A remote sensing diagnosis method for global vegetation canopy heat regulation capacity

Abstract

The application discloses a remote sensing diagnosis method for global vegetation canopy thermal regulation capacity, and relates to the technical field of ecological remote sensing. The method acquires flux site canopy temperature observation data, site environment driving data, global grid environment driving data, grid air temperature data, vegetation coverage data and snow cover data; a canopy temperature training sample set is constructed, a regional balanced weighted canopy temperature inversion model is established; global canopy temperature spatial distribution data is obtained, canopy thermal regulation slopes of each effective vegetation pixel are calculated, and global vegetation canopy thermal regulation capacity spatial distribution results are generated. The application can solve the problem that discrete flux site canopy temperature observation is difficult to expand into global continuous canopy thermal regulation capacity diagnosis results, and realizes continuous quantitative evaluation of global scale vegetation canopy thermal regulation capacity.

Description

Technical Field

[0001] This invention relates to the field of ecological remote sensing technology, and in particular to a remote sensing diagnostic method for the global vegetation canopy thermal regulation capacity. Background Technology

[0002] Vegetation canopy temperature is an important physical quantity characterizing the hydrothermal state of vegetation, transpiration cooling capacity, canopy energy balance, and the ecosystem's response to climate stress. Compared to air temperature or surface temperature alone, canopy temperature can more directly reflect the heat exchange process between the vegetation canopy and the atmosphere, and is influenced by a variety of factors such as radiation, vapor pressure deficit, soil thermal state, wind speed, air pressure, and vegetation cover.

[0003] Against the backdrop of global warming and frequent extreme drought events, different vegetation types exhibit significant differences in their response to temperature changes. Some vegetation canopies can maintain relatively stable canopy temperatures through processes such as transpiration and cooling, resulting in canopy temperature fluctuations that are smaller than air temperature fluctuations. Other vegetation canopies, under conditions of high vapor pressure deficit, low soil moisture, or limited canopy conductance, may show a phenomenon where canopy temperatures rise synchronously with or even exceed the magnitude of air temperature changes. Therefore, quantitatively characterizing the slope of the canopy temperature response to air temperature changes is of great significance for identifying vegetation thermal regulation capacity, ecosystem water stress status, and vegetation functional stability.

[0004] Existing research and products primarily focus on variables such as leaf area index, vegetation index, land surface temperature, solar-induced chlorophyll fluorescence, total primary productivity, or evapotranspiration, but they are still insufficient to directly provide spatial distribution results of the global-scale, continuous temporal distribution of vegetation canopy thermal regulation capacity. Flux stations can provide canopy temperature observations or observational data related to canopy temperature, but the stations are discretely distributed, making it difficult to directly use them for global-scale continuous mapping. On the other hand, reanalysis meteorological data and remote sensing data have better spatial continuity, but they generally cannot directly characterize canopy temperature and its regulation capacity relative to air temperature changes.

[0005] Therefore, there is a need for a remote sensing diagnostic method for global vegetation canopy thermal regulation capacity that can comprehensively utilize flux station canopy temperature observation data, global grid environment-driven data, vegetation cover data, and snow cover data to construct global canopy temperature spatial distribution data and further calculate the canopy thermal regulation slope. Summary of the Invention

[0006] The purpose of this invention is to provide a remote sensing diagnostic method for the global canopy thermal regulation capacity, which can solve the problem that canopy temperature observations from discrete flux stations are difficult to extend to global continuous canopy thermal regulation capacity diagnostic results, and realize continuous quantitative assessment of the global scale of vegetation canopy thermal regulation capacity.

[0007] To achieve the above objectives, the present invention provides the following solution: A remote sensing diagnostic method for the global vegetation canopy thermal regulation capacity includes: Acquire flux station canopy temperature observation data, station environment-driven data, global grid environment-driven data, grid temperature data, vegetation cover data, and snow cover data within the target spatiotemporal range; A training sample set for canopy temperature was constructed based on canopy temperature observation data from flux stations, station environment-driven data, vegetation cover data, and snow cover data. Based on the canopy temperature training sample set, the sample weights are determined according to the spatial regions to which the training samples belong, and a regionally balanced weighted canopy temperature inversion model is established. The canopy temperature inversion model is used to estimate the canopy temperature based on environmental driving variables. Effective vegetation pixels are determined based on vegetation cover data and snow cover data. Global grid environmental driving data and grid temperature data are unified to the target spatial resolution and target time scale to obtain the set of global grid driving variables corresponding to effective vegetation pixels and the corresponding grid temperature data. The global grid driving variable set is input into the canopy temperature inversion model to obtain global canopy temperature spatial distribution data; Based on global canopy temperature spatial distribution data and corresponding grid temperature data, the canopy thermal regulation slope of each effective vegetation pixel is calculated. The global vegetation canopy thermal regulation capacity is determined based on the canopy thermal regulation slope, and the spatial distribution of global vegetation canopy thermal regulation capacity is generated.

[0008] Preferably, a canopy temperature training sample set is constructed based on flux station canopy temperature observation data, station environment-driven data, vegetation cover data, and snow cover data, including: Canopy temperature observations were determined from flux station canopy temperature observation data, and station environmental driving variables with the same observation time as the canopy temperature observations were determined from station environmental driving data. Based on the location of the flux stations and the observation time, vegetation cover data and snow cover data are matched with canopy temperature observations to obtain vegetation cover status and snow cover status. Candidate samples are formed by taking corresponding canopy temperature observations, site environmental driving variables, vegetation cover status, and snow cover status. The candidate samples were screened for quality to obtain a training sample set for canopy temperature.

[0009] Preferably, a canopy temperature training sample set is constructed based on flux station canopy temperature observation data, station environment-driven data, vegetation cover data, and snow cover data, including: Samples that simultaneously meet the conditions for canopy existence and the condition for valid variables are selected from the canopy temperature observation data of flux stations; When the leaf area index is greater than the preset leaf area index threshold, the air temperature is greater than the preset air temperature threshold, there is no snow cover or the snow cover ratio is not greater than the preset snow cover threshold, and the canopy temperature observation value does not belong to the preset outlier value or the preset outlier range, the corresponding sample is determined to meet the canopy existence condition. When all site environmental driving variables corresponding to the canopy temperature observation values ​​are valid values, the corresponding sample is determined to meet the variable validity condition. The canopy temperature observations that simultaneously satisfy the canopy existence condition and the variable validity condition, along with their corresponding site environmental driving variables, constitute the canopy temperature training sample set.

[0010] Preferably, based on the canopy temperature training sample set, sample weights are determined according to the spatial regions to which the training samples belong, and a regionally balanced weighted canopy temperature inversion model is established. This canopy temperature inversion model is used to estimate canopy temperature based on environmental driving variables, including: The environmental driving variable vector was determined from the training sample set of canopy temperature. The environmental driving variable vector consists of downflow longwave radiation, downflow shortwave radiation, water vapor pressure deficit, soil temperature, wind speed and surface air pressure. An inversion model for canopy temperature was established by using environmental driving variable vectors as model inputs and canopy temperature observations as model training labels. Air temperature, which is used as the independent variable for calculating the slope of canopy thermal regulation, is excluded from the explicit input variables of the canopy temperature inversion model.

[0011] Preferably, based on the canopy temperature training sample set, sample weights are determined according to the spatial regions to which the training samples belong, and a regionally balanced weighted canopy temperature inversion model is established, including: Based on the location of the flux station to which the training samples belong, the training samples in the canopy temperature training sample set are divided into multiple spatial regions; The number of training samples, the number of spatial regions, and the number of training samples within each spatial region were counted. For those belonging to the first The first spatial region training samples, according to Determine the sample weights; where, Indicates the first The sample weights of each training sample. Indicates the number of training samples. Indicates the number of spatial regions. Indicates the first The number of training samples within each spatial region; Based on the sample weights, the canopy temperature training sample set is trained with regional balanced weighting to obtain a regional balanced weighted canopy temperature inversion model.

[0012] Preferably, based on the canopy temperature training sample set, sample weights are determined according to the spatial regions to which the training samples belong, and a regionally balanced weighted canopy temperature inversion model is established, including: A gradient boosting tree regression model was used to establish a mapping relationship between the environmental driving variable vector and the canopy temperature estimated by the model; Sample weights are introduced into the error term between observed canopy temperature and model-estimated canopy temperature, and model parameters are determined by minimizing the weighted error. The generalization ability of the canopy temperature inversion model is evaluated by using one of the following methods: site full containment validation and spatial region full containment validation. When the generalization ability meets the preset accuracy requirements, the trained gradient boosting tree regression model is determined as a regionally balanced weighted canopy temperature inversion model.

