Forest carbon reserve prediction method coupling Logistic growth curve and meteorological data

Abstract

The invention discloses a forest carbon reserve prediction method coupling a Logistic growth curve and meteorological data. The method comprises the following steps: 1) constructing a relation of a forest dynamic Logistic growth curve; 2) fitting a theoretical carbon reserve value based on a Logistic growth curve; 3) a growth period monthly meteorological data calculation model; 4) constructing an actual carbon reserve prediction model coupling the meteorological data and the theoretical carbon reserve; and 5) predicting the future forest carbon reserve by using the multi-scene meteorological data. According to the method, the influence of long-time-span monthly-scale meteorological data on forest carbon reserve value prediction is considered, the forest carbon reserve prediction method coupling the Logistic growth curve and the meteorological data is constructed for forest carbon reserve prediction, and a brand new method is provided for accurate carbon reserve prediction.

Description

Technical Field

[0001] The present invention uses meteorological data, land use type data and carbon storage data obtained by remote sensing technology. For forest ecosystems, the difference between the ideal carbon storage value and the actual carbon storage value is calculated by correcting the logistic growth curve using meteorological data. By establishing a random forest model and inputting future independent variable parameters, the annual carbon storage value of the future forest is predicted, belonging to the field of remote sensing inversion technology. Background Art

[0002] With the impact of global warming, the content of greenhouse gases in the atmosphere is increasing. Among these gases, carbon dioxide is the most closely linked to human activities and vegetation growth. Vegetation has a mechanism for absorbing and storing greenhouse gases, acting as a carbon sink. Carbon sinks refer to processes, activities, or mechanisms that absorb atmospheric carbon dioxide through measures such as afforestation and vegetation restoration, thereby reducing the concentration of greenhouse gases in the atmosphere. Ecosystems closely related to human activities and carbon sinks are primarily forests, farmlands, and grasslands. These are also the three most accessible ecosystems that absorb carbon dioxide. Therefore, using random forest modeling based on existing remote sensing data and inputting future exogenous variables to predict carbon reserves is of great significance to understanding and recognizing future global climate change and greenhouse gas absorption.

[0003] However, there are currently few products predicting future forest carbon storage. Therefore, we propose a technique for predicting future carbon storage using random forest modeling based on existing data. Furthermore, given that forests are perennial vegetation, we introduce a logistic growth curve to simulate the effect of age-related changes in theoretical carbon storage. Meteorological data is then used to correct the difference between theoretical and actual carbon storage, thereby achieving the goal of predicting future carbon storage. Summary of the Invention

[0004] In order to solve the problem of predicting future forest carbon reserves, the present invention proposes a forest carbon reserve prediction method that couples the logistic growth curve with meteorological data, combines it with the random forest model for data modeling, and solves the problem of predicting perennial forest carbon reserves.

[0005] The technical solution adopted by the present invention to solve its technical problem is:

[0006] A method for predicting forest carbon reserves by coupling a logistic growth curve with meteorological data, the method comprising the following steps:

[0007] 1) The relationship between the dynamic logistic growth curve of the forest is constructed as follows:

[0008] Five landform types, namely evergreen coniferous forest, evergreen broad-leaved forest, deciduous coniferous forest, deciduous broad-leaved forest, and mixed forest, were uniformly classified as forest. Based on the relationship between forest carbon storage and growth indicators with tree age, the growth curve of forest carbon storage with tree age was fitted using the logistic function to dynamically simulate the temporal variation of forest carbon storage.

[0009] For perennial forests, the Logistic growth function is used to fit the theoretical carbon storage values ​​at different tree ages. The formula is shown in (1):

[0010]

[0011] Where B is the theoretical forest biomass density, in Mg C / ha; t is the tree age, in years; k, w, a are constants to be fitted;

[0012] 2) Fitting the theoretical carbon storage value based on the Logistic growth curve, the process is as follows:

[0013] The tree age of the forest in the predicted year is used as the input variable, and the theoretical carbon storage value of different tree ages in the predicted year is calculated using formula (1), which is used as the initial value for predicting forest carbon storage without climate influence.

