CN111080173B - Estimation method of carbon flux of forest system - Google Patents
Estimation method of carbon flux of forest system Download PDFInfo
- Publication number
- CN111080173B CN111080173B CN201911418927.6A CN201911418927A CN111080173B CN 111080173 B CN111080173 B CN 111080173B CN 201911418927 A CN201911418927 A CN 201911418927A CN 111080173 B CN111080173 B CN 111080173B
- Authority
- CN
- China
- Prior art keywords
- data
- carbon flux
- model
- biome
- physiological
- Prior art date
- Legal status (The legal status 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 status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION 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
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION 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
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
Abstract
A method for estimating carbon flux of a forest system relates to the technical field of carbon flux estimation. The method aims to solve the problems of large calculated amount and low accuracy of an estimated result of the existing method for estimating the carbon flux of the forest system. According to the method, a Biome-BGC model based on a point scale is used for carrying out sensitivity analysis on vegetation physiological and ecological parameters and carbon flux, determining sensitive vegetation physiological and ecological parameters, then carrying out correlation analysis, and rejecting data types which are correlated with data types in the sensitive vegetation physiological and ecological parameters; determining a Biome-BGC model parameter of a point scale by combining meteorological data and taking the Biome-BGC model parameter as a model parameter of an area scale Biome-BGC model, determining an error by combining an error neural network model, and finally determining an estimation result of the carbon flux of the forest system. The method is mainly used for estimating the carbon flux of the forest system.
Description
Technical Field
The invention relates to a method for estimating carbon flux. Belongs to the technical field of forest system carbon flux estimation.
Background
With the development of human activities and industry, especially the use of petrochemical fuels in industry, the human ecological environment has raised a series of problems, among which the "greenhouse effect" problem is prominent. The relevant data show that: the atmospheric CO2 concentration was 280ppm at the early stage of the industrial revolution, and by 2017, the CO2 concentration had been rapidly increased to 406.42ppm, and greenhouse gases were one of the causes of global temperature rise and caused other environmental problems.
The forest ecosystem is the largest carbon reservoir of the land ecosystem, the exchange capacity of the forest and the atmospheric carbon accounts for more than 90% of the total exchange capacity of the land ecosystem and the atmospheric carbon, and the carbon flux of the forest ecosystem plays an important role in global carbon circulation and carbon balance. On the basis of large-scale range research, under the condition that various disturbances such as felling, fire, plant diseases and insect pests are not generally considered, the carbon flux of the forest ecosystem is approximate to NEP, the NEP represents the net carbon flux between the land ecosystem and the atmosphere, and the NEP directly and quantitatively describes the carbon source/sink function and the carbon balance state of the forest ecosystem. When NEP value is greater than 0, forest carbon sink; when NEP value is less than 0, the forest carbon source is obtained; when NEP value is 0, forest carbon balance is achieved. People can help develop deep research of greenhouse effect according to the actual forest carbon flux estimation result, thereby providing a series of corresponding improvement measures.
A very large number of expert scholars have conducted relatively intensive research on the carbon flux of forest systems, which has become a system of systematic research relatively. The carbon flux estimation by using various carbon cycle models becomes a necessary means in global carbon cycle research, the existing research results can realize the estimation and prediction of the carbon flux of the forest system, and along with the gradual improvement of the estimation models, the carbon flux estimation accuracy of the forest system is also improved year by year, so that more prominent results are obtained. In the research of the carbon flux of the forest system, the carbon flux factors influencing the forest system are very many, although the more factors considered by the model, the better the estimation precision of the model theoretically. But the carbon flux research of forest systems still has the following problems:
1. the improvement of the consideration factors is pursued, which results in that the estimation time is greatly increased, the requirement on hardware is very high, the estimation overhead is too large, and the cost is increased.
2. Although theoretically, the estimation accuracy of the model is better as the factors are more, the contribution rate of some factors to the accuracy of the model is small, and if the factors are taken into consideration in the relevant research, the accuracy of the model is not greatly influenced, but the running speed of the estimation model is severely slowed down. If these factors are not considered, selection and selection of factors are involved, but the existing research on the aspect is relatively little, incomplete and not mature.
3. The estimation model can be universal theoretically, but because a forest ecosystem is different along with the change of a geographical position and the climate at the position, factors influencing estimation are changed, but factors in a certain place have larger influence on the precision of the model, and the influence degree of the factors is reduced. What is more important is that the research of people does not have a relatively accurate and uniform system, the estimation factors lack reference significance, the selection of the influencing factors becomes difficult, the development of the estimation model is seriously influenced, and the accuracy of the estimation model is also seriously influenced.