[0013] Preferably, effective vegetation pixels are determined based on vegetation cover data and snow cover data, and global grid environment-driven data and grid temperature data are unified to the target spatial resolution and target time scale to obtain the set of global grid driving variables corresponding to the effective vegetation pixels and the corresponding grid temperature data, including: Unify global grid environment-driven data and grid temperature data to the target spatial resolution and target time scale; Vegetation cover is determined for grid cells based on vegetation cover data, and snow cover is determined for grid cells based on snow cover data. Grid cells that meet the vegetation cover conditions, have no snow cover or a snow cover ratio not exceeding the preset snow cover threshold, and have effective environmental driving variables are identified as effective vegetation cells. A global grid driving variable set is formed from the environmental driving variables corresponding to the effective vegetation pixels, and grid temperature data corresponding to the effective vegetation pixels in terms of spatial location and time scale are retained.

[0014] Preferably, based on global canopy temperature spatial distribution data and corresponding gridded air temperature data, the canopy thermal regulation slope of each effective vegetation pixel is calculated, including: Regarding the first Each effective vegetation pixel is used to extract the canopy temperature time series from global canopy temperature spatial distribution data and the temperature time series from the corresponding grid temperature data. When the When the number of effective time steps in the set of time steps used for calculation by each effective vegetation pixel is not less than a preset minimum number, and the sum of squared deviations of the temperature time series is not less than a preset denominator threshold, the formula is applied. Calculate the first The canopy thermal regulation slope of each effective vegetation pixel; among which... Indicates the first Canopy thermal regulation slope of each effective vegetation pixel Indicates the first The set of time steps in which each effective vegetation pixel participates in the calculation. Indicates the first The effective vegetation pixel in the first Canopy temperature at each time step Indicates the first The effective vegetation pixel in the first Temperature at that time step Indicates the first The average of the canopy temperature time series of each effective vegetation cell. Indicates the first The average of the time series temperature of each effective vegetation cell; When the If the number of valid time steps in the time step set used for calculation by the effective vegetation pixels is less than a preset minimum number, or if the sum of squared deviations of the temperature time series is less than a preset denominator threshold, then the first... The canopy thermal regulation slope of each effective vegetation pixel is set to null.

[0015] Preferably, based on global canopy temperature spatial distribution data and corresponding gridded air temperature data, the canopy thermal regulation slope of each effective vegetation pixel is calculated, including: By removing the multi-year monthly average from the global canopy temperature spatial distribution data, an abnormal canopy temperature sequence was obtained; The average value of the same month over many years is removed from the corresponding grid temperature data to obtain the temperature anomaly sequence; When the canopy temperature anomaly sequence and the air temperature anomaly sequence meet the judgment conditions for the validity of the deseasonalization slope calculation, the deseasonalization canopy thermal regulation slope is calculated based on the canopy temperature anomaly sequence and the air temperature anomaly sequence. The validity criteria for deseasonalized slope calculation include: the number of valid time steps is not less than the preset minimum number; the number of valid month types is not less than the preset minimum number of month types; the sum of squares of the deviations of the temperature anomaly sequence is not less than the preset denominator threshold; and the goodness of fit between the canopy temperature anomaly sequence and the temperature anomaly sequence is not less than the preset goodness of fit threshold, or the significance test result of the deseasonalized canopy thermal regulation slope meets the preset significance level. The global vegetation canopy thermal regulation capacity was determined based on the canopy thermal regulation slope and the deseasonalized canopy thermal regulation slope.

[0016] Preferably, the global vegetation canopy thermal regulation capacity is determined based on the canopy thermal regulation slope, generating spatial distribution results of global vegetation canopy thermal regulation capacity, including: The first preset slope threshold and the second preset slope threshold are determined based on the canopy thermal regulation slope and slope tolerance parameter; the first preset slope threshold is 1 minus the slope tolerance parameter, and the second preset slope threshold is 1 plus the slope tolerance parameter; When the canopy thermal regulation slope is less than the first preset slope threshold, the corresponding effective vegetation pixel is determined as a thermal buffer type vegetation pixel. When the canopy thermal regulation slope is greater than or equal to the first preset slope threshold and less than or equal to the second preset slope threshold, the corresponding effective vegetation pixel is determined as a synchronous vegetation pixel. When the canopy thermal regulation slope is greater than the second preset slope threshold, the corresponding effective vegetation pixel is determined as a thermally magnified vegetation pixel. Based on thermal buffer vegetation pixels, synchronous vegetation pixels, and thermal amplification vegetation pixels, the spatial distribution of global vegetation canopy thermal regulation capacity is generated.

[0017] The present invention discloses the following beneficial effects: This invention constructs a canopy temperature training sample set using flux station canopy temperature observation data, station environmental driving data, vegetation cover data, and snow cover data. Based on the spatial region to which the training samples belong, sample weights are determined, and a regionally balanced weighted canopy temperature inversion model is established. This reduces the impact of uneven spatial distribution of training samples on the training process of the canopy temperature inversion model, enabling discrete flux station canopy temperature observation data to be used for global-scale canopy temperature estimation. Furthermore, this invention unifies global grid environmental driving data and grid temperature data to the target spatial resolution and target time scale, and determines effective vegetation pixels based on vegetation cover data and snow cover data, ensuring that the canopy temperature estimation process is consistent with the temperature response analysis process in terms of spatial location and time scale. By inputting the global grid driving variable set into the canopy temperature inversion model, global canopy temperature spatial distribution data is obtained. Then, combined with the corresponding grid temperature data, the canopy thermal regulation slope of each effective vegetation pixel is calculated, transforming the response relationship of canopy temperature to temperature changes into a spatially comparable quantitative indicator. Determining the global vegetation canopy thermal regulation capacity based on the canopy thermal regulation slope and generating spatial distribution results of the global vegetation canopy thermal regulation capacity can enable the expansion from site observations to continuous global spatial diagnostic results, thereby improving the spatial continuity and scale consistency of the global vegetation canopy thermal regulation capacity assessment. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A flowchart of the method provided in an embodiment of the present invention; Figure 2 The verification results map of the site for the canopy temperature inversion model in this embodiment of the invention is fully reserved; Figure 3 The verification result map is completely reserved for the region of the canopy temperature inversion model in this embodiment of the invention; Figure 4 The canopy thermal regulation slope, represented by row and column numbers, is the embodiment of this invention. Grid diagram; Figure 5 This is a schematic diagram of a canopy thermal regulation type grid represented by row and column numbers in an embodiment of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0022] Figure 1 The method flowchart provided in the embodiments of the present invention is as follows: Figure 1 As shown, this embodiment provides a remote sensing diagnostic method for the global vegetation canopy thermal regulation capacity, including: Step 100: Acquire flux station canopy temperature observation data, station environment-driven data, global grid environment-driven data, grid temperature data, vegetation cover data, and snow cover data within the target spatiotemporal range; Step 200: Construct a canopy temperature training sample set based on flux station canopy temperature observation data, station environment-driven data, vegetation cover data, and snow cover data; Step 300: Based on the canopy temperature training sample set, determine the sample weights according to the spatial regions to which the training samples belong, and establish a regionally balanced weighted canopy temperature inversion model; the canopy temperature inversion model is used to estimate canopy temperature based on environmental driving variables; Step 400: Determine effective vegetation pixels based on vegetation cover data and snow cover data, and unify global grid environment driving data and grid temperature data to the target spatial resolution and target time scale to obtain the set of global grid driving variables corresponding to effective vegetation pixels and the corresponding grid temperature data. Step 500: Input the global grid driving variable set into the canopy temperature inversion model to obtain global canopy temperature spatial distribution data; Step 600: Based on global canopy temperature spatial distribution data and corresponding grid air temperature data, calculate the canopy thermal regulation slope of each effective vegetation pixel; Step 700: Determine the global vegetation canopy thermal regulation capacity based on the canopy thermal regulation slope, and generate the spatial distribution results of the global vegetation canopy thermal regulation capacity.

[0023] This embodiment uses canopy temperature observation data from flux stations as the training sample source and global grid environment-driven data as the spatial generalization data source to establish a regionally balanced weighted canopy temperature retrieval model, generate global canopy temperature spatial distribution data, and further calculate the canopy thermal regulation slope. And canopy thermal regulation type. Grid temperature data is used to extract the temperature time series of each effective vegetation pixel within the target period, and serves as the data source for the independent variable calculation of the canopy thermal regulation slope.

[0024] In step 100, the canopy temperature observation data of the flux stations are the canopy temperature observation fields provided in the publicly available flux station data files, for example, the field name is or with A canopy temperature observation field with equivalent physical meaning is used, and the unit of the canopy temperature observation values ​​is unified to degrees Celsius. The canopy temperature observation field is directly read as the canopy temperature variable already given in the flux station data product; in a preferred embodiment, the temperature obtained by conversion from emitted longwave radiation, energy balance residue, or remote sensing surface temperature products is not used as the canopy temperature training label.

