A multi-hazard land-atmosphere coupling mode collaborative simulation method

Abstract

The application discloses a kind of multi-disaster land-air coupling mode collaborative simulation methods, belong to land-air coupling technical field.The application is by aiming at the target area of existing fire point, first to meteorological grid and land surface parameter do land surface energy balance calculation to obtain land surface state, and do confidence weighting to multi-source observation fire point, according to this, calculate multi-disaster trigger margin and update by fire point data bayesian fusion, screen out encrypted grid that meets trigger condition and do gust fluctuation reconstruction to obtain high-resolution driving field, finally input multi-disaster coupling model to the shared resource account of each disaster module is conserved distribution, generates multi-disaster concurrent simulation result, overcomes the repeated occupation and conservation destruction of same ground energy and material balance when multi-disaster concurrent, solves the same ground energy and material in different disaster calculation in prior art is repeatedly occupied, destroys the energy and mass conservation relation of system, causes the problem of the decline of physical consistency of multi-disaster collaborative simulation.

Description

Technical Field

[0001] This invention relates to the field of land-atmosphere coupling technology, and in particular to a collaborative simulation method for multi-hazard land-atmosphere coupling models. Background Technology

[0002] Sudden ecological damage events, such as forest fires, grassland fires, and dust storms, often cause severe ecological damage and environmental impacts within a short period. To predict and assess the occurrence, development, and ecological consequences of these events, simulations using land-atmosphere coupled numerical models are necessary. Current techniques typically employ global climate system models to simulate large-scale, long-term meteorological background processes to obtain a meteorological background field covering the study area. This background field is then spatially downscaled using a mesoscale model to obtain a high-spatial-resolution meteorological driving field for the region of interest. Based on this, functional modules such as land surface process models, atmospheric chemical transport models, and fire behavior models are coupled to perform simulation calculations for different types of sudden ecological damage events. This multi-model nesting and coupling approach can, to a certain extent, balance the continuity of large-scale background processes with the refinement of regional processes, and has become a common technical approach for conducting simulation assessments of ecological damage events.

[0003] However, in practical applications, the aforementioned existing technologies still have significant limitations. Global climate system models are geared towards long-term meteorological processes, and their temporal resolution is relatively coarse, with the output meteorological background field being smoothed over time. When such a background field drives regional models, rapidly changing processes within the boundary layer and short-term gust pulse information are weakened or even erased during the time smoothing process. The key evolutionary processes of sudden ecological damage events occur precisely on shorter timescales, and their occurrence and development are highly sensitive to instantaneous changes in near-surface wind fields and surface energy and matter. Since sand-raising processes depend on surface friction velocity, and fire behavior depends on instantaneous wind speed and heat transport, when the pulse information in the driving field is smoothed out, surface friction velocity and instantaneous wind speed are systematically underestimated, leading to lower simulation results for sand-raising intensity and fire spread processes, making it difficult to accurately reflect the severity of sudden events.

[0004] More importantly, when multiple sudden ecological damage events occur concurrently in the same region and at the same time, the smoothed driving field masks the competitive relationships between different disaster types on the same surface energy and material budget at fine time scales. During their occurrence and development, different disaster types all require the consumption or alteration of energy and materials in the same surface unit, exhibiting physical relationships of mutual promotion or inhibition. In existing technologies, each disaster type module often independently extracts and calculates the energy and material budget of the same surface unit, lacking a mechanism for unified constraint and coordination of this budget across multiple disaster types. This results in the same surface energy and material being repeatedly used in the calculations of different disaster types, disrupting the system's energy and mass conservation relationship, ultimately leading to a decrease in the physical consistency of multi-hazard co-simulation, distorted co-assessment results, and difficulty in providing reliable support for the comprehensive assessment of sudden ecological damage events.

[0005] Therefore, in the process of land-atmosphere coupled simulation, how to avoid underestimating the intensity of sudden events due to time scale smoothing, and how to overcome the repeated occupation and conservation of the same surface energy and material balance when multiple disasters occur simultaneously, is the core technical problem that the current collaborative simulation technology for sudden ecological damage events urgently needs to solve. Summary of the Invention

[0006] To address the technical problem in existing technologies where the same surface energy and matter are repeatedly used in calculations for different hazard types, thus disrupting the energy and mass conservation relationship of the system and leading to a decrease in the physical consistency of multi-hazard co-simulation, this invention provides a multi-hazard land-atmosphere coupled model co-simulation method. The technical solution is as follows:

[0007] The process involves acquiring meteorological grids, land surface parameters, and multi-source observed fire point data for the target area; performing land surface energy balance calculations using the meteorological grids and land surface parameters to generate land surface state data; weighting the multi-source observed fire point data based on confidence levels to generate weighted fire point observation data; calculating multi-hazard trigger margins characterizing the trigger distance of sudden ecological damage events in the target area based on the meteorological grids and land surface state data, and updating the multi-hazard trigger margins using Bayesian fusion based on the weighted fire point observation data; determining a set of encrypted grids from the meteorological grids that meet preset triggering conditions based on the updated multi-hazard trigger margins; reconstructing the meteorological grids corresponding to the encrypted grid sets using gust pulses to obtain high-resolution driving field data; inputting the high-resolution driving field data into a multi-hazard coupling model, and within a preset time window, performing conservation allocation calculations of shared resource ledgers for each hazard module within the multi-hazard coupling model using a loosely coupled operator splitting mechanism to generate multi-hazard concurrent simulation results.

[0008] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0009] This invention provides a multi-hazard land-atmosphere coupled model collaborative simulation method, which achieves high-fidelity collaborative simulation and assessment of sudden ecological damage events through multi-stage collaboration. The method first obtains land surface state data reflecting the dryness, wetness, temperature, and humidity of the underlying surface through land surface energy balance calculations using meteorological grids and land surface parameters. Then, meteorological and surface characteristics are normalized and aggregated into multi-hazard trigger margins, with a single dimensionless criterion characterizing the proximity of event triggering. Based on this, only the denser grid nearing triggering recovers the gusts of wind that have been smoothed out over time, improving the simulation fidelity for instantaneous wind speed-sensitive processes such as sandstorms and fire spread, while avoiding the computational burden of full-domain densification. Finally, a loosely coupled operator splitting mechanism conserves shared resources across multiple hazards, reflecting the mutual promotion and inhibition relationships between hazards while ensuring energy and material conservation. This overcomes the problems of underestimating the intensity of sudden events and violating conservation principles in multi-hazard collaborative simulation, improving the physical self-consistency and assessment reliability of the collaborative simulation.

[0010] This invention employs a loosely coupled operator splitting mechanism to perform a one-time, non-iterative proportional allocation of shared resources across multiple hazards. This feature involves uniformly registering the total amount of energy and matter available for allocation on the land surface at the start of a preset time window. Each hazard module independently calculates its demand without actually occupying it, and then allocates resources according to the normalized demand based on the hazard interaction matrix, with the total allocation rigidly constrained to within the available total. Thus, explicit conservation constraints replace traditional iterative solutions. Without requiring invasive modifications to the core of the coupled framework or introducing uncertainty in iterative convergence, this fundamentally avoids the conservation disruption caused by repeated occupation of the same land surface resources by multiple hazards, while ensuring the allocation process has a definite computational cost.