The Biome-BGC model has interpretability, can realize the explanation of the principle of carbon flux, and is widely applied to the estimation of the carbon flux, thereby realizing the analysis of CO2 concentration and greenhouse effect in the global or regional range. However, the Biome-BGC model used at present has the problems, and the simulation effect is not ideal.
Disclosure of Invention
The method aims to solve the problems of large calculated amount and low accuracy of an estimated result of the existing method for estimating the carbon flux of the forest system. And further provides an estimation method of the carbon flux of the forest system.
A method for estimating the carbon flux of a forest system comprises the following steps:
training process:
w1, acquiring meteorological data, basic geographic data, vegetation physiological and ecological parameter data, soil data and carbon flux observation data in an estimation area range;
physiological and ecological parameters of vegetation
i＝1,2,3,......,N1，N1The number of specific parameter types contained in the vegetation physiological and ecological parameters;
w2, determining a Biome-BGC model of a point scale and a region scale Biome-BGC model according to the initialized Biome-BGC model;
w3, and a Biome-BGC model based on point scale, for performing sensitivity analysis on vegetation physiological and ecological parameters and carbon flux, determining vegetation physiological and ecological parameters sensitive to carbon flux observation data, and recording as sensitive vegetation physiological and ecological parameters
j＝1,2,3,......,N2，N2≤N1；
Performing correlation analysis on the physiological and ecological parameters of the sensitive vegetation, and rejecting the parameter types in the physiological and ecological parameters of the sensitive vegetation
Parameter types with dependencies
k＝1,2,3,......,N2And k is not equal to j, and obtaining physiological and ecological parameters of the non-relevant sensitive vegetation
l ＝1,2,3,......,N3，N3≤N2；
W4 physiological and ecological parameters of non-relevant sensitive vegetation
The method comprises the steps that data and a Biome-BGC model of meteorological data input point scale for carbon flux observation are used, the carbon flux observation data are used as verification data, and an EnKF method is used for optimizing the carbon flux observation data to obtain optimized model parameters;
w5, taking the optimized model parameters as the model parameters of the area scale Biome-BGC model; operating an area scale Biome-BGC model so as to obtain a carbon flux simulation result Y of the forest ecosystem;
w6 physiological and ecological parameters of non-relevant sensitive vegetation
As input values of the neural network, the normalized data of (a) and the normalized data of (Y); making a difference between the carbon flux simulation result of the forest ecological system and the carbon flux observation data, and recording as E; taking E as a target value of the neural network; training the neural network model to obtain a trained neural network model, and recording as an error neural network model;
the neural network model adopts a fully connected neural network,
number N of neurons in input layer3+1, wherein N3The input of each neuron is the parameter type of the physiological and ecological parameters of the non-relevant sensitive vegetation
Normalized data of (1), input of 1 neuron is normalized data of Y;
the output layer is an estimated value E' of the difference between the carbon flux simulation result of the forest ecosystem and the carbon flux observation data;
and (3) estimation process:
simulating the carbon flux of the forest system in the area range by using an area scale Biome-BGC model, and simulating a result Y of the carbon flux of the forest ecological system;
carrying out error estimation by using an error neural network model;
and taking the sum of the simulation result Y of the carbon flux of the forest ecological system and the estimation result of the error as the estimation result of the carbon flux of the forest system in the area range.
Further, the number of hidden layers in the neural network model is 3-7, preferably 5.
Further, the specific parameter types contained in the vegetation physiological and ecological parameters include: transfer growth to growth season ratio, withering to growth season ratio, leaf and fine root annual turnover ratio, live wood annual turnover ratio, annual plant total death and withering ratio, fire plant death and withering ratio, new fine root C: new leaf C distribution, new stem C: new leaf C allocation, new wood C: new total wood C allocation, new coarse heel C: new stem C allocation, current growth: storage growth, leaf C: ratio N, litter C: n ratio, thin root C: n ratio, wood C: n ratio, dead wood C: n ratio, litter easily-decomposed substance content, litter cellulose content, litter lignin content, fine easily-decomposed substance content, fine cellulose content, fine lignin content, dead wood cellulose content, dead wood lignin content, canopy water retention coefficient, canopy extinction coefficient, and total leaf area: projection leaf area, canopy average specific leaf area, yang-born leaf SLA: the nitrogen content of the leaves in the YingshenSLA and the Rubisco enzyme, the maximum leaf stomatal conductance, the stratum corneum conductance, the boundary layer conductance, the leaf water potential when the stomatal conductance is opened, the leaf water potential when the stomatal conductance is closed and the saturated water-vapor pressure difference when the stomatal conductance is opened.
Further, the area scale Biome-BGC model is a Biome-BGC model established by utilizing area grid basic geographic data, area grid vegetation physiological and ecological parameter data, area grid soil data and area weather grid data.
