Fire combustion area prediction method, prediction device, equipment and medium

Abstract

The invention provides a fire combustion area prediction method and device, equipment and a medium, and the method comprises the steps: obtaining input variable data needed for constructing a fire combustion area prediction model, and constructing a fire feature parameter based on the input variable data; constructing a prediction sample data set based on the fire characteristic parameters, and establishing a fire combustion area prediction model by using the prediction sample data set; and for each to-be-predicted region, inputting the meteorological prediction data of the to-be-predicted region in each prediction time period into the fire combustion area prediction model to obtain a fire combustion area prediction value of the to-be-predicted region in each prediction time period. Through the method and the device, the fire combustion area prediction model has the capability of predicting the inter-annual trend under the driving of the future estimated weather data, and the quantitative evaluation of the overall development trend of the specific combustion area in each fire season in a plurality of years in the future is realized.

Description

Technical Field

[0001] The present application relates to the technical field of fire prediction, and in particular to a method, device, equipment and medium for predicting the burning area of ​​a fire. Background Art

[0002] As global climate change intensifies, wildfires have become a major environmental threat to ecosystems and human safety. In recent years, extreme weather events, coupled with rising global temperatures, have led to a significant increase in the frequency, scale, and destructiveness of wildfires. Especially under the combined extreme weather conditions of high temperatures, drought, and strong winds, wildfires often exhibit sudden onset, rapid spread, and difficulty in control, posing a significant challenge to disaster prevention and mitigation efforts.

[0003] Current mainstream wildfire prediction and risk assessment technologies are primarily based on fire weather indices. These systems use a combination of real-time meteorological observations such as temperature, humidity, and wind speed to estimate the probability of a wildfire occurring in a specific area. However, these methods are essentially meteorological assessments of the likelihood of a wildfire occurring and are unable to directly characterize actual wildfire characteristics such as the actual burned area and fire intensity. Summary of the Invention

[0004] In view of this, the purpose of this application is to provide a fire burning area prediction method, prediction device, equipment and medium. The fire burning area prediction model that integrates atmospheric circulation driving factors and weather process evolution mechanisms has the ability to make interannual trend predictions driven by future estimated weather data. Based on future meteorological forecast data, it is possible to quantitatively evaluate the overall development trend of each fire season in a specific burning area in the next few years.

[0005] In a first aspect, an embodiment of the present application provides a method for predicting a fire burning area, the method comprising:

[0006] Obtaining input variable data required for constructing a fire burn area prediction model, and constructing fire characteristic parameters based on the input variable data; wherein the input variable data includes historical ground meteorological variables, historical atmospheric circulation variables, historical ground ecological variables, and historical fire burn areas in historical fire burning seasons;

[0007] Constructing a prediction sample data set based on the fire characteristic parameters, and establishing the fire burning area prediction model using the prediction sample data set;

[0008] For each area to be predicted, the meteorological forecast data of the area to be predicted in each prediction time period is input into the fire burning area prediction model to obtain the fire burning area prediction value of the area to be predicted in each prediction time period.

[0009] Furthermore, the fire characteristic parameters include a previous lag factor, a key factor of the burning season, atmospheric circulation variables, potential burning types of vegetation, a composite interaction term, and the burned area of ​​historical fires; the composite interaction term is determined by the following steps:

[0010] The product of atmospheric thermal intensity and air dryness, the product of relative humidity and saturated water vapor pressure difference, and the product of fire spreading power and soil moisture in the key factors of the burning season are used as the composite interaction terms.

[0011] Furthermore, constructing a prediction sample data set based on the fire characteristic parameters includes:

[0012] For each historical time period, the previous lag factor, burning season key factor, atmospheric circulation variable, vegetation potential combustion type, and composite interaction term corresponding to the historical time period are obtained from the fire characteristic parameters as input features, and the historical fire burned area corresponding to the historical time period is obtained from the fire characteristic parameters as output features;

[0013] Based on the input features and output features corresponding to each historical time period, sample data corresponding to each historical time period is obtained, and based on the sample data of each historical time period, a predicted sample data set is obtained.

[0014] Furthermore, after obtaining the predicted value of the fire burning area of ​​the area to be predicted in each prediction time period, the prediction method further includes:

[0015] A fire prediction trend graph corresponding to the area to be predicted is generated based on the fire burning area prediction value of the area to be predicted in each preset time period.

