A forest fire spreading automatic prediction method fusing laser radar detection and micro-meteorological determination

Abstract

The present application belongs to the field of meteorological safety technology, and discloses a forest fire spreading automatic prediction method fusing laser radar detection and micro-meteorological determination. The method realizes high space-time resolution real-time monitoring of forest fire points through a laser radar system, and reverses micro-meteorological elements such as particle concentration, wind speed and wind direction around the fire point based on radar echo data, forming a real-time micro-meteorological determination mechanism. The determination result automatically drives WRF-SFIRE numerical fire spreading simulation, and combines high-resolution numerical weather prediction data to automatically predict and dynamically update the fire spreading trend. The overall closed-loop intelligent prediction process of "fire point detection-micro-meteorological determination-automatic triggering simulation-trend prediction output" is constructed, which significantly improves the early detection and spreading trend prediction ability of forest fires.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of meteorological safety, and relates to an automatic forest fire spreading prediction method combining laser radar detection and micro-meteorological determination. BACKGROUND

[0002] With the continuous intensification of global climate change, the earth's climate system shows a warming trend, and the frequency and intensity of extreme weather events are increasing. The interweaving of high temperature, drought, and enhanced wind conditions creates a favorable environment for the frequent occurrence and spread of forest fires. Forest fires, as a strong natural disaster, not only directly damage forest resources and the ecological environment, but also have a profound impact on human life safety, regional air quality, and even global carbon cycling.

[0003] Forest fires release a large amount of smoke and harmful gases, causing air pollution and health risks, as well as triggering secondary ecological disasters such as soil erosion and water resource deterioration. Fire smoke can be transported through the atmosphere, affecting the environment and public health across regions and even across national boundaries, becoming an important cross-border environmental risk factor under the background of global climate change. Therefore, improving the early detection capability and accurate prediction of forest fire spreading trends is an important part of the disaster prevention and reduction system and ecological security strategy.

[0004] Despite the continuous progress of modern remote sensing technology and meteorological numerical prediction systems, the monitoring and prediction of forest fires still face many challenges. On the one hand, the suddenness and locality of forest fires require rapid and accurate real-time monitoring; on the other hand, the spread of forest fires is highly dependent on complex and variable local meteorological conditions and terrain vegetation factors, requiring fine physical process simulation capabilities and future trend prediction capabilities. The current technology still cannot effectively integrate observation data, meteorological determination, and numerical simulation, limiting the timeliness and accuracy of fire warning.

[0005] The occurrence and spread of forest fires are significantly influenced by local micro-meteorological conditions, with factors such as wind speed, wind direction, temperature, and humidity directly affecting the spread path and speed of the fire. Most existing fire monitoring systems use empirical models such as the Fire Weather Index (FWI) to assess fire danger levels by integrating ground-based meteorological factors. However, these models are essentially empirical statistical methods with several key limitations. First, FWI and similar indices reflect fire potential rather than actual fire conditions, lacking quantitative predictions of actual fire spread area and intensity. Second, ground-based observations of meteorological factors are difficult to reflect the spatial heterogeneity of micro-meteorological conditions around the fire point and cannot capture the dynamic meteorological environment at the fire site in real time. Third, traditional models lack physical mechanisms to support large-scale atmospheric circulation systems, making it difficult to explain the meteorological dynamics of extreme fire events. In addition, fire weather index models often ignore the spatial and temporal accumulation and lag effects of meteorological factors, lacking the ability to predict long-term trends and future climate change scenarios.

[0006] In terms of micro-meteorological data acquisition, traditional methods rely on meteorological observation stations and numerical prediction products, which are limited by uneven spatial distribution and temporal resolution, making it difficult to meet the high time efficiency and spatial detail requirements of fire spread. As an active remote sensing technology, LiDAR (Light Detection and Ranging) has the advantages of all-weather, high spatial resolution, and fast response, enabling three-dimensional fire point monitoring and high-frequency inversion of micro-meteorological fields (especially wind speed and direction) in forest areas, providing new technical means for dynamic fire monitoring and environmental determination.

