Forest fire risk assessment method combining microwave remote sensing and fire spreading simulation

Abstract

The invention discloses a forest fire risk assessment method combining microwave remote sensing and fire spreading simulation, and the method comprises the following steps: integrating microwave remote sensing data, and selecting a driving factor which affects the occurrence probability of a forest fire; establishing a time series data set based on the selected driving factor, and constructing a forest fire occurrence probability prediction model based on deep learning; simulating a fire spreading process through a cellular automaton-Monte Carlo simulation method, and calculating a potential combustion probability; and comprehensively evaluating the forest fire risk by combining the fire occurrence probability and the potential combustion probability. The forest fire risk comprehensive assessment method provided by the invention has the advantages of high accuracy, strong predictability and the like, the model can capture trends and changes of factors such as meteorology and vegetation within a period of time before a fire occurs, important features in driving factors can be better concerned, and the risk of the forest fire can be comprehensively assessed. The method for comprehensively evaluating the forest fire risk is more accurate in identification of high-risk fire areas, and more scientific and accurate information is provided for forest fire management.

Description

Technical Field

[0001] The invention belongs to the field of forest fire prevention, and specifically relates to a forest fire risk assessment method combining microwave remote sensing and fire spread simulation. Background Art

[0002] Forest fires are one of the most serious threats to the global ecosystem, posing a huge threat to the ecological environment, economic development, and human life and property. Fires not only destroy forest structure, affect nutrient cycles and vegetation succession patterns, but also lead to a decline in biodiversity, disrupt ecological balance, and have a negative impact on the long-term health of forests. Every year, tens of millions of hectares of land are lost to forest fires around the world, causing widespread ecological damage, economic losses of billions of dollars, and posing a major threat to human life and property.

[0003] Effective forest fire risk assessment and management are of great significance for protecting the ecological environment and reducing fire losses. At present, forest fire risk assessment mainly relies on remote sensing technology and geographic information system (GIS) technology. Optical remote sensing technology is often used for fire monitoring and assessment. It obtains vegetation indices (such as the normalized difference vegetation index NDVI) and combines meteorological data with terrain data to assess the probability of forest fires. In addition, advanced machine learning technology and deep learning models have also been applied to forest fire risk assessment to predict fires by analyzing complex environmental data.

[0004] At present, the assessment of forest fire risk based on optical remote sensing data and deep learning methods includes the following basic steps: first, collect fire history data and related driving factor data to build a complete data set; second, apply different deep learning models to predict the risk of forest fire or simulate the spread of fire; finally, generate forest fire risk assessment results by comprehensively analyzing the model prediction results. Although the existing fire risk assessment methods have made significant progress, there are still some shortcomings. The main problems are as follows: First, optical remote sensing is usually used to obtain relevant data when selecting vegetation driving factors. However, optical remote sensing technology has limited effect in cloud cover or smoky conditions, and the optical vegetation index is limited in dense forests. The optical vegetation index cannot accurately reflect the surface vegetation status due to the saturation of canopy mass, which affects the model effect; second, the fire probability prediction model is mainly based on pre-fire historical data and environmental factors, but fails to consider the impact of the spread probability after the fire occurs. Therefore, relying solely on the probability of fire occurrence cannot provide a comprehensive and accurate fire risk assessment. How to more accurately and predictively determine areas with higher fire risks through forest fire risk assessment has become a technical problem that technicians in this field need to solve. Summary of the invention

[0005] In view of the above, the purpose of the present invention is to provide a forest fire risk assessment method combining microwave remote sensing and fire spread simulation to solve the above problems. The method comprises the following steps:

[0006] 1. A forest fire risk assessment method combining microwave remote sensing and fire spread simulation, characterized in that the method comprises the following steps:

[0007] Step 1: Integrate microwave remote sensing data to select driving factors that affect the probability of forest fires;

[0008] Step 2: Establish a time series data set based on the selected driving factors and build a forest fire probability prediction model based on deep learning;

[0009] Step 3, using the Cellular Automation-Monte Carlo simulation method, simulate the fire spread process of the fire point in the area with a higher probability of occurrence and calculate the potential combustion probability;

[0010] Step 4: Comprehensively assess forest fire risk by combining fire occurrence probability and potential burning probability.