[0016] Furthermore, after obtaining the fire burning area prediction value of each to-be-predicted region in each prediction time period, the prediction method further includes:

[0017] For each preset time period, the fire risk level corresponding to each to-be-predicted area is determined based on the fire burning area prediction value of each to-be-predicted area within the preset time period;

[0018] Obtain a regional spatial distribution map, generate a regional fire risk map based on the fire risk level corresponding to each area to be predicted and the regional spatial distribution map, and display the fire risk levels of different areas to be predicted within the preset time period in the regional fire risk map.

[0019] Furthermore, the fire burning area prediction model is constructed using the prediction sample data set, including:

[0020] Using a cross-validation algorithm, a plurality of training sample sets and validation sample sets are determined based on the prediction sample data set;

[0021] Using multiple sets of training sample sets to perform model training to train multiple fire burning area original prediction models;

[0022] The verification sample set is used to verify each trained original fire burning area prediction model in turn, and the fire burning area original prediction model that passes the verification is determined as the fire burning area prediction model.

[0023] In a second aspect, an embodiment of the present application further provides a fire burning area prediction device, the prediction device comprising:

[0024] A characteristic parameter determination module is used to obtain input variable data required to construct a fire burn area prediction model and construct fire characteristic parameters based on the input variable data; wherein the input variable data includes historical ground meteorological variables, historical atmospheric circulation variables, historical ground ecological variables, and historical fire burn areas in historical fire burning seasons;

[0025] A model building module, configured to build a prediction sample data set based on the fire characteristic parameters, and establish the fire burning area prediction model using the prediction sample data set;

[0026] The prediction module is used to input the meteorological prediction data of each area to be predicted in each prediction time period into the fire burning area prediction model to obtain the fire burning area prediction value of the area to be predicted in each prediction time period.

[0027] Furthermore, the fire characteristic parameters include a previous lag factor, a key factor of the burning season, atmospheric circulation variables, potential burning types of vegetation, a composite interaction term, and the historical fire burn area; the characteristic parameter determination module is further configured to determine the composite interaction term through the following steps:

[0028] The product of atmospheric thermal intensity and air dryness, the product of relative humidity and saturated water vapor pressure difference, and the product of fire spreading power and soil moisture in the key factors of the burning season are used as the composite interaction terms.

[0029] In a third aspect, an embodiment of the present application further provides an electronic device comprising: a processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor and the memory communicate via the bus, and when the machine-readable instructions are executed by the processor, the steps of the fire burning area prediction method as described above are performed.

[0030] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the fire burning area prediction method as described above are executed.

[0031] The embodiments of the present application provide a method, device, equipment, and medium for predicting fire burning area. First, input variable data required for constructing a fire burning area prediction model is obtained, and fire characteristic parameters are constructed based on the input variable data; wherein the input variable data include historical ground meteorological variables, historical atmospheric circulation variables, historical ground ecological variables, and historical fire burning areas in historical fire burning seasons; then, a prediction sample data set is constructed based on the fire characteristic parameters, and the fire burning area prediction model is established using the prediction sample data set; finally, for each area to be predicted, the meteorological prediction data of the area to be predicted in each prediction time period is input into the fire burning area prediction model to obtain a predicted value of the fire burning area of ​​the area to be predicted in each prediction time period.

[0032] The prediction method provided in the embodiment of the present application takes physical driving factors as the main line, combines machine learning algorithms to model nonlinear response relationships, integrates the dynamic characteristics of the weather system with multi-source meteorological data, and combines ground ecological information to achieve an end-to-end closed-loop prediction chain between weather, fuel, and fire, and realizes quantitative prediction of the overall fire trend in the target area in future years. The fire burn area prediction model that integrates atmospheric circulation driving factors and weather process evolution mechanisms has the ability to make interannual trend predictions driven by future estimated weather data. Based on future meteorological forecast data, it can achieve a quantitative assessment of the overall development trend of each fire season in a specific burning area in the next few years, providing scientific support for medium- and long-term wildfire risk management and emergency resource planning.

[0033] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without creative work.

