Tree species total primary productivity fine simulation method and system

Abstract

The invention provides a fine simulation method and system for total primary productivity of tree species in the technical field of ecological remote sensing and carbon cycle simulation. The method comprises the following steps: S1, acquiring leaf hyperspectral reflectivity data, photosynthetic physiological parameter data and canopy hyperspectral remote sensing image data of different tree species in a target area; s2, extracting characteristic wave bands based on the maximum carboxylation rate in the hyperspectral reflectivity data of the leaves and the photosynthetic physiological parameter data, constructing a characteristic wave band combination, and constructing a Vcmax hyperspectral inversion model based on the characteristic wave band combination; step S3, tree species classification is carried out by using the canopy hyperspectral remote sensing image data, and a dynamic Vcmax distribution diagram is generated in combination with the Vcmax hyperspectral inversion model; and S4, replacing static Vcmax parameters of the ecological process model based on the dynamic Vcmax distribution diagram, and inputting driving data into the ecological process model to simulate a GPP space-time pattern. The method has the advantages that the accuracy and practicability of tree species total primary productivity simulation are greatly improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of ecological remote sensing and carbon cycle simulation, and particularly discloses a tree species gross primary productivity fine simulation method and system. BACKGROUND

[0002] Gross primary productivity (GPP) of terrestrial ecosystems is a key indicator of the ability of vegetation to fix carbon dioxide in the atmosphere through photosynthesis, and is also the basis for understanding the process of terrestrial carbon cycle. Due to the high spatial heterogeneity of terrestrial ecosystems, accurate and fine simulation of the spatiotemporal pattern of GPP of different tree species has important scientific and practical significance for assessing the carbon sequestration capacity of regional ecosystems, optimizing ecological management measures, and developing effective ecosystem sequestration strategies.

[0003] Ecological process models are currently the main tool for assessing the carbon sequestration capacity of regional ecosystems, but there is still a lot of uncertainty in their simulation accuracy. Among them, the maximum carboxylation rate (Vcmax) is one of the key parameters that leads to model uncertainty. Therefore, accurately estimating the maximum carboxylation rate of different tree species is a core step to improve the accuracy and fineness of GPP simulation. In recent years, researchers have estimated the spatiotemporal variation of Vcmax based on leaf nitrogen content, chlorophyll content, solar-induced fluorescence (SIF), and ecological optimization theory. However, the above methods still cannot achieve fine characterization of the spatiotemporal pattern of Vcmax and corresponding GPP at the species scale.

[0004] Hyperspectral remote sensing technology has the potential to identify different tree species due to its high spectral and spatial resolution. Vcmax is closely related to multiple leaf traits such as leaf nitrogen content, chlorophyll content, leaf water content, and leaf age, and hyperspectral reflectance can comprehensively reflect these trait information. Therefore, hyperspectral remote sensing provides a new technical path for directly and finely estimating Vcmax of different species. With the continuous development of remote sensing platforms, vegetation hyperspectral information can be efficiently obtained through airborne or spaceborne sensors. In particular, unmanned aerial vehicle (UAV) hyperspectral remote sensing has the advantages of flexible operation, fast imaging, strong timeliness, and high spatial resolution, and shows broad application prospects in the fine simulation of carbon cycle at the species scale of terrestrial ecosystems. Based on hyperspectral remote sensing, the inversion of Vcmax has become a new research direction to improve the simulation accuracy of GPP at the species level.

[0005] However, the existing technology still faces the following key problems in achieving fine simulation of GPP at the tree species scale: (1) Model parameter settings are rough, and interspecific differences are ignored: In existing ecological process models, Vcmax is usually set as an empirical constant related to vegetation functional types (e.g. "subtropical evergreen coniferous forest"). However, there are significant differences in physiological characteristics and photosynthetic capacity among different tree species (such as Pinus massoniana, Pinus elliottii, Cunninghamia lanceolata, etc.) under the same functional type, and Vcmax itself has the characteristics of dynamic change with season, environmental stress and leaf age replacement. The use of static and unified Vcmax parameters cannot accurately reflect the heterogeneity among tree species and over time, resulting in large uncertainty and systematic bias in GPP simulation results.

[0006] (2) The estimation method of Vcmax is indirect, and the mechanism is weakly related: In order to improve the estimation of Vcmax, some studies attempt to establish statistical relationships between Vcmax and leaf nitrogen content, chlorophyll content and other factors. However, such relationships lack direct physiological mechanism support, and have poor stability and generalization ability under different tree species and environmental conditions. In addition, large-scale and high-precision acquisition of leaf biochemical parameters also faces technical challenges.

[0007] (3) The spatio-temporal scale is not matched, which restricts the fine application: Although the traditional Vcmax measurement method (such as A / Ci curve method) has high accuracy, it is complicated and costly to operate, and is only suitable for sample point scale, which is difficult to promote to regional application. The conventional satellite remote sensing data is limited by spatial resolution, which is difficult to effectively identify different tree species within the canopy, and cannot meet the demand of fine simulation of GPP at species level.

[0008] Therefore, how to provide a tree species gross primary productivity fine simulation method and system to improve the accuracy and practicality of tree species gross primary productivity simulation has become a technical problem to be solved. SUMMARY

[0009] The technical problem to be solved by the present application is to provide a tree species gross primary productivity fine simulation method and system to improve the accuracy and practicality of tree species gross primary productivity simulation.

[0010] In a first aspect, the present application provides a tree species gross primary productivity fine simulation method, comprising the following steps: Step S1, collecting leaf hyperspectral reflectance data, photosynthetic physiological parameter data and canopy hyperspectral remote sensing image data of different tree species in a target area; Step S2, based on the maximum carboxylation rate in the leaf hyperspectral reflectance data and the photosynthetic physiological parameter data, extracting feature bands to construct a feature band combination, and constructing a Vcmax hyperspectral inversion model for distinguishing tree species based on the feature band combination; Step S3: Use the canopy hyperspectral remote sensing image data to classify tree species, and combine the Vcmax hyperspectral inversion model to generate a high spatial resolution dynamic Vcmax distribution map; Step S4: Replace the static Vcmax parameters of the ecological process model with the dynamic Vcmax distribution map, obtain driving data, and input the driving data into the updated ecological process model to simulate the high spatiotemporal resolution GPP spatiotemporal pattern of the corresponding tree species.