[0011] This invention injects satellite-observed fire points into the trigger identification process beforehand, weighted by confidence level, and drives on-demand encryption. This feature uses a multiplication of three factors—cloud contamination, radiative power, and time persistence—to suppress false fire points. It then penalizes the trigger margin by multiplying the observation confidence level and historical hit rate, ensuring that even grids with credible fire points, though not yet meeting meteorological and land surface conditions, are still identified as triggered. This compensates for the difficulty in deterministically predicting random ignition processes based on land-atmosphere conditions. Simultaneously, it employs dual-threshold hysteresis and relaxation calculations to suppress trigger state jitter, allowing refined computing power to be precisely focused on real hotspot grids, balancing simulation fidelity and computational efficiency. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0013] Figure 1 A flowchart illustrating a multi-hazard land-atmosphere coupled model co-simulation method provided in this application embodiment;

[0014] Figure 2 A schematic diagram illustrating the principle of on-demand reconfiguration of gust pulsation provided in this application embodiment. Detailed Implementation

[0015] Embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of the present disclosure are shown in the drawings, it should be understood that embodiments of the present disclosure may be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure.

[0016] It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure. In the description of the embodiments of this disclosure, the term "comprising" and similar terms should be understood as open-ended inclusion, i.e., "including but not limited to". The term "based on" should be understood as "at least partially based on". The term "one embodiment" or "this embodiment" should be understood as "at least one embodiment". The terms "first", "second", etc., may refer to different or the same objects.

[0017] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0018] This embodiment provides a multi-hazard land-atmosphere coupled simulation method, applied to the concurrent simulation of multiple hazards in a target area with fire points within a preset time window. This method overcomes the problems of underestimating the intensity of sudden events and violating conservation principles in multi-hazard co-simulation by recovering near-surface gusts that have been time-smoothed during the land-atmosphere coupled simulation and by uniformly allocating energy and material competition among multiple hazards on the same surface unit under conservation constraints. Figure 1 The diagram shown is a flowchart illustrating a multi-hazard land-atmosphere coupled model co-simulation method provided in an embodiment of this application. The method includes the following steps:

[0019] First, acquire meteorological grids, land surface parameters, and multi-source observation fire point data for the target area.

[0020] This step serves as the data preparation stage for the entire simulation process, providing a unified input foundation for subsequent land surface calculations, trigger identification, and hazard coupling. The meteorological grid originates from the regional meteorological field output by the global climate system model after spatial downscaling, containing discrete values ​​of physical quantities such as wind field, temperature, specific humidity, and geopotential height on a regular grid. This data can also be provided in real-time by numerical weather prediction systems or obtained through back-calculation using historical reanalysis data. Land surface parameters include data characterizing underlying surface attributes such as vegetation function type, leaf area index, soil texture, and topographic elevation, which can be obtained from land cover remote sensing products and topographic databases. Multi-source fire point data comes from fire point detection products mounted on different satellite platforms. To balance temporal coverage and spatial resolution, fire point detection results from polar-orbiting and geostationary satellites can be aggregated. By uniformly acquiring the above-mentioned multi-source and multi-type data, this step ensures the consistency of input criteria for subsequent calculation stages, laying a data foundation for multi-hazard collaborative simulation.

[0021] Based on the above data, land surface energy balance calculations are performed using meteorological grids and land surface parameters to generate land surface state data.

[0022] This step characterizes the energy exchange process between the land surface and the atmosphere, providing crucial pre-existing surface conditions for triggering identification. Specifically, the land surface energy balance calculation is based on the closed-loop distribution relationship of net surface radiation among various energy fluxes, the core relationship of which is shown in the following equation:

[0023] ;

[0024] In the formula, Net radiation at the Earth's surface represents the net radiant energy received per unit area of ​​the Earth's surface. Sensible heat flux, which characterizes the sensible heat transferred from the Earth's surface to the atmosphere; Latent heat flux represents the energy transported during water evaporation. Soil heat flux represents the heat transferred downwards into the soil. This equation is designed to distribute the net radiant energy received by the land surface in a closed loop across three pathways: sensible heat, latent heat, and soil heat conduction. This allows for the derivation of state variables such as surface temperature, soil moisture content, and combustible material moisture content under the constraint of energy conservation. Consequently, the generated land surface state data accurately reflects the dryness, wetness, and temperature of the underlying surface, providing a reliable physical basis for subsequently determining whether a specific grid possesses the conditions for a sudden ecological damage event. It should be noted that this land surface energy balance calculation is a relatively mature method in this field, and its solution can be achieved using conventional methods such as iterative approximation or hierarchical parameterization.

[0025] Considering the differences in reliability among multi-source observed fire points and the observation delay in real-time scenarios, confidence-weighted processing is applied to the multi-source observed fire point data to generate weighted fire point observation data. This processing specifically includes the following steps:

[0026] First, identify the timestamps corresponding to the multi-source observed fire point data. Since the multi-source observed fire point data comes from different satellite platforms, their detection time and data arrival time may be different. Therefore, identifying the timestamp corresponding to each observed fire point data can provide a basis for subsequent judgment on whether there is a delay in the observed data relative to the current simulation process.

[0027] Secondly, the current system operating mode is determined. If the current operating mode is real-time early warning mode, and the timestamp is in an observation delay relative to the current simulation time, then high-resolution driving field data from the previous time step is acquired. Short-term extrapolation calculations are performed using a preset fire spread model and the high-resolution driving field data from the previous time step to generate short-term extrapolated prior fire point data. This step is a key design feature of this method to address observation delays. Due to the transit interval and data processing delay in satellite fire point detection, it is often impossible to obtain the current observed fire point at the time required for simulation in real-time early warning scenarios. If fire point information is lacking during this period, trigger identification will be interrupted. Therefore, this method first estimates the fire point location using extrapolation to bridge the time difference between the observation arrival time and the time required for simulation.

[0028] The preset fire spread model is a process model that characterizes the fire line expansion rate based on the type of combustible material, terrain conditions, and near-surface wind field. It can be constructed using publicly available empirical formulas on fire behavior in this field, and the expansion rate coefficients are fitted and calibrated using the actual spread trajectories of historical fires in the target area before being called up. The high-resolution driving field data of the previous time step is obtained from the gust pulse reconstruction stage described later, and the recovered instantaneous wind field information contained therein provides the driving basis for fire spread extrapolation. Thus, this stage maintains a continuous supply of fire point information during the observation gap period, effectively avoiding interruption of trigger identification due to observation delay.

[0029] It should be noted that, in the first time step of system operation, or when a grid is first included in the densified grid set without high-resolution driving field data from the previous time step, to avoid the cold start problem of being unable to perform short-term extrapolation due to lack of driving input, one of the following methods can be used to provide the initial driving basis for the short-term extrapolation calculation: First, directly use the average wind speed field data resolved from the regional meteorological grid obtained by the aforementioned spatial downscaling at the current time step as the driving force for fire spread extrapolation, that is, in the first time step, replace the high-resolution driving field with the average wind field that has not been reconstructed by fluctuations, and wait for subsequent time steps... First, high-resolution driving field data is available before invocation. Second, a gust pulse reconstruction is performed in real-time using meteorological grid data at the current time step, and the resulting high-resolution driving field data is used for this short-term extrapolation, thus obtaining the driving basis containing instantaneous wind field information in the first time step. Third, before the system obtains any reliable observed fire points, short-term extrapolation is not activated. Instead, the previously weighted fire point observation data is maintained or attenuated in a preset manner during the observation gap period as temporary input. After the first reliable observed fire point arrives, the short-term extrapolation is driven by high-resolution driving field data from subsequent time steps. Thus, by providing the above initial driving method, the short-term extrapolation process can be executed normally or smoothly transitioned during the system cold start phase, avoiding process interruption due to the lack of driving field data from the previous time step in the first time step.