Further, the regional grid basic geographic data, the regional grid vegetation physiological and ecological parameter data and the regional grid soil data are obtained by regional rasterization according to the basic geographic data, the vegetation physiological and ecological parameter data and the soil data.
Further, the regional weather grid data is obtained as follows:
acquiring meteorological grid point data according to meteorological data; inputting the meteorological lattice point data into an MT-CLIM model for climate simulation to obtain day-by-day meteorological data required by a Biome-BGC model;
and performing Kriging interpolation operation on the meteorological element simulation result by utilizing ArcGIS software to obtain a meteorological grid data set of the area scale Biome-BGC model, and recording the meteorological grid data set as the area meteorological grid data set to obtain area meteorological grid data.
Has the advantages that:
when the method is used for estimating the carbon flux of the forest system, correlation analysis is further performed on sensitive vegetation physiological and ecological parameter data on the basis of sensitivity analysis on the vegetation physiological and ecological parameter data and the carbon flux observation data, the calculated amount is greatly reduced, and for a Biome-BGC model, the accuracy is not reduced (the accuracy is reduced in theory), but the estimation accuracy of the carbon flux of the forest system is improved. Compared with the prior art, the method greatly reduces the calculated amount of the whole model, reduces the estimation time of the carbon flux of the forest system and improves the efficiency.
The method not only ensures the estimation accuracy of the carbon flux of the forest system, but also has wide application range. The method can be applied to the estimation of the carbon flux of the forest system in various environments and any areas (including dimensional ranges), is hardly limited by the vegetation type of the forest ecosystem, and can be correspondingly adjusted along with the forest environments in various environments and any areas, so that the method does not need to be limited by the conditions of the application range, and the method has wider applicability and also has wider applicability. Compared with the existing model which is a method for estimating the carbon flux aiming at a specific forest environment (the accuracy rate is seriously deteriorated and is further reduced by directly using the forest environment in other ranges), the method provided by the invention can have higher estimation accuracy rate when being applied to any specific forest environment.
Drawings
FIG. 1 is a block flow diagram of the present invention;
FIG. 2 is a line graph of NPP maximum values determined by various methods;
FIG. 3 is a line graph of NPP minimum values determined by various methods;
FIG. 4 is a line graph of the NPP mean values determined by the respective methods.
Detailed Description
The first embodiment is as follows: the present embodiment is described in detail with reference to figure 1,
the embodiment is a method for estimating the carbon flux of a forest system, which comprises the following steps:
step one, training process:
s1, acquiring meteorological data, basic geographic data, vegetation physiological and ecological parameter data (actually, vegetation physiological and ecological data comprise various vegetation physiological and ecological parameters, so the vegetation physiological and ecological parameter data are called as vegetation physiological and ecological parameter data), soil data and carbon flux observation data in an estimation area range;
in the process of obtaining data, part of meteorological data, soil data and the like can be directly obtained according to corresponding data obtaining devices which are actually arranged, and can also be obtained according to existing data, for example, observation data of an observation station is directly used (the observation data of the observation station is obtained and analyzed according to the corresponding data obtaining devices of the observation station);
all or part of the data of the vegetation physiological and ecological parameters can be obtained by utilizing an inversion technology according to the obtained actual remote sensing image; or directly using the observation data of the observation station (the observation data of the observation station is actually obtained by inversion analysis of the remote sensing image, and/or is obtained by analysis according to the data obtained by the corresponding data acquisition device);
the basic geographic data can be obtained from existing data, such as observation data of an observation station or data of a related geographic investigation structure directly used;
carbon flux observations the observations from the observation station were used.
In fact, the most fundamental way of obtaining the above data is by a measuring device and an analyzing device, which are all in the prior art, and the present invention is not particularly limited, but may be obtained in the way available in the art.
Since the cost of actually setting up the measuring device is very expensive and the workload is very large, the present embodiment directly uses the observation data of the observation station, that is: acquiring meteorological data, basic geographic data, vegetation physiological and ecological parameter data, soil data and carbon flux observation data in an estimation area range according to observation data of an observation station;
physiological and ecological parameters of vegetation
i＝1,2,3,......,N1，N1The number of specific parameter types contained in the vegetation physiological and ecological parameters;
in some embodiments, the specific parameter types of the vegetation physiological-ecological parameters include: transfer growth to growth season ratio, withering to growth season ratio, leaf and fine root annual turnover ratio, live wood annual turnover ratio, annual plant total death and withering ratio, fire plant death and withering ratio, new fine root C: new leaf C distribution, new stem C: new leaf C allocation, new wood C: new total wood C allocation, new coarse heel C: new stem C allocation, current growth: storage growth, leaf C: ratio N, litter C: n ratio, thin root C: n ratio, wood C: n ratio, dead wood C: n ratio, litter easily-decomposed substance content, litter cellulose content, litter lignin content, fine easily-decomposed substance content, fine cellulose content, fine lignin content, dead wood cellulose content, dead wood lignin content, canopy water retention coefficient, canopy extinction coefficient, and total leaf area: projection leaf area, canopy average specific leaf area, yang-born leaf SLA: the nitrogen content of the leaves in the YingshenSLA and the Rubisco enzyme, the maximum leaf stomatal conductance, the stratum corneum conductance, the boundary layer conductance, the leaf water potential when the stomatal conductance is opened, the leaf water potential when the stomatal conductance is closed and the saturated water-vapor pressure difference when the stomatal conductance is opened. The vegetation physiological and ecological parameters comprise but are not limited to the specific parameter types, if the specific parameter types are more, the calculated amount advantage of the vegetation physiological and ecological parameters is more obvious, and the accuracy can be improved to a certain extent.