[0011] Furthermore, step S1 specifically includes: Using an ASD spectrometer calibrated with a whiteboard, and with the spectrometer's field of view vertically downward at a 10° angle, hyperspectral reflectance data of leaves from different tree species within the target area were collected. During the collection of the leaf hyperspectral reflectance data, the leaf was measured three times at different locations and the average value was taken as the final leaf hyperspectral reflectance data. Within the target area, photosynthetic physiological parameters data for different tree species, including at least the maximum carboxylation rate, chlorophyll content, and nitrogen content, were collected. The estimation process for the maximum carboxylation rate was as follows: leaf photosynthetic parameters, including net photosynthetic rate and intercellular CO2 concentration, were measured using a LI-COR 6400 portable photosynthetic analyzer, and the maximum carboxylation rate was estimated using an A / Ci curve fitting tool. Using a full-band hyperspectral imager mounted on a drone, hyperspectral remote sensing image data of canopies of different tree species are collected within the target area, and the canopy hyperspectral remote sensing image data is converted into leaf-scale reflectance images.

[0012] Furthermore, step S2 specifically includes: Feature bands are extracted from the leaf hyperspectral reflectance data using a continuous projection algorithm or a partial least squares method for variable projection importance. Feature band combinations are then constructed by selecting the feature bands most correlated with the maximum carboxylation rate in the photosynthetic physiological parameter data. A Vcmax hyperspectral inversion model for distinguishing tree species is constructed based on a machine learning regression algorithm. The constructed Vcmax hyperspectral inversion model is trained by combining the feature bands. The machine learning regression algorithm adopts random forest, support vector machine regression or BP neural network.

[0013] Furthermore, step S3 specifically includes: The leaf hyperspectral reflectance data is transformed by first-order derivative and logarithmic derivative operations to highlight the differences in reflectance and absorption spectral characteristics of different tree species. Principal component analysis is then used to reduce the dimensionality of the transformed leaf hyperspectral reflectance data to obtain the optimal feature band set for identifying different tree species. Based on the optimal feature band set and canopy hyperspectral remote sensing image data, a machine learning classification method is used to identify and classify tree species, generating a high-resolution spatial distribution map of tree species. Combining the spatial distribution map of the tree species and the leaf-scale reflectance image converted from the canopy hyperspectral remote sensing image data, for each pixel in the canopy hyperspectral remote sensing image data, according to the tree species, the maximum carboxylation rate of each pixel in the Vcmax hyperspectral inversion model is called; traversing all pixels, a dynamic Vcmax distribution map with high spatial resolution covering the target area is finally generated.

[0014] Furthermore, step S4 specifically includes: The static Vcmax parameter of the ecological process model is replaced based on the dynamic Vcmax distribution map, and the ecological process model adopts the BEPS model. Acquire driving data including meteorological data and leaf area index data, and input the driving data into the updated ecological process model to simulate the high spatiotemporal resolution GPP spatiotemporal pattern of the corresponding tree species; the meteorological data includes at least temperature, precipitation, relative humidity and radiation intensity.

[0015] Secondly, the present invention provides a fine simulation system for total primary productivity of tree species, comprising the following modules: The data acquisition module is used to collect hyperspectral reflectance data of leaves, photosynthetic physiological parameter data, and canopy hyperspectral remote sensing image data of different tree species within the target area. The Vcmax hyperspectral inversion model construction module is used to extract feature bands based on the maximum carboxylation rate in the leaf hyperspectral reflectance data and photosynthetic physiological parameter data to construct a feature band combination, and to construct a Vcmax hyperspectral inversion model for distinguishing tree species based on the feature band combination. The dynamic Vcmax distribution map generation module is used to classify tree species using the canopy hyperspectral remote sensing image data and generate a high spatial resolution dynamic Vcmax distribution map by combining the Vcmax hyperspectral inversion model. The GPP spatiotemporal pattern simulation module is used to replace the static Vcmax parameter of the ecological process model based on the dynamic Vcmax distribution map, obtain driving data, and input the driving data into the updated ecological process model to simulate the high spatiotemporal resolution GPP spatiotemporal pattern of the corresponding tree species.

[0016] Furthermore, the data acquisition module is specifically used for: Using an ASD spectrometer calibrated with a whiteboard, and with the spectrometer's field of view vertically downward at a 10° angle, hyperspectral reflectance data of leaves from different tree species within the target area were collected. During the collection of the leaf hyperspectral reflectance data, the leaf was measured three times at different locations and the average value was taken as the final leaf hyperspectral reflectance data. Within the target area, photosynthetic physiological parameters data for different tree species, including at least the maximum carboxylation rate, chlorophyll content, and nitrogen content, were collected. The estimation process for the maximum carboxylation rate was as follows: leaf photosynthetic parameters, including net photosynthetic rate and intercellular CO2 concentration, were measured using a LI-COR 6400 portable photosynthetic analyzer, and the maximum carboxylation rate was estimated using an A / Ci curve fitting tool. Using a full-band hyperspectral imager mounted on a drone, hyperspectral remote sensing image data of canopies of different tree species are collected within the target area, and the canopy hyperspectral remote sensing image data is converted into leaf-scale reflectance images.

[0017] Furthermore, the Vcmax hyperspectral inversion model construction module is specifically used for: Feature bands are extracted from the leaf hyperspectral reflectance data using a continuous projection algorithm or a partial least squares method for variable projection importance. Feature band combinations are then constructed by selecting the feature bands most correlated with the maximum carboxylation rate in the photosynthetic physiological parameter data. A Vcmax hyperspectral inversion model for distinguishing tree species is constructed based on a machine learning regression algorithm. The constructed Vcmax hyperspectral inversion model is trained by combining the feature bands. The machine learning regression algorithm adopts random forest, support vector machine regression or BP neural network.