[0030] Third, before the arrival of the multi-source observation fire point data corresponding to the timestamp, the short-term extrapolated prior fire point data is output as the weighted fire point observation data, so that the subsequent update of the trigger margin still has usable fire point input during the observation window and maintains temporal continuity.

[0031] Fourth, after the arrival of multi-source observed fire point data, the cloud pollution factor, radiation power factor, and time persistence factor are extracted from the multi-source observed fire point data. Based on the extracted three factor data, the confidence level of the multi-source observed fire point data is calculated using a confidence multiplication method, and the calculation result is used as the new weighted fire point observation data. This step performs quality screening on the actual observed fire points after arrival, and one form of confidence weighting calculation is shown in the following formula:

[0032] ;

[0033] In the formula, The weighted fire point observation data are weighted by confidence level. To observe the original signal strength of the fire point, This is a cloud contamination factor, reflecting the degree to which observed pixels are obscured by clouds. The more severe the cloud contamination, the lower the value. The radiation power factor reflects the reasonableness of the fire point radiation power value and is used to eliminate detection results with abnormal deviations in radiation power. The time persistence factor reflects the recurrence of fire points in continuous observations, used to distinguish between persistent real fire points and transient non-fire-source thermal anomalies. The values ​​of each factor range from zero to one and can be determined based on the detection characteristics of each satellite platform and historical verification results. The design logic of this formula is to multiply multiple independent quality factors onto the original fire point signal, so that if the confidence level of any factor is low, the final weight of that fire point will be reduced, thereby effectively suppressing false fire points. Once the real observed fire point data arrives and the above confidence weighting is completed, its calculation result overwrites the previously used short-term extrapolated prior fire point data, completing the correction from extrapolated prior to real observation.

[0034] In summary, this step maintains a continuous supply of fire point information during periods of observation gaps and improves the reliability of fire point information after observations arrive. As an optional implementation, the quality factors used for confidence weighting are not limited to the three mentioned above; other factors reflecting the reliability of fire points can be added based on the characteristics of the observation platform. When the system is in a non-real-time early warning operation mode and the observation time sequence of fire points is known, there is no need to perform short-term extrapolation, and the observed fire point data corresponding to the simulation time can be used directly.

[0035] After obtaining land surface condition data and weighted fire point observation data, multi-hazard trigger margins characterizing the trigger distance of sudden ecological damage events in the target area are calculated based on meteorological grids and land surface condition data. The multi-hazard trigger margins are then updated using Bayesian fusion based on the weighted fire point observation data. The calculation of the trigger margins specifically includes the following steps:

[0036] First, meteorological characteristic variables are extracted from the meteorological grid, and surface characteristic variables are extracted from the land surface state data. Meteorological characteristic variables are physical quantities in the meteorological grid that directly affect the conditions for the occurrence of events, including air temperature, relative humidity, and near-surface wind speed; surface characteristic variables are physical quantities in the aforementioned land surface state data that reflect the dryness and wetness of the underlying surface and the vegetation status, including combustible material moisture content and leaf area index, etc. The extraction scope is consistent with the land surface state data generated by the aforementioned land surface energy balance calculation.

[0037] Secondly, a pre-defined historical fire dataset is obtained, and model fitting and calibration are performed based on this dataset to obtain pre-defined threshold parameters. The pre-defined historical fire dataset contains meteorological and surface characteristic variables corresponding to historical events that occurred and did not occur, along with markers indicating whether they were actually triggered. This dataset can be compiled by matching historical satellite fire point observation records with historical meteorological reanalysis data and land surface status data for the corresponding time periods. By using whether an event was actually triggered as a marker, model fitting and calibration are performed on the relationship between each driving variable and the triggering result, thus obtaining pre-defined threshold parameters adapted to the underlying surface and climate characteristics of the target area. This step establishes the trigger margin criterion based on the statistical regularities of historical events, improving the adaptability and reliability of trigger identification for specific regions. Furthermore, the historical dataset used for calibration can also include observation records of other types of sudden ecological damage events, such as historical sandstorms, to expand the applicable disaster range of the calibrated threshold parameters.

[0038] Next, meteorological and surface characteristic variables are input into a preset normalized weighted distance function and aggregated using preset threshold parameters to output multi-hazard trigger margins. The smaller the trigger distance value represented by the multi-hazard trigger margin, the higher the trigger probability of the target grid area. The preset normalized weighted distance function first normalizes the deviation of each driving variable relative to its preset threshold parameter, and then aggregates them into trigger margins through weighted multiplication. One calculation form is shown in the following formula:

[0039] ;

[0040] In the formula, This is a multi-hazard trigger margin, with a value between zero and one. For the first The values ​​of the driving variables, namely the extracted meteorological and surface characteristic variables. For the first Preset threshold parameters corresponding to each driving variable This is a scaling parameter used to adjust the sensitivity of each driving variable near a threshold. For the first The weights of the driving variables, This is a monotonic mapping function with values ​​between zero and one. This formula normalizes the deviations of each driving variable from its relative threshold to a unified interval using the monotonic mapping function, and then aggregates them through a multiplicative method. This ensures that if any key driving variable approaches its trigger threshold, the aggregation result changes, thus using a single dimensionless quantity to comprehensively characterize the triggering distance under the combined effect of multiple factors. The resulting multi-hazard trigger margin, with a value closer to zero, indicates that the target grid is closer to the triggering condition and has a higher triggering probability; conversely, a value further away indicates that it is farther from the trigger. This step unifies and reduces dispersed and heterogeneous multi-source driving information to a single criterion for easy comparison and judgment. As an optional implementation, the monotonic mapping function used for normalization and the aggregation method of each driving variable can be replaced while maintaining the above-mentioned correspondence between triggering distance and triggering probability.

[0041] It should be noted that the preset threshold parameter Scale parameters and weight All of these can be determined through the aforementioned model fitting and calibration process based on a preset historical fire dataset. That is, using whether or not an event is actually triggered as a marker, the parameters in the normalized weighted distance function are jointly fitted to optimize the output trigger margin for identifying historical events, thereby ensuring that all parameters of the function have an objective calibration basis.

[0042] Furthermore, considering that the ignition process of fires is often random and difficult to predict deterministically from the land-atmosphere state, a Bayesian fusion update is performed on the multi-hazard trigger margin based on weighted fire point observation data. This update specifically includes the following steps:

[0043] Firstly, the multi-hazard trigger margin is used as prior margin data. This is combined with weighted fire point observation data and a preset historical hit rate for penalized fusion calculation, resulting in a fused trigger margin data with a reduced value. The preset historical hit rate represents the probability that an observed fire point is indeed a real event, which can be obtained by statistically analyzing the proportion of real events among historical fire point records and corresponding event verification results. One form of the penalized fusion calculation is shown in the following formula:

[0044] ;

[0045] In the formula, To integrate trigger margin data, As a priori multi-hazard trigger margin, The reliability of fire points as represented by weighted fire point observation data. The historical hit rate is preset, with a value between zero and one, to limit the maximum reduction of prior information caused by observations. The design logic of this formula is to use the product of the observation reliability and the historical hit rate as a penalty factor to reduce the prior margin. This means that when a grid has a highly reliable observed fire point, its fusion trigger margin is reduced accordingly, so that even if the meteorological and land surface conditions of that grid have not yet reached the trigger threshold, it can still be judged as close to or at the trigger state. Conversely, when there is no observed fire point or its reliability is low, the reduction tends to disappear, and the fusion result degenerates to the original prior margin. This process organically combines deterministic physical criteria with the random ignition information reflected by observations, reducing missed triggers caused by the unpredictability of random ignition sources.