[0035] Figure 1 A flowchart of a method for predicting fire burning area provided in an embodiment of the present application;

[0036] Figure 2 This is one of the structural schematic diagrams of a fire burning area prediction device provided in an embodiment of the present application;

[0037] Figure 3 This is a second structural diagram of a fire burning area prediction device provided in an embodiment of the present application;

[0038] Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0039] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application for which protection is claimed, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, each other embodiment obtained by those skilled in the art without making creative work falls within the scope of protection of the present application.

[0040] First, the application scenarios to which this application is applicable are introduced. This application can be applied in the field of fire prediction technology.

[0041] As global climate change intensifies, wildfires have become a major environmental threat to ecosystems and human safety. In recent years, extreme weather events, coupled with rising global temperatures, have led to a significant increase in the frequency, scale, and destructiveness of wildfires. Especially under the combined extreme weather conditions of high temperatures, drought, and strong winds, wildfires often exhibit sudden onset, rapid spread, and difficulty in control, posing a significant challenge to disaster prevention and mitigation efforts.

[0042] The damage caused by wildfires is not limited to vegetation destruction and increased carbon emissions; they can also trigger secondary disasters such as soil erosion, water pollution, landslides and mudslides, resulting in long-term ecological and social impacts. More seriously, the smoke particles (such as PM2.5) and toxic gases (such as CO and NOx) released by large-scale wildfires can be transported long distances through the atmosphere, causing long-range and cross-border impacts on air quality and human health in downwind areas. Therefore, wildfire disasters have become a secondary meteorological risk that cannot be ignored in the context of global climate change, and are ranked alongside floods, typhoons, and droughts as one of the major natural disaster types facing modern society.

[0043] Research has found that currently, wildfire forecasting relies primarily on empirical risk assessment models, such as the Fire Weather Index (FWI). These indices use a weighted calculation of surface meteorological factors, such as temperature, relative humidity, wind speed, and precipitation, to measure the meteorological suitability of a given area for fire at a specific moment or time period. These indices play a crucial role in daily fire risk classification and short-term fire warnings. However, these models exhibit a series of difficult-to-overcome structural flaws in interannual forecasting, primarily in the following areas:

[0044] (a) They only reflect the potential fire risk level and lack the ability to quantify actual fires. The FWI and similar indices focus on the degree to which meteorological conditions support the "possibility" of fire occurrence, rather than directly estimating key factors such as the area, intensity, and duration of actual wildfires. This makes them inadequate for quantitative trend prediction.

[0045] (b) Ignoring the regulatory mechanisms of atmospheric circulation systems on fire weather. Fire weather is often driven by large-scale circulation anomalies, such as mid- and high-latitude jet streams, high-pressure areas, and trough-ridge systems. However, existing fire indices, which are mostly based on ground-based meteorological observations, fail to capture these key regulatory factors, making it difficult to explain or predict the causes and evolution of extreme fire events.

[0046] (c) There is a lack of a modeling mechanism for the cumulative lag effects of meteorological factors. The development of wildfires is often associated with the superposition of meteorological processes such as prolonged drought and persistent high temperatures. Exponential models generally only consider meteorological conditions within a short time window and cannot reflect these lags and cumulative effects.

[0047] (d) Inability to extrapolate trends. FWI indices are based solely on current or historical meteorological data and cannot simulate or extrapolate future climate scenarios. Therefore, they lack the ability to predict wildfire activity trends.

[0048] In summary, current wildfire trend forecasting technology has significant shortcomings. In particular, traditional fire risk assessment models cannot meet the needs of trend prediction in interannual applications, and their ability to model the driving mechanisms of weather processes is limited.

[0049] Based on this, an embodiment of the present application provides a method for predicting fire burning area, so as to achieve a quantitative assessment of the overall development trend of each fire season in a specific burning area in the next few years.

[0050] See also Figure 1 , Figure 1 This is a flow chart of a method for predicting the burning area of ​​a fire provided in an embodiment of the present application. Figure 1 As shown in , the prediction method provided in the embodiment of the present application includes:

[0051] S101, obtaining input variable data required for constructing a fire burning area prediction model, and constructing fire characteristic parameters based on the input variable data.