[0014] 3) Calculation model for monthly meteorological data during the growing season: the process is as follows:

[0015] The monthly average precipitation and average temperature for the same months of all years between the year when the tree is 2 years old (calculation starts from the year when the tree is 2 years old in order to ignore the influence and error caused by the cultivation and transplanting of seedlings in the first two years) and the target year, that is, the calculation model of the monthly average meteorological data during the growing season is a multi-year average from January to December, which is used to characterize the meteorological influence on the growth of the forest;

[0016] 4) Construct an actual carbon storage prediction model that couples meteorological data with theoretical carbon storage. The process is as follows:

[0017] The theoretical carbon storage value calculated by the Logistic growth function and meteorological data were input as independent variables, and the actual carbon storage value was used as the dependent variable to construct a carbon storage prediction model.

[0018] Y=f(Y1,T,P) (2)

[0019] Where Y is the predicted forest carbon stock in the future year, T and P are the average monthly temperature and average precipitation from 2 years after growth to the target year, respectively;

[0020] 5) Use multi-scenario meteorological data to predict future forest carbon storage. The process is as follows:

[0021] For perennial forest ecosystems, the future time is substituted into the theoretical annual carbon storage value of the future target year calculated by the Logistic growth function. The monthly average precipitation and average temperature from the tree age of 2 to the future target year are calculated. The theoretical annual carbon storage value of the target year and the future monthly average precipitation and average temperature are substituted into the constructed forest carbon storage prediction model to predict the actual annual carbon storage of the forest in the future years.

[0022] Furthermore, in 4), the model used for fitting modeling adopts an empirical model or machine learning.

[0023] Preferably, the empirical model is a random forest model, and the machine learning is a support vector machine model.

[0024] More preferably, the model accuracy evaluation method in 4) is to use the bootstrap sampling method unique to the random forest model and use the out-of-bag (OOB) data as the validation set.

[0025] In the above 1), the growth indicators include tree height and diameter at breast height, etc.

[0026] The technical concept of the present invention is to use meteorological data to correct the difference between the ideal carbon storage value and the actual carbon storage value calculated by the logistic growth curve, fit the impact of meteorological data on carbon storage prediction by existing long-term span data, solve the carbon storage prediction problem of perennial forests based on the logistic growth curve, and predict the future carbon storage of forests based on existing meteorological data and annual carbon storage data.

[0027] The beneficial effects of the present invention are mainly manifested in: making up for the limitations of traditional aboveground biomass statistics that rely on field measurement data, introducing for the first time a prediction method for perennial forest carbon storage values ​​using theoretical carbon storage values ​​calculated by using a modified logistic function with meteorological data, and then inputting meteorological data of the target year into the established random forest model to predict the aboveground biomass of the forest in the future target year. A forest carbon storage prediction method that couples the logistic curve with meteorological data is proposed, which successfully and efficiently inverts the forest carbon storage value. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 This is a flow chart of the forest carbon storage prediction method that couples the logistic growth curve with meteorological data.

[0029] Figure 2 This is a comparison chart of the actual aboveground biomass of MODIS products and the aboveground biomass of forest ecosystems predicted using a random forest model. DETAILED DESCRIPTION

[0030] The present invention will be further described below with reference to the accompanying drawings using Ningbo, Zhejiang Province.

[0031] Reference Figure 1 A method for predicting forest carbon reserves by coupling a logistic growth curve with meteorological data comprises the following steps:

[0032] 1) Based on the data preprocessing of forest pixels, the relationship between the dynamic logistic growth curve of the forest is constructed. The process is as follows:

[0033] Five landform types, namely evergreen coniferous forest, evergreen broad-leaved forest, deciduous coniferous forest, deciduous broad-leaved forest, and mixed forest, were uniformly classified as forest. Based on the relationship between forest carbon storage and growth indicators with tree age, the growth curve of forest carbon storage with tree age was fitted using the logistic function to dynamically simulate the temporal variation of forest carbon storage.