S2, acquiring meteorological grid point data according to the meteorological data;
the process of acquiring the meteorological grid point data can be acquired by utilizing the existing meteorological grid point data processing mode. Because the meteorological grid data volume of the multi-time sequence is huge, the meteorological grid point vector point data in the preprocessing result based on the 0.5 degrees multiplied by 0.5 degrees meteorological grid point data set is obtained by considering the factors such as the running time cost of the model.
Inputting the meteorological lattice point data into an MT-CLIM model for climate simulation to obtain day-by-day meteorological data required by a Biome-BGC model;
performing Kriging interpolation operation on the meteorological element simulation result by utilizing ArcGIS software to obtain a meteorological grid data set of the area scale Biome-BGC model, and recording the meteorological grid data set as an area meteorological grid data set; the spatial resolution of the meteorological raster data of the area scale Biome-BGC model is 1km, and the time resolution is a daily value.
In the operation of the MT-CLIM model, the initialization data (· ini) of the model mainly includes: inputting/outputting file basic information, control parameters and model parameters; the input data (. mtcin) of the MT-CLIM model, namely the measured meteorological data of the base station, mainly comprises: the highest air temperature, the lowest air temperature and the precipitation; the output data (. mtc43) of the MT-CLIM model mainly includes: maximum air temperature, minimum air temperature, precipitation, daytime average air temperature, precipitation, humidity, incident solar short-wave radiation and day length. And respectively verifying the simulation result with the measured data of the ground radiation station and the secondary trend surface statistical model.
Further, meteorological data (maximum temperature, minimum temperature, and precipitation) of carbon flux observation are used as meteorological input, and MT-CLIM model simulation is used to obtain other daily value meteorological element data required by the Biome-BGC model: saturated vapor pressure difference VPD, solar incident short-wave radiation srad and day length are used for simulating GPP and Re by a site scale Biome-BGC model.
The carbon flux observation data used in the subsequent Biome-BGC model is day-by-day carbon flux observation data, and the day-by-day carbon flux observation data is obtained according to the carbon flux observation data.
Obtaining area grid basic geographic data, area grid vegetation physiological and ecological parameter data and area grid soil data according to the area grid processing of the basic geographic data, the vegetation physiological and ecological parameter data and the soil data;
s3, determining a Biome-BGC model of a point scale and a region scale Biome-BGC model according to the initialized Biome-BGC model;
the area scale Biome-BGC model is established by utilizing area grid basic geographic data, area grid vegetation physiological and ecological parameter data, area grid soil data and area meteorological grid data;
s4, analyzing the sensitivity of vegetation physiological and ecological parameters and carbon flux based on a point scale Biome-BGC model, determining vegetation physiological and ecological parameters sensitive to carbon flux observation data, and recording as sensitive vegetation physiological and ecological parameters
j＝1,2,3,......,N2，N2≤N1；
The sensitive vegetation physiological-ecological parameters are all or part of the vegetation physiological-ecological parameters and are determined according to the precision requirement of the model, and the number of the specific sensitive vegetation physiological-ecological parameters can be controlled through a sensitivity judgment index. The invention relates to a point scale-based Biome-BGC model, which is used for carrying out sensitivity analysis on vegetation physiological and ecological parameters and carbon flux and is carried out by utilizing a method for carrying out parameter sensitivity analysis in Biome-BGC model parameter optimization and northeast forest carbon flux estimation research. In fact, the sensitivity can be analyzed directly through data, but the data calculation amount of the method is very large, and the result determined by analyzing the sensitivity for even one month is not necessarily accurate; the invention is based on the analysis of the mechanism model, and not only has small calculated amount, but also has more accurate result.