[0052] With respect to the above-mentioned step S101, in the specific implementation, the input variable data required for constructing the fire burning area prediction model is first obtained. Here, the input variable data include historical ground meteorological variables, historical atmospheric circulation variables, historical ground ecological variables, and historical fire burning areas in historical fire burning seasons to ensure the consistency of multi-source information integration. Then, based on the obtained input variable data, it is converted into fire characteristic parameters representing key mechanisms such as weather processes, vegetation responses, and dryness accumulation, and the original observation or simulation data is converted into highly expressive modeling features to reflect the real physical mechanism of wildfire development regulated by meteorological conditions, providing a highly expressive input feature set for subsequent machine learning modeling. In this way, the obtained input variable data integrates the input system of high-altitude atmospheric circulation variables and ground meteorological variables, and the subsequent model can enhance the ability to explain and predict fire development from the perspective of the dynamic mechanism of weather processes.

[0053] Specifically, according to the embodiment provided in the present application, historical ground meteorological variables may include 2m temperature (T), precipitation (Pre), surface evaporation (Evap), 2m relative humidity (RH), 10m saturated water vapor pressure difference (VPD), 10m wind speed (WS10), 0-10cm soil moisture (SM10); historical atmospheric circulation variables may include 500hPa potential height (H500); historical ground ecological variables may include underlying surface cover type (Land_Type). The historical fire burning season may be June to August in the historical year, and the historical fire burning area is the wildfire burning area (BA) in June to August in the historical year, which can be obtained using the MCD64A1 (500m) product. When obtaining the historical fire burning area, the space is cropped to the boundary of the target area, and the spatial weighted average is used to aggregate to a resolution of 0.25°, and the unit is unified as km. 2 , as the regression target. All collected variables are standardized according to the sample dimension Z-score; the KNN interpolation method is used to fill in the missing points in space; the time series linear interpolation is used to fill in the missing points throughout the year, and all variables are output as annual sample units (1 / year / region), which are finally used for sample set construction. The existing technology mostly relies on ground meteorological variables (such as temperature, humidity, and wind speed) to calculate the fire risk index, which is difficult to reflect the large-scale dynamic process behind extreme weather. This application introduces atmospheric circulation indicators such as 500hPa potential height intensity, which effectively improves the ability to characterize the meteorological background of extreme fires.

[0054] Alternatively, historical atmospheric circulation variables, in addition to 500hPa geopotential height, can be substituted or supplemented with other atmospheric circulation variables, such as 700hPa temperature anomalies, sea level pressure anomalies, the Northern Hemisphere Annular Mode, or the Pacific Decadal Oscillation Index, to reflect the large-scale circulation background at mid- and high-latitudes. Historical surface meteorological variables, in addition to VPD (vapor pressure deficit), can also include indicators such as the dryness index, the number of high-temperature days, and the frequency of extreme high temperatures to replace some meteorological driving factors. Soil moisture can be measured using observations or reanalysis products at different depths (e.g., 0–10 cm, 10–40 cm) as a surrogate for fuel dryness.

[0055] Specifically, according to the embodiments provided in this application, the fire characteristic parameters include a previous lag factor, a key factor of the burning season, atmospheric circulation variables, potential burning types of vegetation, compound effect interaction terms, and historical fire burning areas.

[0056] Here, according to the embodiments provided herein, the pre-season lag factors (March-May) include: Tmax_MAM: daily maximum temperature, representing fuel heating potential; SPEI_MAM: Standardized Precipitation Evaporation Index, calculated by subtracting evaporation from precipitation, reflecting the cumulative nature of the drying process; SM_MAM: soil moisture, reflecting the degree of pre-season fuel dryness. Key factors for the burning season (June-August) include: Tmax_JJA: daily maximum temperature, representing fuel heating potential; RH_JJA and VPD_JJA: relative humidity and saturation vapor pressure difference, respectively, representing atmospheric dryness; Tmax_P90_JJA: frequency of extreme high temperatures (number of days above the 90th percentile), used to simulate nonlinear outbreak conditions; WS_JJA: provides the dynamics of fire spread and can be weighted by the prevailing wind direction if necessary; SM_JJA: soil moisture, reflecting fuel dryness. Atmospheric circulation variables (June-August) include: H500_JJA: 500hPa geopotential height, which can identify areas controlled by high-pressure ridges. Potential burning types of vegetation (June-August) include: Land_Type_JJA: underlying surface vegetation type, which characterizes different vegetation types during fire burning. The historical fire burning area is BA_JJA: the fire burning area in the historical fire burning season, which is the target variable involved in modeling. The fire characteristic parameters include key meteorological factors in the early stage of the fire season, such as the cumulative drought and high temperature indicators in spring (March-May), to model their lagged effects on the development of the fire season. Most existing technologies ignore the lagged relationship between the occurrence of fires and previous meteorological conditions. This application incorporates meteorological factors in the early stage of the fire season (such as cumulative drought and high temperature in spring), which improves the ability to identify leading signals of fire development. In addition to using fixed seasonal windows (such as spring versus summer), a sliding window method can also be introduced to dynamically determine the most relevant time period.