[0007] However, the integration of LiDAR data and fire numerical simulation is still not perfect, lacking automatic fire point recognition and triggering simulation mechanisms, limiting its application potential in intelligent fire prediction systems. Specifically:

[0008] (1) Fire point recognition technology based on satellite remote sensing

[0009] Satellite remote sensing is one of the main means of forest fire monitoring, especially relying on mid-infrared and thermal infrared bands to identify fire points and intensity. For example, MODIS (Moderate Resolution Imaging Spectrometer) and VIIRS (Visible Infrared Imaging Radiometer Suite) satellite products have been widely used in global fire monitoring. These systems use thermal anomaly detection algorithms to locate fire sources and scale.

[0010] Advantages: wide coverage, capable of monitoring large-scale and hard-to-reach areas; and transparent data, easy to integrate and analyze.

[0011] The shortcomings are: the satellite revisit cycle limits the timeliness of monitoring, especially in the early stages when small-scale fires are difficult to capture in time; the spatial resolution is limited, making it difficult to accurately locate specific fires; clouds, smoke, and meteorological conditions have a significant impact on the observation results; and the lack of integration with real-time micro-meteorological data makes it impossible to dynamically determine the conditions for fire spread.

[0012] (2) Fire risk assessment based on empirical statistical models

[0013] The Fire Weather Index (FWI) system forms the basis of fire risk assessments in many countries worldwide. By integrating meteorological data such as temperature, relative humidity, wind speed, and precipitation, it estimates fuel moisture content and fire spread potential. The FWI and its derivative indices quantify fire risk levels, supporting daily fire risk warnings.

[0014] Despite its ease of use and computational simplicity, the FWI system is limited to risk assessment, cannot directly simulate the fire development process, and has limited predictive timeliness, making it difficult to adapt to dynamic changes in fire conditions. It relies on meteorological station observation data, which has uneven spatial coverage and fails to reflect local meteorological changes around the fire point. The model does not consider the influence of atmospheric circulation and long-term meteorological trends, lacking modeling of the physical evolution of fire processes. (Van Wagner, CE (1987). Development and structure of the Canadian Forest Fire Weather Index System.)

[0015] (3) Fire prediction based on machine learning and statistical methods

[0016] In recent years, scholars have attempted to use statistical regression and machine learning techniques, combined with multi-source meteorological and vegetation data, to predict the frequency and area of ​​fires. For example, methods such as random forests, support vector machines, XGBoost, and long short-term memory neural networks (LSTM) have been used for fire risk modeling.

[0017] These methods improve prediction accuracy through big data analysis, but still face challenges such as reliance on near-ground weather station data, difficulty in capturing complex microclimate environments around fire points, insufficient time matching between input variables and response variables, affecting model robustness, short-term prediction orientation, difficulty in extrapolating seasonal or interannual trends, limited understanding of physical mechanisms of meteorological elements, and difficulty in providing explanatory fire development mechanisms. (Abatzoglou, J.T., & Williams, A.P. (2016). Impact of anthropogenic climate change on wildfire across western US forests. PNAS. and Di Giuseppe, F., et al. (2025). Global data-driven prediction of fire activity. Nature Communications.)

[0018] (4) Fire spread simulation based on physical numerical model

[0019] WRF-SFIRE is a fire spread simulation tool based on the WRF (Weather Research and Forecasting) model, integrating fire physical processes and atmospheric dynamics, capable of simulating fire spread path, speed, and interaction between fire and weather. Its advantages are complete physical process simulation, high spatial and temporal resolution.

[0020] However, the application of WRF-SFIRE is still limited by the following factors: fire source initialization often relies on manual input, lacking real-time automatic triggering mechanism; high demand for high-resolution meteorological driving data, limiting real-time application; insufficient coupling with fire point detection and microclimate observation system, difficult to form an automatic early warning closed loop. (Mandel, J., Beezley, J.D., & Kochanski, A.K. (2011). Coupled atmosphere-wildland fire modeling with WRF 3.3 and SFIRE 2011. Geoscientific Model Development.)