Performing correlation analysis on the physiological and ecological parameters of the sensitive vegetation, and rejecting the parameter types in the physiological and ecological parameters of the sensitive vegetation
Parameter types with dependencies
k＝1,2,3,......,N2And k is not equal to j, and obtaining physiological and ecological parameters of the non-relevant sensitive vegetation
l ＝1,2,3,......,N3，N3≤N2；
The physiological and ecological parameters of the non-relevant sensitive vegetation are all or part of the physiological and ecological parameters of the sensitive vegetation, and are determined according to the model precision requirement and the calculated quantity requirement, and the number of the specific physiological and ecological parameters of the non-relevant sensitive vegetation can be controlled through the significant index of the relevance. The correlation analysis may be performed by using a conventional correlation analysis method, and the present invention is not particularly limited.
S5, applying physiological and ecological parameters of non-relevant sensitive vegetation
The method comprises the steps that a Biome-BGC model of meteorological data input point scale for carbon flux observation is used, carbon flux observation data are used as verification data, and an EnKF method is used for optimizing the carbon flux observation data to obtain optimized model parameters;
s6, taking the optimized model parameters as the model parameters of the area scale Biome-BGC model; operating an area scale Biome-BGC model, and realizing the processing of the model in a batch processing and parallel computing mode and the output of a grid result by means of an ArcGIS platform and an ENVI/IDL platform, thereby obtaining a carbon flux simulation result Y of a forest ecosystem, wherein the carbon flux simulation result Y can be direct carbon flux data or one or more of carbon flux related indexes, and mainly comprises the following steps: total primary productivity GPP, net primary productivity NPP, net ecosystem productivity NEP, plant maintenance respiration MR, plant growth respiration GR and heterotrophic respiration HR;
s7, applying physiological and ecological parameters of non-relevant sensitive vegetation
As input values of the neural network, the normalized data of (a) and the normalized data of (Y); making a difference between the carbon flux simulation result of the forest ecological system and the carbon flux observation data, and recording as E; taking E as a target value of the neural network; training the neural network model to obtain a trained neural network model, and recording as an error neural network model;
the neural network model adopts a fully-connected neural network, and specifically comprises the following steps:
number N of neurons in input layer3+1, wherein N3The input of each neuron is the parameter type of the physiological and ecological parameters of the non-relevant sensitive vegetation
Normalized data of (1), input of 1 neuron is normalized data of Y;
the number of layers of the hidden layer is set according to the actual precision requirement and the time overhead, and is generally 3-7 layers, preferably 5 layers;
the output layer is an estimated value E' of the difference between the carbon flux simulation result of the forest ecosystem and the carbon flux observation data;
step two, estimation process:
simulating the carbon flux of the forest system in the area range by using an area scale Biome-BGC model, and simulating a result Y of the carbon flux of the forest ecological system;
carrying out error estimation by using an error neural network model;
and taking the sum of the simulation result Y of the carbon flux of the forest ecological system and the estimation result of the error as the estimation result of the carbon flux of the forest system in the area range.
The method is not disclosed in the invention, is all the prior art, and can be carried out according to the method in Biome-BGC model parameter optimization and northeast forest carbon flux estimation research.
In the invention, because the point-scale Biome-BGC model uses relatively accurate data and is relatively easy to optimize (calculation amount and calculation time can be saved), the Biome-BGC model parameters determined by using the point-scale data are relatively accurate, the model parameters determined by using the relatively accurate point-scale Biome-BGC model are relatively accurate, and the operation mechanism of the model can be relatively accurately explained. The relatively accurate point scale data is equivalent to the regional scale data, and the regional scale data is obtained through MT-CLIM model simulation, interpolation, other operation and other means, so that the real data cannot be directly replaced to a certain extent, the point scale data which can be directly obtained is relatively accurate relative to the regional scale data, and the model can be ensured to have more accurate model parameters on the whole. In fact, there are some errors in the above process, which are caused by two main aspects: firstly, inputting errors existing in a Biome-BGC model by using parameters; particularly, for carbon flux simulation of different climates at different geographical positions (Biome-BGC model parameters are different), the carbon flux simulation of different climates at different positions by using the model has more errors; secondly, in order to improve the accuracy of the model and reduce the whole calculation amount, the number of state variables (parameter types) which have large actual influence (high sensitivity) on the Biome-BGC model is reduced through correlation analysis on the basis of sensitivity analysis, and the result of the Biome-BGC model is theoretically certainly deteriorated.
However, the invention uses a processing mode which is different from the existing processing thought through the analysis, research and experiment of data, namely: in the methods of the prior scholars and experts, a large error always exists in the simulation through the Biome-BGC model, the method is not processed from the perspective of reducing the error to improve the simulation precision, but temporarily acknowledges and allows the existence of the large error, tolerates the theoretically larger error (the error caused by the number of state variables which actually affect the Biome-BGC model to a large extent is further reduced through correlation analysis on the basis of sensitivity analysis), and compensates all the errors as a whole error.
