Search

CN-121980942-A - Carbon source amount measuring method and system for alpine wetland ecosystem

CN121980942ACN 121980942 ACN121980942 ACN 121980942ACN-121980942-A

Abstract

The invention provides a method and a system for measuring carbon source quantity of an ecological system of a high-cold wetland, wherein the method comprises the following steps of 1, obtaining carbon flux data observed through experiments, 2, numerically simulating space-time distribution of carbon flux, 3, inverting the carbon flux by satellite remote sensing high-cold wetland, 4, calculating and analyzing past space-time scale carbon storage of the high-cold wetland, and 5, constructing a carbon flux machine learning model and estimating future trend of the carbon flux machine learning model. According to the invention, an observation test-numerical simulation-machine learning algorithm is taken as a core technical thought, an observation and simulation data is utilized to establish and train a machine learning estimated carbon flux model, the carbon source/sink and the space-time variation of reserves of the alpine wetland ecosystem are estimated, the space-time distribution of the carbon flux of the alpine wetland is calculated and analyzed, and the carbon reserve trend under the climate change background is estimated. Breaks through the technical bottlenecks of lack and low precision of the carbon source/sink and reserve investigation technology of the alpine wetland ecosystem, and can be applied to investigation and evaluation of the carbon reserve of the alpine wetland ecosystem.

Inventors

  • LIU BEIBEI
  • LIU CHENTAO
  • LAI XIN
  • WEN JUN

Assignees

  • 成都信息工程大学

Dates

Publication Date
20260505
Application Date
20260123

Claims (10)

  1. 1. The method for measuring the carbon source of the alpine wetland ecosystem is characterized by comprising the following steps of: step 1, obtaining carbon flux data observed through experiments; step 2, space-time release of numerical simulation carbon flux: Obtaining space-time continuous carbon flux data of the high wetland through a weather research and forecast mode coupled with a greenhouse gas module, and correcting by utilizing the experimentally observed carbon flux data obtained in the step 1; step3, inverting the satellite remote sensing carbon flux: Acquiring carbon flux data by using satellite inversion data, correcting and checking the carbon flux data observed in the experiment in the step 1, and realizing expansion of time and space by using satellite remote sensing inversion carbon flux; Step4, calculating past space-time scale carbon storage: Obtaining past space-time scale carbon storage by utilizing standardized carbon flux data-time-space three-dimensional integral-unit conversion and uncertainty quantification based on the carbon flux data set obtained in the step 2 and the step 3; And 5, predicting future trend by using a machine learning model: Dividing the carbon flux data set obtained in the step 2 and the step 3 into a training data set and a test data set, constructing and training a machine learning carbon flux estimation model based on the training data set, simulating future climate situation data based on a sixth international coupling mode comparison plan, acquiring the variation trend of air temperature, precipitation and humidity in the future period, inputting a trained random forest model as an independent variable by combining land utilization variation situations, outputting carbon flux predicted values of wetlands in different periods in the future through the random forest model, and obtaining a past time-space scale carbon storage test data set for verification and evaluating a predicted result by utilizing the step 4.
  2. 2. The method of claim 1, comprising obtaining cold-dampness observation station land-gas carbon flux observation test data, the test data comprising carbon flux data and environmental driver data.
  3. 3. The method according to claim 2, wherein the pre-treatment and quality control are performed on the carbon flux observation test data between land and air of the alpine wetland observation station in step 1.
  4. 4. The method of claim 1, wherein the weather research and forecast model coupled with the greenhouse gas module of step 2 includes the steps of dual nested grid design and physical process parameterization: The double nested grid design comprises an outermost simulation area, a second simulation area, a first simulation area, a second simulation area and a third simulation area, wherein the outermost simulation area is used for analyzing the remote influence of western wind rapid flow and quaternary wind water vapor transmission large-scale circulation on carbon exchange; Physical parametrization, namely selecting a fast radiation transmission mode for long-wave radiation, adopting short-wave radiation parameters for short-wave radiation, adopting a road surface process for a boundary layer, and adopting a Norian road surface process to simulate and calculate the temperature and humidity of soil, a snow cover and the energy flux of the ground surface.
  5. 5. The method of claim 1, wherein step3 comprises feature variable screening and statistical model construction: Screening a characteristic variable, namely screening a core driving factor with the highest interpretation rate of carbon flux from satellite remote sensing parameters and meteorological factors based on a Pearson correlation analysis and classification and regression decision tree algorithm; and (3) constructing a statistical model, namely constructing a carbon flux-driving factor regression model by using a random forest integration algorithm.
  6. 6. The method of claim 5, wherein the core driving factors include an enhanced vegetation index that characterizes vegetation productivity, a quantified surface water index that characterizes moisture conditions, a soil temperature that characterizes soil respiration drive, and an average inter-daily air temperature that characterizes net exchange control as key variables; and (3) constructing a statistical model, namely taking the carbon flux observed in the step (1) as a dependent variable, taking the surface water index of the enhanced vegetation index/quantized surface water representing the vegetation productivity inverted by the same-period satellite and the global climate atmospheric analysis meteorological data as independent variables, dividing a training set and a verification set through a 7:3 ratio, and cross-verifying and optimizing model parameters.
  7. 7. The method of claim 5, wherein step 4 comprises normalizing the carbon flux data to space-time continuous net ecosystem exchange data; Combining time dynamics of net ecosystem exchange amount with space distribution, and realizing conversion from carbon flux to carbon storage through integration; And unit conversion and uncertainty quantification, namely obtaining the annual carbon storage uncertainty range through Monte Carlo simulation analysis of data errors.
  8. 8. The method of claim 1, wherein the trained random forest model in step 5 uses the carbon flux observation data in step 1 as a dependent variable, three meteorological elements including air temperature, soil temperature and air relative humidity are selected, two remote sensing parameter independent variables including enhanced vegetation index and surface water index are selected, key driving factors influencing carbon flux are identified through classification and regression algorithm of decision trees, and a random forest integrated learning method is adopted to construct a plurality of decision trees and integrate prediction results.
  9. 9. The method of claim 1, wherein the trained random forest model in step 5 is divided into a training set and a verification set according to the ratio of 7:3 by using the carbon flux data of the complete step 2 numerical simulation and the carbon flux data set of satellite remote sensing inversion after space-time expansion in step 3 in a model training and verification link, the core parameters of the random forest are optimized, the stability and generalization capability of the model are evaluated by adopting a cross verification method, and the model precision uses root mean square error and a decision coefficient as evaluation indexes to obtain two random forest models.
  10. 10. A computer system comprising a processor and a memory, the memory storing a computer program, wherein the computer program, when processed by the processor, implements the steps of a method for measuring the amount of carbon source in an ecosystem of a alpine wetland according to any one of claims 1 to 9.