[0046] Secondly, the multi-hazard trigger margin of the previous time step is obtained, and the fused trigger margin data is compared with the boundary thresholds of the preset dual-threshold hysteresis interval. The preset dual-threshold hysteresis interval is defined by a lower boundary threshold and a higher boundary threshold, forming a hold interval between them. The boundary thresholds can be determined statistically from the aforementioned historical dataset. By comparing the fused trigger margin data with the boundary thresholds of this hysteresis interval, the triggering and de-triggering conditions can be distinguished.

[0047] Third, if the comparison result indicates that the fused trigger margin data is within the preset dual-threshold hysteresis range, then a preset relaxation calculation is performed on the multi-hazard trigger margin of the previous time step, and the calculation result is used as the updated multi-hazard trigger margin to suppress trigger state jitter. The preset relaxation calculation uses a weighted combination of the multi-hazard trigger margin of the previous time step and the current fused trigger margin data as the update result, and one calculation form is shown in the following formula:

[0048] ;

[0049] In the formula, For the updated multi-hazard trigger margin; This represents the multi-hazard trigger margin for the previous time step; This is the current fusion trigger margin data; The relaxation coefficient, with a value between zero and one, is used to adjust the weight of historical and current values ​​in the update result. This formula, by weighting the transition between historical and current values, makes the change in trigger margin more gradual over time, thereby suppressing its high-frequency oscillations when the trigger state has not undergone a clear transition.

[0050] Specifically, a smaller relaxation coefficient indicates that the update results rely more heavily on historical values, have a stronger smoothing effect, and exhibit slower state transitions; conversely, a larger relaxation coefficient results in a faster response to new information. Therefore, the dual-threshold hysteresis interval, combined with relaxation calculation, ensures that a clear state transition only occurs when the fused trigger margin exceeds the boundary threshold of the hysteresis interval, while maintaining a smooth transition within the interval. This step effectively suppresses the jitter of the trigger state near the critical condition, avoiding frequent changes in the encrypted mesh set caused by this, thereby improving the stability of subsequent simulation processes. As an optional implementation, the fusion of prior margin and observation information can also be achieved using other probability-based weighted update methods.

[0051] The lower and higher boundary thresholds of the preset dual-threshold hysteresis interval can be determined by statistically analyzing the distribution of trigger margins corresponding to triggered and non-triggered samples in the aforementioned preset historical fire dataset. This ensures that the two thresholds correspond to the critical levels for trigger and de-triggered judgments, respectively. The difference between the two thresholds, i.e., the width of the retention interval, can be set according to the requirements for the stability of the triggering state; a wider retention interval provides stronger suppression of state transitions. Relaxation coefficient. The value is determined based on the trade-off between the response speed and stability of the trigger margin. Alternatively, it can be set according to the strength of the fluctuations in the trigger margin over time in historical events. The more severe the fluctuations, the smaller the relaxation coefficient should be to enhance the smoothing effect.

[0052] After obtaining the updated trigger margin, a set of encrypted grids that meet the preset trigger conditions is determined from the meteorological grid based on the updated multi-hazard trigger margin. This determination process specifically includes the following steps:

[0053] First, candidate grids whose updated multi-hazard trigger margin is not greater than a preset trigger threshold are identified. Since the aforementioned multi-hazard trigger margin indicates a higher trigger probability with a smaller trigger distance value, when the updated trigger margin of a grid is not greater than the preset trigger threshold, it means that the grid is close enough to trigger conditions and should be included in the candidate grids.

[0054] Secondly, the global computing power budget threshold is obtained, and the total number of grids corresponding to all candidate grids is counted. The global computing power budget threshold represents the upper limit of the number of grids allowed for fine-grained processing while meeting computational timeliness requirements, and can be predetermined based on available computing resources and timeliness requirements. By comparing the total number of candidate grids with the global computing power budget threshold, a decision is made as to whether further selection of candidate grids is needed.

[0055] Secondly, if the total number of grid cells exceeds the global computing power budget threshold, then based on weighted fire point observation data, preset spread risk data, and preset vulnerability exposure data, a risk priority value is calculated for each candidate grid cell. One form of calculating the risk priority value is shown in the following formula:

[0056] ;

[0057] In the formula, The risk priority value for candidate grids. The reliability of fire points as represented by weighted fire point observation data. To pre-determine the spread risk data and characterize the potential for further development and expansion of an event within a grid, the data can be comprehensively determined by the continuity of combustible materials, topographic slope, and wind field conditions within that grid. To pre-determine vulnerability data, characterizing the exposure and vulnerability of the grid and the disaster-affected objects within its influence range, data on the distribution of population and ecological assets downwind of the grid can be used. This formula uses the product of three factors as a comprehensive measure, so that a low value for any factor will lower the priority of the grid, thereby prioritizing limited computing resources to grids that have experienced events, have high development potential, and suffer severe disaster consequences.

[0058] Finally, candidate grids, selected in descending order of risk priority, are chosen in numbers matching the global computing power budget threshold to generate an encrypted grid set. The remaining candidate grids are not included in this round of encryption. Thus, the generated encrypted grid set is both within the limits of the computing power budget and prioritizes grids with higher risk levels. This step, under limited computing power, enables the concentrated allocation of refined computing resources to high-risk areas, balancing the accuracy of characterizing critical areas with the timeliness of overall computation.

[0059] It should be noted that when the total number of candidate grids does not exceed the global computing power budget threshold, there is no need to perform the above risk priority sorting and selection, and all candidate grids can be directly used as the encrypted grid set; in addition, the risk priority value can also be obtained by weighted synthesis after assigning different weights to each factor.

[0060] After determining the encrypted grid set, the meteorological grids corresponding to the encrypted grid set are subjected to gust pulse reconstruction processing to obtain high-resolution driving field data. This processing specifically includes the following steps:

[0061] First, mean wind speed field data is extracted from the meteorological grid corresponding to the encrypted grid set, and the boundary layer stability corresponding to the encrypted grid set is obtained. The mean wind speed field data reflects the average wind field state preserved under the time smoothing effect of the meteorological background field; the boundary layer stability characterizes the strength of turbulence development within the atmospheric boundary layer and can be calculated from the aforementioned land surface state data and the corresponding surface heating and vertical temperature gradient information in the meteorological grid. Specifically, the boundary layer stability can be characterized using commonly used stability criteria in this field, such as the Richardson number or the Moning-Obukhov length to measure the stable, neutral, or unstable state of near-surface turbulence, and then used to look up a pre-defined fluctuation variance mapping table. Since the strength of near-surface gust fluctuations is closely related to the stability state of the boundary layer, obtaining the boundary layer stability corresponding to the encrypted grid can provide a physical basis for determining the fluctuation intensity.

[0062] Secondly, the target fluctuation variance is obtained by querying a pre-defined fluctuation variance mapping table based on boundary layer stability. This pre-defined fluctuation variance mapping table establishes a correspondence between boundary layer stability and fluctuation variance, reflecting the statistical regularity of near-surface wind speed fluctuation amplitude under different stability states. Specifically, a larger fluctuation variance corresponds to an unstable boundary layer with vigorous turbulence development, while a smaller fluctuation variance corresponds to a stable boundary layer with suppressed turbulence. This mapping table can be statistically established from historical atmospheric boundary layer turbulence observation data or large eddy simulation results. By querying this mapping table using boundary layer stability, the target fluctuation variance adapted to the current atmospheric state can be obtained. The technical advantage of this process is that it bases the determination of fluctuation intensity on the statistical regularity of boundary layer turbulence, avoiding arbitrary setting of fluctuation amplitude. As an alternative implementation, the correspondence between boundary layer stability and fluctuation variance can be determined not only through a mapping table but also through a parameterized formula.