[0034] For perennial forests, the Logistic growth function is used to fit the theoretical carbon storage values ​​at different tree ages. The formula is shown in (1):

[0035]

[0036] Where B is the theoretical forest biomass density, in Mg C / ha; t is the tree age, in years; k, w, a are constants to be fitted;

[0037] In the present embodiment, precipitation data is the Chinese 1km resolution monthly precipitation data set (1901-2021) published by Peng Shouzhang on the spatiotemporal environment big data platform in 2020, and the precipitation data from 2002 to 2021 are selected. This data set is China's monthly precipitation data, with a spatial resolution of 0.0083333 ° (about 1 km) and a time of 1901.1-2021.12. The data format is NETCDF, i.e., .nc format. This data set is based on the global 0.5 ° climate data set released by CRU and the global high-resolution climate data set released by WorldClim, and is generated by downscaling in China through the Delta spatial downscaling scheme. In addition, 496 independent meteorological observation point data are used for verification, and the verification results are credible. The geographical space scope included in this data set is the main land of the country (including Hong Kong, Macao and Taiwan), excluding areas such as South China Sea islands and reefs. For ease of storage, the data are all int16 type stored in nc files, and the precipitation units are 0.1mm respectively.

[0038] The predicted precipitation dataset uses the 1km-resolution, multi-scenario, multi-model monthly precipitation dataset for China, published by Peng Shouzhang in 2022 on the Spatiotemporal Environmental Big Data Platform. This dataset contains multi-scenario, multi-model monthly precipitation data for China at a spatial resolution of 0.0083333° (approximately 1km) and covers the period from January 2021 to December 2100. The data are in NETCDF format. The data are generated using the Delta spatial downscaling scheme over China, based on the global climate model dataset (>100km) released by the IPCC Coupled Model Intercomparison Project Phase 6 (CMIP6) and the global high-resolution climate dataset released by WorldClim. The data are based on the latest IPCC SSP scenarios (SSP119, SSP245, and SSP585). Each scenario includes climate data from three GCMs (EC-Earth3, GFDL-ESM4, and MRI-ESM2-0). The dataset covers the main landmass of China and excludes islands and reefs in the South China Sea. The unit is 0.1mm.

[0039] The temperature data is the monthly average temperature dataset (1901-2021) with a resolution of 1 km for China, which was published by Peng Shouzhang on the spatiotemporal environment big data platform in 2019. The data is the monthly average temperature data for China with a spatial resolution of 0.0083333° (about 1 km) and a time period of 1901.1-2021.12. The data format is NETCDF, i.e., .nc format. The data unit is 0.1°C. This dataset was generated by downscaling the global 0.5° climate dataset released by CRU and the global high-resolution climate dataset released by WorldClim using the Delta spatial downscaling scheme in China. In addition, data from 496 independent meteorological observation points were used for verification, and the verification results are credible. The geographical space covered by this dataset is the main land area of ​​the country (including Hong Kong, Macao and Taiwan), and does not include areas such as islands and reefs in the South China Sea.

[0040] The predicted temperature dataset uses the 1km-resolution, multi-scenario, multi-model monthly mean temperature dataset for China, published by Peng Shouzhang in 2022 on the Spatiotemporal Tripolar Environmental Big Data Platform. This dataset contains multi-scenario, multi-model monthly mean temperature data for China at a spatial resolution of 0.0083333° (approximately 1km) and covers the period from January 2021 to December 2100. The data are in NETCDF format. The data are generated using the Delta spatial downscaling scheme over China, based on the global climate model dataset (>100km) released by the IPCC Coupled Model Intercomparison Project Phase 6 (CMIP6) and the global high-resolution climate dataset released by WorldClim. The data are based on the latest IPCC SSP scenarios (SSP119, SSP245, and SSP585). Each scenario includes climate data from three GCMs (EC-Earth3, GFDL-ESM4, and MRI-ESM2-0). The dataset covers the main landmass of China and excludes islands and reefs in the South China Sea. The unit is 0.1°C.

[0041] Land use data is from the MODIS MCD12Q1 product, with a resolution of 500 m. The Terra and Aqua Moderate Resolution Imaging Spectroradiometer (MODIS) Land Cover Classification (MCD12Q1) Version 6.1 data product provides annual global land cover types. The MCD12Q1 Version 6.1 data product is derived using supervised classification of MODIS Terra and Aqua reflectance data. Land cover types are derived from the International Geosphere-Biosphere Programme (IGBP) and the University of Maryland (UMD) Leaf Area Index (LAI), Biome-Biogeochemical Cycle (BGC), and Plant Functional Type (PFT) classification schemes. The supervised classification is then followed by additional post-processing, incorporating prior knowledge and auxiliary information, to further refine specific classes. The Food and Agriculture Organization (FAO) Land Cover Classification System (LCCS) provides an additional layer of land cover attribute assessment for land cover, land use, and surface hydrology. In Google Earth Engine, the current land use data were cropped according to the area of ​​interest, and the meteorological data were resampled to 1 km to obtain the current land use status of Ningbo from 2002 to 2021.