[0025] Site-driven environmental data, after time matching with canopy temperature observations, serves as the training input variable for the canopy temperature inversion model. Global grid-driven environmental data serves as the input variable for the canopy temperature inversion model running on a spatially continuous grid. Vegetation cover data and snow cover data are used for canopy presence detection and snow masking, respectively. Grid-based air temperature data is used to extract the air temperature time series for each effective vegetation pixel within the target time period and is used for calculating the canopy thermal regulation slope.

[0026] The representativeness of the canopy temperature observation field is determined by the site metadata of the corresponding flux station dataset. This site metadata includes the station's latitude and longitude, land surface type, observation time range, and site data version information. For different flux stations, the sensor deployment height, footprint range, or canopy representation range is based on the original flux station metadata. In this embodiment, when constructing training samples, the station number, data source, original file name, time range, latitude and longitude, and land surface type are recorded to trace the data source and spatial representation information of the canopy temperature observations.

[0027] Both site-driven data and global grid-driven data include environmental driving variables. These variables include radiation variables, atmospheric moisture variables, soil thermal state variables, near-surface dynamic variables, and air pressure variables. Radiation variables include downwave longwave radiation and downwave shortwave radiation; atmospheric moisture variables include vapor pressure deficit, relative humidity, or actual vapor pressure; soil thermal state variables include soil temperature; near-surface dynamic variables include wind speed or wind speed components; and air pressure variables include surface air pressure. In a preferred embodiment, the environmental driving variable vector includes downwave longwave radiation, downwave shortwave radiation, vapor pressure deficit, soil temperature, wind speed, and surface air pressure. Downwave longwave radiation and downwave shortwave radiation characterize canopy radiation input; vapor pressure deficit characterizes atmospheric moisture conditions related to canopy transpiration and cooling; soil temperature characterizes near-surface thermal state; wind speed characterizes the exchange conditions between the canopy and the near-surface atmosphere; and surface air pressure characterizes atmospheric background pressure conditions.

[0028] To avoid repeated participation of air temperature variables in the canopy temperature inversion process and the calculation of the canopy thermal regulation slope, and to reduce the direct transmission effect of air temperature input on the canopy thermal regulation slope, air temperature, which is the independent variable for calculating the canopy thermal regulation slope, is not included as an explicit input variable in the canopy temperature inversion model. The canopy temperature inversion model retrieves the absolute canopy temperature based on environmental driving variables, including radiation input, atmospheric moisture conditions, near-surface thermal state, and exchange conditions.

[0029] In step 200, when constructing the canopy temperature training sample set, the canopy temperature observation data of flux stations are matched with the corresponding site environmental driving data using the station number, station longitude, station latitude, and observation time as indices to obtain initial station samples. Each initial station sample includes at least the station number, station longitude, station latitude, observation time, canopy temperature observation value, downdraft longwave radiation, downdraft shortwave radiation, vapor pressure deficit, soil temperature, wind speed, air pressure, air temperature, leaf area index, and snow cover percentage.

[0030] Furthermore, spatial matching was performed based on the location of flux stations with vegetation cover and snow cover data to obtain the vegetation cover and snow cover status of each flux station at the corresponding observation time. Then, invalid values ​​and outliers were removed to ensure that the canopy temperature observations, station environmental driving variables, vegetation cover status and snow cover status corresponded to each other in terms of observation time and station location.

[0031] The canopy presence condition is used to screen effective samples that characterize the thermal state of the vegetation canopy from flux station canopy temperature observation data. The canopy presence condition includes: leaf area index greater than a preset leaf area index threshold, air temperature greater than a preset air temperature threshold, no snow cover or snow cover percentage not greater than a preset snow cover threshold, and canopy temperature observations not belonging to preset outliers or preset outlier ranges. The variable validity condition includes that all station environmental driving variables corresponding to the canopy temperature observations are valid values. The canopy temperature observations that simultaneously meet both the canopy presence condition and the variable validity condition, along with their corresponding station environmental driving variables, form the canopy temperature training sample set.

[0032] In one implementation, the preset leaf area index threshold is: The preset temperature threshold is The default abnormal value is The aforementioned thresholds are used to reduce the impact of bare ground, low vegetation cover, low-temperature dormancy, snow cover, and abnormal canopy temperature records on model training. These thresholds are exemplary parameters and do not constitute a limitation on the scope of protection of this invention; they can be adjusted according to the study area, vegetation type, data source, and data quality.

[0033] In this embodiment, the canopy temperature observation value is derived from... and Canopy temperature observation fields in flux site data files, for example Field or Fields with equivalent physical meanings, and their units are standardized to degrees Celsius. Among them, The data uses daily-scale flux data files. Data is collected from the corresponding site's flux data files and aggregated to a daily scale; data version information for different sites is based on the original data file name and data list record. The data list must at least record the data network name, site number, original file name, start and end dates, surface type, longitude, and latitude.

[0034] Furthermore, quality control is performed on the canopy temperature observation field. If the original flux station data file provides a quality identifier for the canopy temperature observation field, only records whose quality identifier meets the preset quality requirements are retained. If the original flux station data file does not provide a specific quality identifier for the canopy temperature observation field, canopy temperature observation records that are not missing, not filled, and not equal to preset outliers are retained, and further filtering is performed based on the canopy presence conditions. For model input variables, if a flux station environment-driven field with a quality identifier is used, records with a quality identifier that is measured or meets the preset quality requirements are retained. If there is no corresponding quality identifier, records with missing values, filled values, and values ​​exceeding the physical range are removed.

[0035] In this embodiment, the initial site sample includes Each flux station Candidate samples were selected. The initial site samples were then screened as follows: leaf area index was retained. Greater than The sample retains the temperature Greater than The sample retains a snow cover ratio of less than or equal to The sample, excluding canopy temperature observations, equals The samples were selected, and samples with missing or invalid model input variables in downlink longwave radiation, downlink shortwave radiation, vapor pressure deficit, soil temperature, wind speed, and air pressure were removed. Samples were determined to be valid site samples when all the above screening conditions were met. After the above screening, the following samples were obtained: Each flux station Valid site samples, with a sample time range of [number missing]. Year moon From the date Year moon The leaf area index range of the screened samples is [number]. to The median is Temperature range is to The median is The valid site samples cover four land surface types: farmland, grassland, shrubland / savanna, and forest.

[0036] Table 1. Statistics of Training Samples for the Canopy Temperature Inversion Model

[0037] In Table 1, "Land Surface Type" indicates the vegetation type group to which the flux stations participating in model training belong; "Number of Stations" indicates the number of flux stations under each land surface type that entered the canopy temperature training sample set; "Number of Valid Daily Samples" indicates the number of daily-scale samples retained after simultaneously meeting the conditions of leaf area index threshold, air temperature threshold, no snow cover, valid canopy temperature observations, and complete model input variables; and "Time Range" indicates the start and end months of valid samples for the corresponding type of station. As shown in Table 1, this embodiment includes... Each throughput site 1,000 valid daily samples, including farmland stations 1, valid daily sample One, forest station 1, valid daily sample One, grassland station 1, valid daily sample One, shrub / savanna site 1, valid daily sample indivual.

[0038] In step 300, a canopy temperature inversion model is trained based on the selected valid site samples, and the model is validated. Specifically, the selected valid site samples are combined into a canopy temperature training sample set, which is represented as follows:

[0039] in, This represents the training sample set for canopy temperature. Indicates the first training samples, Indicates the number of training samples. Indicates the first The environment-driving variable vector corresponding to each training sample Indicates the first The canopy temperature observation values ​​corresponding to each training sample Indicates the first The sample weights of each training sample.

[0040] In this embodiment, It consists of six variables: descending longwave radiation, descending shortwave radiation, water vapor pressure deficit, soil temperature, wind speed, and air pressure. Descending longwave radiation characterizes the atmospheric longwave radiation input received by the canopy, and its unit is [unit missing]. Downward shortwave radiation is used to characterize the solar shortwave radiation input received by the canopy, and the unit is ; Vapor pressure deficit is used to characterize the degree of air dryness, and the unit is _____. Soil temperature is used to characterize the near-surface thermal state, and the unit is degrees Celsius; wind speed is used to characterize the aerodynamic exchange conditions near the canopy, and the unit is... Atmospheric pressure is used to characterize atmospheric background pressure conditions, and its unit is . Air temperature is not used as an input variable in the canopy temperature inversion model; it is only used in step 600 to calculate the canopy thermal regulation slope. .