[0011] 2. The method according to claim 1, characterized in that step 2 comprises the following sub-steps:

[0012] Step 2-1, build a time series data set containing fire points and non-fire points;

[0013] Step 2-2, use the data set to train the test model and predict the probability of forest fires.

[0014] 3. The method according to claim 1, characterized in that step 3 comprises the following sub-steps:

[0015] Step 3-1, select fire points with a higher probability of forest fire occurrence to simulate the fire spread, and use the cellular automation method to simulate the fire spread process;

[0016] Step 3-2, perform multiple simulations using the Monte Carlo method and use the formula to calculate the potential combustion probability (BP).

[0017] Furthermore, the combining of microwave remote sensing data is to introduce vegetation optical depth (VOD) of microwave remote sensing data into vegetation driving factors.

[0018] Furthermore, the driving factors selected by the present invention include meteorological factors such as rainfall, temperature, humidity, and wind speed; terrain factors such as slope aspect, slope gradient, and elevation; vegetation factors such as the Normalized Difference Vegetation Index (NDVI) and Vegetation Optical Depth (VOD); and human factors such as the distance from roads and railways.

[0019] Furthermore, the time series data set obtains time series data containing driving factors by collecting and processing fire point and non-fire point data, so that the model can capture temporal changes and trends.

[0020] Furthermore, the deep learning model is based on the long short-term memory network (LSTM) and attention mechanism, which can effectively process time series data and capture the changing trends of factors such as meteorology and vegetation before a fire occurs.

[0021] Furthermore, the formula for calculating the potential combustion probability is: Where N b Indicates the number of times a certain point burns during the process of cellular automaton simulation of fire spread, N s Indicates the number of simulations using the Monte Carlo method.

[0022] Furthermore, the forest fire risk is obtained by combining the fire occurrence probability and the potential burning probability, using the formula FR=P(fire)×BP(%), where FR represents the forest fire risk, P(fire) represents the forest fire occurrence probability, and BP represents the potential burning probability obtained by potential spread simulation.

[0023] The beneficial effects of the present invention include:

[0024] The present invention introduces microwave remote sensing data when selecting data, takes vegetation optical depth as one of the vegetation driving factors, more accurately reflects the moisture content of vegetation, enhances the sensitivity of the model to vegetation conditions, and thus improves the accuracy of the prediction; combines the use of forest fire probability and fire spread simulation technology, and uses deep learning models to more accurately predict the probability of fire and simulate the potential impact of fire, thereby obtaining a comprehensive forest fire risk assessment result, providing a scientific basis for disaster prevention and mitigation. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 It is a schematic diagram of the steps of the method of the present invention;

[0026] Figure 2 It is a framework diagram of a deep learning model for predicting the probability of forest fire occurrence of the present invention;

[0027] Figure 3 It is the probability image of forest fire in the study area;

[0028] Figure 4 It is the land cover classification map in the study area;

[0029] Figure 5 It is the potential burning probability image of forest fire in the study area;

[0030] Figure 6It is an image of forest fire risk assessment in the study area. DETAILED DESCRIPTION

[0031] The present invention will be further described below in detail with reference to the accompanying drawings and through specific implementation methods.

[0032] The risk of forest fires is related to a variety of environmental factors, including human characteristics, vegetation, terrain and meteorological factors. These influencing factors are usually referred to as driving factors. Vegetation factors, as a representation of the state of combustibles in forest fires, are one of the main factors affecting the occurrence and development of fires. The present invention selects two vegetation driving factors - the Normalized Difference Vegetation Index (NDVI) based on optical remote sensing information and the Vegetation Optical Depth (VOD) based on microwave remote sensing information to jointly reflect the moisture content information of combustibles. Microwave remote sensing data is added to solve the impact of canopy and other obstructions on vegetation data. Combined with the predicted probability of forest fire occurrence and potential spread, a comprehensive forest fire risk assessment result is obtained.