[0018] Furthermore, the dynamic Vcmax distribution map generation module is specifically used for: The leaf hyperspectral reflectance data is transformed by first-order derivative and logarithmic derivative operations to highlight the differences in reflectance and absorption spectral characteristics of different tree species. Principal component analysis is then used to reduce the dimensionality of the transformed leaf hyperspectral reflectance data to obtain the optimal feature band set for identifying different tree species. Based on the optimal feature band set and canopy hyperspectral remote sensing image data, a machine learning classification method is used to identify and classify tree species, generating a high-resolution spatial distribution map of tree species. Combining the spatial distribution map of the tree species and the leaf-scale reflectance image converted from the canopy hyperspectral remote sensing image data, for each pixel in the canopy hyperspectral remote sensing image data, according to the tree species, the maximum carboxylation rate of each pixel in the Vcmax hyperspectral inversion model is called; traversing all pixels, a dynamic Vcmax distribution map with high spatial resolution covering the target area is finally generated.

[0019] Furthermore, the GPP spatiotemporal pattern simulation module is specifically used for: The static Vcmax parameter of the ecological process model is replaced based on the dynamic Vcmax distribution map, and the ecological process model adopts the BEPS model. Acquire driving data including meteorological data and leaf area index data, and input the driving data into the updated ecological process model to simulate the high spatiotemporal resolution GPP spatiotemporal pattern of the corresponding tree species; the meteorological data includes at least temperature, precipitation, relative humidity and radiation intensity.

[0020] The advantages of this invention are: 1. By collecting leaf hyperspectral reflectance data, photosynthetic physiological parameter data, and canopy hyperspectral remote sensing image data of different tree species within the target area; based on the maximum carboxylation rate in the leaf hyperspectral reflectance data and photosynthetic physiological parameter data, characteristic bands are extracted to construct characteristic band combinations, and a Vcmax hyperspectral inversion model for distinguishing tree species is constructed based on the characteristic band combinations; then, tree species are classified using canopy hyperspectral remote sensing image data, and a high spatial resolution dynamic Vcmax distribution map is generated in conjunction with the Vcmax hyperspectral inversion model; finally, the static Vcmax parameters of the ecological process model are replaced based on the dynamic Vcmax distribution map, and the driving data is input into the updated ecological process model. This method simulates the spatiotemporal patterns of GPP (Gross Primary Productivity) for corresponding tree species with high spatiotemporal resolution. Specifically, by collecting hyperspectral data of leaves and canopy, a hyperspectral inversion model of Vcmax (vitamin Cmax) capable of distinguishing tree species is directly constructed using machine learning algorithms, generating a dynamic Vcmax distribution map with high spatial resolution. This replaces the original static and uniform Vcmax parameter in the ecological process model, thereby refining the key physiological parameters of the model from a coarse vegetation functional scale to a specific tree species scale, and realizing the spatiotemporal dynamics of the parameters. This technical approach bypasses the instability of traditional indirect estimation methods and solves the scale mismatch problem by using UAV hyperspectral remote sensing, ultimately greatly improving the accuracy and practicality of the simulation of total primary productivity of tree species.

[0021] 2. By constructing a dynamic Vcmax (maximum carboxylation rate) distribution map to replace the static Vcmax parameter in traditional ecological process models, the accuracy of GPP simulation is significantly improved. In traditional methods, Vcmax is usually given a fixed value or a rough estimate, which cannot reflect the differences in tree species and spatiotemporal variations. However, this invention retrieves Vcmax based on hyperspectral data and generates a dynamic distribution map by combining tree species classification. This allows the model to capture the influence of environmental factors (such as light and temperature) and tree species physiological characteristics in real time. This dynamic parameterization method reduces simulation errors and is especially suitable for forest areas with high heterogeneity, providing more reliable data support for carbon cycle research and climate change assessment.

[0022] 3. By organically combining hyperspectral remote sensing data, photosynthetic physiological parameters, and machine learning algorithms, a multi-level model system (such as the Vcmax inversion model and tree species classification model) was constructed, demonstrating significant technological innovation. Specifically, by extracting feature bands (such as using continuous projection algorithms or partial least squares methods) and utilizing machine learning regression algorithms such as random forests, efficient Vcmax inversion was achieved. At the same time, the hyperspectral data was transformed (such as by first-order derivative calculation) and dimensionality reduced, enhancing the distinguishability of tree species identification. This data fusion method breaks through the limitations of traditional remote sensing models, solves the problem of integrating multi-source heterogeneous data, and provides a new paradigm for ecological modeling.

[0023] 4. Data acquisition is carried out using a drone equipped with a hyperspectral imager, combined with automated processing procedures (such as machine learning classification and inversion model training), which greatly improves the efficiency of data acquisition and processing. Compared with traditional ground measurement, drone remote sensing can quickly cover a large area, reducing manpower costs and time consumption. The generation of Vcmax distribution maps and GPP simulation based on pixels can be batch processed and easily extended to different geographical scales or tree species. This high efficiency makes this invention suitable for long-term monitoring or large-scale ecological research, and has good engineering application prospects.

[0024] 5. Standardized procedures (such as whiteboard calibration and averaging repeated measurements) ensure data quality and compatibility with various ecological process models (such as the BEPS model), enhancing its practicality. Flexible input of driving data (such as meteorological data and leaf area index) allows the scheme to be adapted to different regions and environmental conditions. The design of tree species classification and Vcmax inversion model takes into account the characteristics of diverse tree species and avoids model overfitting. This versatility makes it have broad application potential in forestry management, carbon sink assessment and biodiversity conservation, meeting the requirements of industrial practicality.

[0025] 6. The automation of the entire chain from data collection to GPP simulation reduces the uncertainty caused by subjective intervention. For example, the spatial distribution map of tree species is automatically generated by machine learning classification, and the parameters of each pixel are dynamically calculated by combining the Vcmax inversion model, which improves the reliability and consistency of the results. This highly integrated and systematic design not only lowers the operation threshold, but also leaves an interface for subsequent technical improvements (such as the introduction of deep learning algorithms).