Description

Carbon source amount measuring method and system for alpine wetland ecosystem Technical Field The invention belongs to the technical fields of observation tests, numerical simulation, machine learning algorithms, data processing and the like, and particularly relates to a carbon source amount measuring method and system for an ecological system of a alpine wetland. Background In the global climate change background, the research of the wetland ecosystem-atmospheric carbon exchange process has become a hot spot in the international academy. The heating rate of the Qinghai-Tibet plateau serving as a third pole is more than twice that of the low-altitude area, and the high-cold wetland serving as a long-term stable huge carbon warehouse is critical to the global carbon balance adjustment in the carbon circulation process. If the cover wetland is used as the high-cold peat area with the largest area in China, the peat development history is long, the carbon storage and climate regulation are irreplaceable ecological functions, but the superposition interference of climate change and human activities seriously threatens the stability of a carbon warehouse, and the carbon exchange process and the regulation mechanism of the cover wetland become important scientific demands for coping with the climate change. At present, research on carbon flux at home and abroad forms a technical system combining direct observation and indirect estimation. The visual observation takes vortex related technology as a core and is widely applied to the international general network, so that the multi-scale continuous observation can be realized. However, in the alpine region, the number of observation stations is small, the data time span is short, the carbon flux spatial heterogeneity can be caused by the difficulty in single-point observation to cover complex terrains, the data loss rate is high due to the severe alpine environment, and the deep application analysis is deficient. Although the indirect estimation from the global scale model to the mesoscale atmosphere-vegetation coupling model realizes technical innovation and part of models are combined with multisource data assimilation to improve the precision, the special simulation research for the carbon flux between the cover wetland and the atmosphere is less, the simulation precision is insufficient, and the space-time pattern of carbon exchange is difficult to accurately describe. Meanwhile, the analysis of the carbon flux driving mechanism by the existing research is insufficient, the ecological process and the human influence are not fully coupled in the future trend prediction, and the accurate carbon management is difficult to support. Disclosure of Invention The invention aims to solve the defects in the prior art and provide a method for measuring the carbon source of the ecological system of the alpine wetland, which is characterized by extracting the characteristics of multi-layer data, checking and improving the parameterization scheme, establishing, training, checking a machine learning model and the like, and finally calculating the space-time distribution of the carbon reserves of the ecological system of the alpine wetland and predicting the future change trend of the carbon reserves. The invention integrates multi-source data and a method, builds an observation-simulation-machine learning multi-source data fusion system on an observation level and combines a distributed sampling method, breaks through single-point limitation and improves data reliability. The simulation layer is used for carrying out localized improvement on weather research and forecast modes coupled with a greenhouse gas module, an empirical orthogonal function analysis preparation-moisture space-time coupling mechanism is used for adapting to special ecological and topographic characteristics of the alpine wetland, a scale expansion and trend prediction layer is used for realizing accurate expansion of carbon flux by a random forest algorithm, and a classification regression tree and a random forest model are combined for overcoming the defect that a traditional research driving mechanism is not used for analyzing, and providing scientific support basis for wetland carbon sink assessment, rising repair and climate change adaptation strategies. The invention adopts the following technical scheme: a method for measuring carbon source of an ecological system of a high-cold wetland comprises the following steps: step 1, obtaining carbon flux data observed through experiments; step 2, space-time release of numerical simulation carbon flux: Obtaining space-time continuous carbon flux data of the high wetland through a weather research and forecast mode coupled with a greenhouse gas module, and correcting by utilizing the experimentally observed carbon flux data obtained in the step 1; step3, inverting the satellite remote sensing carbon flux: Acquiring carbon flux data by using satellite inversi