[0063] Next, gust pulsation sampling data following a normal distribution are generated according to the target pulsation variance. This gust pulsation sampling data is then overlaid and reconstructed with the average wind speed field data over a time dimension to obtain high-resolution driving field data. One form of this overlay and reconstruction calculation is shown in the following equation:

[0064] ;

[0065] In the formula, This represents the instantaneous wind speed as it changes over time after reconstruction. To analyze the average wind speed from the meteorological grid, This is a superimposed sampled data of gusts and pulsations. This indicates that the mean is zero and the variance is the target fluctuation variance. The normal distribution is used. This formula decomposes the wind field into mean and pulsating components. The zero mean ensures that the superimposed pulsations do not change the overall magnitude of the original mean wind field, while the target pulsation variance controls the amplitude of the pulsation fluctuations, thereby reconstructing gust fluctuations that match the real atmospheric process in the time dimension. Since the sand-raising process depends on the surface friction velocity and fire behavior depends on instantaneous wind speed and heat transport, the high-resolution driving field data obtained through this process can more realistically reflect the instantaneous changes of the near-surface wind field on a fine time scale. Its technical effect is to improve the simulation fidelity of processes sensitive to instantaneous wind speed, such as sand-raising intensity and fire spread, and overcome the underestimation of the intensity of sudden events caused by time smoothing.

[0066] It is worth noting that this processing is only performed on the encrypted grid set, thus improving the temporal accuracy of the driving field in key areas while avoiding the computational burden of reconstructing the entire domain's fluctuations. As an optional implementation, the distribution form of the fluctuation components can be replaced, provided that the zero mean and predetermined variance constraints are met. The fluctuations of multiple variables such as wind speed, temperature, and humidity can also be generated collaboratively in a way that maintains their physical correlation. Furthermore, the high-resolution driving field data obtained in this step will serve as the input for the subsequent multi-hazard coupling model and, under the observation delay state in real-time early warning mode, will provide the driving basis for the aforementioned short-term extrapolation calculation at the previous time step.

[0067] See Figure 2 The diagram below illustrates the principle of on-demand reconstruction of gust pulsations provided in this application embodiment, which can be explained by comparing the wind speed time series before and after smoothing. Figure 2 The top column shows the average wind speed retained after time smoothing of the meteorological background field. Its changes over time are relatively gradual, and the information of near-ground gusts is weakened or even erased during the time smoothing process. Figure 2 The bottom column shows the reconstructed instantaneous wind speed obtained by superimposing the fluctuating component on the same mean wind speed baseline. Its satisfaction Among them, the pulsating component Follows a pattern with a mean of zero and a variance equal to the target fluctuation variance. normal distribution The reconstructed instantaneous wind speed maintains its average state while recovering the time fluctuations that are consistent with the statistical characteristics of atmospheric turbulence. The range of its fluctuation amplitude is indicated by the shaded band in the figure. Figure 2 The right side further illustrates how the target fluctuation variance is obtained: by querying a preset fluctuation variance mapping table using the boundary layer stability. This mapping table establishes a correspondence between different boundary layer stability states (strongly unstable, neutral, stable, and strongly stable) and their corresponding fluctuation variances. The more unstable the boundary layer and the more vigorous the turbulence development, the larger the corresponding fluctuation variance; conversely, the more stable the boundary layer and the more suppressed the turbulence, the smaller the corresponding fluctuation variance. Figure 2 It is evident that the high-resolution driving field data reconstructed by gust pulses can recover the instantaneous wind speed fluctuations that were erased by time smoothing, thus providing a driving force that is closer to the real atmospheric process for simulating instantaneous wind speed-sensitive processes such as sandstorm intensity and fire spread.

[0068] Finally, the high-resolution driving field data is input into the multi-hazard coupling model. Within a preset time window, a shared resource ledger conservation allocation calculation is performed on each hazard module within the multi-hazard coupling model using a loosely coupled operator splitting mechanism, generating multi-hazard concurrent simulation results. This step is the core of this method to overcome the conservation violation in multi-hazard collaborative simulation, and it specifically includes the following steps:

[0069] First, at the start of a preset time window, a shared resource ledger is initialized. This ledger records the total available resources for each hazard within the grid during that time window. Resources listed in flux within the ledger are converted to their total available allocation within that time window by integrating the duration of the time window and the grid area before being recorded. The multi-hazard coupling model contains modules corresponding to different hazards, and each module consumes or alters the energy and materials of the same surface unit during its calculations. To prevent the same surface resource from being repeatedly occupied by multiple hazards, this step centrally records all available resources for allocation within the grid during that time period in the shared resource ledger as the total amount within the preset time window.

[0070] Specifically, the total available energy registered in the shared resource ledger is obtained by integrating the net surface radiation flux over the duration of the preset time window and the grid area, representing the net radiation energy available for allocation on the surface during that period; the total available momentum is obtained by integrating the momentum flux input from the near-surface wind field to the surface over the duration of the preset time window and the grid area, representing the momentum available for allocation during that period; and the total amount of dust-raising material is the mass of dust material available for release from the surface of that grid during that period. All of the above resource totals are determined uniformly from the aforementioned land surface state data and high-resolution driving field data.

[0071] It should be further explained that the total energy, total momentum, and total amount of sand-moving material in the shared resource ledger belong to different categories of physical quantities. Their conservation and allocation are carried out independently within their respective resource categories. The demand of each disaster module for a certain type of resource is also measured by the total amount of that type of resource within the window, so that the allocation of each type of resource is completed on its own dimensional basis, and resources of different categories are not added to each other.

[0072] Secondly, each disaster-specific module independently calculates its resource requirements within a preset time window, generating corresponding resource requirement data for each module. In this stage, each module, corresponding to different disaster types, independently calculates the energy and material quantities to be consumed or changed based on its own physical processes. At this point, each module only calculates its own needs without actually occupying resources, thus separating the resource demands of each disaster type from the actual allocation of resources. This creates conditions for coordinating resource competition among multiple disaster types under unified constraints. The resource requirements calculated by each module represent the total demand of that module for the corresponding category of resources within the preset time window, using the same dimensional basis as the total available amount of that category of resources in the shared resource ledger.

[0073] Next, a preset disaster type interaction matrix is ​​obtained. Based on the resource demand data of each module and the preset disaster type interaction matrix, a one-time non-iterative proportional allocation calculation is performed on the shared resource ledger to generate allocated resource boundary data. The total resource amount in the allocated resource boundary data is rigidly constrained to be no greater than the total available resource amount in the shared resource ledger. The preset disaster type interaction matrix is ​​used to characterize the mutual promotion or mutual inhibition relationship between different disaster types in terms of resource occupation. Its matrix elements reflect the direction and degree of influence of the occurrence of one disaster type on the resource demand of another disaster type. The sign and magnitude of each element can be determined based on physical mechanisms such as fire promoting sandstorms and strong winds inhibiting fire behavior. Positive values ​​represent promotion, and negative values ​​represent inhibition. This proportional allocation calculation first uses the disaster type interaction matrix to interactively correct the original resource demand of each disaster type module to obtain the resource demand after interactive correction. One calculation form is shown in the following formula:

[0074] ;

[0075] In the formula, For the first Resource requirements for each disaster module after interactive correction. For the first The original resource requirements of each disaster type module; To represent the first type of disaster in the preset disaster interaction matrix The first type of disaster The element affected by the disaster type is represented by a positive value. The first type of disaster The resource demand for each type of disaster has a promoting effect, while negative values ​​indicate a suppressive effect; This indicates taking the larger of zero and the value within parentheses, used to ensure that the resource demand after interactive correction is always non-negative. The formula uses "1 plus the sum of the interactive effects of each disaster on this disaster" as a modulating factor for the original demand for this disaster. This ensures that when the disaster as a whole is promoted, its corrected demand increases accordingly, and when it is suppressed, it decreases accordingly. Furthermore, by using the non-negative operation, it ensures that the corrected demand will not result in a physically meaningless negative value when the suppression effect is strong.