[0021] (5) Application of lidar technology in fire monitoring and microclimate determination

[0022] As an active remote sensing technology, laser radar can achieve high spatial resolution real-time monitoring of forest fire points. Through Doppler laser radar technology, local particulate matter concentration, wind speed, wind direction and other micro-meteorological parameters can be inverted to reflect the meteorological conditions around the fire point, providing high spatial and temporal resolution data for fire spread environment determination.

[0023] Currently, laser radar has shown high sensitivity and timeliness advantages in fire monitoring, but the automatic integration with numerical simulation systems is still immature, and most of them are offline or semi-automatic processing, lacking real-time triggering mechanism.

[0024] Reference:

[0025] In summary, the prior art has the following deficiencies in several key links:

[0026] (1) Limited fire point detection timeliness and accuracy: satellite remote sensing is lagging and lacks resolution, traditional ground observation is insufficient, laser radar has obvious advantages, but an automatic early warning system has not yet been formed.

[0027] (2) Insufficient integration of micro-meteorological data: lack of high spatial and temporal resolution particulate matter concentration, wind speed, and wind direction data based on radar inversion, which cannot accurately determine the meteorological conditions of fire spread.

[0028] (3) Lack of automatic triggering numerical simulation mechanism: WRF-SFIRE and other physical models lack real-time automatic starting mechanism and rely on manual input of fire source, affecting practicality.

[0029] (4) Early warning process is fragmented and lacks closed-loop intelligence: observation, determination, simulation, and warning are not effectively coupled, making it difficult to achieve rapid response and trend dynamic prediction.

[0030] Based on the above background, the present application innovatively integrates laser radar detection technology with micro-meteorological determination based on radar inversion to build an integrated triggering mechanism for automatic identification of fire points and real-time determination of micro-meteorology, automatically coupling WRF-SFIRE numerical fire simulation to achieve rapid, accurate, and automated prediction of fire spread, significantly improving forest fire monitoring and early warning capabilities, filling the gaps in existing technologies, and having important application value and promotion prospects. SUMMARY

[0031] The application provides an automatic forest fire spreading prediction method fusing laser radar detection and micro-meteorological judgment.

[0032] The technical scheme of the application is as follows:

[0033] The automatic forest fire spreading prediction method fusing laser radar detection and micro-meteorological judgment is realized through the following modules:

[0034] (1) a laser radar fire point real-time detection module;

[0035] (2) a micro-meteorological judgment module based on radar inversion;

[0036] The micro-meteorological judgment module based on radar inversion relies on the echo Doppler shift data captured by the laser radar system in the sweeping process, and obtains the particle concentration, near-ground wind speed and wind direction around the fire point through an inversion algorithm;

[0037] (3) an automatic triggering WRF-SFIRE numerical simulation module;

[0038] The automatic triggering WRF-SFIRE numerical simulation module integrates WRF-SFIRE fire spreading simulation software;

[0039] (4) a fire spreading early warning and automatic release module;

[0040] The automatic triggering WRF-SFIRE numerical simulation module simulation result is used as the core, and the simulation result analysis unit, risk area identification unit, early warning information generation unit and automatic release interface unit are included.

[0041] The application has the following beneficial effects:

[0042] (1) Real-time fire point high-precision identification is realized, and the response efficiency is significantly improved;

[0043] The prior art mainly relies on infrared remote sensing or ground camera to monitor the fire point, and has the problems of low spatial resolution, great influence of weather, information lag and the like. The application adopts laser radar scanning technology, can realize high-temporal and high-spatial resolution detection of the fire point in a complex forest environment, and can update the target area state in real time, so as to provide accurate initial conditions for subsequent prediction.

[0044] (2) Build an automatic triggering mechanism based on micro-meteorological thresholds to avoid redundant computing resource consumption;

[0045] Traditional models often use preset time or manual judgment to determine whether to conduct fire simulation, which is difficult to adapt to rapidly changing environmental conditions. The present application uses the particle concentration, wind speed, wind direction and other near-surface micro-meteorological elements retrieved by laser radar to set scientific threshold criteria. When the weather conditions for rapid fire spread are met, the system automatically starts WRF-SFIRE simulation, with stronger target specificity and system intelligence.

