a specific land use type.
  • the value of the lth neuron in the output layer will generate a value, which represents the probability of occurrence of the lth land use type in the grid cell. The higher the value, the greater the probability of occurrence of the target land use type in a specific grid cell.
  • the probability of occurrence of land use type k on grid cell p at time t is recorded as p(p, k, t):
  • wj ,k is the weight parameter between the hidden layer and the output layer, similar to wi,j ; the final p(p,k,t) is the distribution probability of various land use types in the simulation area calculated by the trained neural network.
  • Step 7 Construct a scanning window through the neighborhood function, and then count the number of pixels of each type in the scanning window to measure the spatial interaction of various land use types.
  • the definition of the neighborhood function is as follows:
  • the FLUS model uses adaptive inertia to represent the inheritance of previous land use types and defines the inertia coefficient according to three cases:
  • Conversion cost is another factor that affects land use dynamics.
  • the conversion cost in FLUS is fixed.
  • the land use conversion cost from c to k is expressed as sc c ⁇ k , with a value range of (0,1);
  • the comprehensive probability formula for a particular land use type occupying a plot is calculated by combining the occurrence probability, neighborhood effect, inertia coefficient and conversion cost, that is, the global total probability synthesis formula of the roulette wheel is:
  • P p,k is the occurrence probability of land use type k on grid cell p
  • P p,k is the neighborhood effect of land use type k on grid cell p at iteration time t
  • sc c ⁇ k is the conversion cost of the original land use type c to the target type k.
  • Step 8 Input the initial land use data ( Figure 2) and the total distribution probability synthesized in step 7 and the restriction data constraining land use changes, set the number of iterations to 300, and stop when the number of iterations is reached.
  • the output result is the final simulation result of land use ( Figure 4, Figure 5, Figure 6).
  • Step 9 Compare with the actual land use distribution map of City A in 2020 to verify the accuracy of the prediction ( Figure 7).
  • Step 10 Input the four major carbon pool data calculated in S23 ( Figure 8) and the land use data simulated in S34 into the InVEST model to obtain the distribution of land carbon storage in the corresponding scenario study area ( Figure 9, Figure 10, Figure 11); ultimately, it provides an effective reference for further controlling carbon emissions in the study area.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Business, Economics & Management (AREA)
  • Data Mining & Analysis (AREA)
  • Mathematical Physics (AREA)
  • Strategic Management (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Human Resources & Organizations (AREA)
  • Tourism & Hospitality (AREA)
  • General Engineering & Computer Science (AREA)
  • General Health & Medical Sciences (AREA)
  • Economics (AREA)
  • Software Systems (AREA)
  • Molecular Biology (AREA)
  • Evolutionary Computation (AREA)
  • General Business, Economics & Management (AREA)
  • Computational Linguistics (AREA)
  • Operations Research (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Computational Mathematics (AREA)
  • Marketing (AREA)
  • Artificial Intelligence (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Computing Systems (AREA)
  • Pure & Applied Mathematics (AREA)
  • Development Economics (AREA)
  • Quality & Reliability (AREA)
  • Databases & Information Systems (AREA)
  • Algebra (AREA)
  • Probability & Statistics with Applications (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Primary Health Care (AREA)
  • Educational Administration (AREA)

Abstract

本发明公开一种乡村生态系统碳储量预测方法,属于生态环境技术领域;预测方法包括:S1,获取研究区域的初始土地利用数据和气候情景数据,建立土地利用和气候情景数据库,气候情景数据为共享社会经济路径;对研究区域进行等距网格划分,在气候情景中选取驱动力因子组成驱动力数据;并计算模拟区域内栅格到土地利用变化驱动因子的距离;S2,构建CARA模型,计算出生物量的年总增量和四大碳库数据,并与初始土地利用数据进行线性回归分析,得到一定时空下土地碳储量的演变模型;S3,构建FLUS模型,并得到土地利用模拟结果;S4,构建InVEST模型,并将四大碳库数据和土地利用模拟结果输入InVEST模型,得出对应年份研究区域土地碳储量的分布情况。