[0026] 7. Through the innovative integration of hyperspectral remote sensing and machine learning, a leap has been achieved in the simulation of Gross Primary Productivity (GPP) from static and extensive to dynamic and refined: On the one hand, dynamic Vcmax parameters are inverted based on tree species-specific spectral characteristics, replacing the fixed values ​​in traditional ecological models, which significantly improves the accuracy and adaptability of carbon cycle simulation; on the other hand, with the help of UAV remote sensing and automated classification inversion technology, large-scale, high spatiotemporal resolution GPP pattern generation has been achieved, which greatly improves monitoring efficiency while ensuring scientific rigor, and provides a scalable and reusable technical paradigm for ecosystem management and climate change research.

[0027] 8. Significantly improves the accuracy of GPP simulation: The use of a species-differentiated dynamic Vcmax based on hyperspectral remote sensing inversion replaces the coarse, static empirical constants in traditional models, making the input parameters closer to the actual physiological state of vegetation. This fundamentally improves the precision, accuracy, and reliability of GPP simulation; Enables refined assessment at the tree species level: By combining high-resolution remote sensing imagery for tree species classification, it is possible to precisely simulate the contribution of different tree species in mixed forests to the total carbon sink of the ecosystem, providing unprecedented quantitative scientific evidence for refined forest resource management, selection of dominant tree species, and formulation of regional carbon sequestration and enhancement strategies; Has a stronger mechanistic basis: Based on the intrinsic physical connection between leaf spectral characteristics and vegetation photosynthetic physiological functions, it is more direct and robust than traditional methods that rely on indirect biochemical parameters, providing an innovative technical paradigm for ecosystem carbon cycle research. Attached Figure Description

[0028] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0029] Fig. 1 This is a flowchart of a method for fine simulation of total primary productivity of tree species according to the present invention.

[0030] Fig. 2 This is a schematic diagram of the structure of a fine simulation system for total primary productivity of tree species according to the present invention. Detailed Implementation

[0031] The overall approach of the technical solution in this application is as follows: By collecting hyperspectral data of leaves and canopy, a hyperspectral inversion model of Vcmax that can distinguish tree species is directly constructed using machine learning algorithms, and a dynamic Vcmax distribution map with high spatial resolution is generated. This replaces the original static and uniform Vcmax parameter in the ecological process model, thereby refining the key physiological parameters of the model from a coarse vegetation functional scale to a specific tree species scale, and realizing the spatiotemporal dynamics of the parameters. This technical approach bypasses the instability of traditional indirect estimation methods and uses UAV hyperspectral remote sensing to solve the scale mismatch problem, thereby improving the accuracy and practicality of the simulation of total primary productivity of tree species.

[0032] Please refer to Figs. 1-2 As shown, a preferred embodiment of the method for fine simulation of total primary productivity of tree species according to the present invention includes the following steps: Step S1: Collect leaf hyperspectral reflectance data, photosynthetic physiological parameter data, and canopy hyperspectral remote sensing image data of different tree species within the target area; Step S2: Based on the maximum carboxylation rate in the leaf hyperspectral reflectance data and photosynthetic physiological parameter data, feature bands are extracted to construct a feature band combination, and a Vcmax hyperspectral inversion model for distinguishing tree species is constructed based on the feature band combination. Step S3: Use the canopy hyperspectral remote sensing image data to classify tree species, and combine the Vcmax hyperspectral inversion model to generate a high spatial resolution dynamic Vcmax distribution map; Step S4: Replace the static Vcmax parameters of the ecological process model with the dynamic Vcmax distribution map, obtain driving data, and input the driving data into the updated ecological process model to simulate the high spatiotemporal resolution GPP spatiotemporal pattern of the corresponding tree species.

[0033] The core idea of ​​this invention is to break through the limitations of the static Vcmax parameter in traditional ecological process models and establish an innovative technical path from "hyperspectral reflectance" to "key physiological parameter (Vcmax)" and then to "ecosystem carbon flux (GPP)". Specifically, firstly, based on field measurement data and machine learning methods, a machine learning model is constructed to connect the hyperspectral characteristics of leaves of different tree species (such as Masson pine, slash pine, and Chinese fir) with their core photosynthetic capacity parameter Vcmax. Then, canopy hyperspectral remote sensing image data of the target area is acquired using a full-band hyperspectral imager mounted on an unmanned aerial vehicle (UAV). This canopy hyperspectral remote sensing image data is converted into leaf-scale reflectance images using a radiative transfer model, and the images are then finely classified at the tree species level using machine learning algorithms. Finally, the classification results are combined with the Vcmax inversion model to generate a high spatiotemporal resolution dynamic Vcmax distribution map that distinguishes tree species within the study area. This map serves as key input data to drive the ecological process model, thereby achieving accurate, dynamic, and refined simulation of the spatiotemporal patterns of GPP for different tree species.

[0034] Step S1 specifically involves: Using an ASD spectrometer (spectral measurement range of 350nm-2500nm, spectral resolution of 8nm) calibrated with a whiteboard, hyperspectral reflectance data of leaves from different tree species within the target area were collected based on a 10° field of view angle vertically downwards (the probe was vertically downwards during measurement, and the vertical distance for sample acquisition was determined by the observation position falling within the field of view). During the acquisition of the leaf hyperspectral reflectance data, measurements were repeated three times at different locations on the leaf, and the average value was taken as the final leaf hyperspectral reflectance data to reduce observation errors. Within the target area, photosynthetic physiological parameters data for different tree species, including at least maximum carboxylation rate, chlorophyll content, and nitrogen content, were collected. The estimation process for the maximum carboxylation rate was as follows: leaf photosynthetic parameters, including net photosynthetic rate and intercellular CO2 concentration, were measured using a LI-COR 6400 portable photosynthesis system, and the maximum carboxylation rate was estimated using an A / Ci curve fitting tool; A represents net photosynthetic rate, and Ci represents intercellular CO2 concentration; under saturated light conditions, the photosynthetic photon flux density (PPFD) was set to 1500 µmol m⁻² s⁻¹, and each CO2 response curve measurement lasted approximately 40 minutes.