[0046] (3) Establish a closed-loop forecasting system to cover the whole process of "detection - judgment - simulation - warning";

[0047] Compared with the traditional method of only showing the spread results, the present application realizes automatic closed-loop coupling between the monitoring module, the weather judgment module, the simulation module and the warning module. The information of each subsystem is shared and worked together in real time, significantly improving the continuity, automation and practicality of the spread forecast, and is suitable for the routine deployment needs of forest fire prevention operations.

[0048] (4) Good deployment flexibility and technical expandability;

[0049] The present application is based on a mobile laser radar platform and a standardized numerical model interface design, which supports rapid deployment in different forest areas and can seamlessly interface with mesoscale weather forecast products, with good regional adaptability and future potential for functional expansion. Compared with existing systems that rely on static observations and manual intervention, the present application is more intelligent, efficient and reliable. DETAILED DESCRIPTION

[0050] The following further describes the specific embodiments of the present application in conjunction with the technical solutions.

[0051] An automatic forest fire spread forecasting method combining laser radar detection and micro-meteorological judgment is realized through the following modules:

[0052] (1) Laser radar fire point real-time detection module;

[0053] The ground-deployed long-range laser radar (LiDAR) is composed of the following components:

[0054] Laser emitting unit: Emit high-frequency pulsed laser with wavelength of 1064nm or 1550nm, frequency of 20~100kHz, support Doppler measurement and intensity retrieval;

[0055] Receiving unit: APD receiver or PMT receiver, achieving nanosecond-level time resolution;

[0056] Three-dimensional scanning system: Dual-axis mechanical scanning or MEMS micromirror is used to achieve a full view of 0~360° horizontally and −5°~+40° vertically;

[0057] Data acquisition and control module: responsible for synchronous control, signal digitization and preliminary filtering;

[0058] Communication and data interface: including optical fiber / 5G module, realizing data docking with the back-end system;

[0059] 1.1) Fire point preliminary screening determination;

[0060] Set the intensity threshold I thresh , combined with the spatial continuity factor C s (the connectivity between adjacent pixels) to determine the fire point candidate area; if the continuous multiple detection signals of the laser radar meet I>I thresh , C s >0.8, it is determined that the detected area is a fire point candidate area;

[0061] 1.2) Dynamic echo filter set:

[0062] Use wavelet filtering + adaptive Kalman filtering combination to process the detection signal and eliminate non-target disturbances (such as insects, wind blown branches and leaves, etc.);

[0063] 1.3) Fire point coordinate inversion and three-dimensional construction:

[0064] Use laser radar ranging + scanning angle to locate the fire point position in real time, with an accuracy of 1~3 meters;

[0065] The fire point position is a three-dimensional fire point set (Lat, Lon, Alt) Time , forming a "time-space-intensity" joint data stream, which is transmitted to the radar inversion-based microclimate determination module in real time;

[0066] (2) Radar inversion-based microclimate determination module;

[0067] The radar inversion-based microclimate determination module relies on the echo Doppler shift data captured by the laser radar system during the sweeping process to obtain the particle concentration around the fire point, the near-ground wind speed and direction through inversion algorithm;

[0068] 2.1) Doppler shift inversion wind speed:

[0069] Use the Doppler shift generated by the relative motion of laser pulses and aerosol particles to calculate the radial wind speed :

[0070]

[0071] where, is the laser frequency, is the speed of light, is the Doppler shift of the target echo signal relative to the transmitted laser light;

[0072] 2.2) Wind speed and direction determination condition setting:

[0073] Wind speed threshold: radial wind speed V > 2.5 m / s (2.5 m / s is the minimum fire spread wind speed);

[0074] Wind direction stability: standard deviation of wind direction change (Fire spread simulation and decision-making require relatively stable wind direction, otherwise the direction of the fire is uncertain and it is difficult to form effective triggering; representing that the wind direction changes less in the past period of time);

[0075] Horizontal shear: wherein, is the distance coordinate in a certain spatial direction; (The wind field is relatively uniform; this threshold setting ensures that the wind direction and wind speed are consistent in the triggering judgment area, avoiding false triggering due to local turbulence or observation errors);