Description

一种乡村生态系统碳储量预测方法 技术领域
本发明属于生态环境技术领域,具体涉及一种乡村生态系统碳储量预测方法。
背景技术
生态系统通过增加和减少大气中的温室气体来调节地球气候。其中,森林、草原、泥炭沼泽和其他陆地生态系统储存的碳比大气储存的多得多。通过将碳储存在木材、其他生物质和土壤中,生态系统会将导致气候变化的CO2排除在大气之外,从而影响气候。火灾、疾病或植被转换等会扰乱生态系统,从而释放大量的CO2;而其他管理变化,如森林恢复或替代农业实践,则会储存大量CO2。因此,管理生态系统的方式对于调节影响CO2驱动的气候至关重要。
气候变化是影响未来生态系统碳储量变化的重要因素。气候变化情景是人们对未来社会经济发展、温室气体变化趋势以及全球升温幅度等协同演进场景的一种刻画和表述;其中,社会经济发展与气候情景的关联可通过共享社会经济路径(Shared Socioeconomic Pathways,SSPs)反映;基于人类适应和缓解气候变化两个维度,SSPs确立了可持续发展(SSP1)、中度发展(SSP2)、局部或不一致发展(SSP3)、不均衡发展(SSP4)、常规发展(SSP5)五个可能影响未来社会经济发展趋势的基础路径;
生态系统服务的定量评估方法分为直接评估和间接评估,直接评估有主观和客观评估法,间接评估则分为最终、中间物质转换法和能值转换法。能值转换法目前的指标体系与转换方法并不完善,中间物质转换法由于空间数据、方 法等的选取存在不确定性,最终物质转换法和直接评估法需要克服以点代面、空间分辨率低的问题。
在最终物质转换法中,InVEST模型(Integrated Valuation of Ecosystem Services and Trade-offs,生态系统服务和权衡的综合评估模型)是当前广泛使用的工具,通常采用生产功能方法来量化和评估生态系统服务。与其他更复杂的工具相比,InVEST模型具有较低的数据要求,可以对陆地、淡水及海洋生态系统服务价值实现动态且可持续的评估。
InVEST碳储存和封存模型是InVEST模型中的一个子模型,根据当前或未来的土地利用/土地覆被(LULC)地图汇总四个碳库(地上生物量、地下生物量、土壤和死亡有机物)中储存的生物物理碳量,从而估计当前储存的碳量;但InVEST模型中的土地情景模拟较少,无法灵活满足研究要求。
FLUS模型(Future Land Use Simulation,未来土地利用变化情景模拟模型)是当前广泛使用的模拟人类活动与自然影响下的土地利用变化以及未来土地利用情景的模型,耦合了“自上而下”的系统动力学模型和“自下而上”的元胞自动机(CA)模型。FLUS模型采用神经网络算法(ANN)结合土地利用数据与包含人类活动与自然效应的多种驱动力因子,计算各用地类型在研究范围内的适宜性概率,更适用于乡村复杂多变的土地情景。此外,在土地变化模拟过程中,FLUS模型提出一种基于轮盘赌选择的自适应惯性竞争机制,该机制能有效处理多种土地利用类型在自然作用与人类活动共同影响下发生相互转化时的不确定性与复杂性,使FLUS模型具有较高的模拟精度并且能获得与现实土地利用分布相似的结果。
现有研究技术中,已存在采用FLUS-InVest模型耦合进行碳储量计算的方法,但往往使用的是已知数据,在政策和建设变化日新月异的当下时效不足,未对未来碳库变化进行模拟。
发明内容
针对现有技术的不足,本发明的目的在于提供一种乡村生态系统碳储量预测方法,解决了背景技术中提到的技术问题。
本发明的目的可以通过以下技术方案实现:
一种乡村生态系统碳储量预测方法,其特征在于,包括以下步骤:
S1,获取研究区域的初始土地利用数据和气候情景数据,建立土地利用和气候情景数据库,气候情景数据为共享社会经济路径;对研究区域进行等距网格划分,在气候情景中选取驱动力因子组成驱动力数据;
S2,构建CARA模型,计算出生物量的年总增量和四大碳库数据,并与S11中的初始土地利用数据进行线性回归分析,得到一定时空下土地碳储量的演变模型;
S3,构建FLUS模型,并得到土地利用模拟结果;
S4,构建InVest模型;并将S2得到的四大碳库数据和S3得到的土地利用模拟结果输入InVEST模型,得出对应情景(SSP1、SSP2、SSP5)研究区域土地碳储量的分布情况。
进一步地,构建CARA模型的步骤包括:
S21,在CARA模型中建立净初级生产力与共享社会经济路径中归一化植被指数的关系模型,应用驱动力数据计算得出对应年份土地像元的净初级生产力,之后对得出的研究区域各年份净初级生产力进行线性回归分析,建立研究区域 土地净初级生产力的时空演变模型;
S22,应用S21中计算出的各年份土地净初级生产力,建立土地净初级生产力与生物量的关系模型,并计算出生物量的年总增量;
S23,建立土地净初级生产力、生物量的年总增量与生物量总碳储量之间的关系模型;之后根据植物不同部位生物量碳库的比例,计算出研究区域土地各碳库数据;
S24,将S22中计算出的生物量年总增量与S1中的初始土地利用数据进行线性回归分析,得到一定时空下土地碳储量的演变模型。
本发明的有益效果:能够较好结合当前的新排放情景,较为有效地集成可收集度较高的数据,对研究区域土地碳储量分布作出较为准确的时空演变模拟与预测,从而为优化地区控制碳排放策略提供有效参考。
附图说明
为了更清楚地说明本发明实施例或现有技术中的技术方案,下面将对实施例或现有技术描述中所需要使用的附图作简单地介绍,显而易见地,对于本领域普通技术人员来讲,在不付出创造性劳动的前提下,还可以根据这些附图获得其他的附图。
图1是在SSPs情景下的乡村生态系统碳储量预测方法的流程图;
图2是2010年A省初始土地利用分类图;
图3是SSP1情景下各类土地利用变化的驱动因子集合图;
图4是SSP1情景下模拟2020年A省土地利用分布图;
图5是SSP2情景下模拟2020年A省土地利用分布图;
图6是SSP5情景下模拟2020年A省土地利用分布图;
图7是比对验证模拟结果的2020年A省实际土地利用分布图;
图8是CARA模型中输出的2020年A省乡村地区各类用地四大碳库数值结果图;
图9是SSPs1情景下本发明实施例输出的2020年A省乡村地区满足情景目标的碳储量评估结果图;
图10是SSPs2情景下本发明实施例输出的2020年A省乡村地区满足情景目标的碳储量评估结果图;
图11是SSPs5情景下本发明实施例输出的2020年A省乡村地区满足情景目标的碳储量评估结果图。
具体实施方式
下面将结合本发明实施例中的附图,对本发明实施例中的技术方案进行清楚、完整地描述,显然,所描述的实施例仅仅是本发明一部分实施例,而不是全部的实施例。基于本发明中的实施例,本领域普通技术人员在没有作出创造性劳动前提下所获得的所有其它实施例,都属于本发明保护的范围。
如图1所示,一种乡村生态系统碳储量预测方法包括以下步骤:
S1,建立数据库,并对数据预处理,具体包括:
S11,获取研究区域的初始土地利用数据和气候情景数据,建立土地利用和气候情景数据库;初始土地利用数据首先通过获取研究区域初始土地高分影像并进行预处理,其次对预处理后的高分影像解译得到分类后的影像,最后从分类后的影像获取初始土地利用数据;气候情景数据为共享社会经济路径(SSPs)。
