CN-121683295-B - Forest fire situation sensing system and method based on optical active and passive remote sensing data

CN121683295BCN 121683295 BCN121683295 BCN 121683295BCN-121683295-B

Abstract

The application provides a forest fire situation sensing system and a forest fire situation sensing method based on optical active and passive remote sensing data, which relate to the technical field of natural disaster management, wherein the system comprises an optical active remote sensing subsystem, a remote sensing system and a remote sensing system, wherein the optical active remote sensing subsystem is used for acquiring real-time atmospheric observation data above a target area; the system comprises an optical passive remote sensing subsystem, a ground processing and modeling unit, a ground modeling unit and a ground modeling unit, wherein the optical passive remote sensing subsystem is used for acquiring real-time ground surface observation data of a target area, the ground processing and modeling unit comprises a processor and a memory, the processor is configured to construct an augmentation state vector in the memory, generate a predicted value based on an ensemble of an augmentation state vector operation coupling model, calculate residual errors between the real-time observation data and the predicted value, update the vector, and generate a final product based on the updated vector operation coupling model. And the multi-mode cooperative data of the optical active remote sensing technology and the optical passive remote sensing technology are utilized to synchronously estimate and adaptively optimize the internal key physical parameters and the change trend of the internal key physical parameters of a coupling physical model, so that the prediction performance of the model is improved.

Inventors

XIE CHENBO
CHEN JIANFENG
JI JIE
LU JIE
ZHANG ZICHEN
NI YUGUO

Assignees

中国科学院合肥物质科学研究院

Dates

Publication Date: 20260508
Application Date: 20260210

Claims (3)

1. Forest fire situation awareness system based on optical active and passive remote sensing data is characterized by comprising: The optical active remote sensing subsystem comprises at least one Doppler anemometry laser radar and is used for acquiring real-time atmospheric observation data above a target area, wherein the real-time atmospheric observation data comprises a turbulence kinetic energy profile or a vertical velocity variance; the optical passive remote sensing subsystem comprises a satellite remote sensing platform and a real-time ground surface monitoring subsystem, wherein the satellite remote sensing platform is used for acquiring real-time ground surface observation data of a target area, and the real-time ground surface observation data comprises real-time live wire positions calculated by utilizing thermal infrared and short-wave infrared wave bands; a ground processing and modeling unit comprising a processor and a memory, the processor configured to perform the steps of: Constructing an augmentation state vector in a memory, and operating an ensemble of a coupling model based on the augmentation state vector to generate a model predicted value, wherein the augmentation state vector comprises a dynamic state variable of the coupling model, an adjustable physical parameter of at least one atmospheric physical module, an adjustable physical parameter of at least one earth surface physical module and a parameter variation trend item of at least one adjustable physical parameter, the coupling model at least comprises one atmospheric physical module and one earth surface physical module, the coupling model comprises a WRF-Fire model comprising the atmospheric module and a Fire behavior module, the adjustable parameter of the atmospheric physical module at least comprises a turbulent mixing length or a TKE closure constant in a planetary boundary layer parameterization scheme, the parameter variation trend item corresponding to the adjustable parameter of the atmospheric physical module comprises a time variation rate of the turbulent mixing length, the adjustable parameter of the earth surface physical module at least comprises a balanced water content deviation item of a combustible water content time lag model, and the parameter variation trend item corresponding to the adjustable parameter of the earth surface physical module comprises a time variation rate of the balanced water content deviation item; Calculating residual errors between the real-time atmospheric observation data and the real-time surface observation data and the predicted value, and updating the augmented state vector according to the residual errors by using a data assimilation algorithm; And generating a forest fire situation awareness product based on the updated augmentation state vector operation coupling model.
2. A forest fire situation sensing method based on optical active and passive remote sensing data is characterized by comprising the following steps of: An augmented state vector is constructed in a memory, a coupling model is operated based on the augmented state vector to generate a model predicted value, wherein the augmented state vector comprises a dynamic state variable of the coupling model, adjustable physical parameters of at least one atmospheric physical module, adjustable physical parameters of at least one earth surface physical module and parameter variation trend items of at least one adjustable physical parameter, the coupling model at least comprises one atmospheric physical module and one earth surface physical module, the coupling model comprises a WRF-Fire model comprising the atmospheric module and a Fire behavior module, the adjustable parameters of the atmospheric physical module at least comprise turbulent flow mixing length or TKE closing constant in a planetary boundary layer parameterization scheme, the parameter variation trend items corresponding to the adjustable parameters of the atmospheric physical module comprise time variation rates of turbulent flow mixing length, the parameter variation trend items corresponding to the adjustable parameters of the earth surface physical module at least comprise balance water content deviation items of a combustible water content time lag model, and the parameter variation trend items corresponding to the adjustable parameters of the earth surface physical module comprise time variation rates of the balance water content deviation items; Acquiring real-time atmospheric observation data and real-time earth surface observation data, calculating residual errors between the real-time atmospheric observation data and the real-time earth surface observation data and a predicted value, and updating an augmentation state vector according to the residual errors by using a data assimilation algorithm, wherein the updated augmentation state vector comprises updated adjustable physical parameters and corresponding parameter change trend items; And generating a forest fire situation awareness product based on the updated augmentation state vector operation coupling model.
3. The method of claim 2, wherein generating a forest fire situation awareness product based on the updated augmented state vector running coupling model comprises: extracting the updated adjustable physical parameters and corresponding parameter change trend items of the updated augmented state vector; Constructing a dynamic parameter prediction model according to the updated adjustable physical parameters and the corresponding parameter variation trend items; And injecting the dynamic parameter model into the coupling model to generate a self-adaptive optimized coupling model, and generating a forest fire situation sensing product based on an output result of the self-adaptive optimized coupling model.