[0076] After obtaining the resource requirements for each disaster type after cross-correction, the resources are allocated according to the normalized proportions of each corrected requirement among the disaster types. One calculation method is shown in the following formula:

[0077] ;

[0078] The above proportional allocation is performed independently for each type of resource in the shared resource ledger, where, To be assigned to the The total amount of resources for each disaster module after interactive correction. This represents the total amount of this type of resource available within a preset time window. For the first The total demand for each disaster-specific module after interactive correction, denominator This formula represents the sum of the demands for each disaster type after cross-correction, used to normalize the allocation results. Based on the promoting and inhibiting relationships among disaster types, it allocates available resources according to the proportion of each disaster type's cross-corrected demand to the total demand, ensuring that the amount of resources allocated to each disaster type is always non-negative, and that the sum of the allocated resources for each disaster type is always equal to the total available resources. This ensures that the total amount of resources does not exceed the available total, thus always satisfying the conservation constraint. Unlike methods that approximate convergence through repeated iterations, this step allocates resources in a single, non-iterative proportion within a time window, replacing the iterative solution process with explicit conservation constraints. This approach guarantees the conservation of energy and matter among multiple hazards without requiring invasive modifications to the core of the coupled framework or introducing uncertainty in iterative convergence, avoiding the repeated occupation and destruction of surface resources. Simultaneously, by imposing non-negative constraints on the interaction correction factor, it ensures that the amount of resources allocated to each hazard remains non-negative even in the presence of inhibitory interactions (such as the suppression of fire behavior by strong winds), preventing physically meaningless negative values ​​from appearing in the allocation results due to inhibitory effects. This gives the allocation process definite computational overhead and clear physical meaning.

[0079] Specifically, the values ​​of each element in the pre-defined disaster interaction matrix can be determined in the following ways: First, the sign and magnitude range of the interaction between disasters are determined based on the physical parameterization relationship between them. For example, the interaction magnitude of fire-induced strong winds promoting sand formation is determined based on the relationship between wind speed and sand-forming flux. Second, the magnitude range is constrained by historical examples of multiple disasters occurring simultaneously in the target area. Third, considering that the sample size of multiple disasters occurring simultaneously is relatively small and it is difficult to accurately determine the values ​​of each element, each element can be configured as a sensitivity parameter within the aforementioned physical magnitude range and take values ​​within that range. Thus, even when the number of co-occurrence samples is limited, the values ​​of the disaster interaction matrix can be guaranteed to have a clear physical basis and a definite value range.

[0080] Finally, the allocated resource boundary data is distributed to the corresponding disaster-type modules for model calculation, generating multi-hazard concurrent simulation results. Each disaster-type module performs model calculations with the allocated resource quota as a constraint. Because the calculation results are subject to a unified conservation allocation constraint, there is no duplicate resource occupation between them, thus reflecting the mutual promotion and inhibition effects of multiple disasters under the premise of conservation. This step improves the physical self-consistency of multi-hazard collaborative simulation, providing a reliable foundation for subsequent collaborative assessment of ecological damage. As an optional implementation, the resource types included in the shared resource ledger can be added or removed according to the types of disasters being simulated. The values ​​of each element in the disaster interaction matrix can be adjusted according to different underlying surface and climatic conditions, and the specific form of proportional allocation can also be varied while satisfying the resource total conservation constraint.

[0081] Furthermore, to reconcile conflicts between the fire size anchored by the observation site and the energy available from the shared resource ledger, the proportional allocation calculation also includes the following processing steps:

[0082] First, determine whether the energy consumption of the burned area anchored by the weighted fire point observation data is greater than the total available energy in the shared resource ledger. Since the weighted fire point observation data can be used to anchor the burned area of ​​a fire, and the corresponding energy consumption can be calculated from the burned area, if it exceeds the total available energy, directly adopting the fire scale indicated by the observations will exceed the energy conservation boundary of the grid. Second, if it is greater, then based on the preset model uncertainty variance and the preset observation uncertainty variance, a weighted fusion calculation is performed on the model-calculated burned area and the observed burned area to obtain the scaled burned area. This step does not simply discard the observation information by truncation, but rather performs a weighted fusion of the two based on the uncertainty of each of the model estimate and the observation estimate. One calculation form is shown in the following formula:

[0083] ;

[0084] In the formula, To reduce the area affected by the fire, Calculate the burned area for the model. To observe the burned area, The pre-defined model uncertainty variance characterizes the reliability of the model's calculation of the burned area. To predetermine the observation uncertainty variance, which characterizes the reliability of the observed burned area, both can be determined by the model's historical backcalculation error statistics and the accuracy index of the observation products, respectively. The design logic of this formula lies in using the reciprocals of their respective uncertainty variances as weights for fusion, giving greater weight to the side with lower uncertainty and higher reliability in the fusion result. This allows for a burned area estimate that balances model and observation within the framework of conservation constraints. This processing avoids directly discarding observational information due to observations exceeding conservation boundaries, ensuring that the adjusted burned area falls within the conservation-allowed range while maximizing the preservation of fire scale information carried by the observations.

[0085] After obtaining the scaled-down fire area, a positive correlation function between combustion efficiency and total available energy is constructed. Based on this positive correlation function and the scaled-down fire area, scaled emission source strength data is calculated. Since fires often exhibit incomplete combustion under energy constraints, and combustion efficiency decreases with decreasing available energy, this step establishes a positive correlation between combustion efficiency and total available energy. The scaling of the area is explained by the physical mechanism of incomplete combustion under energy constraints. The calculation form of the scaled emission source strength data is shown in the following formula:

[0086] ;

[0087] In the formula, For the first The scaling emission source strength of the emissions, To reduce the area affected by the fire, Combustible material load per unit area For combustion efficiency, For the first Emission factors of various pollutants This is a positive correlation function between combustion efficiency and total available energy. The design logic of this formula is that it does not treat the scaling of the burned area in isolation for emission calculations, but rather simultaneously reduces combustion efficiency in tandem with the total available energy, thereby making the resulting emission source strength physically consistent with the constrained energy state. The above process overcomes the emission calculation distortion that may be caused by unilateral scaling of the burned area, maintaining physical consistency between emission source strength and energy constraints.

[0088] In the above formula, the combustible material load per unit area It can be determined by the type and load of combustibles in the target area; emission factors Values ​​corresponding to the type of combustible material and the type of emissions can be obtained from publicly available emission factor databases in this field; the positive correlation function between combustion efficiency and total available energy. It can adopt a function that increases monotonically with the total available energy, such as a linear or piecewise increasing relationship, so that the combustion efficiency decreases accordingly as the total available energy decreases, but does not exceed its upper limit of complete combustion under sufficient energy conditions.

[0089] Finally, the scaled emission source strength data is updated in the allocated resource boundary data to align the fire scale and emissions, anchored by observations and adjusted for conservation constraints, with the energy constraints in the shared resource ledger, thereby maintaining the energy conservation state among the multi-hazard modules. As an optional implementation, the fusion weights between model estimates and observation estimates, as well as the positive correlation between combustion efficiency, can be represented using other appropriate methods.