S12,根据国家统计局印发的《统计用区划代码和城乡划分代码编制规则》(国统字〔2009〕91号)选出乡村用地,对研究区域进行等距网格划分;在该气候情景中选取影响土地利用变化的驱动力因子(坡度、湿度、温度、人口、 数字高程、到市区的距离等)组成驱动力数据;并对初始土地利用数据规定好模拟区域的范围与标准栅格影像大小,用欧式距离公式计算模拟区域内栅格到土地利用变化驱动因子的距离,在ArcMap(地理信息系统编辑软件)中生成与标准栅格影像图幅大小一致的栅格距离数据;根据现有植被指数、树叶、树根和树干生物量比例,建立植被自身碳库比例数据集;其中采用空间欧式距离公式计算空间栅格到驱动力因子的距离的计算公式为:
Figure PCTCN2022132094-appb-000001
其中(x 0,y 0)表示驱动力因子的坐标,(x n,y n)表示空间栅格的坐标,dis e表示计算的的欧式距离。
S2,构建农村地区碳积累CARA(Carbon Accumulation in Rural Area)模型,具体步骤为:
S21,在CARA模型中建立净初级生产力(NPP)与共享社会经济路径(SSPs)中归一化植被指数(NDVI)的关系模型,应用驱动力数据计算得出对应年份土地像元的净初级生产力(NPP),之后对得出的研究区域各年份净初级生产力(NPP)进行线性回归分析,建立研究区域土地净初级生产力(NPP)的时空演变模型;具体计算过程为:
NPP x,t=APAR x,t×ε x,t   (2)
APAR x,t为像元x在t月份吸收的光合有效辐射;ε x,t为像元x在t月份的实际光能利率;
ε x,t=f1 x,t×f2 x,t×W x,t×εmax   (3)
f1 x,t和f2 x,t表示低温和高温对光能利用率的胁迫作用(无单位),W x,t为水分胁迫影响系数(无单位)反映水分条件的影响,εmax是理想条件下的最大光利用率(单位:gC·MJ-1),f1,f2和W的值为常数;
植被吸收的光合有效辐射取决于总太阳辐射和光合有效辐射分量(FPAR),模型如下:
APAR x,t=S x,t×FPAR x,t×0.5   (4)
S x,t为像元x在t月份的总太阳辐射量;FPAR x,t为植被层对入射光合有效辐射的吸收比例;常数0.5表示植被所能利用的太阳有效辐射(波长为0.38~0.71μm)占总太阳辐射的比例;其中FPAR和NDVI、比值植被指数(SR)表现出较好的线性关系;
Figure PCTCN2022132094-appb-000002
SR min和SR max分别对应各植被类型NDVI的5%和95%下侧百分位数;SR x,t可由NDVI x,t求得:
Figure PCTCN2022132094-appb-000003
NDVI xt为像元x在t月份的NDVI;由于NDVI来源于NDVIMOD13Q1数据集,本发明可直接获取所需年份的月时序NDVI数据集,从而计算出各个年份的NPP;之后对得出的研究区域各年份净初级生产力(NPP)进行线性回归分析,建立研究区域土地净初级生产力(NPP)的时空演变模型。
S22,应用S21中计算出的各年份土地净初级生产力(NPP),建立土地净初级生产力(NPP)与生物量的关系模型,并计算出生物量的年总增量(GAI);计算公式如下:
建立生物量的年总增量(GAI)与净初级生产力(NPP)之间的关系模型:
GAI=rB*(1-B/K)-Σk ji*B-H   (7)
其中,B是代表种植物的长期生物量(吨C/公顷),r是内在增长率(年-1),K是承载能力(吨C/公顷),ε i和k j代表各自的比例占总生物质和周转 率(年-1)的不同形态(叶、大小分枝、茎、根),H是收获率(吨公顷-1年-1)。
S23,建立土地净初级生产力(NPP)、生物量的年总增量(GAI)与生物量总碳储量之间的关系模型;之后根据植物不同部位生物量碳库的比例,计算出研究区域土地各碳库数据C above、C below、C dead、C soil(C above表示地上生物碳密度;C below地下植物根系碳密度;C soil土壤中的有机碳密度;C dead植物凋零物及死亡生物的碳密度);
研究区域土地各碳库数据的计算步骤包括:
建立生物量总枯落碳储量L total(Mg C·a-1)与生物量的年总增量(GAI)和净初级生产力(NPP)之间的关系模型,计算出生物量总枯落碳储量L total,进而计算得出凋落物碳库:
GAI+L total=NPP   (8)
C dead=L total   (9)
之后根据植物不同部位生物量碳库的比例,建立各碳库与生物量的年总增量(GAI)的比例关系模型:
C above,x=GAI x*(m F+m S)   (10)
C below,x=GAI x*m R   (11)
其中,mF、mR、mS分别为植物叶、根、茎生物量碳库的比例;x表示一类土地利用类型,GAI x表示x类用地的生物量的年总增量;C above,x表示x类用地的地上生物碳密度;C below,x表示x类用地的地下植物根系碳密度。
S24,将S22中计算出的生物量年总增量(GAI)与S11中的初始土地利用数据进行线性回归分析,得到一定时空下土地碳储量的演变模型。
S3,构建FLUS模型,具体步骤为:
S31,对驱动力数据与初始土地利用数据进行随机点采样,获得采样数据,使用采样数据对参数自适应神经网络(ANN)算法进行训练,由训练好的神经网络计算全部的驱动力数据,以获取每种土地利用类型在模拟区域内的分布概率;
获取各种土地利用类型在模拟区域内的分布概率的计算步骤为:
采用均匀采样策略或比例采样策略对驱动力数据和初始土地利用数据进行随机点采样;均匀采样模式中,各类别用地的采样点数相同;而随机采样模式中,各类用地的采样点数量与各类用地所占的比例相关;采样后的样本公式表示为:
X=[x 1,x 2,…,x n] T   (12)
其中x n表示第一个采样点抽取的第n个驱动力因子的变量,T为转置;
使用采样数据对输入参数自适应神经网络算法进行训练之前,需要对采样数据进行归一化处理,归一化处理计算公式为:
Figure PCTCN2022132094-appb-000004
其中max w和min w分别是第w个驱动力因子的最大和最小值;
参数自适应神经网络算法可表示如下:
Figure PCTCN2022132094-appb-000005
其中η(n)是第n次迭代的学习率,E(n)和E(n-1)是相邻两次迭代的神经网络输出的均方根误差,a,b,c是常数,取值范围分别为(1,2)、(0,1)、(1,1.1);
参数自适应神经网络算法包括输入层、隐藏层和输出层,全体驱动力数据通过输入层输入训练好的神经网络,驱动力输入数据经输入层、隐藏层和输出层依次处理后,获得每种土地利用类型在模拟区域内的分布概率;
设输入层第i个神经元为x i,则隐藏层中神经元j在t时刻从网格细胞p上 的所有输入层神经元接收到的信号为:
Figure PCTCN2022132094-appb-000006
w i,j为输入层和隐藏层之间一一对应的参数,也就是两个层级间的权重值,神经自适应网络中的优化器会根据损失函数产生的数值对权值进行训练和校准;x i(p,t)是在时刻t网格细胞p上与输入神经元i相关的第i个变量;
输入层和隐藏层之间的连接由sigmo??激活函数构建:
Figure PCTCN2022132094-appb-000007