[0057] Furthermore, the compound interaction term is determined by the following steps:

[0058] The product of atmospheric thermal intensity and air dryness, the product of relative humidity and saturated water vapor pressure difference, and the product of fire spreading power and soil moisture in the key factors of the burning season are used as the composite interaction terms.

[0059] Here, this application simulates the synergistic amplification effect between meteorological factors by constructing interaction terms. For example, when the temperature is extremely high and the air is extremely dry, the evaporation of vegetation moisture accelerates and the risk of fire outbreak increases nonlinearly.

[0060] Specifically, according to the embodiments provided herein, the composite interaction term includes the product of atmospheric thermal intensity and air dryness, two key factors in the burning season. Specifically, Tmax_JJA × VPD_JJA represents the maximum temperature and saturated vapor pressure difference. Tmax_JJA represents atmospheric thermal intensity, and VPD_JJA represents air dryness. When both values ​​are high, it indicates that the atmosphere has a strong drying and dehydrating capacity, significantly accelerating the rate of water loss from vegetation and surface combustibles. The composite interaction term also includes the product of relative humidity and saturated vapor pressure difference, namely RH_JJA × VPD_JJA. RH_JJA reflects the current moisture content of the air, and VPD_JJA represents the air's ability to absorb water. Together, these two factors determine whether vegetation can retain moisture and how quickly it will be absorbed. The composite interaction term also includes WS_JJA × SM_JJA: high wind speeds can promote fire spread and expand the fire front, while low soil moisture can lead to insufficient moisture in surface vegetation and dry fuels. The product term reflects the synergistic amplification effect between fire spread and surface fuel conditions. Here, all interaction terms are generated year by year and included in the model with equal dimensions as the main variables.

[0061] Optionally, in addition to the traditional product-type interaction terms mentioned above, nonlinear combination functions (such as exponential form, logistic regression transformation, etc.) can also be used to represent the coupling relationship between different variables in the composite interaction terms; multivariate principal component analysis (PCA) or principal factors constructed after dimensionality reduction using empirical orthogonal function (EOF) can also be introduced as an alternative expression of the relationship between variables.

[0062] By introducing interaction terms between key variables, the subsequent prediction model's ability to capture nonlinear responses under extreme fire conditions is enhanced. Existing fire risk index methods mostly rely on linear superposition, but this application constructs interaction terms for key meteorological factors (such as Tmax×VPD and RH×VPD) to more realistically reflect the complex mechanisms that influence fire development.

[0063] S102: constructing a prediction sample data set based on the fire characteristic parameters, and establishing the fire burning area prediction model using the prediction sample data set.

[0064] Regarding step S102 above, during specific implementation, after constructing the fire characteristic parameters in step S101, a prediction sample dataset is constructed based on the fire characteristic parameters, and a fire burn area prediction model is constructed using the prediction sample dataset. As an optional embodiment, the fire burn area prediction model can use an XGBoost regressor to capture high-order nonlinear relationships; the model parameters can be set as: max_depth = 5, n_estimators = 150, subsample = 0.8; and the parameter group is automatically optimized using GridSearchCV. The fire burn area prediction model can also use Random Forest or LSTM (for time series). Random Forest is used to compare model stability; LSTM is suitable for sequence modeling when the input variables are expanded to the monthly scale.

[0065] Optionally, in addition to random forest, XGBoost, and LSTM, the fire burn area prediction model can alternatively use machine learning regression algorithms including but not limited to: LightGBM, CatBoost, support vector regression (SVR), Bayesian regression, neural networks (such as GRU, MLP), etc.; if there is sufficient data support, graph neural networks (GNN) or Transformer structures can also be introduced to model spatial-temporal correlation.