[0042] The tree age data is a 30m resolution map of young forests in China compiled by Xiao et al. for 2020. Mapping young forest age at a fine spatial resolution is important for improving the accuracy of estimating land-to-atmosphere carbon fluxes and guiding forest management practices. In recent decades, China has actively implemented afforestation and forest protection programs, laying the foundation for achieving carbon neutrality. However, datasets on young forest ages across China at fine spatial resolution are scarce. The 2020 map of young forest ages in China (1-31 years) achieved an overall accuracy of 90.28%. This forest age map provides fundamental data for studying forest carbon cycling and forest ecosystem services, and is of great guidance to government agencies. The tree age data was clipped to the area of ​​interest in Google Earth Engine and resampled to 1km to obtain the 2020 tree age data for Ningbo, from which the current tree age status for other years can be extrapolated.

[0043] Carbon storage data are based on a 1km-resolution map of aboveground biomass carbon stocks in Chinese forests developed by Chen et al., covering 20 years from 2002 to 2021. This dataset consists of annual maps of aboveground and belowground biomass carbon storage in Chinese forests from 2002 to 2021. This dataset was developed by combining SAR, optical, and passive microwave remote sensing data with extensive field surveys. By cropping the tree age data based on the region of interest in Google Earth Engine, Ningbo's forest carbon storage data for the period 2002-2021 were obtained.

[0044] 2) The process of constructing the parameters of the logistic growth curve and fitting the theoretical carbon storage value is as follows:

[0045] Based on the dynamic relationship between forest carbon storage and tree age, a forest logistic growth curve was constructed. Based on this growth function relationship and the tree age data of each sample point, the theoretical carbon storage value of the forest at each sample point was fitted using formula (1).

[0046] 3) Construction of the monthly meteorological data model during the growing season. The process is as follows:

[0047] Based on the 2020 30m resolution map of young forests in China drawn by Xiao et al., the specific year when the tree age of each sample point was equal to 2 was deduced according to the year. A monthly mean automatic calculation model was constructed to calculate the monthly average precipitation and average temperature of each sample point in the same month of all years between the year when the tree age was equal to 2 and the target year. That is, the calculation model of the monthly average meteorological data during the growing season is the multi-year average from January to December, which is used to characterize the meteorological influence on the growth of the forest.

[0048] 4) Construction of the random forest model of forest carbon storage: the process is as follows:

[0049] Land use and tree age data were cropped and resampled in Google Earth Engine to a 1km resolution, consistent with the resolution of the carbon storage map. After masking these data to make the number of pixels consistent, monthly temperature, monthly precipitation, tree age, land use type, and carbon storage were derived year by year from 2002 to 2021. The tree age data for each year were calculated and modified accordingly based on the 2020 30m resolution map of young forests in China drawn by Xiao et al. Temperature and precipitation data were averaged from the age of 2 to the same month in different years from 2002 to 2021 for each sample point. The impact of the meteorological conditions on carbon storage changes throughout the growth cycle of each sample point was considered, and starting from the year when the tree was 2 years old was to reduce the impact of cultivation and transplanting in the previous two years.

[0050] Import the obtained dataset into Python. Then, random forest modeling was performed based on different landform types (evergreen coniferous forest, evergreen broad-leaved forest, deciduous coniferous forest, deciduous broad-leaved forest, and mixed forest were all classified as forests). The modeling factors of the five forest types were monthly temperature, monthly precipitation, and the theoretical carbon storage value calculated by the Logistic growth function. The model accuracy was evaluated by using the random forest model's unique self-service sampling method, using the out-of-bag data (OOB) as the validation set, and tuning the model hyperparameters. The R of the model was calculated separately. 2 , RMSE and MAPE, compare the accuracy of each model and obtain preliminary research results. The model accuracy results are as follows Figure 2 As shown. R of the forest model 2 The value of the proposed method is 0.5172, the RMSE is 0.000286Tg, and the AMPE is 0.0592%.