[0090] In addition, to ensure the numerical stability of long-term simulations, the conservation allocation calculation also includes the quantitative monitoring and control of conservation deviations.

[0091] First, during the data transfer process between different disaster modules, an area conservation interpolation algorithm is used for grid mapping transformation, and the conservation residuals within each preset time window are calculated. Since different modules corresponding to different disasters may use different grid divisions, grid mapping transformation is required when transferring data between modules. If the transformation process does not meet the conservation requirements, it will cause leakage or false increase of energy and matter during the transfer. Therefore, an area conservation interpolation algorithm is used to keep the total amount of physical quantities before and after mapping consistent, and the deviation between the actual allocation result and the conservation requirements within each preset time window is calculated accordingly, i.e., the conservation residual. The area conservation interpolation is a conventional grid mapping method used in this field to maintain the conservation of the total amount of physical quantities.

[0092] Secondly, the cumulative residuals are obtained by summing up multiple conservation residuals within a preset statistical period to show the cumulative trend of conservation deviation over a long period of time.

[0093] Furthermore, if the cumulative residual is detected to exceed the preset residual boundary threshold, residual out-of-bounds alarm data is generated, and the shared resource ledger is reset. The preset residual boundary threshold defines the maximum range allowed for the cumulative residual. When the cumulative residual exceeds this threshold, it indicates that the conservation state is abnormal. At this time, the shared resource ledger is reset to return the resource baseline to the initial conservation state, thereby constraining the conservation residual to be bounded and avoiding systematic drift of conservation bias.

[0094] Furthermore, if the cumulative residual is detected to be less than the preset residual boundary threshold, the state change of the allocated resource boundary data is limited to the preset single-window amplitude threshold data. The preset single-window amplitude threshold data defines the maximum allowable change of the allocated resource boundary data within a single time window. Limiting the state change to this amplitude threshold allows the resource boundary to transition smoothly between adjacent windows, thereby suppressing numerical oscillations caused by abrupt changes in resource allocation results.

[0095] Finally, if multiple conservation residuals are found to have consistent signs within a preset observation period, diagnostic label data representing a structural bias between the model and observations is generated and output. Under normal conservation conditions, conservation residuals should typically fluctuate between positive and negative without exhibiting a fixed bias. If multiple conservation residuals consistently maintain consistent signs, it indicates a systematic tendency of conservation bias in a single direction, often suggesting a structural bias between the model and observations rather than random numerical errors. Generating and outputting diagnostic label data based on this distinguishes structural bias from random numerical errors, providing guidance for further investigation of systematic problems in model parameters or observation inputs, rather than silently absorbing such biases into the resource ledger. All the aforementioned thresholds and observation period lengths can be adjusted according to different emphases on stability and sensitivity.

[0096] After generating the multi-hazard concurrent simulation results, the method also includes a collaborative assessment of ecological damage. First, independent ecological damage indicators for each individual hazard are extracted from the multi-hazard concurrent simulation results. These independent ecological damage indicators include at least carbon loss calculation data and soil degradation calculation data. The multi-hazard concurrent simulation results generated by the aforementioned conservation allocation contain the simulation states of each hazard under conservation constraints. From these, carbon loss calculation data characterizing the loss of vegetation carbon storage caused by events such as fires, and soil degradation calculation data characterizing the degree of soil degradation caused by events such as dust storms, are extracted. Second, a collaborative damage weight matrix is ​​obtained. This matrix contains interval sensitivity parameters characterizing the physical interactions between hazard types. Considering the relatively scarce interaction samples under multi-hazard co-occurrence scenarios and the difficulty in precise calibration, these interaction parameters are given in the form of interval sensitivity parameters. That is, the direction and magnitude of the interaction between hazard types are characterized by a range of values ​​rather than a single fixed value, thus objectively reflecting the inherent uncertainty in its calibration. Next, the synergistic ecological damage index is generated by linearly weighting multiple independent ecological damage indicators and summing the products of their cross terms using a synergistic damage weight matrix. One calculation form of this index is shown in the following formula:

[0097] ;

[0098] In the formula, To coordinate the ecological damage index, For the first Independent ecological damage indicators for a single disaster type For the first The weight of damage caused by a single disaster. For the first The types of disasters and the first The product of the cross terms of damage from each type of disaster, To characterize the first The types of disasters and the first The interval sensitivity parameter of the physical interaction between disaster types. The formula summarizes the independent damage contribution of each disaster type with a linear weighting term, and then uses the product summation term of the cross term to characterize the additional damage caused by the interaction between each disaster type. This makes the comprehensive damage measurement not only cover the individual effects of each disaster type, but also reflect the amplification effect caused by the synergy of disaster types.

[0099] Among them, the weight of damage for each type of disaster The values ​​can be pre-set based on the relative severity of damage from each type of disaster or the specific assessment focus. Before participating in the above weighted and cross-term calculations, the independent ecological damage indicators for each single disaster type can be normalized or otherwise dimensionless processed to ensure the comparability of each indicator in the comprehensive measurement.

[0100] Since the interaction parameters in the cross terms are given in the form of interval sensitivity parameters, the generated synergistic ecological damage index is correspondingly presented as a range of values ​​corresponding to that interval. Finally, based on the synergistic ecological damage index, a comprehensive assessment report with confidence interval indicators is generated and output; since the synergistic ecological damage index is presented as a range of values, the comprehensive assessment report generated accordingly includes confidence interval indicators to indicate the range of values ​​for the assessment results rather than a single, definitive value.

[0101] Based on this step, the assessment conclusion clearly conveys the inherent uncertainty of the results while stating the degree of damage, avoiding exaggerating the accuracy of the assessment with a single definite value, thus facilitating the rational use of the assessment results in actual decision-making. As an optional implementation, the unified processing method of the dimensions of each single disaster damage index and the presentation method of the comprehensive assessment report can be selected from different forms such as graphics, tables, or text, depending on the application needs.

[0102] Through the above description of the implementation methods, those skilled in the art can clearly understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the above functions can be divided into different functional modules to complete all or part of the functions described above.

[0103] In the embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the embodiments described above are merely illustrative; for instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or indirect coupling or communication connection between devices or units, and may be electrical, mechanical, or other forms.

[0104] The units described as separate components may or may not be physically separate. A component shown as a unit can be one or more physical units, located in one place or distributed in multiple different locations. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0105] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0106] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, or the parts that contribute to the solution, or all or part of the technical solution, can be embodied in the form of a software product. This software product is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0107] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A multi-hazard land-atmosphere coupled model co-simulation method, applied to multi-hazard concurrent simulation of a target area with fire points within a preset time window, characterized in that, Includes the following steps: Acquire meteorological grids, land surface parameters, and multi-source observation fire point data for the target area; The meteorological grid and the land surface parameters are used to perform land surface energy balance calculations to generate land surface state data. The multi-source fire point observation data are subjected to confidence weighting to generate weighted fire point observation data; The multi-hazard trigger margin, which characterizes the trigger distance of sudden ecological damage events in the target area, is calculated based on the meteorological grid and the land surface state data, and the multi-hazard trigger margin is updated by Bayesian fusion based on the weighted fire point observation data. Based on the updated multi-hazard trigger margin, a set of encrypted grids that meet the preset trigger conditions is determined from the meteorological grid; The meteorological grid corresponding to the encrypted grid set is subjected to gust pulse reconstruction processing to obtain high-resolution driving field data; The high-resolution driving field data is input into the multi-hazard coupling model. Within the preset time window, the shared resource ledger of each hazard module in the multi-hazard coupling model is conserved and allocated through a loosely coupled operator splitting mechanism to generate multi-hazard concurrent simulation results.