[0066] Specifically, with respect to the above step S102, constructing a prediction sample data set based on the fire characteristic parameters includes:

[0067] Step 1021: For each historical time period, the previous lag factor, burning season key factor, atmospheric circulation variable, vegetation potential combustion type, and composite interaction term corresponding to the historical time period are obtained from the fire characteristic parameters as input features, and the historical fire burning area corresponding to the historical time period is obtained from the fire characteristic parameters as output features.

[0068] Regarding step 1021 above, during the specific implementation, for each historical time period, the previous lag factor, burning season key factor, atmospheric circulation variable, vegetation potential combustion type, and compound effect interaction term corresponding to the historical time period are obtained from the fire characteristic parameters as input features, and the historical fire burn area corresponding to the historical time period is obtained from the fire characteristic parameters as output features. In this way, the pairing logic between X (input) and Y (output) in the sample set is clearly constructed to ensure that the physical logic is valid and the time structure is reasonable. As an example, when performing causal pairing logic, the input feature time period can be spring (March-May) + summer (June-August), and the output feature time period is the cumulative fire burn area from June to August; all variables are sorted by year, and the causal order is strictly maintained.

[0069] Step 1022 : obtaining sample data corresponding to each historical time period based on the input features and output features corresponding to each historical time period, and obtaining a predicted sample data set based on the sample data of each historical time period.

[0070] Regarding step 1022 above, during specific implementation, sample data corresponding to each historical time period is obtained based on the input features and output features corresponding to each historical time period, and a predicted sample data set is obtained based on the sample data for each historical time period. In this way, each constructed sample data piece is input with a feature vector (the fire characteristic parameters constructed in the above steps) and a response variable (fire burned area). When constructing the sample data, each historical time period is traversed, which can be a year. Input features and output features are extracted for each year to form a complete sample set. In this way, the actual fire burned area is used as the prediction target, and the input factors and response variables are time-aligned to ensure the consistency of the physical process.

[0071] Specifically, with respect to the above step S102, constructing the fire burning area prediction model using the prediction sample data set includes:

[0072] A: Using a cross-validation algorithm, multiple sets of training sample sets and validation sample sets are determined based on the prediction sample data set.

[0073] Here, as an example, a K=5 cross-validation algorithm may be adopted.

[0074] Regarding step A above, during implementation, a cross-validation algorithm is used to determine multiple training and validation sample sets based on the prediction sample dataset. Here, 70% of the prediction sample dataset is divided into training samples, and 30% of the prediction sample dataset is divided into validation samples. When K=5 cross-validation is used, the training samples are divided into five training sample sets, and five training cycles are performed.

[0075] B: Use multiple sets of the training sample sets to perform model training to train multiple original fire burning area prediction models.

[0076] Regarding the above step B, during specific implementation, the determined multiple groups of training sample sets are used to perform model training to train multiple original prediction models for fire burning area.

[0077] C: using the verification sample set to sequentially verify each trained original fire burning area prediction model, and determining the fire burning area original prediction model that passes the verification as the fire burning area prediction model.

[0078] Regarding the above step C, in the specific implementation, the validation sample set is used to sequentially validate each trained original prediction model of the fire burning area. The evaluation index may include the determination coefficient R 2 , root mean square error RMSE, mean absolute error MAE. Then, the original fire burning area prediction model that has passed the verification is determined as the fire burning area prediction model.

[0079] S103 , for each area to be predicted, inputting the meteorological forecast data of the area to be predicted in each prediction time period into the fire burning area prediction model to obtain a fire burning area prediction value of the area to be predicted in each prediction time period.

