Remote sensing evaluation method for forest drought death by coupling variable relationship and data learning model

Abstract

The invention provides a forest drought death remote sensing evaluation method for coupling a variable relationship and a data learning model, and relates to the field of ecological disaster and remote sensing technologies. According to the method, firstly, a structural equation model is constructed based on sample point data, and direct and indirect relations between the forest death rate caused by drought and key meteorological conditions, forest internal physiological parameters and external greenness features are quantified; and then, integrating a random forest regression technology, performing spatial prediction on the selected key variables, realizing wide coverage from sample point observation to global scale, and finally generating a forest drought-caused death rate in a global range. Compared with a traditional single evaluation means, the method has the advantages that the multivariate influence factors and the multi-dimensional prediction model are integrated, the evaluation precision and the space coverage capability of the death rate of the global forest due to drought are improved, and more direct and practical technical support is provided for global forest health monitoring and ecological system management.

Description

Technical Field

[0001] The present invention relates to the field of ecological disasters and remote sensing technology, and in particular to a method for evaluating the mortality rate of global forests caused by drought using statistical analysis combined with artificial intelligence technology. Background Art

[0002] Drought is one of the most common and costly natural disasters worldwide, and its frequency and intensity are expected to increase further due to global warming. Drought poses a significant threat to forest ecosystems and may increase the pressure on forest growth and survival. Therefore, revealing the spatiotemporal pattern of drought and its impact on forests, and assessing the damage to global forests caused by drought have become key tasks in responding to and mitigating global forest drought risks. Understanding the response mechanism of forests to drought, especially through remote sensing technology, has become an urgent need for formulating global water resources planning and ecological drought mitigation measures.

[0003] Drought monitoring and assessment usually rely on meteorological and hydrological indicators, such as precipitation, soil moisture, vapor pressure deficit and the Palmer Drought Severity Index. These indicators play a certain role in characterizing meteorological drought, but often ignore the complexity of forest ecosystems and the dynamic changes of drought processes. The response of forests to drought conditions varies depending on their ecological characteristics. There are significant differences in the adaptability of tropical forests and northern forests to drought, and these differences have been confirmed in multiple studies.

[0004] Vegetation indices, such as the Normalized Difference Vegetation Index, Enhanced Vegetation Index, and Leaf Area Index, are often used to assess forest growth and carbon sequestration capacity and are sensitive to water responses. However, changes in these vegetation indices are not only affected by drought, but may also be driven by factors such as fire, pests and diseases, or human intervention. Therefore, a single vegetation indicator may not accurately reflect the comprehensive impact of forest drought.

[0005] In recent years, the study of vegetation moisture content and solar-induced chlorophyll fluorescence has provided a new method for large-scale monitoring of forest drought. Vegetation optical depth monitors vegetation moisture content through microwave remote sensing technology, which can effectively characterize the total moisture content of aboveground biomass and has little impact on atmospheric conditions. As a direct indicator of photosynthesis, SIF's immediate response to light and water stress makes it a new tool for measuring the functional status of plants. Studies have shown that water stress can cause the stomata of vegetation to close, thereby reducing the SIF signal.

[0006] Although existing meteorological and vegetation response research provides an important basis for forest drought monitoring, it still faces significant challenges. Although the relationship between meteorological indicators and soil moisture and tree growth is direct, it is not completely equivalent to the actual situation of vegetation drought. Changes in vegetation greenness index may also be affected by other factors. The application of microwave remote sensing and SIF needs further exploration and verification. In order to effectively monitor forest drought, it is urgent to build a comprehensive drought monitoring framework that integrates the relationship between vegetation, meteorological indicators and forest drought mortality to improve the robustness and accuracy of drought monitoring. Summary of the invention

[0007] Traditional forest drought mortality assessment methods often rely on a single data source or model, which limits the comprehensive understanding of the complex environmental variable relationship, thus affecting the accuracy and reliability of the assessment. To overcome this limitation, the present invention proposes a method of coupling a structural equation model with a data learning model. This method first clarifies the relationship between forest drought mortality and environmental variables through a structural equation model, and then combines these relationships with machine learning technology. Through this comprehensive method, the accuracy of global forest drought mortality assessment can be significantly improved, providing more accurate and reliable assessment results.