[0035] Using a full-band hyperspectral imager mounted on a drone (such as the HG-HyerUAV2500, with a spectral range of 400-2500nm, spectral resolution ≤7.0nm, and 480 spectral channels), hyperspectral remote sensing image data of canopies for different tree species within the target area are acquired. This canopy hyperspectral remote sensing image data is then converted into leaf-scale reflectance images using a radiative transfer model (4-scale). This process eliminates the influence of factors such as soil background, canopy structure, and illumination geometry. In practice, the data acquired by the full-band hyperspectral imager undergoes preprocessing such as format conversion and image stitching to obtain hyperspectral orthophotos of the flux contribution area (canopy hyperspectral remote sensing image data).

[0036] Step S2 specifically involves: Feature bands are extracted from the leaf hyperspectral reflectance data using the continuous projection algorithm (SPA) or the variable projection importance method (VIP) of partial least squares (PLSR), and feature band combinations are constructed by selecting the feature bands that are most correlated with the maximum carboxylation rate in the photosynthetic physiological parameter data (and have the lowest redundancy). A Vcmax hyperspectral inversion model for distinguishing tree species is constructed based on a machine learning regression algorithm. The constructed Vcmax hyperspectral inversion model is trained by combining the feature bands. The machine learning regression algorithm adopts random forest, support vector machine regression or BP neural network.

[0037] Step S3 specifically involves: The leaf hyperspectral reflectance data is transformed by first-order derivative and logarithmic derivative operations to highlight the differences in reflectance and absorption spectral characteristics of different tree species. Principal component analysis is then used to reduce the dimensionality of the transformed leaf hyperspectral reflectance data to obtain the optimal feature band set for identifying different tree species. Based on the optimal feature band set and canopy hyperspectral remote sensing image data, a machine learning classification method (random forest, support vector machine or convolutional neural network) is used to identify and classify tree species, generating a high-resolution spatial distribution map of tree species. Combining the spatial distribution map of the tree species and the leaf-scale reflectance image converted from the canopy hyperspectral remote sensing image data, for each pixel in the canopy hyperspectral remote sensing image data, according to the tree species, the maximum carboxylation rate of each pixel in the Vcmax hyperspectral inversion model is called; traversing all pixels, a dynamic Vcmax distribution map with high spatial resolution covering the target area is finally generated.

[0038] Step S4 specifically involves: The static Vcmax parameter of the ecological process model is replaced based on the dynamic Vcmax distribution map, and the ecological process model adopts the BEPS model. Acquire driving data including meteorological data and leaf area index (LAI) data, and input the driving data into an updated ecological process model to simulate the high spatiotemporal resolution GPP spatiotemporal pattern of the corresponding tree species. The meteorological data includes at least temperature, precipitation, relative humidity, and radiation intensity. That is, simulate and output the GPP value of each pixel within the target area, ultimately obtaining a high spatiotemporal resolution GPP spatiotemporal pattern map distinguishing tree species.

[0039] A preferred embodiment of the present invention provides a fine simulation system for total primary productivity of tree species, comprising the following modules: The data acquisition module is used to collect hyperspectral reflectance data of leaves, photosynthetic physiological parameter data, and canopy hyperspectral remote sensing image data of different tree species within the target area. The Vcmax hyperspectral inversion model construction module is used to extract feature bands based on the maximum carboxylation rate in the leaf hyperspectral reflectance data and photosynthetic physiological parameter data to construct a feature band combination, and to construct a Vcmax hyperspectral inversion model for distinguishing tree species based on the feature band combination. The dynamic Vcmax distribution map generation module is used to classify tree species using the canopy hyperspectral remote sensing image data and generate a high spatial resolution dynamic Vcmax distribution map by combining the Vcmax hyperspectral inversion model. The GPP spatiotemporal pattern simulation module is used to replace the static Vcmax parameter of the ecological process model based on the dynamic Vcmax distribution map, obtain driving data, and input the driving data into the updated ecological process model to simulate the high spatiotemporal resolution GPP spatiotemporal pattern of the corresponding tree species.

[0040] The core idea of ​​this invention is to break through the limitations of the static Vcmax parameter in traditional ecological process models and establish an innovative technical path from "hyperspectral reflectance" to "key physiological parameter (Vcmax)" and then to "ecosystem carbon flux (GPP)". Specifically, firstly, based on field measurement data and machine learning methods, a machine learning model is constructed to connect the hyperspectral characteristics of leaves of different tree species (such as Masson pine, slash pine, and Chinese fir) with their core photosynthetic capacity parameter Vcmax. Then, canopy hyperspectral remote sensing image data of the target area is acquired using a full-band hyperspectral imager mounted on an unmanned aerial vehicle (UAV). This canopy hyperspectral remote sensing image data is converted into leaf-scale reflectance images using a radiative transfer model, and the images are then finely classified at the tree species level using machine learning algorithms. Finally, the classification results are combined with the Vcmax inversion model to generate a high spatiotemporal resolution dynamic Vcmax distribution map that distinguishes tree species within the study area. This map serves as key input data to drive the ecological process model, thereby achieving accurate, dynamic, and refined simulation of the spatiotemporal patterns of GPP for different tree species.

[0041] The data acquisition module is specifically used for: Using an ASD spectrometer (spectral measurement range of 350nm-2500nm, spectral resolution of 8nm) calibrated with a whiteboard, hyperspectral reflectance data of leaves from different tree species within the target area were collected based on a 10° field of view angle vertically downwards (the probe was vertically downwards during measurement, and the vertical distance for sample acquisition was determined by the observation position falling within the field of view). During the acquisition of the leaf hyperspectral reflectance data, measurements were repeated three times at different locations on the leaf, and the average value was taken as the final leaf hyperspectral reflectance data to reduce observation errors. Within the target area, photosynthetic physiological parameters data for different tree species, including at least maximum carboxylation rate, chlorophyll content, and nitrogen content, were collected. The estimation process for the maximum carboxylation rate was as follows: leaf photosynthetic parameters, including net photosynthetic rate and intercellular CO2 concentration, were measured using a LI-COR 6400 portable photosynthesis system, and the maximum carboxylation rate was estimated using an A / Ci curve fitting tool; A represents net photosynthetic rate, and Ci represents intercellular CO2 concentration; under saturated light conditions, the photosynthetic photon flux density (PPFD) was set to 1500 µmol m⁻² s⁻¹, and each CO2 response curve measurement lasted approximately 40 minutes.