[0033] Example:

[0034] A forest fire risk assessment study was conducted using prefecture-level cities in the border area between Sichuan Province and Yunnan Province (including Liangshan Prefecture, Panzhihua City in Sichuan Province, and Diqing Prefecture, Lijiang City, Dali City, Chuxiong City and Kunming City in Yunnan Province) as examples.

[0035] A forest fire risk assessment method combining microwave remote sensing and fire spread simulation includes the following steps: Figure 1 As shown:

[0036] Step 1: Integrate optical remote sensing data and microwave remote sensing data to select four types of data as driving factors affecting forest fires;

[0037] In this embodiment, microwave remote sensing data is combined to introduce vegetation optical depth (VOD) of microwave remote sensing data into vegetation driving factors. The data used is the VODCA product provided by the University of Texas Wayne, which integrates and reprocesses vegetation optical depth data from a variety of microwave remote sensing.

[0038] Step 2: Establish a time series data set based on the selected driving factors and build a forest fire probability prediction model based on deep learning;

[0039] The driving factors selected in this embodiment mainly consider rainfall, temperature, relative humidity and wind speed in meteorological factors, and use data from the global atmospheric reanalysis data set provided by the European Centre for Medium-Range Weather Forecasts (ECMWF); the elevation, slope and aspect are mainly considered in terrain factors, and the data used is from the Shuttle Radar Topography Mission (SRTM); the normalized difference vegetation index (NDVI) and vegetation optical depth (VOD) are mainly considered in vegetation factors, and the normalized vegetation index data used is from the MODIS product MOD13A2; the distance from roads and railways is mainly considered in the human activity impact factor, and the data used is calculated by setting a certain range of spatial buffers.

[0040] Wherein, step 2 includes the following sub-steps:

[0041] Step 2-1, build a time series data set containing fire points and non-fire points;

[0042] Specifically, this embodiment collects and processes fire point and non-fire point data, matches driving factors with fire point and non-fire point data according to time and location, and obtains a time series data set containing driving factors so that the model can capture temporal changes and trends.

[0043] Step 2-2, use the data set to train the test model and predict the probability of forest fires.

[0044] Specifically, in this embodiment, the data is input into Figure 2 The deep learning model framework shown in the figure uses the long short-term memory network (LSTM) and attention mechanism to effectively process time series data, capture the changing trends of factors such as meteorology and vegetation before a fire occurs, realize the learning of the forest fire probability model, and obtain the forest fire probability map in the study area, that is, Figure 3 .

[0045] Step 3, using the Cellular Automation-Monte Carlo simulation method, simulate the fire spread process of the fire point in the area with a higher probability of occurrence and calculate the potential combustion probability;

[0046] Wherein, step 3 includes the following sub-steps:

[0047] Step 3-1, select fire points with a higher probability of forest fire occurrence to simulate the fire spread, and use the cellular automation method to simulate the fire spread process;

[0048] Specifically, in this embodiment, random points in areas with high and extremely high fire probability are selected from the forest fire probability results predicted by the deep learning model, and the random points are selected according to the following formula: Figure 4 The land cover classification shown excludes points located in non-combustible areas such as residential areas and water bodies as fire points, and the fire spread is simulated using the cellular automation method.

[0049] Step 3-2, perform multiple simulations using the Monte Carlo method and use the formula to calculate the potential combustion probability (BP).

[0050] Specifically, this embodiment uses the Monte Carlo method to perform cellular automaton fire spread simulation multiple times, and uses the proportion of the number of combustion times in the number of simulation times as the potential combustion probability (BP). The basic formula is shown in formula (1), and the following is obtained: Figure 5 Map of forest fire burn potential within the study area shown.