[0008] The method for assessing global forest drought mortality by coupling a structural equation model with a data learning model of the present invention comprises the following steps:

[0009] Step S1: Data collection and preprocessing: Collect various meteorological and hydrological data and vegetation data, and preprocess the data to ensure data consistency and availability;

[0010] Step S2: Calculate canopy mortality sample data: collect geographical location and time data of forest deaths due to drought around the world, and use high-resolution global forest cover change data as background data to calculate canopy mortality;

[0011] Step S3: Training the structural equation model: Combining the data processed in step S1 and the canopy mortality calculated in step S2, the partial least squares path model PLS-PM is used as an implementation tool to construct a structural equation model;

[0012] Step S4: Establishing a mortality prediction function: combining the structural equation model constructed in step S3, using the death sample data to train and verify the structural equation model to identify significant environmental characteristics, and constructing a drought mortality prediction function based on the optimal structural equation model, wherein the environmental characteristics include vegetation and meteorological characteristics;

[0013] Step S5: Spatial extrapolation of environmental characteristics: spatially extrapolate the environmental characteristics obtained in step S4 by using RF technology to generate a global spatial distribution map of latent variables;

[0014] Step S6: Calculation of forest drought mortality rate in each region of the world: Substitute the latent variable distribution result obtained in step S5 into the prediction function obtained in step S4 to calculate the forest drought mortality rate in each region of the world;

[0015] Step S7: Draw a global forest drought mortality map: Based on the mortality results obtained in step S6, draw a visualization map of the global forest drought mortality rate to show the degree of drought impact in each region.

[0016] A storage device stores instructions and data for implementing the forest drought death remote sensing assessment method of the coupling variable relationship and data learning model.

[0017] A remote sensing assessment device for forest drought death based on coupled variable relationship and data learning model comprises: a processor and a storage device; the processor loads and executes instructions and data in the storage device to implement the remote sensing assessment method for forest drought death based on coupled variable relationship and data learning model.

[0018] The beneficial effects provided by the present invention are:

[0019] (1) Deepened the understanding of forest drought response mechanism: Through the analysis of structural equation model, this paper clarifies the impact of meteorological environment, vegetation water content change and photosynthesis activity on the complex process of forest death due to drought, and quantifies the specific response mechanism of forest under drought conditions. This research result not only improves the accuracy of understanding of forest drought phenomenon, but also lays the foundation for accurate assessment of forest drought mortality worldwide.

[0020] (2) Improved the accuracy of remote sensing assessment of forest drought mortality by coupling variable relationships and data learning models: This paper integrates the physiological mechanism of forest drought with machine learning technology to construct an efficient and accurate hybrid model for assessing the distribution and severity of forest mortality caused by drought disasters around the world. This method not only significantly improves the accuracy of the assessment, but also draws a detailed spatial distribution map of global forest drought mortality, allowing forest managers to more intuitively grasp the distribution characteristics of global drought risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a technical flow chart for estimating global forest drought mortality using a hybrid model;

[0022] Figure 2 It is the result of the PLS-PM model based on sample information;

[0023] Figure 3 It is the overall impact of latent variables calculated by inputting external manifest variables such as meteorology and hydrology, vegetation water content, chlorophyll fluorescence (i.e. photosynthesis capacity) and vegetation greenness on forest drought mortality;

[0024] Figure 4 It is the spatial distribution of latent variables obtained by spatial extrapolation of random forest regression;

[0025] Figure 5 are scatter plots of reference mortality and model-estimated mortality: (a) the prediction model using only random forests directly, and (b) the hybrid model combining the information quantified by PLS-PM and the RF extrapolated prediction;

[0026] Figure 6 This is a map of global forest mortality due to drought from 2014 to 2018 estimated using a hybrid model. (a) to (e) correspond to five different years (2014-2018).

[0027] Figure 7 It is a schematic diagram of the working of the hardware device. DETAILED DESCRIPTION

[0028] To make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0029] The present invention first explains the relevant basic concepts and the core points of the present application as follows, and then elaborates on the technical solution of the present invention in detail.