[0042] Using a full-band hyperspectral imager mounted on a drone (such as the HG-HyerUAV2500, with a spectral range of 400-2500nm, spectral resolution ≤7.0nm, and 480 spectral channels), hyperspectral remote sensing image data of canopies for different tree species within the target area are acquired. This canopy hyperspectral remote sensing image data is then converted into leaf-scale reflectance images using a radiative transfer model (4-scale). This process eliminates the influence of factors such as soil background, canopy structure, and illumination geometry. In practice, the data acquired by the full-band hyperspectral imager undergoes preprocessing such as format conversion and image stitching to obtain hyperspectral orthophotos of the flux contribution area (canopy hyperspectral remote sensing image data).

[0043] The Vcmax hyperspectral inversion model construction module is specifically used for: Feature bands are extracted from the leaf hyperspectral reflectance data using the continuous projection algorithm (SPA) or the variable projection importance method (VIP) of partial least squares (PLSR), and feature band combinations are constructed by selecting the feature bands that are most correlated with the maximum carboxylation rate in the photosynthetic physiological parameter data (and have the lowest redundancy). A Vcmax hyperspectral inversion model for distinguishing tree species is constructed based on a machine learning regression algorithm. The constructed Vcmax hyperspectral inversion model is trained by combining the feature bands. The machine learning regression algorithm adopts random forest, support vector machine regression or BP neural network.

[0044] The dynamic Vcmax distribution map generation module is specifically used for: The leaf hyperspectral reflectance data is transformed by first-order derivative and logarithmic derivative operations to highlight the differences in reflectance and absorption spectral characteristics of different tree species. Principal component analysis is then used to reduce the dimensionality of the transformed leaf hyperspectral reflectance data to obtain the optimal feature band set for identifying different tree species. Based on the optimal feature band set and canopy hyperspectral remote sensing image data, a machine learning classification method (random forest, support vector machine or convolutional neural network) is used to identify and classify tree species, generating a high-resolution spatial distribution map of tree species. Combining the spatial distribution map of the tree species and the leaf-scale reflectance image converted from the canopy hyperspectral remote sensing image data, for each pixel in the canopy hyperspectral remote sensing image data, according to the tree species, the maximum carboxylation rate of each pixel in the Vcmax hyperspectral inversion model is called; traversing all pixels, a dynamic Vcmax distribution map with high spatial resolution covering the target area is finally generated.

[0045] The GPP spatiotemporal pattern simulation module is specifically used for: The static Vcmax parameter of the ecological process model is replaced based on the dynamic Vcmax distribution map, and the ecological process model adopts the BEPS model. Acquire driving data including meteorological data and leaf area index (LAI) data, and input the driving data into an updated ecological process model to simulate the high spatiotemporal resolution GPP spatiotemporal pattern of the corresponding tree species. The meteorological data includes at least temperature, precipitation, relative humidity, and radiation intensity. That is, simulate and output the GPP value of each pixel within the target area, ultimately obtaining a high spatiotemporal resolution GPP spatiotemporal pattern map distinguishing tree species.

[0046] In summary, the advantages of this invention are as follows: 1. By collecting leaf hyperspectral reflectance data, photosynthetic physiological parameter data, and canopy hyperspectral remote sensing image data of different tree species within the target area; based on the maximum carboxylation rate in the leaf hyperspectral reflectance data and photosynthetic physiological parameter data, characteristic bands are extracted to construct characteristic band combinations, and a Vcmax hyperspectral inversion model for distinguishing tree species is constructed based on the characteristic band combinations; then, tree species are classified using canopy hyperspectral remote sensing image data, and a high spatial resolution dynamic Vcmax distribution map is generated in conjunction with the Vcmax hyperspectral inversion model; finally, the static Vcmax parameters of the ecological process model are replaced based on the dynamic Vcmax distribution map, and the driving data is input into the updated ecological process model. This method simulates the spatiotemporal patterns of GPP (Gross Primary Productivity) for corresponding tree species with high spatiotemporal resolution. Specifically, by collecting hyperspectral data of leaves and canopy, a hyperspectral inversion model of Vcmax (vitamin Cmax) capable of distinguishing tree species is directly constructed using machine learning algorithms, generating a dynamic Vcmax distribution map with high spatial resolution. This replaces the original static and uniform Vcmax parameter in the ecological process model, thereby refining the key physiological parameters of the model from a coarse vegetation functional scale to a specific tree species scale, and realizing the spatiotemporal dynamics of the parameters. This technical approach bypasses the instability of traditional indirect estimation methods and solves the scale mismatch problem by using UAV hyperspectral remote sensing, ultimately greatly improving the accuracy and practicality of the simulation of total primary productivity of tree species.

[0047] 2. By constructing a dynamic Vcmax (maximum carboxylation rate) distribution map to replace the static Vcmax parameter in traditional ecological process models, the accuracy of GPP simulation is significantly improved. In traditional methods, Vcmax is usually given a fixed value or a rough estimate, which cannot reflect the differences in tree species and spatiotemporal variations. However, this invention retrieves Vcmax based on hyperspectral data and generates a dynamic distribution map by combining tree species classification. This allows the model to capture the influence of environmental factors (such as light and temperature) and tree species physiological characteristics in real time. This dynamic parameterization method reduces simulation errors and is especially suitable for forest areas with high heterogeneity, providing more reliable data support for carbon cycle research and climate change assessment.

[0048] 3. By organically combining hyperspectral remote sensing data, photosynthetic physiological parameters, and machine learning algorithms, a multi-level model system (such as the Vcmax inversion model and tree species classification model) was constructed, demonstrating significant technological innovation. Specifically, by extracting feature bands (such as using continuous projection algorithms or partial least squares methods) and utilizing machine learning regression algorithms such as random forests, efficient Vcmax inversion was achieved. At the same time, the hyperspectral data was transformed (such as by first-order derivative calculation) and dimensionality reduced, enhancing the distinguishability of tree species identification. This data fusion method breaks through the limitations of traditional remote sensing models, solves the problem of integrating multi-source heterogeneous data, and provides a new paradigm for ecological modeling.