输出层的每个神经元对应一个特定的土地利用类型,输出层第l个神经元的值将生成一个值,表示网格单元第l个土地使用类型的发生概率;数值越高,说明目标土地利用类型特定网格单元发生的概率越大;时刻t时网格单元p上土地利用类型k的发生概率记为p(p,k,t):
Figure PCTCN2022132094-appb-000008
w j,k是隐藏层和输出层之间的权重参数,和w i,j相似;最终的p(p,k,t)即为训练好的神经网络计算所得的各种土地利用类型在模拟区域内的分布概率。
S32,设定好邻域大小、转换限制矩阵和每种用地类型的像元个数,将S31输出的分布概率与S11中的初始土地利用数据在土地利用模拟模块中进行迭代;迭代扫描初始土地利用数据的像元,计算每个像元在邻域内包含的土地利用类型和在邻域内所占的比例,与S31输出的分布概率、转换限制矩阵共同合成每个像元上各类土地利用类型的总分布概率;
每个像元上各类土地利用类型的总分布概率的具体计算步骤为:
通过邻域函数构建一个扫描窗口,统计扫描窗口内的各类像元的数量来衡量各种土地利用类型在空间上的相互影响,邻域函数的定义如下:
Figure PCTCN2022132094-appb-000009
Figure PCTCN2022132094-appb-000010
是在第t次迭代时特定网格单元p处土地利用类型k的邻域开发密度,
Figure PCTCN2022132094-appb-000011
表示在最后迭代时间t-1时以像元p为中心的N·N窗口内为土地利用类型为k的网格细胞的总数,con是条件函数,
Figure PCTCN2022132094-appb-000012
表示邻域内的当前被扫描到的像元,
Figure PCTCN2022132094-appb-000013
表示检测邻域内的当前被扫描的像元类型是否为第k类,w k是不同土地利用类型间的变权值,因为不同土地利用类型存在不同的邻域效应;
FLUS模型利用自适应惯性表示以往土地利用类型的继承,根据三种情况定义惯性系数:
Figure PCTCN2022132094-appb-000014
Figure PCTCN2022132094-appb-000015
表示在时刻t时土地利用类型k的惯性系数,
Figure PCTCN2022132094-appb-000016
表示土地利用类型k在迭代时间t-1时宏观需求和分配数量之间的差异;三种情况分别为:
如果特定的土地利用类型k的发展趋势满足宏观需求,即
Figure PCTCN2022132094-appb-000017
则迭代时刻t的惯性系数保持不变;
如果宏观需求具体土地利用类型k小于当前分配数量,并且土地利用类型k的发展趋势与宏观需求相矛盾,即
Figure PCTCN2022132094-appb-000018
则惯性系数迭代时间t将略有减少,将之前的系数乘
Figure PCTCN2022132094-appb-000019
如果特定土地利用类型k的宏观需求大于当前的分配数量,且土地利用类型k的发展趋势与宏观需求相矛盾,即
Figure PCTCN2022132094-appb-000020
则迭代时刻t的惯性系数将之前的系数乘
Figure PCTCN2022132094-appb-000021
会略有增加;
如此通过对CA迭代中各土地利用类型的惯性系数进行动态调整,实现不同 土地利用类型的分配相互竞争,从而使各土地利用类型的分配与宏观土地利用需求相匹配;
转换成本是影响土地利用动态的另一个因素,FLUS中的转换成本是固定不变的,对于每个土地利用对c和k,由c变化到k的土地利用转换成本表示为sc c→k,取值范围为(0,1);
综合发生概率、邻域效应、惯性系数和转换成本,计算某一特定土地利用类型占用小区的综合概率计算公式,即构成轮盘赌的全局总概率合成公式为:
Figure PCTCN2022132094-appb-000022
Figure PCTCN2022132094-appb-000023
表示在迭代时刻t时网格单元p从原始土地利用类型转换到目标类型k的组合概率,P p,k为网格单元p上土地利用类型为k的发生概率,
Figure PCTCN2022132094-appb-000024
为迭代时间t时土地利用类型k对网格单元p的邻域效应,
Figure PCTCN2022132094-appb-000025
为迭代时刻土地利用类型k的惯性系数,sc c→k原土地利用类型c向目标类型k的转换成本。
S33,将每个像元上的各类土地利用类型的总分布概率构成轮盘,通过轮盘赌的方法,使区域内各种土地利用类型在像元上竞争,竞争获胜的土地利用类型占据该像元。
S34,转到S32,直到迭代完一副影像的全部有效像元,所述有效像元即土地利用数据中像元值不为空值的像元,然后返回S31刷新初始影响进入下一次迭代;输入初始土地利用数据和S31中合成的总分布概率和约束用地变化的限制数据,设置迭代次数到达迭代次数后停止,输出的结果即为土地利用的最终模拟结果。
S4,构建InVest模型;将S23中计算得出的四大碳库数据(Cabove、Cbelow、Cdead、Csoil)与S34中模拟的土地利用数据输入InVEST模型,得出对应年份研究区域土地碳储量的分布情况;最终为研究区域进一步控制碳排放提供有效 参考。
下面结合具体实施例来阐述本发明的实施过程:
本发明中的研究对象为A省的乡村地区;本研究区域中采用的数据为:2010年的A省乡村的四大碳库数据(Cabove、Cbelow、Cdead、Csoil)、土地利用数据以及气候情景数据,建立碳储量数据库;其中,2010年A省乡村碳储量数据,由2010年A省总碳储量通过A省农村城市植物覆盖比例估算各算得出,2010年A省乡村四大碳库数据通过A省乡村碳储量数据依照比例计算得出。A省归一化植被指数(normalized difference vegetation index,NDVI)来源于NDVI MOD13Q1数据集,本发明可直接获取所需年份的月时序NDVI数据集,从而计算出A省2010-2019年份NPP数据;最终通过叠加计算得到2020年份A省乡村四大碳库数据;采用2020年的A省乡村四大碳库数据作为验证数据,2020年A省乡村四大碳库数据来源与2010年A省四大碳库来源方式相同;
土地利用数据由初始土地高分影像解译而成,分为耕地、林地、草地、水体、建设用地和未利用地6类;采用2020年的初始土地高分影像解译的土地利用变化数据作为验证数据,2010年的A省土地利用变化数据与2020年的数据具有相同的土地利用分类系统;根据土地利用变化模拟的研究经验结合A省乡村的历史数据和现有数据,本次实例共选取11种土地利用土地覆盖变化的驱动力因子。驱动力因子分别为DEM、坡度、行政区划矢量、降水、气温、人口密度、到高速公路的距离、到区县中心的距离、到省道的距离、到市中心的距离、到铁路的距离、到县道的距离;InVEST碳储存和封存模型估计景观中当前储存的碳量,并评估随时间推移的固碳量;该模型首先根据土地利用/土地覆被(LULC)汇总四个碳库中储存的生物物理碳量,其次根据未来的LULC估计碳储量随时间推移的变化。
气候情景数据为共享社会经济路径(SSPs)作为模拟的初始数据,探讨A省碳储量的变化趋势,选择可持续发展路径(SSP1)、中间发展路径(SSP2)、区域竞争路径(SSP3)以及化石燃料发展路径(SSP5)四种共享社会经济情景,需要指出的是目前中国面临着巨大的发展潜力和减排挑战,以低减排和高适应挑战为特征的SSP4情景与中国实际国情并不相符,因此本文没有选取SSP4情景。