[0030] Please refer to Figure 1 , Figure 1 The present invention is a flowchart of a method; a remote sensing evaluation method for forest drought mortality based on a coupled variable relationship of a coupled structural equation model and a data learning model and a data learning model, comprising the following:

[0031] Step S1: Data collection and preprocessing: A variety of meteorological, hydrological and vegetation data were collected, and preprocessed by unifying spatial resolution, averaging and calculating annual percentage changes to ensure data consistency and availability;

[0032] Step S2: Calculate canopy mortality sample data; collect geographical location and time data of forest deaths due to drought around the world, and use the high-resolution global forest cover change data provided by Hansen's team as background data to calculate canopy mortality;

[0033] Step S3: Training the structural equation model: Combining the data collected and calculated in steps S1 and S2, PLS-PM is used as an implementation tool to build a theoretical model to describe and quantify the direct and indirect effects of meteorological and hydrological drought on forest mortality;

[0034] Step S4: Establishing a mortality prediction function; combining the theoretical model constructed in step S3, using the death sample data to train and verify the model, to identify significant vegetation and meteorological features, and to construct a drought mortality prediction function;

[0035] Step S5: spatial extrapolation of environmental characteristics; at the same time, the latent variable scores (environmental characteristics) obtained in step S3 are spatially extrapolated by RF technology to generate a global spatial distribution map of the latent variables;

[0036] Step S6: construct a mixed model of global forest drought mortality; substitute the latent variable distribution result obtained in step S5 into the prediction function obtained in step S4 to calculate the forest drought mortality rate in each region of the world;

[0037] Step S7: Draw a global forest mortality map due to drought; visualize the global forest mortality due to drought based on the mortality results obtained in step S6.

[0038] In the method of the present invention, in step S1, the meteorological and hydrological variables include precipitation, temperature, SPEI, PDSI, VPD and SM, which are from TerraClimate and SPEIbase. The vegetation response variables include VOD, SWIR, NDWI, SIF, SIFyield, NDVI and LAI, which are respectively from VODCA, PROBA-V, MODIS and other data sets. Data preprocessing includes, in order to ensure the comparability between different data sets, converting data of different resolutions into a consistent spatial grid by interpolation method for subsequent analysis. Calculate the average values ​​of all meteorological and hydrological variables in the three months of summer, and calculate the annual percentage change for each vegetation response variable.

[0039] In step S2, calculating the canopy mortality sample data includes three steps: ① data collection, extracting global forest death geographic location and time data from several existing research papers. In this embodiment, the data comes from the research of Allen et al. (2010), Caudullo and Barredo (2019), Gazol and Camarero (2022) and Hammond et al. (2022). These records include specific geographic coordinates and time information of death, providing basic data for research. And using the high-resolution global forest cover change data provided by Hansen et al. (2013) as background data, this data set provides global forest cover, increase, and decrease, with high spatial resolution; ② spatially superimposing the collected drought-induced forest death location data with the high-resolution forest cover change data, and dividing them according to the grid units of the unified spatial resolution size. In this embodiment, the spatial resolution is 0.25°. In this way, each death record can be classified into the corresponding grid cell, thus achieving structured processing of the data; ③ For each grid cell, the ratio of the dead canopy area to the total forest cover area in the grid is calculated to determine the forest mortality rate. Specifically, the dead canopy area in each grid is first identified, and then compared with the total forest cover area in the same grid to calculate the corresponding mortality rate, which is expressed as a percentage.

[0040] In step S3, training the structural equation model includes four steps: ① constructing a theoretical model. According to the data processed in step S1 and the canopy mortality calculated in step S2, the partial least squares path model PLS-PM is used as an implementation tool to construct a theoretical model, namely, a structural equation model, which aims to describe the direct impact of meteorological and hydrological drought on forest mortality and the indirect impact brought about by mediating variables (such as soil moisture and plant physiological state). In the model, the independent variables include a variety of meteorological and hydrological variables, and the dependent variable is forest mortality; ② standardization and stability check. In order to ensure the stability and comparability of the model, all variables are standardized and converted into standard deviation units. This process helps to eliminate the interference of different dimensions and units on the results and ensure the validity of the path coefficient and intercept; ③ checking the unity of indicators: using Cronbach's alpha index to check the unity between the indicators in the model to evaluate the reliability of measurement. By calculating the loading value of the indicators under each potential variable (i.e., the correlation between them), we identify those variables whose loading values ​​are lower than the preset loading value, and delete or adjust these low loading value variables. In this embodiment, the preset loading value is 0.5; ④ Model training and verification. After the processing and adjustment of the above steps, the model is trained and verified. Part of the sample data is used for model training, and the other part is used for model verification to evaluate the model's fit and prediction ability to select the optimal model.