[0049] 4. Data acquisition is carried out using a drone equipped with a hyperspectral imager, combined with automated processing procedures (such as machine learning classification and inversion model training), which greatly improves the efficiency of data acquisition and processing. Compared with traditional ground measurement, drone remote sensing can quickly cover a large area, reducing manpower costs and time consumption. The generation of Vcmax distribution maps and GPP simulation based on pixels can be batch processed and easily extended to different geographical scales or tree species. This high efficiency makes this invention suitable for long-term monitoring or large-scale ecological research, and has good engineering application prospects.

[0050] 5. Standardized procedures (such as whiteboard calibration and averaging repeated measurements) ensure data quality and compatibility with various ecological process models (such as the BEPS model), enhancing its practicality. Flexible input of driving data (such as meteorological data and leaf area index) allows the scheme to be adapted to different regions and environmental conditions. The design of tree species classification and Vcmax inversion model takes into account the characteristics of diverse tree species and avoids model overfitting. This versatility makes it have broad application potential in forestry management, carbon sink assessment and biodiversity conservation, meeting the requirements of industrial practicality.

[0051] 6. The automation of the entire chain from data collection to GPP simulation reduces the uncertainty caused by subjective intervention. For example, the spatial distribution map of tree species is automatically generated by machine learning classification, and the parameters of each pixel are dynamically calculated by combining the Vcmax inversion model, which improves the reliability and consistency of the results. This highly integrated and systematic design not only lowers the operation threshold, but also leaves an interface for subsequent technical improvements (such as the introduction of deep learning algorithms).

[0052] 7. Through the innovative integration of hyperspectral remote sensing and machine learning, a leap has been achieved in the simulation of Gross Primary Productivity (GPP) from static and extensive to dynamic and refined: On the one hand, dynamic Vcmax parameters are inverted based on tree species-specific spectral characteristics, replacing the fixed values ​​in traditional ecological models, which significantly improves the accuracy and adaptability of carbon cycle simulation; on the other hand, with the help of UAV remote sensing and automated classification inversion technology, large-scale, high spatiotemporal resolution GPP pattern generation has been achieved, which greatly improves monitoring efficiency while ensuring scientific rigor, and provides a scalable and reusable technical paradigm for ecosystem management and climate change research.

[0053] 8. Significantly improves the accuracy of GPP simulation: The use of a species-differentiated dynamic Vcmax based on hyperspectral remote sensing inversion replaces the coarse, static empirical constants in traditional models, making the input parameters closer to the actual physiological state of vegetation. This fundamentally improves the precision, accuracy, and reliability of GPP simulation; Enables refined assessment at the tree species level: By combining high-resolution remote sensing imagery for tree species classification, it is possible to precisely simulate the contribution of different tree species in mixed forests to the total carbon sink of the ecosystem, providing unprecedented quantitative scientific evidence for refined forest resource management, selection of dominant tree species, and formulation of regional carbon sequestration and enhancement strategies; Has a stronger mechanistic basis: Based on the intrinsic physical connection between leaf spectral characteristics and vegetation photosynthetic physiological functions, it is more direct and robust than traditional methods that rely on indirect biochemical parameters, providing an innovative technical paradigm for ecosystem carbon cycle research.

[0054] While specific embodiments of the present invention have been described above, those skilled in the art should understand that the specific embodiments described are merely illustrative and not intended to limit the scope of the present invention. Equivalent modifications and variations made by those skilled in the art in accordance with the spirit of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for fine simulation of total primary productivity of a tree species, characterized in that: Includes the following steps: Step S1: Collect leaf hyperspectral reflectance data, photosynthetic physiological parameter data, and canopy hyperspectral remote sensing image data of different tree species within the target area; Step S2: Based on the maximum carboxylation rate in the leaf hyperspectral reflectance data and photosynthetic physiological parameter data, feature bands are extracted to construct a feature band combination, and a Vcmax hyperspectral inversion model for distinguishing tree species is constructed based on the feature band combination. Step S3: Use the canopy hyperspectral remote sensing image data to classify tree species, and combine the Vcmax hyperspectral inversion model to generate a high spatial resolution dynamic Vcmax distribution map; Step S4: Replace the static Vcmax parameters of the ecological process model with the dynamic Vcmax distribution map, obtain driving data, and input the driving data into the updated ecological process model to simulate the high spatiotemporal resolution GPP spatiotemporal pattern of the corresponding tree species.

2. The method for fine simulation of total primary productivity of a tree species as described in claim 1, characterized in that: Step S1 specifically involves: Using an ASD spectrometer calibrated with a whiteboard, and with the spectrometer's field of view vertically downward at a 10° angle, hyperspectral reflectance data of leaves from different tree species within the target area were collected. During the collection of the leaf hyperspectral reflectance data, the leaf was measured three times at different locations and the average value was taken as the final leaf hyperspectral reflectance data. Within the target area, photosynthetic physiological parameters data for different tree species, including at least the maximum carboxylation rate, chlorophyll content, and nitrogen content, were collected. The estimation process for the maximum carboxylation rate was as follows: leaf photosynthetic parameters, including net photosynthetic rate and intercellular CO2 concentration, were measured using a LI-COR 6400 portable photosynthetic analyzer, and the maximum carboxylation rate was estimated using an A / Ci curve fitting tool. Using a full-band hyperspectral imager mounted on a drone, hyperspectral remote sensing image data of canopies of different tree species are collected within the target area, and the canopy hyperspectral remote sensing image data is converted into leaf-scale reflectance images.

3. The method for fine simulation of total primary productivity of a tree species as described in claim 1, characterized in that: Step S2 specifically involves: Feature bands are extracted from the leaf hyperspectral reflectance data using a continuous projection algorithm or a partial least squares method for variable projection importance. Feature band combinations are then constructed by selecting the feature bands most correlated with the maximum carboxylation rate in the photosynthetic physiological parameter data. A Vcmax hyperspectral inversion model for distinguishing tree species is constructed based on a machine learning regression algorithm. The constructed Vcmax hyperspectral inversion model is trained by combining the feature bands. The machine learning regression algorithm adopts random forest, support vector machine regression or BP neural network.