本发明研究对象选择A省乡村地区的原因主要在于:近年来A省社会经济发展迅速。城市化进程的加快对乡村建设用地的需求增大,会侵占大量的土地资源,导致A省乡村的景观格局发生了较大变化,景观格局的改变对A省的生态环境产生了重要影响;乡村生态环境受到了剧烈的负面影响,大气中二氧化碳含量逐年增加,土地碳储量也随之变化,给乡村种植业与建设都带来了负面的影响;在碳中和的背景下,通过SSPs几种模式下,预测未来A省形态及土地利用变化,并评估土地利用变化对生碳储量的影响,对A省的乡村建设规划具有重要的意义。
如图1所示,预测方法包括以下步骤:
第1步:参照国家统计局印发的《统计用区划代码和城乡划分代码编制规则》(国统字〔2009〕91号),根据国家统计局官网上查询的城乡分类三位数代码,首位代码为1的区域划定为城镇,首位代码为2的区域划定为乡村。在Arcgis10.7中对A省城市和乡村区域进行划分;
第2步:利用SSPs情景,从SSP1、SSP2、SSP3、SSP4及SSP5五种共享社会经济情景中,在考虑气候政策的条件下,选取SSP1、SSP2、SSP5作为应用的情景。之后在选定的气候情景中选取温度,湿度,辐射,人口,排放,相关政策六种影响土地利用变化的驱动力因子组成驱动力数据。在GIS中生成与 标准栅格影像图幅大小一致的土地利用图(图2)和驱动因子数据(图3)。根据现有NDVI、树叶、树根和树干生物量比例,建立植被自身碳库比例数据集。其中采用空间欧式距离公式计算空间栅格到驱动力因子的距离计算公式为:
Figure PCTCN2022132094-appb-000026
其中(x0,y0)表示驱动力因子的坐标,(xn,yn)表示空间栅格的坐标,dise表示计算的的欧式距离。
第3步:选取单位像元内3~4种优势种群植物作为代表种,建立净初级生产力(NPP)与光合有效辐射的关系模型,应用SSPs情景下相关数据计算得出对应年份土地像元的净初级生产(NPP):
NPP x,t=APAR x,t×ε x,t   (2)
APAR x,t为像元x在t月份吸收的光合有效辐射;ε x,t为像元x在t月份的实际光能利率;
ε x,t=f1 x,t×f2 x,t×W x,t×εmax   (3)
f1 x,t和f2 x,t表示低温和高温对光能利用率的胁迫作用(无单位),W x,t为水分胁迫影响系数(无单位)反映水分条件的影响,εmax是理想条件下的最大光利用率(单位:gC·MJ-1),f1,f2和W的值为常数;
植被吸收的光合有效辐射取决于总太阳辐射和光合有效辐射分量(FPAR),模型如下:
APAR x,t=S x,t×FPAR x,t×0.5   (4)
S x,t为像元x在t月份的总太阳辐射量;FPAR x,t为植被层对入射光合有效辐射的吸收比例;常数0.5表示植被所能利用的太阳有效辐射(波长为0.38~0.71μm)占总太阳辐射的比例;其中FPAR和NDVI、比值植被指数(SR)表现出较好的线性关系;
Figure PCTCN2022132094-appb-000027
SR min和SR max分别对应各植被类型NDVI的5%和95%下侧百分位数;SR x,t可由NDVI x,t求得:
Figure PCTCN2022132094-appb-000028
将2010年A省NDVI数据置入ArcGIS工具统计计算,通过累计百分比确定一个置信区间,一般是2%-90%,这里分别取累计百分比在5%和95%时的NDVI值作为最小值和最大值,如表1所示:
Figure PCTCN2022132094-appb-000029
表1
将表1中的NDVI代入式(6),得表2:
Figure PCTCN2022132094-appb-000030
表2
将表2中的SR相关数据代入式(3)、式(4),得到表3:
Figure PCTCN2022132094-appb-000031
Figure PCTCN2022132094-appb-000032
表3
NDVIxt为像元x在t月份的NDVI。由于各土地像元类型固定,且NDVI(归一化植被指数:NDVI=(Nir-Rad)/(Nir+Rad)Nir近红外波段,Rad红光波段)一般每年随季节呈周期性变化,由于一定地域内每年相同时间点的气候情况基本相同,一定地域内每年同一时间点的同种类型土地像元的NDVI基本保持不变,因此在此发明的后续预测中将NDVI视为常数。
由于本发明可直接获取所需年份的月时序SSPs情境(125)下的温度和湿度数据(即不同年份、不同排放情景下的f1、f2与W常数不同),重复第3步上述内容,根据公式(3)(2)从而计算出各个年份的NPP,如表4所示:
Figure PCTCN2022132094-appb-000033
表4 SSPs情境(1、2、5)下各参数及变化
之后对得出的各年份的研究区域净初级生产力(NPP)进行线性回归分析,建立研究区域土地净初级生产力(NPP)的时空演变模型;
第4步:建立生物量的年总增量(GAI)的计算模型及其与净初级生产力(NPP)之间的关系模型:
GAI=rB*(1-B/K)-Σk ji*B-H   (7)
其中B是代表种植物的长期生物量(吨C/公顷),r是内在增长率(年-1), K是承载能力(吨C/公顷),εi和kj代表各自的比例占总生物质和周转率(年-1)的不同形态(叶、大小分枝、茎、根),H是收获率(吨公顷-1年-1);
根据建立的生物量总枯落碳储量L total(Mg C·a-1)与生物量的年总增量(GAI)和净初级生产力(NPP)之间的关系模型,计算出生物量总枯落碳储量L total,进而计算得出凋落物碳库:
GAI+L total=NPP   (8)
C dead=L total   (9)
之后根据植物不同部位生物量碳库的比例,建立各碳库与生物量的年总增量(GAI)的比例关系模型:
C above,x=GAI x*(m F+m S)   (10)
C below,x=GAI x*m R   (11)
其中,mF、mR、mS分别为植物叶、根、茎生物量碳库的比例;x表示一类土地利用类型,GAI x表示x类用地的生物量的年总增量;C above,x表示x类用地的地上生物碳密度;C below,x表示x类用地的地下植物根系碳密度。
将A省每种类型的土地像元内所选出的3-4种代表种植物的长期生物量、内在增长率、承载能力等数据代入式(7),计算得出A省乡村各类土地像元的生物量年总增量;再将步骤3中得出的各年份NPP值代入式(8),进而通过式(9)、(10)、(11)计算得出2010-2020年A省乡村各土地利用类型的碳密度(Mg/ha),如图8所示:
Figure PCTCN2022132094-appb-000034
Figure PCTCN2022132094-appb-000035
表5 SSPs情景(1)下预测2020年A省乡村各土地利用类型的碳密度(Mg/ha)
Figure PCTCN2022132094-appb-000036
表6 SSPs情景(2)下预测2020年A省乡村各土地利用类型的碳密度(Mg/ha)
Figure PCTCN2022132094-appb-000037
表7 SSPs情景(5)下预测2020年A省乡村各土地利用类型的碳密度(Mg/ha)