[0041] In step S4, the formula of the drought mortality prediction function is as follows:

[0042]

[0043] Where Y represents the estimated forest mortality rate due to drought, LV i represents a significant feature, β i It is LV i , n is the number of significant features detected by PLS-PM, ε i Indicates error.

[0044] In step S5, the spatial extrapolation of environmental characteristics includes three steps: ① Acquisition of latent variable scores, using PLS-PM analysis to obtain latent variable scores related to forest mortality. These latent variables represent different environmental characteristics and reflect the key factors affecting forest drought mortality; ② Application of RF algorithm, using the latent variable scores obtained by the structural equation model as response variables, and using corresponding external explicit variables (such as soil data, NDVI, etc.) as predictors to train the RF model; ③ Spatial extrapolation, using the RF model to predict unobserved areas, by inputting variable data affecting forest drought mortality worldwide, the RF model can estimate the latent variable scores of these areas.

[0045] In step S6, the hybrid model of global forest drought mortality is obtained by bringing the spatial distribution of latent variables generated in step S5 into the drought mortality prediction function established in step S4 to obtain a global distribution of forest mortality caused by drought. The spatial distribution of latent variables includes hydrometeorology, vegetation water content, vegetation photosynthesis capacity and vegetation greenness. This calculation provides a quantitative assessment of forest drought mortality under drought conditions worldwide.

[0046] In step S7, a global forest drought mortality map is drawn. Based on the calculated evaluation results, a global forest drought mortality distribution map is visualized. These maps clearly show the predicted mortality rates in different regions, allowing users to identify high-risk areas and potential forest mortality hotspots. The mortality rates of the six major forest types in the world and the mortality distribution at different latitudes are statistically displayed.

[0047] The embodiment takes the global high temperature and drought years from 2014 to 2018 as an example to further describe the technical solution of the method for assessing global forest drought mortality based on microwave remote sensing of the present invention. The embodiment is used to illustrate the present invention, but is not used to limit the scope of application of the present invention, and is also applicable to different regions or time periods.

[0048] The implementation flow chart of the forest drought remote sensing assessment method based on the coupled variable relationship of the coupled structural equation model and the data learning model of the method of the present invention is as follows: Figure 1 As shown, the specific steps are as follows:

[0049] (1) Collection and preprocessing of basic data;

[0050] In this example, a SEM was constructed to further study the impact of meteorological and hydrological environmental variables on forest moisture content, growth status and drought mortality. Several key driving variables were included, including precipitation, temperature, SPEI, PDSI, VPD and SM. SPEI evaluates the impact of precipitation on the hydrological system by combining precipitation and evapotranspiration, and is particularly suitable for analyzing the characteristics of drought events; PDSI further integrates factors such as precipitation evaporation, temperature and soil moisture; VPD measures the moisture content in the atmosphere and reflects the degree to which the air is close to saturation; and SM directly reflects the moisture status in the soil and is an important indicator for monitoring vegetation water stress. All of these data, except SPEI, are derived from the TerraClimate dataset, with a spatial resolution of 1 / 24 degrees and a monthly time scale. SPEI data comes from the SPEIbase database with a resolution of 0.5 degrees, which is also monthly data. The average value of the three summer months was selected as the model input because summer is the season when forests face the greatest drought pressure.

[0051] Several vegetation response variables were also introduced, including VOD, SWIR, NDWI, SIF, SIFyield, NDVI and LAI. These variables reflect the moisture content, biomass, photosynthetic activity and health of vegetation from different perspectives. For example, VOD reflects the moisture and biomass of vegetation by measuring the absorption and scattering intensity of microwave radiation by the plant canopy; SWIR is sensitive to vegetation moisture content and can be used to monitor vegetation moisture conditions; SIF, as a fluorescence signal re-emitted by chlorophyll, is an effective tool for assessing the health and efficiency of plant photosynthesis; NDVI and LAI characterize the growth status and density of vegetation through optical properties and leaf area, respectively. These response variables come from data sources such as VODCA, PROBA-V, LT SIFc and MODIS. After preprocessing, the data reflects the changes in forests under summer water stress in the form of annual percentage changes. The calculation formula is as follows:

[0052]

[0053] Where VI after Represents the vegetation index of the current summer observation year, VI before Represents the vegetation index in the previous summer.

[0054] In addition, two derived indices, relative water content (RWC) and SIFyield, were calculated.