4. The method for fine simulation of total primary productivity of a tree species as described in claim 1, characterized in that: Step S3 specifically involves: The leaf hyperspectral reflectance data is transformed by first-order derivative and logarithmic derivative operations to highlight the differences in reflectance and absorption spectral characteristics of different tree species. Principal component analysis is then used to reduce the dimensionality of the transformed leaf hyperspectral reflectance data to obtain the optimal feature band set for identifying different tree species. Based on the optimal feature band set and canopy hyperspectral remote sensing image data, a machine learning classification method is used to identify and classify tree species, generating a high-resolution spatial distribution map of tree species. Combining the spatial distribution map of the tree species and the leaf-scale reflectance image converted from the canopy hyperspectral remote sensing image data, for each pixel in the canopy hyperspectral remote sensing image data, according to the tree species, the maximum carboxylation rate of each pixel in the Vcmax hyperspectral inversion model is called; traversing all pixels, a dynamic Vcmax distribution map with high spatial resolution covering the target area is finally generated.

5. The method for fine simulation of total primary productivity of a tree species as described in claim 1, characterized in that: Step S4 specifically involves: The static Vcmax parameter of the ecological process model is replaced based on the dynamic Vcmax distribution map, and the ecological process model adopts the BEPS model. Acquire driving data including meteorological data and leaf area index data, and input the driving data into the updated ecological process model to simulate the high spatiotemporal resolution GPP spatiotemporal pattern of the corresponding tree species; the meteorological data includes at least temperature, precipitation, relative humidity and radiation intensity.

6. A refined simulation system for total primary productivity of a tree species, characterized in that: Includes the following modules: The data acquisition module is used to collect hyperspectral reflectance data of leaves, photosynthetic physiological parameter data, and canopy hyperspectral remote sensing image data of different tree species within the target area. The Vcmax hyperspectral inversion model construction module is used to extract feature bands based on the maximum carboxylation rate in the leaf hyperspectral reflectance data and photosynthetic physiological parameter data to construct a feature band combination, and to construct a Vcmax hyperspectral inversion model for distinguishing tree species based on the feature band combination. The dynamic Vcmax distribution map generation module is used to classify tree species using the canopy hyperspectral remote sensing image data and generate a high spatial resolution dynamic Vcmax distribution map by combining the Vcmax hyperspectral inversion model. The GPP spatiotemporal pattern simulation module is used to replace the static Vcmax parameter of the ecological process model based on the dynamic Vcmax distribution map, obtain driving data, and input the driving data into the updated ecological process model to simulate the high spatiotemporal resolution GPP spatiotemporal pattern of the corresponding tree species.

7. The refined simulation system for total primary productivity of a tree species as described in claim 6, characterized in that: The data acquisition module is specifically used for: Using an ASD spectrometer calibrated with a whiteboard, and with the spectrometer's field of view vertically downward at a 10° angle, hyperspectral reflectance data of leaves from different tree species within the target area were collected. During the collection of the leaf hyperspectral reflectance data, the leaf was measured three times at different locations and the average value was taken as the final leaf hyperspectral reflectance data. Within the target area, photosynthetic physiological parameters data for different tree species, including at least the maximum carboxylation rate, chlorophyll content, and nitrogen content, were collected. The estimation process for the maximum carboxylation rate was as follows: leaf photosynthetic parameters, including net photosynthetic rate and intercellular CO2 concentration, were measured using a LI-COR 6400 portable photosynthetic analyzer, and the maximum carboxylation rate was estimated using an A / Ci curve fitting tool. Using a full-band hyperspectral imager mounted on a drone, hyperspectral remote sensing image data of canopies of different tree species are collected within the target area, and the canopy hyperspectral remote sensing image data is converted into leaf-scale reflectance images.

8. The refined simulation system for total primary productivity of a tree species as described in claim 6, characterized in that: The Vcmax hyperspectral inversion model construction module is specifically used for: Feature bands are extracted from the leaf hyperspectral reflectance data using a continuous projection algorithm or a partial least squares method for variable projection importance. Feature band combinations are then constructed by selecting the feature bands most correlated with the maximum carboxylation rate in the photosynthetic physiological parameter data. A Vcmax hyperspectral inversion model for distinguishing tree species is constructed based on a machine learning regression algorithm. The constructed Vcmax hyperspectral inversion model is trained by combining the feature bands. The machine learning regression algorithm adopts random forest, support vector machine regression or BP neural network.

9. The refined simulation system for total primary productivity of a tree species as described in claim 6, characterized in that: The dynamic Vcmax distribution map generation module is specifically used for: The leaf hyperspectral reflectance data is transformed by first-order derivative and logarithmic derivative operations to highlight the differences in reflectance and absorption spectral characteristics of different tree species. Principal component analysis is then used to reduce the dimensionality of the transformed leaf hyperspectral reflectance data to obtain the optimal feature band set for identifying different tree species. Based on the optimal feature band set and canopy hyperspectral remote sensing image data, a machine learning classification method is used to identify and classify tree species, generating a high-resolution spatial distribution map of tree species. Combining the spatial distribution map of the tree species and the leaf-scale reflectance image converted from the canopy hyperspectral remote sensing image data, for each pixel in the canopy hyperspectral remote sensing image data, according to the tree species, the maximum carboxylation rate of each pixel in the Vcmax hyperspectral inversion model is called; traversing all pixels, a dynamic Vcmax distribution map with high spatial resolution covering the target area is finally generated.

10. A refined simulation system for total primary productivity of a tree species as described in claim 6, characterized in that: The GPP spatiotemporal pattern simulation module is specifically used for: The static Vcmax parameter of the ecological process model is replaced based on the dynamic Vcmax distribution map, and the ecological process model adopts the BEPS model. Acquire driving data including meteorological data and leaf area index data, and input the driving data into the updated ecological process model to simulate the high spatiotemporal resolution GPP spatiotemporal pattern of the corresponding tree species; the meteorological data includes at least temperature, precipitation, relative humidity and radiation intensity.

Images