图8
第5步(验证):将第3步得出的NPP数据与将第4步中计算出的生物量年总增量(GAI)与第一步中2010年A省乡村土地碳储量数据进行线性回归分析,得出2020年A省乡村土地碳储量数据,并与实测2020年A省乡村土地碳储量数据进行比对;如表8所示:
Figure PCTCN2022132094-appb-000038
Figure PCTCN2022132094-appb-000039
表8实测2020年A省乡村各土地利用类型的碳密度(Mg/ha)
第6步:采用均匀采样策略或比例采样策略对初始土地利用数据(图2)和驱动力数据(图3)进行随机点采样。均匀采样模式中,各类别用地的采样点数相同;而随机采样模式中,各类用地的采样点数量与各类用地所占的比例相关。采样后的样本公式表示为:
X=[x 1,x 2,…,x n] T   (12)
其中x n表示第一个采样点抽取的第n个驱动力因子的变量,T为转置;
使用采样数据对输入参数自适应神经网络算法进行训练之前,需要对采样数据进行归一化处理,归一化处理计算公式为:
Figure PCTCN2022132094-appb-000040
其中max w和min w分别是第w个驱动力因子的最大和最小值;
参数自适应神经网络算法可表示如下:
Figure PCTCN2022132094-appb-000041
其中η(n)是第n次迭代的学习率,E(n)和E(n-1)是相邻两次迭代的神经网络输出的均方根误差,a,b,c是常数,取值范围分别为(1,2)、(0,1)、(1,1.1);
参数自适应神经网络算法包括输入层、隐藏层和输出层,全体驱动力数据通过输入层输入训练好的神经网络,驱动力输入数据经输入层、隐藏层和输出层依次处理后,获得每种土地利用类型在模拟区域内的分布概率;
设输入层第i个神经元为x i,则隐藏层中神经元j在t时刻从网格细胞p上的所有输入层神经元接收到的信号为:
net j(p,t)=∑ iw i,j·x i(p,t)   (15)
w i,j为输入层和隐藏层之间一一对应的参数,也就是两个层级间的权重值,神经自适应网络中的优化器会根据损失函数产生的数值对权值进行训练和校准;x i(p,t)是在时刻t网格细胞p上与输入神经元i相关的第i个变量;
输入层和隐藏层之间的连接由sigmo??激活函数构建:
Figure PCTCN2022132094-appb-000042
输出层的每个神经元对应一个特定的土地利用类型,输出层第l个神经元的值将生成一个值,表示网格单元第l个土地使用类型的发生概率;数值越高,说明目标土地利用类型特定网格单元发生的概率越大;时刻t时网格单元p上土地利用类型k的发生概率记为p(p,k,t):
Figure PCTCN2022132094-appb-000043
w j,k是隐藏层和输出层之间的权重参数,和w i,j相似;最终的p(p,k,t)即为训练好的神经网络计算所得的各种土地利用类型在模拟区域内的分布概率。
第7步:通过邻域函数构建一个扫描窗口,然后统计扫描窗口内的各类像元的数量来衡量各种土地利用类型在空间上的相互影响,邻域函数的定义如下:
Figure PCTCN2022132094-appb-000044
Figure PCTCN2022132094-appb-000045
是在第t次迭代时特定网格单元p处土地利用类型k的邻域开发密度,
Figure PCTCN2022132094-appb-000046
表示在最后迭代时间t-1时以像元p为中心的N·N窗口内为土地利用类型为k的网格细胞的总数,con是条件函数,
Figure PCTCN2022132094-appb-000047
表示邻域内的当前被扫描到的像元,
Figure PCTCN2022132094-appb-000048
表示检测邻域内的当前被扫描的像元类型是否为第k类,w k是不同土地利用类型间的变权值,因为不同土地利用类型存在不同的邻域效应;
FLUS模型利用自适应惯性表示以往土地利用类型的继承,根据三种情况定义惯性系数:
Figure PCTCN2022132094-appb-000049
Figure PCTCN2022132094-appb-000050
表示在时刻t时土地利用类型k的惯性系数,
Figure PCTCN2022132094-appb-000051
表示土地利用类型k在迭代时间t-1时宏观需求和分配数量之间的差异;三种情况分别为:
如果特定的土地利用类型k的发展趋势满足宏观需求,即
Figure PCTCN2022132094-appb-000052
则迭代时刻t的惯性系数保持不变;
如果宏观需求具体土地利用类型k小于当前分配数量,并且土地利用类型k的发展趋势与宏观需求相矛盾,即
Figure PCTCN2022132094-appb-000053
则惯性系数迭代时间t将略有减少,将之前的系数乘
Figure PCTCN2022132094-appb-000054
如果特定土地利用类型k的宏观需求大于当前的分配数量,且土地利用类型k的发展趋势与宏观需求相矛盾,即
Figure PCTCN2022132094-appb-000055
则迭代时刻t的惯性系数将之前的系数乘
Figure PCTCN2022132094-appb-000056
会略有增加;
如此通过对CA迭代中各土地利用类型的惯性系数进行动态调整,实现不同土地利用类型的分配相互竞争,从而使各土地利用类型的分配与宏观土地利用需求相匹配;
转换成本是影响土地利用动态的另一个因素,FLUS中的转换成本是固定不变的,对于每个土地利用对c和k,由c变化到k的土地利用转换成本表示为sc c→k,取值范围为(0,1);
综合发生概率、邻域效应、惯性系数和转换成本,计算某一特定土地利用类型占用小区的综合概率计算公式,即构成轮盘赌的全局总概率合成公式为:
Figure PCTCN2022132094-appb-000057
Figure PCTCN2022132094-appb-000058
表示在迭代时刻t时网格单元p从原始土地利用类型转换到目标类型k的组合概率,P p,k为网格单元p上土地利用类型为k的发生概率,
Figure PCTCN2022132094-appb-000059
为迭代时间t时土地利用类型k对网格单元p的邻域效应,
Figure PCTCN2022132094-appb-000060
为迭代时刻土地利用类型k的惯性系数,sc c→k原土地利用类型c向目标类型k的转换成本。
第8步:输入初始土地利用数据(图2)和第7步中合成的总分布概率和约束用地变化的限制数据,设置迭代次数300,到达迭代次数后停止,输出的结果即为土地利用的最终模拟结果(图4、图5、图6)。
第9步:与2020年A省实际土地利用分布图对比检验预测准确性(图7)。
第10步:将S23中计算得出的四大碳库数据(图8)与S34中模拟的土地利用数据输入InVEST模型,得出对应情景研究区域土地碳储量的分布情况(图9、图10、图11);最终为研究区域进一步控制碳排放提供有效参考。
以上显示和描述了本发明的基本原理、主要特征和本发明的优点。本行业的技术人员应该了解,本发明不受上述实施例的限制,上述实施例和说明书中描述的只是说明本发明的原理,在不脱离本发明精神和范围的前提下,本发明还会有各种变化和改进,这些变化和改进都落入要求保护的本发明范围内。