[0055] RWC is based on VOD data. By normalizing the data, the 5th and 95th percentiles of VOD from 2002 to 2017 (the available time period of the selected VOD) are used to represent dry and wet conditions, respectively. The median relative water content in summer is calculated using the following formula:

[0056]

[0057] Where t represents each summer season, s represents the 0.25° grid unit, median t，s (VOD) represents the median value of VOD in summer each year; 5 th percentile s (VOD) indicates the lowest value of VOD within the preset time period; 95 th percentile s (VOD) indicates the highest value of VOD within a preset time period.

[0058] SIFyield is calculated by the ratio of SIF to photosynthetically active radiation (PAR) and the fraction of PAR absorbed by the vegetation canopy (fPAR). This indicator eliminates the impact of different solar radiation conditions and canopy structures on SIF data and more accurately reflects the utilization efficiency of fluorescence during photosynthesis. The calculation formula is as follows:

[0059]

[0060] Where PAR represents the photosynthetically active radiation reaching the vegetation canopy, and fPAR represents the proportion of all PAR absorbed by the vegetation canopy.

[0061] Table 1 Selected data information

[0062] Table 1. Main data information

[0063]

[0064] (2) Calculate sample data on canopy mortality;

[0065] This paper aims to build a drought-induced forest mortality prediction model, and systematically collects and processes data to generate grid-based canopy mortality sample data, which will serve as the basis for model training and validation. The specific calculation process is as follows:

[0066] First, a total of 464 records of geographic location and time information on forest death events from 2014 to 2018 were extracted from the research results of Allen et al. (2010), Caudullo and Barredo (2019), Gazol and Camarero (2022), and Hammond et al. (2022). This time period was chosen not only because it covers one of the ten hottest years in the records of the National Oceanic and Atmospheric Administration (NOAA), but also because it fully considers the availability and representativeness of the data required for the study.

[0067] Subsequently, the high-resolution global forest cover change data released by Hansen et al. (2013) was used as background information. It is a grid data of forest cover and loss, and the data from 2014 to 2018 were still selected. By superimposing the drought-induced forest death location data with the global forest grid data, the canopy mortality rate monitored by satellite is accurately calculated in each 0.25° grid cell. This calculation process is achieved by comparing the ratio of the dead canopy area to the total forest cover area in each grid.

[0068] Finally, a global drought-induced forest death dataset containing 464 grid sample points was successfully constructed. The dataset was divided into two parts: 325 sample points were used for model training to ensure that the model could learn the complex relationship between drought and forest death; and the remaining 139 sample points were used for model validation to evaluate the model's predictive ability and generalization performance on unknown data.

[0069] (3) Training the structural equation model;

[0070] This paper focuses on training and applying structural equation models, especially using the statistical method of PLS-PM, to deeply analyze the complex causal relationship behind the phenomenon of forest mortality due to drought. A detailed theoretical framework is constructed, aiming to systematically reveal how these key characteristics act alone or in synergy on forest ecosystems, ultimately leading to and reflecting the increase in forest mortality, by comprehensively considering multiple meteorological driving variables (such as drought indicators, precipitation and soil moisture) and vegetation response variables (such as forest water content, photosynthesis capacity and vegetation greenness).

[0071] To construct this model, we first designed a path analysis diagram that included the direct effects of meteorological and hydrological drought and its indirect effects on forest mortality through mediating variables such as soil moisture and plant physiological state. The theoretical model was trained using death sample records collected from the literature. By fitting these sample data, the model identified meteorological and vegetation characteristics that were significantly correlated with forest mortality.

[0072] The overall goodness of fit based on the optimal theoretical model is 0.79. Figure 2 . Arrows indicate unidirectional relationships between latent variables. Solid arrows and dotted arrows indicate positive and negative relationships, respectively. The value on the arrow is the path coefficient, which indicates the magnitude and direction of the direct causal effect between the two variables. Insignificant coefficients in the arrow (α=0.05) are not plotted. The value inside the latent variable is the R-square of the latent variable, which is used to measure the explanatory power or predictive power of the latent variable on its observed variable. The bar chart on the right shows the overall impact (direct + indirect) of the latent variables calculated from various environmental factors on forest drought mortality. It can be seen that vegetation moisture content, vegetation greenness and coverage show a more significant direct relationship than dry and hot conditions and photosynthetic capacity, but the overall impact of hydrological and meteorological conditions is greatly improved after combining the indirect relationship. This total impact is used to construct a prediction function for mortality in the following.