[0051] 5) For the prediction of forest carbon storage in the target year, the process is as follows:

[0052] For a target year, such as 2035, forest carbon storage is predicted. Based on an established model for predicting forest carbon storage based on logistic growth curves coupled with meteorological data, the model's independent variable parameters include the average monthly precipitation and temperature for the same month from the year when trees are 2 years old to the target year. The theoretical forest carbon storage value calculated based on tree age using the logistic growth function is then input. Monthly average temperature and precipitation data are predicted using meteorological data from three scenarios (SSP119, SSP245, and SSP585). This model can predict carbon storage values ​​for the target year under the three meteorological scenarios. The results are shown in Table 1.

[0053] Table 1 shows the forest carbon storage projections for 2025, 2035, 2050, 2075, and 2100;

[0054]

[0055]

[0056] The embodiments of this specification are merely examples of implementations of the invention and are provided for illustrative purposes only. The scope of protection of the present invention should not be considered limited to the specific embodiments described in these embodiments. The scope of protection of the present invention also extends to equivalent technical means that can be conceived by a person of ordinary skill in the art based on the invention.

Claims

1. A method for predicting forest carbon reserves by coupling logistic growth curves with meteorological data, characterized in that: The method comprises the following steps: 1) The relationship between the dynamic logistic growth curve of the forest is constructed as follows: Five landform types, namely evergreen coniferous forest, evergreen broad-leaved forest, deciduous coniferous forest, deciduous broad-leaved forest, and mixed forest, were uniformly classified as forest. Based on the relationship between forest carbon storage and growth indicators with tree age, the growth curve of forest carbon storage with tree age was fitted using the logistic function to dynamically simulate the temporal variation of forest carbon storage. For perennial forests, the Logistic growth function is used to fit the theoretical carbon storage values ​​at different tree ages. The formula is shown in (1): Where B is the theoretical forest biomass density, in Mg C / ha; t is the tree age, in years; k, w, a are constants to be fitted; 2) Fitting the theoretical carbon storage value based on the Logistic growth curve, the process is as follows: The tree age of the forest in the predicted year is used as the input variable, and the theoretical carbon storage value of different tree ages in the predicted year is calculated using formula (1), which is used as the initial value for predicting forest carbon storage without climate influence. 3) Calculation model for monthly meteorological data during the growing season: the process is as follows: The monthly average precipitation and average temperature in the same month of all years from the year when the tree is 2 years old to the target year, that is, the calculation model of the monthly average meteorological data during the growing season is the multi-year average from January to December, which is used to characterize the meteorological influence on the growth of the forest; 4) Construct an actual carbon storage prediction model that couples meteorological data with theoretical carbon storage. The process is as follows: The theoretical carbon storage value calculated by the Logistic growth function and meteorological data were input as independent variables, and the actual carbon storage value was used as the dependent variable to construct a carbon storage prediction model. Y=f(Y1,T,P) (2) Where Y is the predicted forest carbon stock in the future year, T and P are the average monthly temperature and average precipitation from 2 years after growth to the target year, respectively; 5) Use multi-scenario meteorological data to predict future forest carbon storage. The process is as follows: For perennial forest ecosystems, the future time is substituted into the theoretical annual carbon storage value of the future target year calculated by the Logistic growth function. The monthly average precipitation and average temperature from the tree age of 2 to the future target year are calculated. The theoretical annual carbon storage value of the target year and the future monthly average precipitation and average temperature are substituted into the constructed forest carbon storage prediction model to predict the actual annual carbon storage of the forest in the future years.

2. The method for predicting forest carbon reserves by coupling logistic growth curves with meteorological data according to claim 1, characterized in that: In the above 4), the model used for fitting modeling adopts an empirical model or machine learning.

3. The method for predicting forest carbon reserves by coupling logistic growth curves with meteorological data according to claim 2, characterized in that: In the above 4), the empirical model is a random forest model, and the machine learning is a support vector machine model.

4. A method for predicting forest carbon reserves by coupling logistic growth curves with meteorological data according to any one of claims 1 to 3, characterized in that: In 4), the model accuracy is evaluated by using the bootstrap sampling method of the random forest model, and the out-of-bag data OOB is used as the validation set.

5. A method for predicting forest carbon reserves by coupling logistic growth curves with meteorological data according to any one of claims 1 to 3, characterized in that: In the above 1), the growth indicators include tree height and diameter at breast height.

Images