[0041] During model training, regional balancing weights are assigned to samples from different spatial regions. Specifically, all training samples are first divided into regions based on the latitude and longitude of flux stations or preset spatial partitioning rules. A spatial region; statistics of the first Number of training samples in each spatial region And count the total number of training samples. For those belonging to the first The first spatial region Each training sample has a sample weight. Determined according to the following formula:

[0042] in, Indicates the first The sample weights of each training sample. This represents the total number of training samples. Indicates the number of spatial regions. Indicates the first The number of training samples within each spatial region. After setting the sample weights as described above, each sample within the same spatial region has the same weight, and the total weight of samples in each spatial region is the same. In spatial regions with a larger number of samples, the weight of an individual sample is relatively smaller, while in spatial regions with a smaller number of samples, the weight of an individual sample is relatively larger, thereby reducing the impact of uneven spatial distribution of training samples on model parameter estimation. express and The product of and is used to clarify the sample weights. The denominator is the number of spatial regions and the product of and . The product of the number of training samples in each spatial region.

[0043] In this embodiment, training samples are grouped according to the spatial region where the flux stations are located. These spatial regions include North America, Central Europe, Western Europe, Eastern Mediterranean-Asia, and the Southern Hemisphere. (This belongs to the...) The first spatial region The weights of each training sample are determined according to the above regional equilibrium weight formula. and will Substituting the weighted squared error loss function reduces the dominant role of regions with large sample sizes in model training.

[0044] The gradient boosting tree regression model is preferred for canopy temperature retrieval. This model takes the environmental driving variable vector from the canopy temperature training sample set as input and the observed canopy temperature value as the output label. The nonlinear relationship between the environmental driving variables and canopy temperature is represented by the summation of multiple regression trees. The gradient boosting tree regression model is expressed as follows:

[0045] in, Indicates the first The model estimates canopy temperature for each training sample. Represents the canopy temperature inversion function. Indicates the number of regression trees. Indicates the first A regressive tree.

[0046] The sample weights are incorporated into the model training objective function. The model parameters are determined by minimizing the weighted error between the observed canopy temperature and the model's estimated canopy temperature. The objective function is expressed as:

[0047] in, This represents the model training objective function. Indicates the first Canopy temperature observations for each training sample. Indicates the first The model estimates canopy temperature using a training sample. This represents a model complexity constraint. The model complexity constraint is used to constrain the number of regression trees, tree depth, leaf node weights, or the number of leaf nodes, in order to reduce the risk of model overfitting.

[0048] Based on the trained gradient boosting tree regression model, the canopy temperature inversion model is obtained:

[0049] in, This represents the vector of environmental driving variables for the canopy temperature to be estimated. This indicates that the model estimates canopy temperature. This represents the canopy temperature inversion function obtained through training.

[0050] In one exemplary parameter setting, the number of regression trees is... The maximum tree depth is The learning rate is The sample subsampling ratio is The feature subsampling ratio is , The regularization coefficient is , The regularization coefficient is The weight of the smallest child node is The above model parameters are preferred implementation parameters and do not constitute a limitation on the scope of protection of this invention.

[0051] In both site-wide and region-wide full-leave validation, each round of model training is based solely on the current training set for statistical analysis. , and And recalculate the sample weights. The above processing ensures that the total contribution of different spatial regions to the model training objective function remains consistent, avoiding the excessive influence of regions with large sample sizes on model training.

[0052] Furthermore, the canopy temperature retrieval model underwent site-wide full hold-out validation and spatial region-wide full hold-out validation. Site-wide full hold-out validation involves using all samples from a single flux site as test samples and the remaining samples from other sites as training samples to evaluate the model's predictive ability for unknown sites. Spatial region-wide full hold-out validation involves using all samples from a single spatial region as test samples and the remaining samples from other spatial regions as training samples to evaluate the model's generalization ability to unknown spatial regions.

[0053] In this embodiment, site-wide pass / fail verification is replaced by site-by-site cross-validation: each round will... All valid samples from each throughput station are used as the test set, and the valid samples from the remaining throughput stations are used as the training set. This process is repeated until each throughput station is selected once. The number of valid stations after selection is... If there are 100, then the site will completely reserve verification for execution. Round; the number of test sites in each round is There are [number] training sites. One. (The rest is missing) After merging the off-site prediction results obtained from each round of validation, the overall coefficient of determination, root mean square error, mean absolute error, and bias are calculated. Therefore, the number of test samples in the full site retention validation is the cumulative number of test samples from each round, and the number of test sites is the total number of sites retained in each round.

[0054] like Figure 2 As shown, Figure 2 The site verification results map is fully reserved for the canopy temperature inversion model. Figure 2 In the middle, the horizontal axis represents the measured canopy temperature. The unit is degrees Celsius; the vertical axis represents the predicted canopy temperature. The unit is degrees Celsius; the scatter plot represents the model prediction results of samples from each reserved site; the dashed line represents the reference line where the measured canopy temperature is equal to the predicted canopy temperature; the scatter plot color changes from purple to green and then yellow, indicating the change in sample point density from low to high. Figure 2 The coefficient of determination is marked in the upper left corner. for Root mean square error for Mean absolute error for The number of reserved sites is . Figure 2 The scatter points are mainly distributed along the reference line where the measured and predicted values ​​are equal, indicating that the canopy temperature inversion model has good predictive ability for sites that did not participate in the training.

[0055] The complete regional validation is performed as a spatial region-by-spatial cross-validation: first, the flux stations are divided into spatial regions based on their latitude and longitude; in each round, the... All station samples within a spatial region are used as the test set, and station samples from the remaining spatial regions are used as the training set. This process is repeated until each spatial region is left out once. The prediction results outside the regions in each round are merged, and the overall model accuracy is calculated. Therefore, the number of test samples in the full region leave-out validation is the cumulative number of regional test samples in each round, and the number of test stations is the total number of stations included in the test regions in each round, not the number of stations in a single left-out region.

[0056] like Figure 3 As shown, Figure 3 The region is fully reserved for the verification results map of the canopy temperature inversion model. Figure 3 In the middle, the horizontal axis represents the measured canopy temperature. The unit is degrees Celsius; the vertical axis represents the predicted canopy temperature. The unit is degrees Celsius; the scatter plot represents the model prediction results of each reserved area sample; the dashed line represents the reference line where the measured canopy temperature is equal to the predicted canopy temperature; the scatter plot color changes from purple to green and then yellow, indicating the change in sample point density from low to high. Figure 3 The coefficient of determination is marked in the upper left corner. for Root mean square error for Mean absolute error for The number of reserved areas is . Figure 3 The scatter points are still mainly distributed along the reference line where the measured and predicted values ​​are equal, indicating that the canopy temperature inversion model has a good generalization ability to regions not involved in the training.

[0057] In this embodiment, a total of valid samples were selected for model training and validation. Item, from One throughput site. The site fully reserves verification for execution. After merging the off-site prediction results of each round, the cumulative test sample is: The total number of test sites is [number]. One, coefficient of determination for The root mean square error is The mean absolute error is The validation process involves leaving out all spatial regions one by one. After merging the prediction results from each round outside the regions, the cumulative test sample is [number missing]. The total number of test sites is [number]. One, coefficient of determination Approximately The root mean square error is approximately The mean absolute error is approximately The accuracy indicators mentioned above are overall evaluation results after merging the results of each round of leaveout verification predictions, and not the results of a single leaveout round. Based on the above verification results, this canopy temperature inversion model is selected as the subsequent global grid canopy temperature inversion model.

[0058] In other embodiments, while maintaining the technical approach of establishing a canopy temperature inversion model based on site canopy temperature training samples and further calculating the canopy thermal regulation slope, the canopy temperature inversion model can employ a random forest model, a neural network model, a support vector regression model, or a generalized additive model. This embodiment uses a region-balanced weighted gradient boosting tree regression model as the preferred model.

[0059] In step 400, global grid environment-driven data for the target time period is acquired and processed uniformly. In this embodiment, the target time period is... Year The target timescale is year, the target spatial resolution is month, and the target spatial resolution is [missing information]. The target spatial reference is latitude and longitude coordinates. The global grid driving variables use the following data sources: downlink shortwave radiation... Daily scale Products are aggregated into a monthly average; when the target month or target cell... When data is missing, use Supplement with similar products or reanalysis of shortwave radiation data, and record the data source identifier; for downlink longwave radiation, air temperature, dew point temperature, wind speed, surface air pressure, and soil temperature, use... The lunar-scale reanalysis product has an original spatial resolution of [missing value]. Leaf area index is adopted Leaf area index product, original time resolution is Tian or Heaven, the original spatial resolution is or And aggregate or resample to a monthly scale. Snow cover adopts Moon-scale snow cover products, with an original spatial resolution of [missing information]. Or a global grid resolution consistent with the data publishing product, and unified to Grid.