[0073] Table 2. PLS-PM model results based on sample information

[0074]

[0075] (4) Establish a mortality prediction function;

[0076] This paper uses sample data of forest drought mortality to quantify the overall impact of key characteristic variables such as vegetation and meteorology on forest drought mortality through rigorous statistical analysis and model building processes. Figure 3 It is the overall impact of latent variables calculated by inputting external explicit variables such as meteorological hydrology, vegetation water content, chlorophyll fluorescence (i.e. photosynthesis capacity) and vegetation greenness on forest drought mortality. Specifically, PLS-PM was used as an analysis tool to deeply explore and accurately identify latent variables that significantly affect forest drought mortality, which mainly include specific vegetation response characteristics and meteorological drought conditions.

[0077] During the model training and validation phase, the core role of these latent variables in explaining and predicting forest mortality was verified. They indirectly or directly affect mortality through complex interaction mechanisms within the model. Based on the quantitative results of the model, a mortality prediction function was constructed. This function closely integrates the core results of PLS-PM analysis, which not only reflects the influence of key latent variables, but also ensures the high accuracy and reliability of the prediction results. The formula of the established mortality prediction function is as follows:

[0078]

[0079] Where Y represents the estimated forest mortality rate due to drought, LV i represents a significant feature, βi is LV i , n is the number of significant features detected by PLS-PM, ε i Indicates error.

[0080] (5) spatial extrapolation of environmental characteristics;

[0081] The present invention not only quantifies the complex influence relationship between latent variables through PLS-PM, but also accurately extracts the scores of these latent variables. These latent variable scores, as quantitative expressions of the intrinsic attributes of key features, deeply reveal the interactions between different characteristic elements and their contributions to forest drought mortality.

[0082] The present invention uses random forest regression algorithm to perform spatial extrapolation. The input variables of the latent variable hydrometeorology (external explicit variables) include cumulative precipitation, soil moisture, PDSI and SPEI. The input variables of the latent variable vegetation water content (external explicit variables) include ΔVOD, RWC and SWIR. The input variables of the latent variable photosynthesis capacity (external explicit variables) include SIF and SIFyield. The input variables of the latent variable vegetation greenness and coverage (external explicit variables) include EVI, NDVI and LAI. Finally, the latent variable scores obtained based on limited sample point observations are successfully mapped to a continuous grid on a global scale, realizing the spatial expansion from point to surface, and generating four sets of spatially continuous global distribution maps of latent variables to provide data for subsequent calculations, such as Figure 4 .

[0083] (6) Construct a hybrid model of global forest mortality due to drought;

[0084] The present invention substitutes the latent variable distribution results obtained by RF extrapolation as input variables into the previously constructed mortality prediction function. This process realizes the leap from sample point observation to global prediction, enabling the prediction model to output corresponding forest drought mortality prediction values ​​for specific environmental conditions in different regions around the world.

[0085] The results were validated and evaluated using a dedicated validation set, which accounts for 30% of the total dataset and covers a five-year time span to ensure the representativeness of the sample. The prediction accuracy of the hybrid model and the RF model alone for predicting global forest drought mortality was compared. Figure 5 The results shown in (a) and (b) show that the hybrid model performs better in predicting the global forest mortality due to drought, and its determination coefficient (R 2 ) reached 0.74, which was a significant improvement of 18% compared to 0.63 using only the RF model. This result not only shows that the hybrid model can more effectively capture the multi-dimensional and complex factors affecting forest drought mortality, but also further verifies that the accuracy and robustness of the prediction model can be significantly improved by integrating multiple algorithms or data sources.

[0086] (7) Map global forest drought mortality rates from 2014 to 2018;

[0087] Finally, a global map of forest mortality due to drought was successfully drawn between 2014 and 2018 ( Figure 6 ), Figure 6 (a) shows the global forest mortality rate caused by drought in 2014. Figure 6 (b) shows the global forest mortality rate caused by drought in 2015. Figure 6(c) shows the global forest mortality rate caused by drought in 2016. Figure 6 (d) in the figure is the global forest mortality rate caused by drought in 2017. Figure 6 (e) shows the global forest mortality rate due to drought in 2018. The map reveals in detail the geographical distribution and changes of forest canopy mortality due to drought in different years. It is worth noting that the vast majority (more than 98%) of forest areas recorded drought mortality rates below 15% during these five years, while areas with mortality rates exceeding 15% are considered to be severely drought-affected areas worldwide, including but not limited to Yunnan, China in 2014, Nigeria in 2015, eastern Brazil in 2016, Chile in 2017, and southeastern Australia in 2018.