Claims (10)

  1. 一种乡村生态系统碳储量预测方法,其特征在于,包括以下步骤:
    S1,获取研究区域的初始土地利用数据和气候情景数据,建立土地利用和气候情景数据库,气候情景数据为共享社会经济路径;对研究区域进行等距网格划分,在气候情景中选取驱动力因子组成驱动力数据;并计算模拟区域内栅格到土地利用变化驱动因子的距离;
    S2,构建CARA模型,计算出生物量的年总增量和四大碳库数据,并与S11中的初始土地利用数据进行线性回归分析,得到一定时空下土地碳储量的演变模型;
    S3,构建FLUS模型,并得到土地利用模拟结果;
    S4,构建InVest模型;并将S2得到的四大碳库数据和S3得到的土地利用模拟结果输入InVEST模型,得出对应年份研究区域土地碳储量的分布情况。
  2. 根据权利要求1所述的一种乡村生态系统碳储量预测方法,其特征在于,所述S1中,采用空间欧式距离公式计算空间栅格到驱动力因子的距离,计算公式为:
    Figure PCTCN2022132094-appb-100001
    其中(x 0,y 0)表示驱动力因子的坐标,(x n,y n)表示空间栅格的坐标,dis e表示计算的的欧式距离。
  3. 根据权利要求1所述的一种乡村生态系统碳储量预测方法,其特征在于,所述S2中,构建CARA模型的步骤包括:
    S21,在CARA模型中建立净初级生产力与共享社会经济路径中归一化植被 指数的关系模型,应用驱动力数据计算得出对应年份土地像元的净初级生产力,之后对得出的研究区域各年份净初级生产力进行线性回归分析,建立研究区域土地净初级生产力的时空演变模型;
    S22,应用S21中计算出的各年份土地净初级生产力,建立土地净初级生产力与生物量的关系模型,并计算出生物量的年总增量;
    S23,建立土地净初级生产力、生物量的年总增量与生物量总碳储量之间的关系模型;之后根据植物不同部位生物量碳库的比例,计算出研究区域土地各碳库数据;
    S24,将S22中计算出的生物量年总增量与S1中的初始土地利用数据进行线性回归分析,得到一定时空下土地碳储量的演变模型。
  4. 根据权利要求3所述的一种乡村生态系统碳储量预测方法,其特征在于,所述S21中,各年份净初级生产力的计算过程为:
    NPP x,t=APAR x,t×ε x,t  (2)
    APAR x,t为像元x在t月份吸收的光合有效辐射;ε x,t为像元x在t月份的实际光能利率;
    ε x,t=f1 x,t×f2 x,t×W x,t×ε max  (3)
    f1 x,t和f2 x,t表示低温和高温对光能利用率的胁迫作用(无单位),W x,t为水分胁迫影响系数(无单位)反映水分条件的影响,εmax是理想条件下的最大光利用率(单位:gC·MJ-1),f1,f2和W的值为常数;
    植被吸收的光合有效辐射取决于总太阳辐射和光合有效辐射分量,模型如下:
    APAR x,t=S x,t×FPAR x,t×0.5  (4)
    S x,t为像元x在t月份的总太阳辐射量;FPAR x,t为植被层对入射光合有效辐射的吸收比例;常数0.5表示植被所能利用的太阳有效辐射(波长为0.38~0.71μm)占总太阳辐射的比例;其中光合有效辐射分量和植被指数、比值植被指数(SR)表现出较好的线性关系;
    Figure PCTCN2022132094-appb-100002
    SR min和SR max分别对应各植被类型NDVI的5%和95%下侧百分位数;SR x,t可由NDVI x,t求得:
    Figure PCTCN2022132094-appb-100003
    NDVI xt为像元x在t月份的植被指数。
  5. 根据权利要求4所述的一种乡村生态系统碳储量预测方法,其特征在于,所述S22中,计算生物量的年总增量(GAI)的公式如下:
    GAI=rB*(1-B/K)-Σk ji*B-H  (7)
    其中,B是代表种植物的长期生物量(吨C/公顷),r是内在增长率(年-1),K是承载能力(吨C/公顷),ε i和k j代表各自的比例占总生物质和周转率(年-1)的不同形态(叶、大小分枝、茎、根),H是收获率(吨公顷-1年-1)。
  6. 根据权利要求5所述的一种乡村生态系统碳储量预测方法,其特征在于,所述S23中,研究区域土地各碳库数据的计算步骤包括:
    建立生物量总枯落碳储量L total(Mg C·a-1)与建立土地净初级生产力(NPP)、生物量的年总增量(GAI)的关系模型,计算出生物量总枯落碳储量L total,进而计算得出凋落物碳库:
    GAI+L total=NPP  (8)
    C dead=L total  (9)
    之后根据植物不同部位生物量碳库的比例,建立各碳库与生物量的年总增量的比例关系模型:
    C above,x=GAI x*(m F+m S)  (10)
    C below,x=GAI x*m R  (11)
    其中,mF、mR、mS分别为植物叶、根、茎生物量碳库的比例;x表示一类土地利用类型,GAI x表示x类用地的生物量的年总增量;C above,x表示x类用地的地上生物碳密度;C below,x表示x类用地的地下植物根系碳密度。
  7. 根据权利要求3所述的一种乡村生态系统碳储量预测方法,其特征在于,所述S3中,得到土地利用模拟结果的步骤包括:
    S31,对驱动力数据与初始土地利用数据进行随机点采样,获得采样数据,使用采样数据对参数自适应神经网络(ANN)算法进行训练,由训练好的神经网络计算全部的驱动力数据,以获取每种土地利用类型在模拟区域内的分布概率;
    S32,设定好邻域大小、转换限制矩阵和每种用地类型的像元个数,将S31输出的分布概率与S1中的初始土地利用数据在土地利用模拟模块中进行迭代;迭代扫描初始土地利用数据的像元,计算每个像元在邻域内包含的土地利用类型和在邻域内所占的比例,与S31输出的分布概率、转换限制矩阵共同合成每个像元上各类土地利用类型的总分布概率;
    S33,将每个像元上的各类土地利用类型的总分布概率构成轮盘,通过轮盘赌的方法,使区域内各种土地利用类型在像元上竞争,竞争获胜的土地利用类型占据该像元;
    S34,转到S32,直到迭代完一副影像的全部有效像元,所述有效像元即土地利用数据中像元值不为空值的像元,然后返回S31刷新初始影响进入下一次迭代;输入初始土地利用数据和S31中合成的总分布概率和约束用地变化的限制数据,设置迭代次数到达迭代次数后停止,输出的结果即为土地利用的最终模拟结果。
  8. 根据权利要求7所述的一种乡村生态系统碳储量预测方法,其特征在于,所述S31中,获取各种土地利用类型在模拟区域内的分布概率的计算步骤为:
    1)对驱动力数据和初始土地利用数据进行随机点采样,采样后的样本公式表示为:
    X=[x 1,x 2,…,x n] T  (12)
    其中x n表示第一个采样点抽取的第n个驱动力因子的变量,T为转置;
    2)使用采样数据对输入参数自适应神经网络算法进行训练之前,需要对采样数据进行归一化处理,归一化处理计算公式为:
    Figure PCTCN2022132094-appb-100004
    其中max w和min w分别是第w个驱动力因子的最大和最小值;
    参数自适应神经网络算法可表示如下:
    Figure PCTCN2022132094-appb-100005
    其中η(n)是第n次迭代的学习率,E(n)和E(n-1)是相邻两次迭代的神经网络输出的均方根误差,a,b,c是常数,取值范围分别为(1,2)、(0,1)、(1,1.1);
    3)参数自适应神经网络算法包括输入层、隐藏层和输出层,全体驱动力数据通过输入层输入训练好的神经网络,驱动力输入数据经输入层、隐藏层和输出层依次处理后,获得每种土地利用类型在模拟区域内的分布概率;
    设输入层第i个神经元为x i,则隐藏层中神经元j在t时刻从网格细胞p上 的所有输入层神经元接收到的信号为:
    Figure PCTCN2022132094-appb-100006
    w i,j为输入层和隐藏层之间一一对应的参数,也是两个层级间的权重值,神经自适应网络中的优化器会根据损失函数产生的数值对权值进行训练和校准;
    x i(p,t)是在时刻t网格细胞p上与输入神经元i相关的第i个变量;
    输入层和隐藏层之间的连接由sigmoid激活函数构建:
    Figure PCTCN2022132094-appb-100007
    输出层的每个神经元对应一个特定的土地利用类型,输出层第l个神经元的值将生成一个值,表示网格单元第l个土地使用类型的发生概率;数值越高,说明目标土地利用类型特定网格单元发生的概率越大;时刻t时网格单元p上土地利用类型k的发生概率记为p(p,k,t):
    Figure PCTCN2022132094-appb-100008
    w j,k是隐藏层和输出层之间的权重参数,和w i,j相似;最终的p(p,k,t)即为训练好的神经网络计算所得的各种土地利用类型在模拟区域内的分布概率。
  9. 根据权利要求8所述的一种乡村生态系统碳储量预测方法,其特征在于,所述S32中,每个像元上各类土地利用类型的总分布概率的具体计算步骤为:
    1)通过邻域函数构建一个扫描窗口,统计扫描窗口内的各类像元的数量来衡量各种土地利用类型在空间上的相互影响,邻域函数如下:
    Figure PCTCN2022132094-appb-100009
    Figure PCTCN2022132094-appb-100010
    是在第t次迭代时特定网格单元p处土地利用类型k的邻域开发密度,
    Figure PCTCN2022132094-appb-100011
    表示在最后迭代时间t-1时以像元p为中心的N·N窗口内为土地利用类型为k的网格细胞的总数,con是条件函数,
    Figure PCTCN2022132094-appb-100012
    表示邻域内的当前 被扫描到的像元,
    Figure PCTCN2022132094-appb-100013
    表示检测邻域内的当前被扫描的像元类型是否为第k类,w k是不同土地利用类型间的变权值;
    2)FLUS模型利用自适应惯性表示以往土地利用类型的继承,根据三种情况定义惯性系数:
    Figure PCTCN2022132094-appb-100014
    Figure PCTCN2022132094-appb-100015
    表示在时刻t时土地利用类型k的惯性系数,
    Figure PCTCN2022132094-appb-100016
    表示土地利用类型k在迭代时间t-1时宏观需求和分配数量之间的差异;
    3)综合发生概率、邻域效应、惯性系数和转换成本,计算某一特定土地利用类型占用小区的综合概率计算公式,即构成轮盘赌的全局总概率合成公式为:
    Figure PCTCN2022132094-appb-100017
    Figure PCTCN2022132094-appb-100018
    表示在迭代时刻t时网格单元p从原始土地利用类型转换到目标类型k的组合概率,P p,k为网格单元p上土地利用类型为k的发生概率,
    Figure PCTCN2022132094-appb-100019
    为迭代时间t时土地利用类型k对网格单元p的邻域效应,
    Figure PCTCN2022132094-appb-100020
    为迭代时刻土地利用类型k的惯性系数,sc c→k为原土地利用类型c向目标类型k的转换成本。
  10. 根据权利要求9所述的一种乡村生态系统碳储量预测方法,其特征在于,在所述S32的第2)步中,三种情况分别为:
    1)特定的土地利用类型k的发展趋势满足宏观需求,即
    Figure PCTCN2022132094-appb-100021
    则迭代时刻t的惯性系数保持不变;
    2)宏观需求具体土地利用类型k小于当前分配数量,并且土地利用类型k的发展趋势与宏观需求相矛盾,即
    Figure PCTCN2022132094-appb-100022
    则惯性系数迭代时间t将略有减少,将之前的系数乘
    Figure PCTCN2022132094-appb-100023
    3)特定土地利用类型k的宏观需求大于当前的分配数量,且土地利用类型k的发展趋势与宏观需求相矛盾,即
    Figure PCTCN2022132094-appb-100024
    则迭代时刻t的惯性系数将之前的系数乘
    Figure PCTCN2022132094-appb-100025
PCT/CN2022/132094 2022-11-10 2022-11-15 一种乡村生态系统碳储量预测方法 WO2024098444A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202211406775.XA CN116151413B (zh) 2022-11-10 2022-11-10 一种乡村生态系统碳储量预测方法
CN202211406775.X 2022-11-10