2. The multi-hazard land-atmosphere coupled model co-simulation method as described in claim 1, characterized in that: The calculation of the multi-hazard trigger margin, based on the meteorological grid and the land surface state data, characterizing the trigger distance of sudden ecological damage events in the target area, includes: Meteorological characteristic variables are extracted from the meteorological grid, and surface characteristic variables are extracted from the land surface state data; Obtain a preset historical fire dataset, and perform model fitting and calibration based on the preset historical fire dataset to obtain preset threshold parameters; The meteorological characteristic variables and the surface characteristic variables are input into a preset normalized weighted distance function, and aggregated and calculated in combination with the preset threshold parameters to output the multi-hazard trigger margin; wherein, the smaller the trigger distance value represented by the multi-hazard trigger margin, the higher the trigger probability of the target grid area.

3. The multi-hazard land-atmosphere coupled model co-simulation method as described in claim 1, characterized in that: The step of updating the multi-hazard trigger margin using Bayesian fusion based on the weighted fire point observation data includes: The multi-hazard trigger margin is used as the prior margin data. The weighted fire point observation data and the preset historical hit rate are combined to perform a penalty fusion calculation to obtain the fused trigger margin data with a reduced value. Obtain the multi-hazard trigger margin of the previous time step, and compare the fused trigger margin data with the boundary threshold of the preset dual-threshold hysteresis interval; If the comparison result indicates that the fusion trigger margin data is within the preset dual-threshold hysteresis interval, then a preset relaxation calculation is performed on the multi-hazard trigger margin of the previous time step, and the calculation result is used as the updated multi-hazard trigger margin to suppress trigger state jitter.

4. The multi-hazard land-atmosphere coupled model co-simulation method as described in claim 1, characterized in that: The step of determining a set of encrypted grids that meet preset triggering conditions from the meteorological grid based on the updated multi-hazard triggering margin includes: Identify candidate grids whose updated multi-hazard trigger margin is not greater than a preset trigger threshold; Obtain the global computing power budget threshold and count the total number of grids corresponding to all the candidate grids; If the total number of grids is greater than the global computing power budget threshold, then the risk priority value of each candidate grid is calculated based on the weighted fire point observation data, the preset spread risk data, and the preset exposure vulnerability data. The number of candidate grids matching the global computing power budget threshold is selected according to the risk priority values ​​from largest to smallest to generate the encrypted grid set.

5. The multi-hazard land-atmosphere coupled model co-simulation method as described in claim 1, characterized in that: The step of performing gust pulse reconstruction processing on the meteorological grid corresponding to the encrypted grid set to obtain high-resolution driving field data includes: The average wind speed field data is parsed from the meteorological grid corresponding to the encrypted grid set, and the boundary layer stability corresponding to the encrypted grid set is obtained. Based on the boundary layer stability, a preset pulsation variance mapping table is queried to obtain the target pulsation variance; Based on the target fluctuation variance, generate gust fluctuation sampling data that follows a normal distribution. Then, perform time-dimensional superposition and reconstruction calculation on the gust fluctuation sampling data and the average wind speed field data to obtain the high-resolution driving field data.

6. The multi-hazard land-atmosphere coupled model co-simulation method as described in claim 1, characterized in that: The step of performing confidence-weighted processing on the multi-source observed fire point data to generate weighted fire point observation data includes: Identify the timestamps corresponding to the multi-source observed fire point data; Determine the current system operating mode. If the current operating mode is real-time early warning mode and the timestamp is in the observation delay state relative to the current simulation time, then obtain the high-resolution driving field data of the previous time step, and use the preset fire spread model and the high-resolution driving field data of the previous time step to perform short-term extrapolation calculation to generate short-term extrapolation prior fire point data. Before the arrival of the multi-source observed fire point data corresponding to the timestamp, the short-term extrapolated prior fire point data is output as the weighted fire point observation data; After the multi-source observation fire point data arrives, the cloud pollution factor, radiation power factor, and time persistence factor of the multi-source observation fire point data are extracted. Based on the extracted three factor data, the confidence multiplication of the multi-source observation fire point data is calculated, and the calculation result is overwritten as the new weighted fire point observation data.

7. The multi-hazard land-atmosphere coupled model co-simulation method as described in claim 1, characterized in that: The method involves performing a conservation allocation calculation of the shared resource ledger for each disaster module within the multi-hazard coupling model using a loosely coupled operator splitting mechanism, generating multi-hazard concurrent simulation results, including: At the start of the preset time window, the shared resource ledger is initialized and established; Each disaster-related module independently calculates its resource requirements within the preset time window, generating corresponding resource requirement data for each module. Obtain a preset disaster type interaction matrix. Based on the resource requirement data of each module and the preset disaster type interaction matrix, perform a one-time non-iterative proportional allocation calculation on the shared resource ledger to generate allocated resource boundary data. The total amount of resources in the allocated resource boundary data is hard-constrained to be no greater than the total amount of available resources in the shared resource ledger. The allocated resource boundary data is sent to the corresponding disaster module for model calculation, generating the multi-disaster concurrent simulation results.

8. The multi-hazard land-atmosphere coupled model co-simulation method as described in claim 7, characterized in that: The step of performing a one-time non-iterative proportional allocation calculation on the shared resource ledger based on the resource requirement data of each module and the preset disaster interaction matrix, and generating allocated resource boundary data, includes: Determine whether the energy consumption of the burned area anchored by the weighted fire point observation data is greater than the total available energy in the shared resource ledger; If it is greater than the preset model uncertainty variance and the preset observation uncertainty variance, the model-calculated fire area and the observed fire area are weighted and fused to obtain the scaled fire area. A positive correlation function between combustion efficiency and the total available energy is constructed, and scaled emission source strength data is calculated based on the positive correlation function and the scaled burn area. Update the scaled emission source strength data to the allocated resource boundary data.

9. The multi-hazard land-atmosphere coupled model co-simulation method as described in claim 7, characterized in that: The step of performing conservation allocation calculations of shared resource ledgers for each disaster module within the multi-hazard coupling model through a loosely coupled operator splitting mechanism to generate multi-hazard concurrent simulation results also includes: During the data transfer process between the various disaster modules, an area conservation interpolation algorithm is used for grid mapping transformation, and the conservation residuals within each preset time window are calculated. The cumulative residual is obtained by summing up the multiple conserved residuals within a preset statistical period. If the cumulative residual is detected to exceed the preset residual boundary threshold, residual out-of-bounds alarm data is generated, and the shared resource ledger is reset; If the cumulative residual is detected to be less than the preset residual boundary threshold, the state change of the allocated resource boundary data is limited to the preset single-window amplitude threshold data. If the positive and negative signs of multiple conserved residuals are found to be consistent within a preset observation period, diagnostic label data indicating a structural deviation between the representation model and the observation is generated and output.

10. The multi-hazard land-atmosphere coupled model co-simulation method as described in claim 1, characterized in that: After generating the multi-hazard concurrent simulation results, the method further includes: Independent ecological damage indicators for each single hazard are extracted from the multi-hazard concurrent simulation results, wherein the independent ecological damage indicators include at least carbon loss measurement data and soil degradation measurement data; Obtain the collaborative damage weight matrix, which contains interval sensitivity parameters characterizing the physical interaction between disaster types; The synergistic ecological damage index is generated by linearly weighting and summing the products of multiple independent ecological damage indicators using the synergistic damage weight matrix; Based on the aforementioned collaborative ecological damage index, a comprehensive assessment report with confidence interval markers is generated and output.

Images