[0088] exist Figure 6 In the lower left corner, the severity of death under the influence of drought in the six major forest types in the world (evergreen coniferous forest ENF, evergreen broad-leaved forest EBF, deciduous coniferous forest DNF, deciduous broad-leaved forest DBF, mixed forest MF and savanna forest Savannas) is further counted and displayed. The analysis results show that savannas have the highest mortality rate, followed by evergreen broad-leaved forests, mixed forests, and other types of forests. This ranking reveals the differences in sensitivity and resistance of different forest types to drought stress. At the same time, Figure 6 The right side shows the mortality statistics of different latitudes, clearly pointing out that tropical, southern temperate and northern coniferous forest regions are areas where forest death due to drought is frequent.

[0089] To verify the prediction effect of the hybrid model, its results were compared year by year with the map drawn using only the random forest algorithm. The comparison results showed that although the hybrid model predicted a wider distribution of deaths, the new part was mainly concentrated in areas with a mortality rate below 10%, while the severe drought areas with a mortality rate of more than 15% were relatively reduced.

[0090] See also Figure 7 , Figure 7 4 is a schematic diagram of the working of the hardware device of an embodiment of the present invention, wherein the hardware device specifically comprises: a remote sensing assessment device 401 for forest drought mortality with a coupled variable relationship and a data learning model, a processor 402 and a storage device 403.

[0091] A remote sensing assessment device 401 for forest drought death using a coupled variable relationship and a data learning model: The remote sensing assessment device 401 for forest drought death using a coupled variable relationship and a data learning model implements a remote sensing assessment method for forest drought death using a coupled variable relationship and a data learning model.

[0092] Processor 402: The processor 402 loads and executes the instructions and data in the storage device 403 to implement the forest drought mortality remote sensing assessment method based on the coupling variable relationship and data learning model.

[0093] Storage device 403: The storage device 403 stores instructions and data; the storage device 403 is used to implement the forest drought mortality remote sensing assessment method based on the coupling variable relationship and data learning model.

[0094] The beneficial effects of the present invention are:

[0095] (1) Deepened the understanding of forest drought response mechanism: Through the analysis of structural equation model, this paper clarifies the impact of meteorological environment, vegetation water content change and photosynthesis activity on the complex process of forest death due to drought, and quantifies the specific response mechanism of forest under drought conditions. This research result not only improves the accuracy of understanding of forest drought phenomenon, but also lays the foundation for accurate assessment of forest drought mortality worldwide.

[0096] (2) Improved the accuracy of remote sensing assessment of forest drought mortality by coupling variable relationships and data learning models: This paper integrates the physiological mechanism of forest drought with machine learning technology to construct an efficient and accurate hybrid model for assessing the distribution and severity of forest mortality caused by drought disasters around the world. This method not only significantly improves the accuracy of the assessment, but also draws a detailed spatial distribution map of global forest drought mortality, allowing forest managers to more intuitively grasp the distribution characteristics of global drought risks.

[0097] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A remote sensing assessment method for forest drought mortality based on coupled variable relationships and data learning models, characterized in that: The following steps are involved: Step S1: Data collection and preprocessing: Collect various meteorological and hydrological data and vegetation data, and preprocess the data to ensure data consistency and availability; Step S2: Calculate canopy mortality sample data: collect geographical location and time data of forest deaths due to drought around the world, and use high-resolution global forest cover change data as background data to calculate canopy mortality; Step S3: Training the structural equation model: Combining the data processed in step S1 and the canopy mortality calculated in step S2, the partial least squares path model PLS-PM is used as an implementation tool to construct a structural equation model; Step S4: Establishing a mortality prediction function: combining the structural equation model constructed in step S3, using the death sample data to train and verify the structural equation model to identify significant environmental characteristics, and constructing a drought mortality prediction function based on the optimal structural equation model, wherein the environmental characteristics include vegetation and meteorological characteristics; Step S5: Spatial extrapolation of environmental characteristics: spatially extrapolate the environmental characteristics obtained in step S4 by using RF technology to generate a global spatial distribution map of latent variables; Step S6: Calculation of forest drought mortality rate in each region of the world: Substitute the latent variable distribution result obtained in step S5 into the prediction function obtained in step S4 to calculate the forest drought mortality rate in each region of the world; Step S7: Draw a global forest drought mortality map: Based on the mortality results obtained in step S6, draw a visualization map of the global forest drought mortality rate to show the degree of drought impact in each region.