Although the neural network model has a relatively strong automatic learning capability and can be used for directly estimating the carbon flux, the neural network model has the characteristic of inaccurate explanation, the interpretability of the carbon flux estimation is damaged by using the neural network model for estimating the carbon flux, and a relatively accurate estimation result cannot be obtained by directly estimating the carbon flux by using the neural network model. The method does not directly use the neural network model to estimate the carbon flux data, but uses the neural network model as an error estimation tool, processes the errors caused by uncertainty and irrecoverability by using the characteristic that the neural network cannot accurately explain, equivalently uses the characteristic that the neural network cannot accurately explain, and simultaneously ensures that the whole process has interpretability. Experiments prove that the method can obtain a good estimation result.
Examples
The simulation is performed according to a specific implementation manner, and a scheme not limited in the specific implementation manner is performed according to a mode in Biome-BGC model parameter optimization and northeast forest carbon flux estimation research (hereinafter, referred to as a paper). To illustrate the effect of the invention by comparison, experiments were also conducted to estimate the carbon flux in the northeast forest. Carbon flux has a close relationship with ecosystem productivity, and both have the same physical meaning under certain assumed conditions. On the basis of large-scale range research, the carbon flux of the ecosystem of a deep forest can not be directly and comprehensively measured, and under the condition that various disturbances such as felling, fire, plant diseases and insect pests are not generally considered, the carbon flux of the ecosystem of the deep forest is approximate to NPP (net primary productivity) for verifying the estimation result of a regional scale model by ignoring alloxyrespiration, and only the NPP is estimated in the embodiment.
In the invention, fixed sample plot data is checked by adopting forest resources in the Xiaoxingan area, which are the same as the paper, and sample plot NPP data in 2003-2006 is estimated by utilizing an indirect forest ground NPP calculation method, so that the estimation of the NPP of the forest in the Xiaoxingan area of the corresponding year by the Biome-BGC model is verified. In fact, the invention utilizes the daily value data of multiple time series to train and obtain NPP estimation, in order to compare with the existing data and convert into annual NPP data, and in order to compare, only show and explain the NPP data of 2003-2006, the NPP determined by each method is as shown in Table 1
TABLE 1
The maximum, minimum and mean values of NPP determined by each method are plotted against the data in table 1.
The NPP maximum line plots determined by the respective methods are shown in FIG. 2. As can be seen from fig. 2, the method of the present invention not only approximates the trend of the same-day censored data with respect to the NPP maximum, but also approximates the same-day censored data in terms of data value as compared with the MODIS and paper methods, so that the simulation result for the NPP maximum in one year is more accurate.
The NPP minimum line plots determined by the various methods are shown in FIG. 3. It can be seen from fig. 3 that the estimated NPP minimum of the present invention is more closely comparable to the data reviewed in the paper approach, and is theoretically superior to the paper approach. In the method of MODIS, the data is unstable, and the data in 2003 and 2004 are closer to the data of the same sample check, but a large value deviation is generated in 2005 and 2006, particularly, in 2005, not only a very large difference is generated in the value, but also the trend does not accord with the data of the same sample check, even a reverse trend is generated, so that the method of the present invention is superior to the MODIS in the stability and the trend of the difference in the value, although the data of the present invention cannot be compared with the MODIS to a certain extent.
The line graph of the NPP mean values determined by the respective methods is shown in fig. 4. As can be seen from fig. 4, the trend of the NPP mean curve obtained by the present invention is similar to the trend of the NPP value of the sample-examined data, and compared with the trend of the paper and the sample-examined data, the trend of the NPP change of the present invention is closer to the sample-examined data, and the data value is closer to the sample-examined data, so the result of the present invention is better. Moreover, the mode of the invention is superior to MODIS in both fold line change trend and the similarity degree of NPP average data and sample-mode clearing data. Therefore, theoretically, the method is more accurate in result and better in simulation estimation effect of the model.
By combining the various conditions, the method has more accurate estimation result of the carbon flux of the northeast forest and better model performance.
Finally, it should be noted that the detailed description and examples are only illustrative and explanatory of the technical solution of the present invention, and the scope of the claims should not be limited thereby. It is intended that all such modifications and variations be included within the scope of the invention as defined in the following claims and the description.