[0060] Specifically, the downwave shortwave radiation is calculated by averaging daily data to obtain the monthly average radiative flux, with units uniformly set to [unit not specified]. Downward longwave radiation is converted from the monthly cumulative or average amount of downward longwave radiation from the Earth's surface to the monthly average radiation flux, with the unit being uniformly set to [unit not specified]. Air temperature, dew point temperature, and soil temperature are all set in degrees Celsius; wind speed is calculated from the east-west and north-south wind speed components near the ground and is uniformly set in degrees Celsius. Surface air pressure is uniformly set to Leaf area index is uniformly set as Snow cover is consistently expressed as a percentage. Vapor pressure deficit is calculated from air temperature and dew point temperature. Specifically, the saturated vapor pressure is first calculated based on the air temperature, and the actual vapor pressure is calculated based on the dew point temperature. The difference between the saturated vapor pressure and the actual vapor pressure is then taken as the vapor pressure deficit, with all units standardized. .

[0061] For leaf area index data, quality control and time-based synthesis were performed first, followed by resampling. Gridding. For snow cover data, it is standardized to a monthly percentage and used for subsequent snow cover masking. Furthermore, spatial alignment and unit standardization are performed on data from different sources. Data is converted using spatial aggregation. ;for Or other spatial resolution data, first by target Grids are used for spatial aggregation or resampling; for those already... The data is pixel-aligned according to the target latitude and longitude grid. For data with a temporal resolution higher than the monthly scale, monthly average, monthly cumulative, or quality-controlled monthly composite methods are used to obtain monthly-scale data.

[0062] After unifying all variables, a global grid driving variable file is generated for each month. This file includes downwind longwave radiation, downwind shortwave radiation, vapor pressure deficit, soil temperature, wind speed, surface air pressure, air temperature, leaf area index, and snow cover percentage. Downwind longwave radiation, downwind shortwave radiation, vapor pressure deficit, soil temperature, wind speed, and surface air pressure serve as input variables for the canopy temperature inversion model; air temperature is used to calculate the canopy thermal regulation slope; and leaf area index and snow cover percentage are used for effective vegetation pixel selection.

[0063] Global grid environment-driven data is masked based on vegetation cover and snow cover data. Grid cells that do not meet the vegetation cover criteria or have snow cover are marked as invalid cells; grid cells that still have valid environment-driven variables after masking are identified as valid vegetation cells. In one implementation, the target time scale is monthly, and the target spatial resolution is [missing information]. Monthly scale and Spatial resolution is merely an exemplary parameter and does not constitute a limitation on the scope of protection of this invention.

[0064] In step 500, the global grid driving variable vector corresponding to each effective vegetation pixel at the effective time step is input into the canopy temperature inversion model to obtain the model-estimated canopy temperature of the effective vegetation pixel at the corresponding time step. For the first... The effective vegetation pixel in the first Environmental driver variable vector for each month Input the environmental driving variable vector into the canopy temperature inversion model obtained in step 300 to obtain the first... The effective vegetation pixel in the first Model estimates of canopy temperature for the month:

[0065] in, Indicates the first The effective vegetation pixel in the first Model estimates of canopy temperature for each month This represents the canopy temperature inversion function obtained from training in step 300. Indicates the number of regression trees. Indicates the first A tree of return, Indicates the first The effective vegetation pixel in the first The vector of environmental driving variables for each month.

[0066] The above processing is repeated for all valid vegetation pixels and all months within the target spatiotemporal range to obtain monthly global canopy temperature spatial distribution data. Global canopy temperature spatial distribution data can be generated at monthly, seasonal, annual, or other time scales; global canopy temperature spatial distribution data can include the canopy temperature estimate for each valid vegetation pixel, the number of valid samples, mask status, and data quality indicators.

[0067] For a given pixel in a given month, if the pixel simultaneously satisfies the leaf area index in that month... Greater than Temperature Greater than Snow cover ratio less than or equal to If the following conditions are met: downlink longwave radiation is valid, downlink shortwave radiation is valid, water vapor pressure deficit is valid, soil temperature is valid, wind speed is valid, and surface air pressure is valid, then the record of the pixel in that month is determined as a valid pixel time record, and the canopy temperature inversion value for the corresponding month is generated. If none of the above conditions are met, then the record of the pixel in that month is set to null and is not included in subsequent calculations. Calculation. Through the above processing, we can avoid misinterpreting bare land, low vegetation cover areas, snow cover areas, or areas lacking key driving variables as vegetation canopy temperature.

[0068] In step 600, the canopy thermal regulation slope is calculated. For each vegetation pixel with a valid time record, the canopy temperature time series for all valid months within the target time period is extracted, and the corresponding month's temperature time series is extracted from the grid temperature data. To ensure the stability of the canopy thermal regulation slope calculation, the calculation... Previously, the validity of slope calculation was judged for each vegetation pixel with a valid pixel time record.

[0069] Specifically, the validity judgment of slope calculation includes: first, the number of valid months involved in the calculation is not less than the preset minimum number of months; second, the number of valid months involved in the calculation covers not less than the preset minimum number of month types; third, the temperature time series involved in the calculation has sufficient amplitude; and fourth, the linear fitting quality between canopy temperature and air temperature meets the preset requirements.

[0070] In this embodiment, the preset minimum number of months is taken as Months, used to guarantee Calculate a valid sample base with at least a one-year timescale; preset minimum number of categories for each month. The term "months" is used to avoid the concentration of effective months in a single short period or a single season; temperature variation conditions include that the standard deviation of the temperature series is not less than a preset temperature standard deviation threshold, or The sum of squared temperature deviations in the calculation formula must not be less than a preset denominator threshold; linear fit quality conditions include a coefficient of determination not less than a preset coefficient of determination threshold, or a slope significance test. The value is not greater than the preset significance level.

[0071] When a valid vegetation pixel simultaneously meets the above conditions, the canopy thermal regulation slope of that pixel is calculated using the following formula. :

[0072] in, Indicates the first Canopy thermal regulation slope of each effective vegetation pixel; Indicates the first The set of time steps in which each effective vegetation cell participates in the calculation; Indicates the first The effective vegetation pixel in the first Canopy temperature at each time step; Indicates the first The effective vegetation pixel in the first Temperature at each time step; Indicates the first The average value of the canopy temperature time series of each effective vegetation pixel; Indicates the first The average value of the temperature time series of each effective vegetation pixel.

[0073] When a pixel has insufficient effective months, insufficient seasonal coverage of effective months, insufficient temperature time series variation, or the fitting quality does not meet preset requirements, the pixel's... Setting this to null indicates that the canopy is not included in the classification of canopy thermal regulation capacity types. The preset minimum number of months, preset minimum number of month types, preset temperature standard deviation threshold, preset denominator threshold, preset coefficient of determination threshold, and preset significance level can all be determined based on the target time period length, time resolution, canopy temperature inversion error, and application accuracy requirements. In one exemplary embodiment, the preset minimum quantity can also be determined based on the target time scale and target time period length; when the target time scale is a monthly scale, the preset minimum quantity can be taken as... , , or The preset denominator threshold is determined based on the distribution of the sum of squared deviations of the temperature time series; it can be taken as the [missing value] of the sum of squared deviations distribution. Percentiles, or take values ​​greater than 100%. The preset positive threshold.

[0074] The canopy thermal regulation slope is used to characterize the intensity of the canopy temperature response to changes in air temperature. When the canopy thermal regulation slope is less than... When the canopy temperature change is less than the air temperature change, the canopy thermal regulation slope is close to... When the canopy temperature changes approximately synchronously with the air temperature, the canopy thermal regulation slope is greater than 1. When the temperature change is greater than the temperature change, it indicates that the canopy temperature changes more significantly than the air temperature changes.

[0075] Furthermore, the multi-year monthly average canopy temperature distribution data and gridded air temperature data can be removed separately to obtain canopy temperature anomaly sequences and air temperature anomaly sequences. Based on these sequences, the deseasonalized canopy thermal regulation slope is calculated. This deseasonalized canopy thermal regulation slope is used to reduce the impact of seasonal cycles on the relationship between canopy temperature and air temperature. Before calculating the deseasonalized canopy thermal regulation slope, the first... The validity of the deseasonalized slope calculation is judged for each effective vegetation pixel; if the number of effective time steps, the number of effective monthly species, the sum of squares of the temperature anomaly sequence, the goodness of fit, or the significance test results of the deseasonalized canopy thermal regulation slope do not meet the preset requirements, the first effective vegetation pixel will be removed from the calculation. The deseasonal canopy thermal regulation slope of each effective vegetation pixel is set to null. Canopy thermal regulation slope and deseasonal canopy thermal regulation slope can be used as diagnostic indicators of canopy thermal regulation capacity.