[0080] Regarding the above steps, during the specific implementation, for each area to be predicted, the meteorological forecast data of the area to be predicted in each forecast time period is input into the fire burn area prediction model to obtain the fire burn area prediction value of the area to be predicted in each forecast time period. Here, the forecast time period can be a certain year in the future, and the annual meteorological forecast data under the CMIP6 climate model is used. In this way, the deployment and deduction of the fire burn area prediction model on the future forecast data are completed through the above steps, and the interannual fire development trend is output. Compared with the traditional fire risk index that represents the "possibility of fire occurrence", this application directly uses the fire burn area as the response variable, so that the model can better reflect the actual development degree of the fire situation. By using a machine learning regression model, the prediction accuracy and generalization ability are improved, so that the model has the ability to make interannual trend predictions driven by future estimated weather data. Based on future meteorological forecast data, it is possible to achieve a quantitative assessment of the overall development trend of each fire season in a specific burning area in the next few years, providing scientific support for medium- and long-term wildfire risk management and emergency resource planning. The existing technology is mostly used for real-time risk assessment and lacks the ability to predict future trends. This application can use climate model prediction data to estimate interannual trends, serving as a tool for medium- and long-term fire risk management. Alternatively, in addition to using "fire burnt area" as a target variable, other realistic fire indicators such as fire duration, number of fire points, and fire impact level can be used as alternative or supplementary target variables.

[0081] As an optional embodiment, after obtaining the fire burning area prediction value of the to-be-predicted area in each prediction time period in step S103, the prediction method provided by the present application further includes:

[0082] A fire prediction trend graph corresponding to the area to be predicted is generated based on the fire burning area prediction value of the area to be predicted in each preset time period.

[0083] In the specific implementation of the above steps, after obtaining the predicted value of the fire burn area in the predicted area within each prediction time period, a fire prediction trend graph corresponding to the predicted area is generated based on the predicted value of the fire burn area in the predicted area within each preset time period to visualize the prediction results. Here, the fire prediction trend graph can be a line graph or a bar graph, which is not specifically limited in this application.

[0084] As an optional embodiment, after obtaining the fire burning area prediction value of each to-be-predicted region in each prediction time period in step S103, the prediction method provided by the present application further includes:

[0085] I: For each preset time period, the fire risk level corresponding to each to-be-predicted area is determined based on the fire burning area prediction value of each to-be-predicted area within the preset time period.

[0086] Regarding step I above, during specific implementation, for each preset time period, the fire risk level corresponding to each area to be predicted is determined based on the predicted fire area value within the preset time period. Here, the fire risk level of each area to be predicted can be automatically graded based on the quantile method.

[0087] II: Obtain a regional spatial distribution map, generate a regional fire risk map based on the fire risk level corresponding to each area to be predicted and the regional spatial distribution map, and display the fire risk levels of different areas to be predicted within the preset time period in the regional fire risk map.

[0088] For the above step II, during the specific implementation, a regional spatial distribution map is obtained, and a regional fire risk map is generated based on the fire risk level corresponding to each area to be predicted and the regional spatial distribution map. Specifically, for each area to be predicted, the location of the area to be predicted is determined in the regional spatial distribution map, and then the fire risk level corresponding to the area to be predicted is displayed in the regional spatial distribution map, and a regional fire risk map can be obtained. As an example, the regional fire risk map can be based on GeoTIFF+Shapefile. In this way, the fire risk levels of different areas to be predicted within the preset time period can be displayed in the regional fire risk map. In this way, a regional-scale fire risk map can be output eventually, and users can compare historical fire risk maps with future fire risk maps to help users quickly understand which years in the future are more likely to have severe fire seasons. In addition, the geographical background can be superimposed to realize the spatial risk layer display of the fire risk level.

[0089] The fire burning area prediction method provided in the embodiment of the present application first obtains the input variable data required for constructing a fire burning area prediction model, and constructs fire characteristic parameters based on the input variable data; wherein the input variable data includes historical ground meteorological variables, historical atmospheric circulation variables, historical ground ecological variables and historical fire burning areas in historical fire burning seasons; then, a prediction sample data set is constructed based on the fire characteristic parameters, and the fire burning area prediction model is established using the prediction sample data set; finally, for each area to be predicted, the meteorological prediction data of the area to be predicted in each prediction time period is input into the fire burning area prediction model to obtain the fire burning area prediction value of the area to be predicted in each prediction time period.

[0090] The prediction method provided in the embodiment of the present application takes physical driving factors as the main line, combines machine learning algorithms to model nonlinear response relationships, integrates the dynamic characteristics of the weather system with multi-source meteorological data, and combines ground ecological information to achieve an end-to-end closed-loop prediction chain between weather, fuel, and fire, and realizes quantitative prediction of the overall fire trend in the target area in future years. The fire burn area prediction model that integrates atmospheric circulation driving factors and weather process evolution mechanisms has the ability to make interannual trend predictions driven by future estimated weather data. Based on future meteorological forecast data, it can achieve a quantitative assessment of the overall development trend of each fire season in a specific burning area in the next few years, providing scientific support for medium- and long-term wildfire risk management and emergency resource planning.