[0076] If at least two of the above three conditions are met at the same time, it is determined that the area has "triggering conditions";

[0077] 2.3) Fire point-wind field superposition decision algorithm:

[0078] The center of the "triggering condition" area, i.e. the center of the fire point, is taken as the decision point;

[0079] A fire point direction-wind direction consistency factor is introduced:

[0080]

[0081] wherein, is the wind direction angle, is the slope angle;

[0082] It is determined whether the wind direction is consistent with the direction of the fire slope, it is considered to have fire spread power;

[0083] 2.4) Decision output mechanism:

[0084] The output quantities of the radar echo inversion include: whether to trigger (YES / NO), particulate matter concentration, wind speed, wind direction, time, and spatial coordinates;

[0085] The decision result is immediately sent to the automatically triggered WRF-SFIRE numerical simulation module as a switch for starting the fire spread simulation;

[0086] (3) automatically triggered WRF-SFIRE numerical simulation module;

[0087] The automatically triggered WRF-SFIRE numerical simulation module is integrated with WRF-SFIRE fire spread simulation software;

[0088] 3.1) Automatic triggering judgment: According to the "triggering condition" set in step 2.2), it is judged whether the triggering threshold of fire development is reached, and the numerical simulation process is automatically activated: download the GFS weather forecast data, process the downloaded GFS weather forecast data using WPS, form the initial field file driving WRF-SFIRE, then set the relevant parameters in the namelist.input of WRF-SFIRE according to the longitude and latitude position of the ignition point, and start the model simulation;

[0089] 3.2) Meteorological driving data acquisition: Call high-resolution numerical weather prediction data (such as ECMWF IFS or China Meteorological Bureau GRAPES model output), extract meteorological elements in the future several hours to several days, including wind speed, wind direction, air temperature and relative humidity, for driving fire spread simulation;

[0090] 3.3) Fire source position initialization: The fire point obtained by laser radar detection is used as the initial fire source position, combined with the current and future meteorological driving data, the WRF-SFIRE numerical simulation module is started, and the spatial spread path, boundary position, spread rate and fire intensity evolution process of the fire are numerically simulated; The fire ignition start position information in the namelist.input file of the WRF-SFIRE numerical simulation module is set by the fire_ignition_start_lon1 and fire_ignition_start_lat1 parameters;

[0091] 3.4) Dynamic rolling simulation: Using a cyclic rolling mechanism, GFS weather forecast data is reacquired and simulation results are updated every set time interval (such as 1 hour or less), achieving dynamic tracking and prediction of fire evolution;

[0092] (4) fire spread early warning and automatic release module;

[0093] Based on the simulation results of the automatically triggered WRF-SFIRE numerical simulation module, it includes a simulation result analysis unit, a risk area identification unit, a warning information generation unit, and an automatic release interface unit;

[0094] 1) Receive the spatial spread path, boundary position, spread rate and fire intensity evolution process of the fire output by the automatically triggered WRF-SFIRE numerical simulation module, and generate visual grid layers according to the preset time limit;

[0095] 2) Identify potential affected areas by combining terrain, vegetation, roads, and population distribution;

[0096] 3) Divide warning levels and generate structured warning information according to spread rate, fire intensity, and impact range;

[0097] 4) Push visual grid layers and corresponding warning information to the warning platform and management terminal to achieve quick response and release.

[0098] Based on simulation results, automatic generation and hierarchical release of warnings replace traditional static warning methods; combining multi-source geographic information to intelligently identify risk areas improves warning accuracy and timeliness; and opening up data channels between simulation output and command platforms achieves intelligent and automated closed-loop fire warning.