Publications (1)

Publication Number Publication Date
WO2024098444A1 true WO2024098444A1 (zh) 2024-05-16

Family

ID=86355168

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/132094 WO2024098444A1 (zh) 2022-11-10 2022-11-15 一种乡村生态系统碳储量预测方法

Country Status (2)

Country Link
CN (1) CN116151413B (zh)
WO (1) WO2024098444A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN118229307A (zh) * 2024-05-23 2024-06-21 中国环境科学研究院 一种基于模型耦合的减污降碳分析方法
CN118228945A (zh) * 2024-05-24 2024-06-21 浙江省测绘科学技术研究院 基于地块分布的碳收支变化强度获取方法、终端及介质

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117892053A (zh) * 2023-07-14 2024-04-16 中国矿业大学(北京) 结合多目标规划生态系统碳储量遥感预测方法及系统
CN116596095B (zh) * 2023-07-17 2023-11-07 华能山东泰丰新能源有限公司 基于机器学习的碳排放量预测模型的训练方法及装置
CN117494860B (zh) * 2023-07-31 2024-07-26 华北水利水电大学 一种基于土地资源的生态系统评价方法及相关设备
CN116976526B (zh) * 2023-09-20 2023-12-15 北京师范大学 耦合ViViT和ANN的土地利用变化预测方法
CN117172140B (zh) * 2023-11-03 2024-01-09 北京师范大学 基于土地系统类型变化的碳储量预测方法及装置
CN117974401B (zh) * 2024-03-29 2024-06-21 南京大学 基于多源数据和模型集成的生态修复区域智能识别方法
CN118097441B (zh) * 2024-04-28 2024-06-18 四川省川建勘察设计院有限公司 一种光伏电站对区域碳储量影响评价方法

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140156549A1 (en) * 2012-11-19 2014-06-05 Syracuse University Method for estimating carbon sequestration in grasslands
CN111488902A (zh) * 2020-01-14 2020-08-04 沈阳农业大学 一种原生滨海湿地生态系统碳储量定量估算方法及系统
CN112836610A (zh) * 2021-01-26 2021-05-25 平衡机器科技(深圳)有限公司 一种基于遥感数据的土地利用变化与碳储量定量估算方法
CN114494865A (zh) * 2022-01-17 2022-05-13 南京信息工程大学 一种山地生态系统未来情景的cset方法

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP2980735A1 (en) * 2009-07-31 2016-02-03 Global Surface Intelligence Ltd Greenhouse gas grid and tracking system
CN113222316B (zh) * 2021-01-15 2023-07-07 中山大学 一种基于flus模型和生物多样性模型的变化情景模拟方法
CN115049164B (zh) * 2022-08-15 2022-10-21 四川师范大学 耦合社会网络模型和碳抵消的生态网络优化方法

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140156549A1 (en) * 2012-11-19 2014-06-05 Syracuse University Method for estimating carbon sequestration in grasslands
CN111488902A (zh) * 2020-01-14 2020-08-04 沈阳农业大学 一种原生滨海湿地生态系统碳储量定量估算方法及系统
CN112836610A (zh) * 2021-01-26 2021-05-25 平衡机器科技(深圳)有限公司 一种基于遥感数据的土地利用变化与碳储量定量估算方法
CN114494865A (zh) * 2022-01-17 2022-05-13 南京信息工程大学 一种山地生态系统未来情景的cset方法

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN118229307A (zh) * 2024-05-23 2024-06-21 中国环境科学研究院 一种基于模型耦合的减污降碳分析方法
CN118228945A (zh) * 2024-05-24 2024-06-21 浙江省测绘科学技术研究院 基于地块分布的碳收支变化强度获取方法、终端及介质

Also Published As

Publication number Publication date
CN116151413A (zh) 2023-05-23
CN116151413B (zh) 2024-04-26

Similar Documents

Publication Publication Date Title
WO2024098444A1 (zh) 一种乡村生态系统碳储量预测方法
Liang et al. Integrated assessment of land-use/land-cover dynamics on carbon storage services in the Loess Plateau of China from 1995 to 2050
Maes et al. Mapping and Assessment of Ecosystems and their Services
Chumachenko et al. Simulation modelling of long-term stand dynamics at different scenarios of forest management for coniferous–broad-leaved forests
CN101916337B (zh) 一种基于地理信息系统的水稻生产潜力动态预测方法
Boudon et al. V-Mango: a functional–structural model of mango tree growth, development and fruit production
CN113222316B (zh) 一种基于flus模型和生物多样性模型的变化情景模拟方法
CN106716479A (zh) 从局部模型导出性能的计算机实现方法
Wang et al. Dynamic characteristics and responses of ecosystem services under land use/land cover change scenarios in the Huangshui River Basin, China
Blanco et al. The effect of forest owner decision-making, climatic change and societal demands on land-use change and ecosystem service provision in Sweden
CN111783360A (zh) 高分辨率土地利用及森林景观过程耦合模拟系统及方法
Wang et al. Estimating regional terrestrial carbon fluxes for the Australian continent using a multiple-constraint approach I. Using remotely sensed data and ecological observations of net primary production
Guan et al. Comparison of big-leaf and two-leaf light use efficiency models for GPP simulation after considering a radiation scalar
CN110503283A (zh) 一种县域高标准基本农田建设综合成效的评估方法
Bai et al. Exploring regional land use dynamics under shared socioeconomic pathways: A case study in Inner Mongolia, China
AU2021102457A4 (en) High-resolution coupling simulation system and method for land use and forest landscape process
Deng et al. Spatially explicit optimization of the forest management tradeoff between timber production and carbon sequestration
Chaudhari et al. Spatial wheat yield prediction using crop simulation model, GIS, remote sensing and ground observed data
Acosta-Michlik et al. An agent-based framework for assessing vulnerability futures
CN114723142A (zh) 一种基于非支配排序遗传算法和flus模型的多目标土地利用模拟系统与方法
Kong et al. Past and projected future patterns of fractional vegetation coverage in China
Wu et al. Effects of leaf age and canopy structure on gross ecosystem production in a subtropical evergreen Chinese fir forest
CN117892550A (zh) 一种生态系统模拟与森林培育优化方法
Lo et al. Linking climate change and forest ecophysiology to project future trends in tree growth: A review of forest models
Tatarinov et al. Effect of climate change and nitrogen deposition on central-European forests: Regional-scale simulation for South Bohemia

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22964911

Country of ref document: EP

Kind code of ref document: A1