[0051]

[0052] Where N b Indicates the number of times a certain point burns during the process of cellular automaton simulation of fire spread, N s Indicates the number of simulations using the Monte Carlo method.

[0053] Step 4: Comprehensively assess forest fire risk by combining fire occurrence probability and potential burning probability.

[0054] Specifically, the inventor believes that the forest fire risk is the comprehensive result of the forest fire occurrence probability and the potential combustion probability. In this embodiment, the forest fire risk is obtained by combining the fire occurrence probability and the potential combustion probability. The formula shown in formula (2) is used to obtain Figure 6 Forest fire risk shown.

[0055] FR=P(fire)×BP(%) (2)

[0056] Among them, FR represents forest fire risk, P(fire) represents the probability of forest fire occurrence, and BP represents the potential burning probability obtained by potential spread simulation.

[0057] The present invention is explained above in conjunction with the accompanying drawings and specific embodiments, but the present invention is not limited to the above-mentioned specific embodiments, which are merely illustrative rather than restrictive. Under the guidance of the present invention, those skilled in the art can make various equivalent substitutions, modifications or improvements to the technical solutions and implementation methods of the present invention without departing from the spirit of the present invention and the scope of protection of the claims, all of which fall within the protection of the present invention.

Claims

1. A forest fire risk assessment method combining microwave remote sensing and fire spread simulation, characterized in that: The method comprises the following steps: Step 1: Integrate optical remote sensing data and microwave remote sensing data to select four types of data as driving factors affecting forest fires: meteorological factors, topographic factors, vegetation factors, and human activity factors; meteorological factors include rainfall, temperature, humidity, and wind speed; topographic factors include slope aspect, slope, and elevation; human activity factors include the distance to roads and railways; and vegetation factors include the Normalized Difference Vegetation Index (NDVI) based on optical remote sensing and the Vegetation Optical Depth (VOD) based on microwave remote sensing. Step 2: Establish a time series data set based on the selected driving factors and build a forest fire probability prediction model based on deep learning; Step 3, using the cellular automaton-Monte Carlo simulation method, simulate the fire spread process of the fire point in the area with a higher probability of occurrence and calculate the potential combustion probability; Step 4: Comprehensively assess forest fire risk by combining fire occurrence probability and potential burning probability.

2. The method according to claim 1, characterized in that: Step 2 includes the following sub-steps: Step 2-1, build a time series data set containing fire points and non-fire points; Step 2-2, use the data set to train the test model and predict the probability of forest fires.

3. The method according to claim 1, characterized in that: Step 3 includes the following sub-steps: Step 3-1, select fire points with a higher probability of forest fire occurrence to simulate the fire spread, and use the cellular automation method to simulate the fire spread process; Step 3-2, perform multiple simulations using the Monte Carlo method and use the formula to calculate the potential combustion probability (BP).

4. The method according to claim 1, characterized in that: In step 2, the deep learning model is based on the long short-term memory network (LSTM) and attention mechanism, which can effectively process time series data and capture the changing trends of factors such as meteorology and vegetation before a fire occurs.

5. The method according to claim 1, characterized in that In step 4, the combination of fire occurrence probability and potential combustion probability is achieved using the formula FR=P(fire)×BP(%), where FR represents forest fire risk, P(fire) represents forest fire occurrence probability, and BP represents potential combustion probability obtained by potential spread simulation.

6. The method according to claim 2, characterized in that In step 2-1, the time series data set is obtained by collecting and processing fire point and non-fire point data to obtain time series data containing driving factors so that the model can capture temporal changes and trends.

7. The method according to claim 3, characterized in that In step 3-2, the formula is Where N b Indicates the number of times a certain point burns during the process of cellular automaton simulation of fire spread, N s Indicates the number of simulations using the Monte Carlo method.

Images