Claims

1. A forest fire spread automatic prediction method that fuses laser radar detection and micro-meteorological determination, characterized by, Realized by the following modules: (1) Laser radar fire point real-time detection module; (1.1) Fire point preliminary screening determination; Setting intensity threshold I thresh , matching spatial continuity factor C s Judging fire point candidate region; if continuous multiple detection signals of the laser radar meet I>I thresh , C s >0.8, the detected region is determined as the fire point candidate region; (1.2) Dynamic echo filter set: Use wavelet filtering + adaptive Kalman filtering combination processing detection signal, eliminate non-target disturbance; (1.3) Fire point coordinate inversion and three-dimensional construction: Use laser radar ranging + scanning angle, real-time positioning of fire point position; The fire point position is a three-dimensional fire point set (Lat, Lon, Alt) Time The "time-space-intensity" combined data stream is formed, and is transmitted to the radar inversion-based microclimate determination module in real time. (2) Microclimate determination module based on radar inversion, relying on the Doppler shift data of the echo captured by the laser radar in the sweeping process, obtaining the particle concentration, wind speed and direction around the fire point through inversion algorithm; (2.1) Doppler shift inversion wind speed: The radial wind speed is calculated by using the Doppler frequency shift generated by the relative motion of the laser pulse and the aerosol particles through the following formula : ； wherein, is the laser frequency, is the speed of light, is the Doppler shift of the target return signal relative to the transmitted laser light; (2.2) Wind speed and direction determination condition setting: Wind speed threshold: radial wind speed V>2.5m / s; Wind direction stability: standard deviation of wind direction changes <25°; Horizontal shear: wherein, is a distance coordinate along a spatial direction; If at least two of the above three conditions are met at the same time, it is determined that the region has "triggering conditions"; (2.3) Fire point-wind field superposition decision algorithm: The center of the "triggering condition" region, that is, the fire point center, is taken as the determination point; Introducing a fire point direction - wind direction consistency factor : ； wherein wind direction angle, is the slope angle; determining whether the wind direction is consistent with the fire slope direction, > 0.5 considered to have fire spread dynamics; (2.4) Determination output mechanism: The output of the laser radar inversion includes whether to trigger, particle concentration, wind speed, wind direction, time limit, spatial coordinates; The determination result is immediately sent to the automatically triggered WRF-SFIRE numerical simulation module as a switch to start the fire spread simulation; (3) Automatically triggered WRF-SFIRE numerical simulation module; The automatically triggered WRF-SFIRE numerical simulation module integrates the WRF-SFIRE fire spread simulation software; (3.1) Automatic triggering judgment: According to the "triggering condition" set in step (2.2), it is judged whether the triggering threshold of fire development is reached, and the numerical simulation process is automatically activated: download GFS weather forecast data, use WPS to process the downloaded GFS weather forecast data, form the initial field file to drive WRF-SFIRE, then set the relevant parameters in the namelist.input of WRF-SFIRE according to the longitude and latitude position of the ignition point, and start the mode simulation; (3.2) Meteorological driving data acquisition: call high-resolution numerical weather prediction data, extract meteorological elements in the future several hours to several days, including wind speed, wind direction, air temperature and relative humidity, for driving fire spread simulation; (3.3) Fire source position initialization: take the fire point obtained by laser radar detection as the initial fire source position, combine the current and future meteorological driving data, start the WRF-SFIRE numerical simulation module, and numerically simulate the spatial spread path, boundary position, spread rate and fire intensity evolution process of the fire; The fire point position information in the namelist.input file of the WRF-SFIRE numerical simulation module is set by the fire_ignition_start_lon1 and fire_ignition_start_lat1 parameters; (3.4) Dynamic rolling simulation: use a rolling mechanism, reacquire GFS weather forecast data and update the simulation results every set time interval to realize dynamic tracking and prediction of fire evolution; (4) Fire spread warning and automatic release module; With the automatic triggered WRF-SFIRE numerical simulation module simulation results as the core, including simulation results analysis unit, risk area identification unit, early warning information generation unit and automatic release interface unit; (1) Receive the spatial spread path, boundary position, spread rate and fire intensity evolution process of the fire output by the automatic triggered WRF-SFIRE numerical simulation module, and generate visual grid layers according to the preset time limit; (2) Identify the potential affected area in combination with the terrain, vegetation, road and population distribution; (3) According to the spread rate, fire intensity and influence range, divide the early warning level and generate structured early warning information; (4) Push the visual grid layer and the corresponding early warning information to the early warning platform and the management terminal to realize rapid response and release.

Images