Claims (6)
1. A method for estimating the carbon flux of a forest system is characterized by comprising the following steps:
training process:
w1, acquiring meteorological data, basic geographic data, vegetation physiological and ecological parameter data, soil data and carbon flux observation data in an estimation area range;
physiological and ecological parameters of vegetation
N1The number of specific parameter types contained in the vegetation physiological and ecological parameters;
w2, determining a Biome-BGC model of a point scale and a region scale Biome-BGC model according to the initialized Biome-BGC model;
w3 Point-based ScaleThe Biome-BGC model analyzes the sensitivity of vegetation physiological and ecological parameters and carbon flux, determines the vegetation physiological and ecological parameters sensitive to carbon flux observation data, and records as sensitive vegetation physiological and ecological parameters
Performing correlation analysis on the physiological and ecological parameters of the sensitive vegetation, and rejecting the parameter types in the physiological and ecological parameters of the sensitive vegetation
Parameter types with dependencies
And k is not equal to j, and obtaining the physiological and ecological parameters X of the non-relevant sensitive vegetation3 3=[Xl 3],l= 1,2,3,......,N3， N3≤ N2；
W4 physiological and ecological parameters of non-relevant sensitive vegetation
The method comprises the steps that data and a Biome-BGC model of meteorological data input point scale for carbon flux observation are used, the carbon flux observation data are used as verification data, and an EnKF method is used for optimizing the carbon flux observation data to obtain optimized model parameters;
w5, taking the optimized model parameters as the model parameters of the area scale Biome-BGC model; operating an area scale Biome-BGC model so as to obtain a carbon flux simulation result Y of the forest ecosystem;
w6 physiological and ecological parameters of non-relevant sensitive vegetation
As input values of the neural network, the normalized data of (a) and the normalized data of (Y); grow forest intoThe difference between the carbon flux simulation result of the state system and the carbon flux observation data is recorded as E; taking E as a target value of the neural network; training the neural network model to obtain a trained neural network model, and recording as an error neural network model;
the neural network model adopts a fully connected neural network,
number N of neurons in input layer3+1, wherein N3The input of each neuron is the parameter type of the physiological and ecological parameters of the non-relevant sensitive vegetation
Normalized data of (1), input of 1 neuron is normalized data of Y;
the output layer is an estimated value E' of the difference between the carbon flux simulation result of the forest ecosystem and the carbon flux observation data;
and (3) estimation process:
simulating the carbon flux of the forest system in the area range by using an area scale Biome-BGC model, and simulating a result Y of the carbon flux of the forest ecological system;
carrying out error estimation by using an error neural network model;
and taking the sum of the simulation result Y of the carbon flux of the forest ecological system and the estimation result of the error as the estimation result of the carbon flux of the forest system in the area range.
2. The method for estimating the carbon flux of the forest system according to claim 1, wherein the number of hidden layers in the neural network model is 3-7.
3. The method for estimating the carbon flux of the forest system according to claim 2, wherein the number of hidden layers in the neural network model is 5.
4. The method for estimating carbon flux of a forest system according to claim 1, 2 or 3, wherein the area scale Biome-BGC model is a Biome-BGC model established by using area grid basic geographic data, area grid vegetation physiological and ecological parameter data, area grid soil data and area weather grid data.
5. The method for estimating the carbon flux of the forest system according to claim 4, wherein the regional grid basic geographic data, the regional grid vegetation physiological and ecological parameter data and the regional grid soil data are obtained by regional rasterization according to the basic geographic data, the vegetation physiological and ecological parameter data and the soil data.
6. A forest system carbon flux estimation method as claimed in claim 4, wherein the regional meteorological grid data is obtained as follows:
acquiring meteorological grid point data according to meteorological data; inputting the meteorological lattice point data into an MT-CLIM model for climate simulation to obtain day-by-day meteorological data required by a Biome-BGC model;
and performing Kriging interpolation operation on the meteorological element simulation result by utilizing ArcGIS software to obtain a meteorological grid data set of the area scale Biome-BGC model, and recording the meteorological grid data set as the area meteorological grid data set to obtain area meteorological grid data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911418927.6A CN111080173B (en) | 2019-12-31 | 2019-12-31 | Estimation method of carbon flux of forest system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911418927.6A CN111080173B (en) | 2019-12-31 | 2019-12-31 | Estimation method of carbon flux of forest system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111080173A CN111080173A (en) | 2020-04-28 |
| CN111080173B true CN111080173B (en) | 2020-11-03 |
Family
ID=70321192