[0076] like Figure 4 As shown, Figure 4 Canopy thermal regulation slope, expressed in row and column numbers Grid diagram. Figure 4 This does not represent a real geographical region, administrative boundary, national border, provincial border, coastline, or specific latitude and longitude range; it is only used to illustrate the form of the result of generating canopy thermal regulation slope on a continuous grid in this invention. Figure 4 The horizontal axis represents the column number of the grid, labeled , , , and The vertical axis represents the grid row number, labeled... , , , and ; Figure 4 The color band below is marked The range of the color band is approximately to The color band has triangular extensions at both ends, used to indicate values ​​below or above the display range. Numerical values. Blue indicates numerical values. Less than This means that the canopy temperature change is less than the air temperature change; the white area indicates... near This means that changes in canopy temperature are basically synchronized with changes in air temperature; red indicates... Greater than This means that the variation in canopy temperature is greater than the variation in air temperature. Figure 4 Different colors in to , to The abstract grid formed by the continuous distribution is used to illustrate the differences in the thermal regulation slope of the canopy at different spatial locations.

[0077] In step 700, based on the canopy thermal regulation slope Generate a global vegetation canopy thermal regulation type layer. In this embodiment, it is based on the canopy thermal regulation slope. With unit slope The degree of deviation between them is used to classify the effective vegetation pixels into types. Unit slope This indicates that canopy temperature changes approximately in sync with air temperature. To avoid this... of Values ​​that are misclassified as thermally buffered or thermally amplified due to canopy temperature inversion errors or fluctuations in regression samples are addressed in this embodiment by setting a slope tolerance parameter. and according to Determine the classification threshold.

[0078] The first preset slope threshold is expressed as:

[0079] The second preset slope threshold is expressed as:

[0080] in, This represents the first preset slope threshold. This indicates the second preset slope threshold. This represents the slope tolerance parameter.

[0081] In a preferred embodiment, Pick Then the first preset slope threshold is The second preset slope threshold is .when Less than When, the corresponding effective vegetation pixels are divided into heat buffer type vegetation pixels; when Greater than or equal to and less than or equal to When, the corresponding effective vegetation pixels are divided into synchronous vegetation pixels; when Greater than At that time, the corresponding valid vegetation pixels are classified as thermally magnified vegetation pixels. This classification is performed on all valid vegetation pixels to generate a global vegetation canopy thermal modulation type layer. It should be noted that... and The preferred classification threshold used in this embodiment is used to represent the value around a unit slope. of The tolerance range is not intended to limit the scope of protection of this invention.

[0082] In other implementations, the slope tolerance parameter Adjustments can be made based on canopy temperature inversion error, standard error of canopy thermal regulation slope, regression confidence interval, or application accuracy requirements. If confidence interval classification is used, then for the... For each effective vegetation pixel, first calculate the canopy thermal regulation slope. and its standard error And determine based on the preset information level The confidence interval. When The upper limit of the confidence interval is less than When, the corresponding pixel is classified as a heat buffer type vegetation pixel; when The confidence interval includes When, the corresponding pixels are classified as synchronous vegetation pixels; when The lower limit of the confidence interval is greater than When the corresponding pixel is classified as a thermally amplified vegetation pixel, the classification of canopy thermal regulation type can simultaneously consider the slope magnitude and the uncertainty of slope estimation.

[0083] like Figure 5 As shown, Figure 5 This is a schematic diagram of a grid representing canopy thermal regulation types, represented by row and column numbers. Figure 5 This does not represent a real geographical region, administrative boundary, national border, provincial border, coastline, or specific latitude and longitude range; it is only used to illustrate the invention based on the canopy thermal adjustment slope. The result format of classifying effective vegetation grids by type. Figure 5 The horizontal axis represents the column number of the grid, labeled , , , and The vertical axis represents the grid row number, labeled... , , , and Each color block corresponds to a valid vegetation grid, and the classification result is determined by the valid vegetation grid. The value is determined. Figure 5 In the legend below, blue represents heat-buffered vegetation pixels, corresponding to... Lighter colors represent synchronous vegetation pixels, corresponding to... Red indicates large-scale vegetation pixels that are thermally amplified, corresponding to... . Figure 5 and Figure 4 The same abstract grid and the same set of effective vegetation grids are used to illustrate the acquisition After obtaining the raster results, the effective vegetation pixels can be further divided into different canopy thermal regulation types according to preset thresholds.

[0084] The spatial distribution results of global vegetation canopy thermal regulation capacity include global canopy thermal regulation slope. Spatial distribution layer, global vegetation canopy thermal regulation type layer, and statistical results of canopy thermal regulation capacity. Among them, the global canopy thermal regulation slope... The spatial distribution layer is used to characterize the intensity of the canopy temperature response to changes in air temperature; the global vegetation canopy thermal regulation type layer is used to classify and express the canopy thermal regulation capacity of different regions; the statistical results of canopy thermal regulation capacity are used to characterize the area proportion or number of pixels of each canopy thermal regulation type in different vegetation types, different spatial regions, or different time periods. The statistical results of canopy thermal regulation capacity can also be statistically analyzed according to global land regions, climate zones, ecological zones, vegetation types, administrative regions, or other spatial divisions.

[0085] It should be noted that the target time period in this embodiment... Year Time resolution at the year and month scales Spatial resolution, leaf area index threshold Temperature threshold Snow cover threshold , Minimum number of valid months Minimum number of month types, gradient boosting tree model and Classification threshold and These are all preferred implementation parameters in this embodiment. Without changing the basis of the invention based on canopy temperature inversion and... When using the technical approach of diagnosing the global vegetation canopy thermal regulation capacity through calculation, the above parameters can be adjusted according to different research areas, data sources, vegetation types, or application needs.

[0086] Compared to existing methods for diagnosing crop water stress based on canopy temperature or canopy-temperature difference, this invention does not rely on a single crop to determine local water stress through reference crops, crops without water deficit, or canopy-temperature difference thresholds. Instead, it constructs a continuous diagnostic process that starts with training on flux station canopy temperature observation data, then moves to global grid environment-driven data inversion, calculates the canopy thermal regulation slope, and finally obtains the spatial distribution results of global vegetation canopy thermal regulation types. Specifically, this invention uses flux station canopy temperature observations and their synchronous environment-driven data as the sample basis. After simultaneously satisfying conditions such as canopy presence, no snow cover, and valid variables, a canopy temperature inversion model is established. Global grid environment-driven data with uniform spatial resolution, time scale, and variable units are used to extend discrete station observations into a continuous global canopy temperature layer. Furthermore, air temperature time series are used as the independent variable for canopy temperature response analysis to calculate the canopy thermal regulation slope. and based on Effective vegetation pixels are classified into thermal buffer type, synchronous type, and thermal expansion type. Therefore, the technical focus of this invention is to form a global-scale, temporally continuous, and spatially consistent diagnostic result of vegetation canopy thermal regulation capacity, rather than judging the local crop moisture status based solely on canopy temperature or canopy temperature difference at a certain moment.

[0087] This embodiment also provides a remote sensing diagnostic system for global vegetation canopy thermal regulation capacity, used to perform the aforementioned remote sensing diagnostic method for global vegetation canopy thermal regulation capacity. The system includes a data acquisition module, a training sample construction module, an inversion model establishment module, a grid data processing module, a canopy temperature inversion module, a thermal regulation slope calculation module, and a thermal regulation capacity generation module.

[0088] The data acquisition module is used to acquire flux station canopy temperature observation data, station environment-driven data, global grid environment-driven data, grid temperature data, vegetation cover data, and snow cover data within the target spatiotemporal range. When acquiring flux station canopy temperature observation data, the data acquisition module reads the canopy temperature observation field from the publicly available flux station data file and records the station number, longitude, latitude, land surface type, observation time range, and data version information. When acquiring station environment-driven data, it establishes a correspondence between station environment-driven variables and canopy temperature observation values ​​over the observation time. When acquiring global grid environment-driven data, grid temperature data, vegetation cover data, and snow cover data, it retains the spatial resolution, time scale, variable units, and data source identifier for each data set.

[0089] The training sample construction module is connected to the data acquisition module and is used to construct a canopy temperature training sample set based on flux station canopy temperature observation data, station environmental driving data, vegetation cover data, and snow cover data. The training sample construction module matches canopy temperature observations, station environmental driving variables, vegetation cover status, and snow cover status according to the flux station location and observation time, and filters valid station samples based on canopy presence conditions and variable validity conditions. The canopy presence conditions include a leaf area index greater than a preset leaf area index threshold, an air temperature greater than a preset air temperature threshold, no snow cover or a snow cover ratio not greater than a preset snow cover threshold, and canopy temperature observations not belonging to preset outliers or preset outlier ranges. The variable validity conditions include that all station environmental driving variables corresponding to the canopy temperature observations are valid values. The training sample construction module forms the canopy temperature training sample set by combining canopy temperature observations that simultaneously meet the canopy presence conditions and variable validity conditions with their corresponding station environmental driving variables.

[0090] The inversion model building module is connected to the training sample construction module. It is used to establish a regionally balanced weighted canopy temperature inversion model based on the canopy temperature training sample set, determining sample weights according to the spatial region to which the training samples belong. The inversion model building module divides the training samples in the canopy temperature training sample set into multiple spatial regions according to the latitude and longitude of flux stations or preset spatial zoning rules, and counts the number of training samples, the number of spatial regions, and the number of training samples within each spatial region. For the j-th training sample belonging to the r-th spatial region, the sample weight is determined according to the regional balanced weighting rule, ensuring that the total contribution of each spatial region to the model training objective function remains consistent. The inversion model building module uses the environmental driving variable vector as the model input and the observed canopy temperature as the model output label. It incorporates the sample weights into the error term between the observed canopy temperature and the model's estimated canopy temperature, training the regionally balanced weighted canopy temperature inversion model. The environmental driving variable vector includes downflow longwave radiation, downflow shortwave radiation, water vapor pressure deficit, soil temperature, wind speed, and surface air pressure. Air temperature, which is used as the independent variable for calculating the canopy thermal regulation slope, is not used as an explicit input variable for the canopy temperature inversion model.

[0091] The grid data processing module is connected to both the data acquisition module and the inversion model establishment module. It is used to determine effective vegetation pixels based on vegetation cover and snow cover data, and to unify global grid environmental driving data and grid temperature data to the target spatial resolution and target time scale, obtaining the set of global grid driving variables corresponding to effective vegetation pixels and the corresponding grid temperature data. The grid data processing module performs spatial alignment, temporal aggregation, and unit unification on global grid environmental driving data from different sources, forming a set of global grid driving variables from downlink longwave radiation, downlink shortwave radiation, water vapor pressure deficit, soil temperature, wind speed, and surface air pressure. It retains the grid temperature data as corresponding to the set of global grid driving variables in terms of spatial location and time scale. Simultaneously, based on vegetation cover and snow cover data, it performs masking processing on grid pixels that do not meet the vegetation cover conditions or have snow cover. Grid pixels that still have effective environmental driving variables after masking are identified as effective vegetation pixels.

[0092] The canopy temperature inversion module is connected to both the inversion model building module and the grid data processing module. It inputs the global grid driving variable set into the canopy temperature inversion model to obtain global canopy temperature spatial distribution data. For each effective vegetation pixel at each effective time step, the canopy temperature inversion module calls the canopy temperature inversion model output by the inversion model building module to calculate the model-estimated canopy temperature for that effective vegetation pixel at the corresponding time step. This process is repeated for all effective vegetation pixels and all effective time steps within the target spatiotemporal range to form global canopy temperature spatial distribution data. This global canopy temperature spatial distribution data includes estimated canopy temperatures for effective vegetation pixels and can be associated with the number of effective samples, mask status, and data quality identifiers.

[0093] The thermal regulation slope calculation module is connected to both the canopy temperature inversion module and the grid data processing module. It calculates the canopy thermal regulation slope for each effective vegetation pixel based on global canopy temperature spatial distribution data and corresponding grid air temperature data. For the i-th effective vegetation pixel, the module extracts the canopy temperature time series from the global canopy temperature spatial distribution data and the air temperature time series from the corresponding grid air temperature data. It calculates the canopy thermal regulation slope for the i-th effective vegetation pixel only when the number of effective time steps, the sum of squared deviations of the air temperature time series, and the effective monthly coverage or fitting quality meet preset requirements. If the slope calculation validity criteria are not met, the canopy thermal regulation slope for the corresponding effective vegetation pixel is set to a null value. The module can also remove the multi-year average for the same month from the global canopy temperature spatial distribution data and the grid air temperature data to obtain canopy temperature anomaly sequences and air temperature anomaly sequences, and calculate the de-seasonalized canopy thermal regulation slope based on these sequences.

[0094] The thermal regulation capacity generation module is connected to the thermal regulation slope calculation module. It is used to determine the global vegetation canopy thermal regulation capacity based on the canopy thermal regulation slope and generate the spatial distribution results of the global vegetation canopy thermal regulation capacity. The thermal regulation capacity generation module determines a first preset slope threshold and a second preset slope threshold based on the canopy thermal regulation slope and slope tolerance parameters. When the canopy thermal regulation slope is less than the first preset slope threshold, the corresponding effective vegetation pixel is identified as a heat-buffered vegetation pixel. When the canopy thermal regulation slope is greater than or equal to the first preset slope threshold and less than or equal to the second preset slope threshold, the corresponding effective vegetation pixel is identified as a synchronous vegetation pixel. When the canopy thermal regulation slope is greater than the second preset slope threshold, the corresponding effective vegetation pixel is identified as a heat-amplifying vegetation pixel. Based on heat-buffered vegetation pixels, synchronous vegetation pixels, and heat-amplifying vegetation pixels, the thermal regulation capacity generation module generates the spatial distribution results of the global vegetation canopy thermal regulation capacity.

[0095] In one optional embodiment, the system further includes a model validation module. The model validation module is connected to the inversion model building module and is used to evaluate the generalization ability of the canopy temperature inversion model using site full hold-out validation and spatial region full hold-out validation. Site full hold-out validation is used to test the predictive ability of the canopy temperature inversion model for sites not used in training; spatial region full hold-out validation is used to test the generalization ability of the canopy temperature inversion model for regions not used in training. The model validation module can evaluate the predictive accuracy of the canopy temperature inversion model based on the coefficient of determination, root mean square error, mean absolute error, and bias, and feed the evaluation results back to the inversion model building module for adjusting at least one model parameter among the following: number of regression trees, tree depth, learning rate, sample subsampling ratio, feature subsampling ratio, regularization parameter, and minimum child node weight.

[0096] In one optional embodiment, the system further includes a results management module. The results management module is connected to the thermal regulation capacity generation module and is used to store and manage global canopy temperature spatial distribution data, canopy thermal regulation slope spatial distribution layers, deseasonalized canopy thermal regulation slope spatial distribution layers, global vegetation canopy thermal regulation type layers, and canopy thermal regulation capacity statistical results. The canopy thermal regulation capacity statistical results can be statistically analyzed according to global land regions, climate zones, ecological zones, vegetation types, administrative regions, or other spatial partitions to characterize the area proportion or number of thermal buffer vegetation pixels, synchronous vegetation pixels, and thermally amplifying vegetation pixels within different spatial partitions.

[0097] Each module in this embodiment can be implemented by a processor in a computer device executing program instructions stored in memory, or it can be implemented collaboratively by one or more processing units in a server, cloud computing platform, or geospatial data processing platform. The data transfer relationship between the modules corresponds to the data processing flow in the above method embodiment. The raw data obtained by the data acquisition module is sequentially transferred to the training sample construction module, the inversion model establishment module, the grid data processing module, the canopy temperature inversion module, the thermal regulation slope calculation module, and the thermal regulation capacity generation module, thereby completing the conversion from flux station canopy temperature observation data to the spatial distribution results of global vegetation canopy thermal regulation capacity.

[0098] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0099] Specific examples have been used to illustrate the principles and implementation methods of this invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this invention. Furthermore, those skilled in the art will recognize that, based on the ideas of this invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A remote sensing diagnostic method for the global vegetation canopy thermal regulation capacity, characterized in that, include: Acquire flux station canopy temperature observation data, station environment-driven data, global grid environment-driven data, grid temperature data, vegetation cover data, and snow cover data within the target spatiotemporal range; Based on the canopy temperature observation data from the flux stations, the environmental driving data from the stations, the vegetation cover data, and the snow cover data, a canopy temperature training sample set is constructed. Based on the canopy temperature training sample set, the sample weights are determined according to the spatial regions to which the training samples belong, and a regionally balanced weighted canopy temperature inversion model is established. The canopy temperature inversion model is used to estimate the canopy temperature based on environmental driving variables. Based on the vegetation cover data and the snow cover data, effective vegetation pixels are determined, and the global grid environment driving data and the grid temperature data are unified to the target spatial resolution and target time scale to obtain the set of global grid driving variables corresponding to the effective vegetation pixels and the corresponding grid temperature data. The global grid driving variable set is input into the canopy temperature inversion model to obtain global canopy temperature spatial distribution data; Based on the global canopy temperature spatial distribution data and the corresponding grid temperature data, the canopy thermal regulation slope of each effective vegetation pixel is calculated. The global vegetation canopy thermal regulation capacity is determined based on the canopy thermal regulation slope, and the spatial distribution results of the global vegetation canopy thermal regulation capacity are generated.

2. The remote sensing diagnostic method for the global vegetation canopy thermal regulation capacity according to claim 1, characterized in that, Based on the canopy temperature observation data from the flux stations, the station environmental driving data, the vegetation cover data, and the snow cover data, a canopy temperature training sample set is constructed, including: Canopy temperature observations are determined from the flux station canopy temperature observation data, and station environmental driving variables with the same observation time as the canopy temperature observations are determined from the station environmental driving data. Based on the location of the flux station and the observation time, the vegetation cover data and the snow cover data are matched with the canopy temperature observation value to obtain the vegetation cover status and the snow cover status. The corresponding canopy temperature observations, the site environmental driving variables, the vegetation cover status, and the snow cover status are used to form candidate samples. The candidate samples are subjected to quality screening to obtain the canopy temperature training sample set.

3. The remote sensing diagnostic method for the global vegetation canopy thermal regulation capacity according to claim 1, characterized in that, Based on the canopy temperature observation data from the flux stations, the station environmental driving data, the vegetation cover data, and the snow cover data, a canopy temperature training sample set is constructed, including: Samples that simultaneously meet the conditions for canopy existence and the condition for valid variables are selected from the canopy temperature observation data of the flux stations; When the leaf area index is greater than the preset leaf area index threshold, the air temperature is greater than the preset air temperature threshold, there is no snow cover or the snow cover ratio is not greater than the preset snow cover threshold, and the canopy temperature observation value does not belong to the preset abnormal value or the preset abnormal range, the corresponding sample is determined to meet the canopy existence conditions. When all site environmental driving variables corresponding to the canopy temperature observations are valid values, the corresponding sample is determined to meet the valid variable conditions. The canopy temperature training sample set consists of canopy temperature observations that simultaneously satisfy the canopy existence condition and the variable validity condition, along with their corresponding site environment driving variables.

4. The remote sensing diagnostic method for the global vegetation canopy thermal regulation capacity according to claim 1, characterized in that, Based on the aforementioned canopy temperature training sample set, sample weights are determined according to the spatial regions to which the training samples belong, and a regionally balanced weighted canopy temperature inversion model is established. This canopy temperature inversion model is used to estimate canopy temperature based on environmental driving variables, including: An environmental driving variable vector is determined from the canopy temperature training sample set. The environmental driving variable vector consists of downflow longwave radiation, downflow shortwave radiation, water vapor pressure deficit, soil temperature, wind speed and surface air pressure. Using the environmental driving variable vector as model input and the canopy temperature observation value as model training label, the canopy temperature inversion model is established. The air temperature, which is used as the independent variable for calculating the slope of the canopy thermal regulation, is excluded from the explicit input variables of the canopy temperature inversion model.

5. The remote sensing diagnostic method for the global vegetation canopy thermal regulation capacity according to claim 1, characterized in that, Based on the aforementioned canopy temperature training sample set, sample weights are determined according to the spatial regions to which the training samples belong, and a regionally balanced weighted canopy temperature inversion model is established, including: Based on the location of the flux station to which the training samples belong, the training samples in the canopy temperature training sample set are divided into multiple spatial regions; The number of training samples, the number of spatial regions, and the number of training samples within each spatial region were counted. For those belonging to the first The first spatial region training samples, according to Determine the sample weights; where, Indicates the first The sample weights of each training sample. Indicates the number of training samples. Indicates the number of spatial regions. Indicates the first The number of training samples within each spatial region; Based on the sample weights, the canopy temperature training sample set is trained using a regional balanced weighting method to obtain the regional balanced weighted canopy temperature inversion model.

6. The remote sensing diagnostic method for the global vegetation canopy thermal regulation capacity according to claim 1, characterized in that, Based on the aforementioned canopy temperature training sample set, sample weights are determined according to the spatial regions to which the training samples belong, and a regionally balanced weighted canopy temperature inversion model is established, including: A gradient boosting tree regression model was used to establish a mapping relationship between the environmental driving variable vector and the canopy temperature estimated by the model; The sample weights are incorporated into the error term between the observed canopy temperature and the model-estimated canopy temperature, and the model parameters are determined by minimizing the weighted error. The generalization ability of the canopy temperature inversion model is evaluated using one of the following methods: site full-leave validation and spatial region full-leave validation. When the generalization ability meets the preset accuracy requirements, the trained gradient boosting tree regression model is determined as the regionally balanced weighted canopy temperature inversion model.

7. The remote sensing diagnostic method for the global vegetation canopy thermal regulation capacity according to claim 1, characterized in that, Based on the vegetation cover data and the snow cover data, effective vegetation pixels are determined, and the global grid environment driving data and the grid temperature data are unified to the target spatial resolution and target time scale to obtain the set of global grid driving variables corresponding to the effective vegetation pixels and the corresponding grid temperature data, including: Unify the global grid environment-driven data and the grid temperature data to the target spatial resolution and the target time scale; The grid cells are judged for vegetation cover based on the vegetation cover data, and the grid cells are judged for snow cover based on the snow cover data. Grid pixels that meet the vegetation cover conditions, have no snow cover or a snow cover ratio not greater than a preset snow cover threshold, and have effective environmental driving variables are determined as the effective vegetation pixels. The global grid driving variable set is formed by the environmental driving variables corresponding to the effective vegetation pixels, and the grid temperature data corresponding to the effective vegetation pixels in terms of spatial location and time scale is retained.

8. The remote sensing diagnostic method for the global vegetation canopy thermal regulation capacity according to claim 1, characterized in that, Based on the global canopy temperature spatial distribution data and the corresponding gridded air temperature data, the canopy thermal regulation slope of each effective vegetation pixel is calculated, including: Regarding the first Each effective vegetation pixel extracts the canopy temperature time series from the global canopy temperature spatial distribution data and extracts the temperature time series from the corresponding grid temperature data; When the first When the number of effective time steps in the set of time steps used for calculation by each effective vegetation pixel is not less than a preset minimum number, and the sum of squared deviations of the temperature time series is not less than a preset denominator threshold, the calculation is performed according to the formula... Calculate the first The canopy thermal regulation slope of each effective vegetation pixel; among which... Indicates the first Canopy thermal regulation slope of each effective vegetation pixel Indicates the first The set of time steps in which each effective vegetation pixel participates in the calculation. Indicates the first The effective vegetation pixel in the first Canopy temperature at each time step Indicates the first The effective vegetation pixel in the first Temperature at that time step Indicates the first The average of the canopy temperature time series of each effective vegetation cell. Indicates the first The average of the time series temperature of each effective vegetation cell; When the first If the number of valid time steps in the time step set for calculating the effective vegetation pixels is less than the preset minimum number, or if the sum of squared deviations of the temperature time series is less than the preset denominator threshold, then the first effective vegetation pixel will be included in the calculation. The canopy thermal regulation slope of each effective vegetation pixel is set to null.

9. The remote sensing diagnostic method for the global vegetation canopy thermal regulation capacity according to claim 1, characterized in that, Based on the global canopy temperature spatial distribution data and the corresponding gridded air temperature data, the canopy thermal regulation slope of each effective vegetation pixel is calculated, including: The global canopy temperature spatial distribution data is modified by removing the multi-year monthly average to obtain the canopy temperature anomaly sequence; The average value of the same month over many years is removed from the corresponding grid temperature data to obtain the temperature anomaly sequence; When the canopy temperature anomaly sequence and the air temperature anomaly sequence meet the validity judgment conditions for the deseasonalization slope calculation, the deseasonalization canopy thermal regulation slope is calculated based on the canopy temperature anomaly sequence and the air temperature anomaly sequence. The validity criteria for the deseasonal slope calculation include: the number of valid time steps is not less than the preset minimum number; the number of valid month types is not less than the preset minimum number of month types; the sum of squares of the deviations of the temperature anomaly sequence is not less than the preset denominator threshold; and the goodness of fit between the canopy temperature anomaly sequence and the temperature anomaly sequence is not less than the preset goodness of fit threshold, or the significance test result of the deseasonal canopy thermal regulation slope meets the preset significance level. The global vegetation canopy thermal regulation capacity is determined based on the canopy thermal regulation slope and the deseasonalized canopy thermal regulation slope.

10. The remote sensing diagnostic method for the global vegetation canopy thermal regulation capacity according to claim 1, characterized in that, The global vegetation canopy thermal regulation capacity is determined based on the canopy thermal regulation slope, generating spatial distribution results of global vegetation canopy thermal regulation capacity, including: A first preset slope threshold and a second preset slope threshold are determined based on the canopy thermal regulation slope and slope tolerance parameter; the first preset slope threshold is 1 minus the slope tolerance parameter, and the second preset slope threshold is 1 plus the slope tolerance parameter; When the canopy thermal regulation slope is less than the first preset slope threshold, the corresponding effective vegetation pixel is determined as a thermal buffer type vegetation pixel. When the canopy thermal regulation slope is greater than or equal to the first preset slope threshold and less than or equal to the second preset slope threshold, the corresponding effective vegetation pixel is determined as a synchronous vegetation pixel. When the canopy thermal regulation slope is greater than the second preset slope threshold, the corresponding effective vegetation pixel is determined as a thermally magnified vegetation pixel. Based on the heat-buffered vegetation pixels, the synchronous vegetation pixels, and the thermally amplified vegetation pixels, the spatial distribution results of the global vegetation canopy thermal regulation capacity are generated.

Images