Description

Forest fire situation sensing system and method based on optical active and passive remote sensing data Technical Field The application relates to the technical field of natural disaster management, in particular to a forest fire situation sensing system and method based on optical active and passive remote sensing data. Background Forest fire situation awareness is the core in Lin Huoguan and requires the commander to have accurate grasp on the current state (awareness), development mechanism (understanding) and future trend (prediction) of the fire scene. However, the prior art has two fundamental technical difficulties in achieving high-fidelity situation awareness: 1. the first difficulty is the fatal defect of the static combustible database The predictive capabilities of a business Fire behavior model (e.g., WRF-Fire) depend to a large extent on the accuracy of the surface combustible data. The update period of the combustible database (such as LANDFIRE) widely used at present is as long as a few years, so that the information (such as combustible type, loading capacity and water content model) in the database is seriously disjointed with the dynamic reality (such as drought and insect damage) of the forest ecological system. 2. Two difficulties are out of control of the fire-gas coupling feedback loop The second major source of uncertainty in forest fire prediction is the bi-directional coupling feedback between the fire and the atmosphere. Large forest fires release large amounts of heat, producing strong updraft and "fire wind". While theoretically coupled with an atmospheric module, existing Fire behavior models (such as WRF-Fire) typically have their initial field and boundary conditions from standard mesoscale weather predictions, whose time-space resolution is far insufficient to resolve the severe atmospheric changes in the Fire site. Meanwhile, WRF built-in planetary boundary layer parameterization schemes (e.g., YSU, MYNN) are designed for normal weather, which often underestimate fire-driven strong turbulent mixing. To solve the above-mentioned problems, those skilled in the art have attempted to introduce real-time observation data into a model using data assimilation techniques. For example: (1) State correction-the live position (FRP or live perimeter) observed by the satellite, the live state predicted by the data assimilation direct "correction" or "override" model (e.g. Level-Set function LFN in WRF-Fire). (2) Atmospheric constraints-parameters observed by active/passive techniques (e.g., U, V, W), atmospheric state variables of the model are "corrected" by data assimilation. Existing solutions are limited from "state correction" to "static parameter estimation". This "state correction" method has the fundamental disadvantage that it only corrects the "symptoms" predicted by the model (e.g. live position LFN errors) and does not address the "etiology" of the errors, i.e. systematic deviations of the model internal physical parameters (e.g. the model parameters of the water content of combustibles or PBL turbulence parameters). More advanced techniques in the art solve this problem by "parameter estimation", i.e. correction of the value of the parameter itself by means of the observation residual. However, this "static parameter estimation", while superior to "state correction", has the fundamental disadvantage that it assumes that the optimized parameters are constant between assimilations. This assumption is wrong in the case of strong fire-gas coupling. Fire is a highly dynamic process in that the fire itself actively and continuously changes its surrounding environmental parameters. For example, heat radiation from a fire can continue to dry the combustible material ahead, resulting in a continuous decrease in the moisture content (FMC) parameter of the combustible material, and a large heat flux from a fire can continue to enhance localized turbulence, resulting in a continuous increase in the turbulent mixing length parameter of the PBL. The existing "static parameter estimation" cannot capture such "parameter variation trend" driven by the fire itself, so its "prediction skill" is still limited. Disclosure of Invention Based on the above, it is necessary to provide a forest Fire situation sensing system and method based on optical active and passive remote sensing data, which utilize multi-mode cooperative data of optical active remote sensing (especially atmospheric detection laser radar) and optical passive remote sensing (especially satellite imaging) technologies to synchronously estimate and adaptively optimize internal key physical parameters and variation trends thereof for a coupled atmospheric-Fire behavior physical model (such as WRF-Fire), so as to improve the prediction performance of the model, thereby generating a situation sensing product with high fidelity for operational decisions. In a first aspect, the present application provides a forest fire situation aware