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911418927.6A Active CN111080173B (en) | 2019-12-31 | 2019-12-31 | Estimation method of carbon flux of forest system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111080173B (en) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111723482B (en) * | 2020-06-17 | 2023-11-21 | 南京大学 | Satellite CO-based 2 Method for inverting surface carbon flux by column concentration observation |
| CN111881407B (en) * | 2020-07-30 | 2021-06-11 | 中国科学院地理科学与资源研究所 | Surface water, heat and carbon flux coupling estimation method based on remote sensing information |
| CN112819365A (en) * | 2021-02-23 | 2021-05-18 | 中国科学院空天信息创新研究院 | Carbon sink detection method and device, storage medium and electronic equipment |
| CN113449976A (en) * | 2021-06-21 | 2021-09-28 | 广东翁源滃江源国家湿地公园管理处 | Forestry carbon metering method based on ecological process model |
| CN115952702B (en) * | 2022-08-30 | 2023-07-07 | 中国气象科学研究院 | Forest NEP calculation method based on FORCHN model and remote sensing data |
| CN116341724B (en) * | 2023-03-08 | 2023-09-15 | 黑龙江省生态气象中心 | Carbon absorption pre-estimating method based on global climate mode driven carbon circulation mechanism model |
| CN116858999B (en) * | 2023-07-05 | 2024-01-16 | 中环宇恩(广东)生态科技有限公司 | Carbon sink statistical method based on monitoring and evaluating carbon sink potential of mangrove ecological system |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102507906A (en) * | 2011-11-15 | 2012-06-20 | 莫路锋 | WSN (Wireless Sensor Network) forest environmental benefit monitoring system based on wide-range soil carbon flux monitoring system |
| CN107014951A (en) * | 2017-02-27 | 2017-08-04 | 北京林业大学 | Forest ecosystem breathes Carbon flux assay method |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107016445A (en) * | 2009-02-02 | 2017-08-04 | 行星排放管理公司 | The system for monitoring each system of greenhouse gases flux |
| CN109636069A (en) * | 2019-01-29 | 2019-04-16 | 平安科技（深圳）有限公司 | Data processing method, device, equipment and storage medium |
-
2019
- 2019-12-31 CN CN201911418927.6A patent/CN111080173B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102507906A (en) * | 2011-11-15 | 2012-06-20 | 莫路锋 | WSN (Wireless Sensor Network) forest environmental benefit monitoring system based on wide-range soil carbon flux monitoring system |
| CN107014951A (en) * | 2017-02-27 | 2017-08-04 | 北京林业大学 | Forest ecosystem breathes Carbon flux assay method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111080173A (en) | 2020-04-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111080173B (en) | Estimation method of carbon flux of forest system | |
| Niu et al. | An increasing trend in the ratio of transpiration to total terrestrial evapotranspiration in China from 1982 to 2015 caused by greening and warming | |
| Oguntunde et al. | Relationship between rice yield and climate variables in southwest Nigeria using multiple linear regression and support vector machine analysis | |
| CN111783987A (en) | Farmland reference crop evapotranspiration prediction method based on improved BP neural network | |
| Shiri et al. | Evaluation of different data management scenarios for estimating daily reference evapotranspiration | |
| CN112906298B (en) | Blueberry yield prediction method based on machine learning | |
| Liu et al. | Linking field survey with crop modeling to forecast maize yield in smallholder farmers’ fields in Tanzania | |
| Briegel et al. | Factors controlling long-term carbon dioxide exchange between a Douglas-fir stand and the atmosphere identified using an artificial neural network approach | |
| Lilleleht et al. | Spatial forest structure reconstruction as a strategy for mitigating edge-bias in circular monitoring plots | |
| Tian et al. | Fusion of multiple models for improving gross primary production estimation with eddy covariance data based on machine learning | |
| CN116894514A (en) | Crop yield prediction method and system based on soil quality index | |
| Richard et al. | Filling gaps in micro-meteorological data | |
| Maurer et al. | Optimizing spatial distribution of watershed-scale hydrologic models using Gaussian Mixture Models | |
| CN113128871B (en) | Cooperative estimation method for larch distribution change and productivity under climate change condition | |
| CN115510991A (en) | Sunlight greenhouse aphid early warning method based on 1DCNN-LSTM multi-source information fusion model | |
| Bagnara et al. | Bayesian calibration of simple forest models with multiplicative mathematical structure: A case study with two Light Use Efficiency models in an alpine forest | |
| Rit | Impact of Climate Change on Agriculture: Empirical Evidence from South Asian Countries | |
| Jiang et al. | Application and evaluation of an improved LSTM model in the soil moisture prediction of southeast chinese tobacco-producing areas | |
| Celis et al. | Simple and Innovative Methods to Estimate Gross Primary Production and Transpiration of Crops: A Review | |
| Qin et al. | Application of least squares vector machines in modelling water vapor and carbon dioxide fluxes over a cropland | |
| Rawat et al. | Regional time series forecasting of chickpea using ARIMA and neural network models in central plains of Uttar Pradesh (India) | |
| Estacio | Simulating the future of the Ifugao rice terraces through observations of the past | |
| Niaghi et al. | Evaluation of Machine Learning Techniques for Daily Reference Evapotranspiration Estimation | |
| Mallick et al. | Soil and atmospheric drought explain the biophysical conductance responses in diagnostic and prognostic evaporation models over two contrasting European forest sites | |
| McCurdy | Characterizing spatiotemporal variation in LAI of Virginia Pine Plantations |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |