A method for predicting canopy combustible moisture content by combining remote sensing and meteorological data

Abstract

This invention belongs to the field of remote sensing technology and relates to a method for predicting canopy combustible moisture content (CFMC) by combining remote sensing CFMC products and meteorological data. Specifically, through global two-stage sensitivity analysis, key input parameters for different tree species in the MEDFATE model are identified. Using these parameters, a cost function is constructed to measure the error between the CFMC simulated by the MEDFATE ecophysical process model and the CFMC measured in the field, achieving species-based localized correction of MEDFATE. Then, through four-dimensional variational data assimilation technology, the remote sensing CFMC product is combined with the localized MEDFATE. This achieves enhanced CFMC simulation, improving the accuracy of MEDFATE CFMC simulation while also increasing the temporal resolution of remote sensing data and the spatial resolution of the MEDFATE model. With accurate meteorological forecast data, effective daily CFMC prediction of vegetation can be achieved.

Description

Technical Field

[0001] This invention belongs to the field of remote sensing technology and relates to a MEDFATE model that combines remote sensing CFMC products and meteorological data to simulate (predict) canopy combustible water content (CFMC). Background Technology

[0002] Wildfires are a crucial ecological process influencing atmospheric and biosphere carbon and nutrient cycles. They naturally shape ecosystems by regulating vegetation growth, nutrient cycling, and biodiversity, but wildfires exacerbated by climate change also pose significant risks. Canopy combustible content (CFMC), defined as the percentage of water content in the vegetation canopy relative to its dry weight, is a key factor influencing fire ignition probability, spread rate, and intensity. Vegetation with high CFMC is less likely to ignite and sustain burning, while dry vegetation with low CFMC becomes highly flammable and contributes to rapidly spreading fires. Therefore, monitoring and predicting CFMC is essential for effective wildfire management, timely early warning systems, and disaster mitigation strategies.

[0003] In recent years, optical and microwave remote sensing technologies have become indispensable tools for large-scale estimation of CFMC (Farm Moisture Content). Platforms such as Sentinel-2, Landsat, and MODIS are widely used to acquire vegetation indices related to CFMC. These sensors provide frequent global observations, supporting near real-time tracking of vegetation moisture dynamics. Microwave sensors, whether active or passive, excel at continuous, all-weather monitoring, unaffected by cloud cover and atmospheric interference. Despite these advantages, active microwave sensing faces challenges such as high cost and sensitivity to heterogeneous vegetation and complex terrain, while passive systems suffer from low spatial resolution. In contrast, meteorological data can alleviate some of the limitations of remote sensing in predicting fuel moisture content. Researchers such as García combined meteorological data with remote sensing observations to improve previous empirical models, thereby enhancing the prediction of shrub and grassland surface vegetation cover.

[0004] However, Ruffault et al. pointed out that due to significant differences in performance across species, scales, and regions, spatial CFMC estimation should not rely solely on drought indices from meteorological data. Furthermore, process-based models such as MEDFATE have been developed, utilizing meteorological inputs to simulate vegetation growth and CFMC dynamics. The MEDFATE model combines vegetation and soil characteristics, providing a comprehensive framework for ecosystem-level CFMC simulation. In addition to meteorological inputs, this model considers soil moisture availability, root water uptake, and plant transpiration, offering detailed insights into vegetation responses to environmental change.

[0005] Recent advancements in the MEDFATE model have improved its accuracy in daily CFMC simulations for specific vegetation types by refining the soil moisture balance model used to estimate pre-dawn leaf water potential. Despite these progresses, accurate CFMC prediction at the species level remains challenging, particularly for tree species. Tree species typically possess deeper root systems and more efficient stomatal control, enabling them to better buffer against short-term fluctuations in shallow water availability. Results indicate that tree species exhibit the least seasonal variation in CFMC compared to grasslands, seedling shrubs, and regenerated shrubs, maintaining relatively stable canopy fuel moisture throughout the fire season. These characteristics continue to pose challenges to accurate modeling and prediction of CFMC at the species level. Summary of the Invention

[0006] This invention aims to provide a method for enhancing daily CFMC simulations by combining remote sensing CFMC products with a localized MEDFATE ecophysical process model through four-dimensional variational data assimilation technology. The method employed in this invention is summarized as follows: First, key input parameters for different tree species within the MEDFATE model are determined through global sensitivity analysis. Furthermore, these parameters guide the creation of a cost function to quantify the differences between MEDFATE simulations and field-measured CFMC, thereby achieving localized model calibration for specific species. Finally, the DEoptim global optimization algorithm, combined with four-dimensional variational data assimilation technology, integrates remote sensing CFMC product values ​​into the MEDFATE model.

[0007] This invention provides a method for predicting canopy combustible moisture content by combining remote sensing and meteorological data, comprising the following steps:

[0008] Step 1: Sensitivity analysis of parameters; obtain key input species parameters that affect the canopy combustible water content (CFMC) simulated by the MEDFATE ecophysical process model, and then perform sensitivity analysis to screen and obtain species parameter vectors;

[0009] Step 2: Species-based localization correction; Using the vegetation simulation ecological rules of the MEDFATE ecophysical process model as constraints, minimize the difference between CFMC simulated values ​​and field measurements, and use an evolutionary algorithm to optimize the selected species parameter vectors. The optimized species parameter values ​​are used as the standard species parameter inputs for the MEDFATE ecophysical process model.

[0010] Step 3: Four-dimensional variational data assimilation; using the standard input species parameters of the MEDFATE ecophysical process model as the state field variables for four-dimensional variational data assimilation, the accuracy of the MEDFATE ecophysical process model in simulating CFMC is improved through data assimilation techniques.

[0011] Step 4: The optimized MEDFATE ecophysical process model is used to make CFMC predictions based on meteorological forecast data.

[0012] Step 1 is described in detail as follows:

[0013] First, key input species parameters affecting the MEDFATE ecophysical process model's simulation of CFMC are obtained, including meteorological data, soil, geographical and vegetation information. Sensitive vegetation parameters for the MEDFATE ecophysical process model's simulation of CFMC include vegetation allometric growth coefficient, vegetation anatomy parameters, radiation balance and water interception parameters, vegetation photosynthetic parameters, as well as related hydraulic and vegetation water storage parameters. The model output is set as the mean and variance of the simulated CFMC.

[0014] Then, sensitivity analysis was performed using the test method based on the mean and variance of the simulated CFMC to obtain the parameter vector of the screened species, including m sensitivity parameters.

[0015] The test method is the extended Fourier amplitude sensitivity test, which selects parameters whose total order sensitivity index shows a significant decrease or is greater than 0.1 as species parameter vectors.

[0016] The evolutionary algorithm is specifically the DEoptim global optimization algorithm, and the formula is as follows:

[0017]

[0018] in, Indicates species parameters as hour, The model in the Japanese simulation ; Indicates the first Japanese Measured value; express Upper bound of the feasible domain for species parameters in the ecological physical process model; express Lower bound of the feasible domain for species parameters in the ecological physical process model; Represents a species parameter vector. For the m-th sensitivity parameter, Indicates the first The default values ​​for each sensitivity parameter, Indicates the first Lower bound of the domain of a sensitivity parameter Indicates the first Upper bound of the domain of each sensitivity parameter; This represents the lower limit of the m-th sensitivity parameter. This represents the upper bound of the m-th sensitivity parameter; n represents... The amount of measured CFMC data used in the localization correction of the ecological physical process model.

[0019] The four-dimensional variational data assimilation optimizes the accuracy of the MEDFATE ecophysical process model in simulating CFMC by constructing the following cost function:

[0020]

[0021] In the formula, The cost function; Represented as a species parameter vector, This represents the value of X in the previous iteration, and is the background field error covariance matrix; Represents the MEDFATE model. These are CFMC remote sensing observations, used only in background field correction and model validation. This is the observation field error matrix.

[0022] The species parameter vector includes eight sensitivity parameters, specifically: maximum live fuel water content, hydraulic vulnerability curve of stem xylem, ratio of leaf-branch weight to leaf weight of 6.35 mm branches, two holographic coefficients of leaf biomass, osmotic potential when leaves are fully expanded, hydraulic vulnerability curve of leaves, and permeable portion of stem tissue.

[0023] The beneficial effects of this invention are as follows:

[0024] This invention provides a model for enhancing daily vegetation CFMC simulations by combining remote sensing CFMC products with the MEDFATE model using four-dimensional variational data assimilation technology. First, a cost function is established to measure the error between the MEDFATE model's simulated CFMC and the field-measured CFMC, enabling localized correction of species-based MEDFATE input species parameters. Then, the remote sensing CFMC product is combined with the localized MEDFATE model using four-dimensional variational data assimilation technology. This achieves enhanced CFMC simulation, improving the accuracy of MEDFATE's simulated CFMC while simultaneously increasing the temporal resolution of the remote sensing data and the spatial resolution of the MEDFATE model. This advancement makes reliable prediction of vegetation CFMC possible, provided accurate weather forecasts are available. Attached Figure Description

[0025] Figure 1 The species parameters are ordinalized according to the higher-order sensitivity index. (a) reflects the overall impact of the input parameters on the model using the mean of the simulated LFMC, and (b) captures the variability of the model's response to the input parameters using the variance of the simulated LFMC.

[0026] Figure 2Two analytical plots are presented for the observation field CFMC product (MODIS-based CFMC product). (a) A line graph shows the R values ​​in the four components of the CFMC product. 2 (a) The changes in RMSE are shown in the inversion values ​​arranged in ascending order. (b) The scatter plot shows the correlation between the absolute error of the remote sensing CFMC value and the measured value and the remote sensing CFMC value.

[0027] Figure 3 A schematic diagram comparing simulated / estimated CFMC with field measurements, where: (a) MEDFATE simulation using default Mediterranean species; (b) accuracy of remote sensing products of the observation field; (c) accuracy of background field simulation; (d) accuracy of MEDFATE simulation after combining remote sensing CFMC products with four-dimensional variational data assimilation techniques. Detailed Implementation

[0028] To better understand the purpose, structure, and function of this invention, the following detailed description of a canopy combustible moisture content prediction method combining remote sensing and meteorological data is provided in conjunction with the accompanying drawings.

[0029] This invention provides a method for predicting canopy flammable material (CFMC) moisture content by combining remote sensing and meteorological data. First, a global two-stage sensitivity analysis is used to identify key input parameters for different tree species in the MEDFATE model. Furthermore, a cost function is constructed using these parameters to quantify the difference between the CFMC values ​​simulated by MEDFATE and the field measurements, thereby achieving model localization based on tree species. Then, by combining the DEoptim global optimization algorithm and four-dimensional variational data assimilation technology, the remote sensing CFMC product values ​​are integrated into the MEDFATE model.

[0030] The method for predicting canopy combustible moisture content by combining remote sensing and meteorological data provided by this invention specifically includes the following steps:

[0031] Step 1: Sensitivity Analysis of Parameters. First, obtain the key input species parameters that affect MEDFATE simulation of CFMC. Then, perform a global sensitivity analysis on the parameters that have a significant impact on MEDFATE simulation of CFMC to enhance the feasibility and universality of the model and improve computational efficiency.

[0032] MEDFATE (version 3.2.0) is a soil-vegetation-atmosphere ecological process model, primarily driven by meteorological data, integrating soil, geographic, and vegetation information to simulate the daily CFMC of specific species. Meteorological inputs include: daily minimum temperature (°C), daily maximum temperature (°C), daily minimum relative humidity (%), daily maximum relative humidity (%), daily precipitation (mm), and daily post-cloud solar radiation (MJ / m²). This invention extracts and processes relevant meteorological data from the ERA5-LAND dataset, setting non-essential climate parameters to MEDFATE's default values. Soil depth plays a crucial role in the formation of vegetation root habitats; this invention divides the soil into four layers with depths of 0-10 cm, 10-20 cm, 20-60 cm, and 60-100 cm. Data on clay content, sand content, organic matter content, and percentage of rock fragments are obtained from the SoilGrids 2.0 system (250-meter resolution) and downsampled to a spatial resolution of 0.1°*0.1°. For fine root depth, values ​​of 20 cm and 100 cm were used for the locations of 50% (Z50) and 95% (Z95) of the roots, respectively. Other species parameters were set to MEDFATE's default settings, supplemented by the most recent plot data from the Third Spanish National Forest Inventory.

[0033] The sensitive vegetation parameters for MEDFATE simulation of CFMC include vegetation allometric growth coefficients, vegetation anatomy parameters, radiation balance and water interception parameters, vegetation photosynthetic parameters, and related hydraulic and vegetation water storage parameters. To ensure that only species parameter variables change and climate data remain consistent, the MEDFATE model's simulation step size is set to one year, and the model output is set as the mean and variance of the simulated CFMC to characterize the model's susceptibility to input parameters. Parameters affecting the MEDFATE simulated CFMC are selected according to the MEDFATE guide, and then a global sensitivity analysis is performed on the selected parameters using the Extended Fourier Amplitude Sensitivity Test. The two sets of sensitivity analysis results—the mean and variance of the CFMC—are sorted according to the overall order sensitivity index. See [link to relevant documentation]. Figure 1The left figure shows the overall impact of input parameters on the simulated CFMC based on the mean, while the right figure reflects the fluctuation impact of input parameters on the simulated CFMC based on variance. Finally, parameters with a significant decrease in the overall sensitivity index or greater than 0.1 were selected as the model state field variables for four-dimensional variational data assimilation. Ultimately, eight species parameters were selected, including maximum live fuel water content (maxFMC), the hydraulic vulnerability curve of the stem xylem (VCsteam_d), the ratio of leaf-branch weight to leaf weight of a 6.35 mm branch (r635), two holographic coefficients of leaf biomass (b_fbt and c_fbt), used to calculate the coefficients of crown and diameter, the osmotic potential when the leaf is fully expanded (LeafPI0), the hydraulic vulnerability curve of the leaf (VCleaf_d), which characterizes the vulnerability or resistance of the plant stem xylem under water stress, and the permeable part in the stem tissue (StemAF).

[0034] The Extended Fourier Amplitude Sensitivity Test (EFAST) is... Others conducted Fourier amplitude sensitivity tests ( Based on ) and combined The proposed method is a global sensitivity analysis method based on variance decomposition. EFAST selects a suitable search curve and runs it in a multidimensional parameter space. It assigns a set of nonlinearly correlated integer frequencies to all input parameters of the model and introduces a function with common independent parameters to the selected parameters in the model, making the model a periodic function with independent parameters, thus reducing the multidimensional integral to a one-dimensional integral. By transforming the objective function into a Fourier series, the Fourier spectrum curves of each frequency are obtained, and the parameters can be calculated from the spectrum curves. The ratio of the resulting model output variance to the total variance is the sensitivity of this parameter.

[0035] Step 2: Species-based localization correction;

[0036] Plant growth is the result of the combined effects of multiple factors, including meteorology, soil, and geography. The MEDFATE model is primarily driven by meteorological data while also considering soil, geographical, and vegetation information to simulate the daily CFMC of a given species. The MEDFATE model cleverly transforms vegetation information into specific species parameters to characterize vegetation growth rate, morphology, and survival ability. The CFMC measurements used in this invention come from different sites, and the ecological environments corresponding to different sites often differ. Moreover, even the same vegetation will exhibit different growth performances at different sites, and these differences are reflected through different values ​​of species parameters. Therefore, localized adjustments to species are crucial for accurately simulating vegetation growth. Since measuring species parameters in different experimental areas is difficult and complex, this invention transforms the problem into an extremum problem using variational thinking to improve model feasibility. By constructing a cost function to describe the difference between simulated and measured CFMC values, and minimizing the difference between simulated and measured CFMC values ​​under dynamic constraints, the simulation of the background field is achieved. Species parameters were used as decision variables. The objective function was defined as the sum of the root mean square errors of the simulated CFMC values ​​from the MEDFATE model and the measured CFMC values ​​from the Globe LFMC 2.0 dataset. MEDFATE's vegetation simulation ecological rules were used as constraints, and the feasible region was defined as 50% to 200% of MEDFATE's default species parameters. The DEoptim global optimization algorithm (differential evolution algorithm) was used for background field simulation, and the optimized species parameter values ​​were used as the standard species parameter input values ​​for the MEDFATE model. The specific formula for the DEoptim algorithm is as follows:

[0037]

[0038] In the formula, Indicates species parameters as hour, The model in the Japanese simulation ; Indicates the first Japanese Actual measured values; express Upper bound of the feasible domain for species parameters; express Lower bound of the feasible region for model species parameters; This represents a vector of species parameters, specifically sensitivity parameters for eight species. Indicates the first Default values ​​for species parameters Indicates the first Lower bound of parameter domain Indicates the first Upper bound of the parameter domain; This represents the lower limit of the m-th sensitivity parameter. This represents the upper bound of the m-th sensitivity parameter; n represents... The amount of measured CFMC data used during model localization correction.

[0039] The MEDFATE model's R package provides default values ​​for some species parameters in the Mediterranean region, but using these default values ​​for CFMC simulations of the target vegetation does not yield very good results. (See...) Figure 3 In (a), R 2 =0.25, RMSE=27.41%, p<0.01. This paper constructs an optimization problem using the difference between simulated and measured CFMC values ​​as the cost function, and performs species-based background field simulations for the species *Quercus faginea* and *Quercus ilex*. By simulating vegetation parameters based on species, the obtained vegetation parameters are used as species parameter input values ​​for the MEDFATE model, improving the accuracy of simulated vegetation CFMC, as shown in Figure 3(c). 2 =0.53, RMSE=19.28%, p<0.01. The obtained simulation values ​​are also used as the background state for subsequent four-dimensional variational data assimilation.

[0040] Step 3: Four-dimensional variational data assimilation;

[0041] Four-dimensional variational data assimilation techniques are commonly used in meteorological models, where the assimilated variables are mostly meteorological parameters. However, the application of four-dimensional variational data assimilation techniques in this invention is quite different. This invention uses the input species parameters of the MEDFATE model as the state field variables for four-dimensional variational data assimilation, improving the accuracy of the MEDFATE model in simulating CFMC through data assimilation techniques. Four-dimensional variational data assimilation mainly consists of the background field, the background field error covariance matrix, the observation field, the observation field error covariance matrix, and the forecast model. The background field and observation field error covariance matrices define the relative weights of background information and observation information, as well as the relative information content projected onto the assimilated variables in each iteration. The cost function of four-dimensional variational data assimilation consists of two parts: a background term and an observation term.

[0042]

[0043] In the formula, The cost function; This represents the model state field vector, which in this embodiment is the species parameter vector, corresponding to 8 sensitivity parameters; This represents the background field state vector, and X represents the value of X in the previous iteration. The background field error covariance matrix; This refers to the forecast model, which in this invention is the MEDFATE model; This is the observation vector, which in this case represents CFMC remote sensing observations and is only used in background field correction and model validation. This is the observation field error matrix.

[0044] For calculating the observation field error covariance matrix, the observation field data comes from remote sensing CFMC products. For remote sensing CFMC inversion values ​​that specify the product, latitude, longitude, and time, the values ​​are often unique. Furthermore, the inversion error of remote sensing products usually represents the overall simulation accuracy, not the error of a single inversion value. Therefore, it is difficult to obtain the observation field error covariance matrix. This invention performs accuracy analysis on the observation field CFMC products. After arranging the observation field CFMC products in ascending order, they are divided into four equal parts (let's call them the first to the fourth parts from low to high values). Calculations show that the accuracy of the observation field CFMC remote sensing products is highest in the first part, and the accuracy gradually decreases in the second, third, and fourth parts. Figure 2 In the example 'a', this invention also calculates the absolute value of the error between the CFMC remote sensing product value and the CFMC measurement value, see [reference 1]. Figure 2 In step b, a significant positive correlation was found between the error and the observed field CFMC product value (r=0.91, ubRMSE=13.49%, p<0.01). Therefore, to enable the observed field error covariance matrix to characterize the error characteristics in the observed data and to reflect the effect of weighted observed data on the model state during data assimilation, while overcoming the difficulty in obtaining the observed field error covariance matrix, this invention simplifies the cost function of four-dimensional variational data assimilation. The observed field error covariance matrix is ​​simplified to the observed field error matrix, assuming all off-diagonal elements are zero, and the diagonal elements are the product of the overall RMSE of the observed field CFMC product and the observed values. Normalization is then used to eliminate the dimensional influence between the assimilation variable (CFMC) and the model state field variables. Although this simplification reduces the input of correct information from the observed and background fields, it significantly reduces computational complexity and improves the model's universality. It also solves the problems of requiring a large amount of historical data and empirical information to calculate the background field covariance matrix and the difficulty in obtaining the observed field error covariance matrix. Figure 3 -b represents the accuracy (Rb) of the observation field CFMC product after four-dimensional variational data assimilation. 2 =0.36, RMSE=37.16%, p<0.01). Simulated values ​​were used as background state values ​​for the background field, CFMC product values ​​as observed field data, and MEDFATE was used as the forecast model. Enhanced CFMC simulations were performed on five sampling points in Spain using the cognitive variational data assimilation technique (e.g., Figure 3 -d), significantly improving the simulation accuracy of MEDFATE's simulated vegetation CFMC (R2 =0.72, RMSE=12.58%, p<0.01).

[0045] It is understood that the present invention has been described through some embodiments, and those skilled in the art will recognize that various changes or equivalent substitutions can be made to these features and embodiments without departing from the spirit and scope of the invention. Furthermore, under the teachings of the present invention, these features and embodiments can be modified to adapt to specific situations and materials without departing from the spirit and scope of the invention. Therefore, the present invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the protection scope of the present invention.

Claims

1. A method for predicting canopy combustible moisture content by combining remote sensing and meteorological data, characterized in that, Includes the following steps: Step 1: Sensitivity analysis of parameters; obtain key input species parameters that affect the canopy combustible water content (CFMC) simulated by the MEDFATE ecophysical process model, and then perform sensitivity analysis to screen and obtain species parameter vectors; Step 2: Species-based localization correction; Using the vegetation simulation ecological rules of the MEDFATE ecophysical process model as constraints, minimize the difference between CFMC simulated values ​​and field measurements, and use an evolutionary algorithm to optimize the selected species parameter vectors. The optimized species parameter values ​​are used as the standard species parameter inputs for the MEDFATE ecophysical process model. Step 3: Four-dimensional variational data assimilation; using the standard input species parameters of the MEDFATE ecophysical process model as the state field variables for four-dimensional variational data assimilation, the accuracy of the MEDFATE ecophysical process model in simulating CFMC is improved through data assimilation techniques; Step 4: The optimized MEDFATE ecophysical process model is used to make CFMC predictions based on meteorological forecast data.

2. The method for predicting canopy combustible moisture content by combining remote sensing and meteorological data according to claim 1, characterized in that, Step 1 is described in detail as follows: First, key input species parameters affecting the MEDFATE ecophysical process model's simulation of CFMC are obtained, including meteorological data, soil, geographical and vegetation information. Sensitive vegetation parameters for the MEDFATE ecophysical process model's simulation of CFMC include vegetation allometric growth coefficient, vegetation anatomy parameters, radiation balance and water interception parameters, vegetation photosynthetic parameters, as well as related hydraulic and vegetation water storage parameters. The model output is set as the mean and variance of the simulated CFMC. Then, sensitivity analysis was performed using the test method based on the mean and variance of the simulated CFMC to obtain the parameter vector of the screened species, including m sensitivity parameters.

3. The method for predicting canopy combustible moisture content by combining remote sensing and meteorological data according to claim 2, characterized in that, The test method is the extended Fourier amplitude sensitivity test, which selects parameters whose total order sensitivity index shows a significant decrease or is greater than 0.1 as species parameter vectors.

4. The method for predicting canopy combustible moisture content by combining remote sensing and meteorological data according to claim 3, characterized in that, The evolutionary algorithm is specifically the DEoptim global optimization algorithm, and the formula is as follows: ； in, Indicates species parameters as hour, The model in the Japanese simulation ; Indicates the first Japanese Measured value; express Upper bound of the feasible domain for species parameters in the ecological physical process model; express Lower bound of the feasible domain for species parameters in the ecological physical process model; Represents a species parameter vector. For the m-th sensitivity parameter, Indicates the first The default values ​​for each sensitivity parameter, Indicates the first Lower bound of the domain of a sensitivity parameter Indicates the first Upper bound of the domain of each sensitivity parameter; This represents the lower limit of the m-th sensitivity parameter. This represents the upper bound of the m-th sensitivity parameter; n represents... The amount of measured CFMC data used in the localization correction of the ecological physical process model.

5. The method for predicting canopy combustible moisture content by combining remote sensing and meteorological data according to claim 4, characterized in that, The four-dimensional variational data assimilation optimizes the accuracy of the MEDFATE ecophysical process model in simulating CFMC by constructing the following cost function: ； In the formula, The cost function; Represented as a species parameter vector, This represents the value of X in the previous iteration, and is the background field error covariance matrix; Represents the MEDFATE model. These are CFMC remote sensing observations, used only in background field correction and model validation. This is the observation field error matrix.

6. The method for predicting canopy combustible moisture content by combining remote sensing and meteorological data according to claim 5, characterized in that, The species parameter vector includes eight sensitivity parameters, specifically: maximum live fuel water content, hydraulic vulnerability curve of stem xylem, ratio of leaf-branch weight to leaf weight of 6.35 mm branches, two holographic coefficients of leaf biomass, osmotic potential when leaves are fully expanded, hydraulic vulnerability curve of leaves, and permeable portion of stem tissue.

7. The method for predicting canopy combustible moisture content by combining remote sensing and meteorological data according to claim 6, characterized in that, The meteorological data inputs for the MEDFATE model include daily minimum temperature, daily maximum temperature, daily minimum relative humidity, daily maximum relative humidity, daily precipitation, and daily solar radiation behind clouds.

Images