2. A remote sensing assessment method for forest drought mortality based on a coupled variable relationship and data learning model as claimed in claim 1, characterized in that: In step S1, the meteorological and hydrological variables include precipitation, temperature, standardized precipitation evapotranspiration index SPEI, Palmer Drought Severity Index PDSI, vapor pressure deficit VPD and soil moisture SM; The vegetation response variables include vegetation optical depth VOD, shortwave infrared reflectance SWIR, normalized water index NDWI, solar induced chlorophyll fluorescence SIF, normalized vegetation index NDVI and leaf area index LAI; relative water content RWC is calculated based on VOD t,s : Where t represents each summer season, s represents the grid unit with the unified spatial resolution; median t,s (VOD) represents the median value of VOD in summer each year; 5 th percentile s (VOD) indicates the lowest value of VOD within the preset time period; 95 th percentile s (VOD) indicates the highest value of VOD within the preset time period; Calculate the fluorescence utilization efficiency SIFyield based on SIF: In the formula, PAR represents the photosynthetically active radiation reaching the vegetation canopy, and fPAR represents the proportion of all PAR absorbed by the vegetation canopy; The preprocessing includes unifying the spatial resolution, averaging, and calculating the annual percentage change.

3. The remote sensing assessment method for forest drought mortality based on coupled variable relationship and data learning model as claimed in claim 1, characterized in that: In step S2, the calculation of canopy mortality sample data includes three contents: Step S2.1: Data collection: extract geographic location and time data of forest drought mortality events worldwide from several studies, and collect high-resolution global forest cover change data as background data; Step S2.2: Data overlay and gridding, specifically: Overlay these forest death location data with forest grid data and divide them globally according to grid cells of a unified spatial resolution; Step S2.3: Calculate canopy mortality, specifically: For each grid cell obtained in step S2.1, the canopy mortality rate is calculated as the ratio of the area of ​​dead forest canopy in the grid cell to the total forest cover area in the region.

4. The remote sensing assessment method for forest drought mortality based on coupled variable relationship and data learning model as claimed in claim 1, characterized in that: In step S3, training the structural equation model includes four steps: Step S3.1: PLS-PM was used as the implementation tool of SEM to construct a structural equation model to describe the process by which meteorological and hydrological drought directly affects forest mortality or affects it through mediating variables, including soil moisture and plant physiological status; Step S3.2: Standardize the variables to standard deviation units to normalize the path coefficients and intercepts and ensure the stability of the structural equation model; Step S3.3: Use Cronbach's alpha index to check the unity between indicators and adjust the structural equation model according to the loading value. Variables with loading values ​​lower than the preset loading value are deleted to improve the accuracy of the structural equation model. Step S3.4: After the processing and adjustment in the above steps, the structural equation model is trained and verified to select the optimal model.

5. The remote sensing assessment method for forest drought mortality based on coupled variable relationship and data learning model as claimed in claim 1, characterized in that: In step S4, a drought mortality prediction function is established through an optimal structural equation model, which is used to convert environmental characteristics into predicted values ​​of forest mortality. The formula of the established drought mortality prediction function is as follows: Where Y represents the estimated forest mortality rate due to drought, LV i Represents significant environmental features, β i It is LV i , n is the number of significant features detected by PLS-PM, ε i Indicates error.

6. The remote sensing assessment method for forest drought mortality based on coupled variable relationship and data learning model as claimed in claim 1, characterized in that: In step S5, the main process of spatial extrapolation of environmental variables includes extracting latent variable scores from the structural equation model, which reflect the environmental characteristics and vegetation status of each sample point; using the RF algorithm, spatially extrapolating the extracted latent variable scores, i.e., environmental characteristics, to unobserved areas around the world.

7. A storage device, characterized in that: The storage device stores instructions and data for implementing the forest drought death remote sensing assessment method based on the coupled variable relationship and data learning model described in any one of claims 1 to 6.

8. A remote sensing assessment device for forest drought mortality using a coupled variable relationship and data learning model, characterized in that: include: A processor and a storage device; the processor loads and executes instructions and data in the storage device to implement the forest drought mortality remote sensing assessment method based on the coupled variable relationship and data learning model described in any one of claims 1 to 6.

Images