[0091] See also Figure 2 and Figure 3 , Figure 2 This is one of the structural diagrams of a fire burning area prediction device provided in an embodiment of the present application. Figure 3 This is a second structural diagram of a fire burning area prediction device provided in an embodiment of the present application. Figure 2 As shown in , the prediction device 200 includes:

[0092] The characteristic parameter determination module 201 is used to obtain input variable data required for constructing a fire burn area prediction model and construct fire characteristic parameters based on the input variable data; wherein the input variable data includes historical ground meteorological variables, historical atmospheric circulation variables, historical ground ecological variables, and historical fire burn areas in historical fire seasons;

[0093] A model building module 202 is configured to build a prediction sample data set based on the fire characteristic parameters, and establish the fire burning area prediction model using the prediction sample data set;

[0094] The prediction module 203 is used to input the meteorological prediction data of each area to be predicted in each prediction time period into the fire burning area prediction model to obtain the fire burning area prediction value of the area to be predicted in each prediction time period.

[0095] Furthermore, the fire characteristic parameters include a previous lag factor, a key factor of the burning season, atmospheric circulation variables, potential vegetation combustion types, a composite interaction term, and the historical fire burn area; the characteristic parameter determination module 201 is further configured to determine the composite interaction term through the following steps:

[0096] The product of atmospheric thermal intensity and air dryness, the product of relative humidity and saturated water vapor pressure difference, and the product of fire spreading power and soil moisture in the key factors of the burning season are used as the composite interaction terms.

[0097] Furthermore, when the model building module 202 is used to build a prediction sample dataset based on the fire characteristic parameters, the model building module 202 is further used to:

[0098] For each historical time period, the previous lag factor, burning season key factor, atmospheric circulation variable, vegetation potential combustion type, and composite interaction term corresponding to the historical time period are obtained from the fire characteristic parameters as input features, and the historical fire burned area corresponding to the historical time period is obtained from the fire characteristic parameters as output features;

[0099] Based on the input features and output features corresponding to each historical time period, sample data corresponding to each historical time period is obtained, and based on the sample data of each historical time period, a predicted sample data set is obtained.

[0100] like Figure 3 As shown, the prediction device 200 further includes a trend graph generating module 204. After obtaining the predicted value of the fire burning area of ​​the to-be-predicted region in each prediction time period, the trend graph generating module 204 is configured to:

[0101] A fire prediction trend graph corresponding to the area to be predicted is generated based on the fire burning area prediction value of the area to be predicted in each preset time period.

[0102] like Figure 3 As shown, the prediction device 200 further includes a fire risk map generation module 205. After obtaining the fire burning area prediction value of each to-be-predicted region in each prediction time period, the fire risk map generation module 205 is configured to:

[0103] For each preset time period, the fire risk level corresponding to each to-be-predicted area is determined based on the fire burning area prediction value of each to-be-predicted area within the preset time period;

[0104] Obtain a regional spatial distribution map, generate a regional fire risk map based on the fire risk level corresponding to each area to be predicted and the regional spatial distribution map, and display the fire risk levels of different areas to be predicted within the preset time period in the regional fire risk map.

[0105] Furthermore, when the model building module 202 is used to build the fire burning area prediction model using the prediction sample data set, the model building module 202 is further used to:

[0106] Using a cross-validation algorithm, a plurality of training sample sets and validation sample sets are determined based on the prediction sample data set;

[0107] Using multiple sets of training sample sets to perform model training to train multiple fire burning area original prediction models;

[0108] The verification sample set is used to verify each trained original fire burning area prediction model in turn, and the fire burning area original prediction model that passes the verification is determined as the fire burning area prediction model.

[0109] See also Figure 4 , Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. Figure 4 As shown in FIG, the electronic device 400 includes a processor 410, a memory 420 and a bus 430.

[0110] The memory 420 stores machine-readable instructions executable by the processor 410. When the electronic device 400 is running, the processor 410 communicates with the memory 420 via the bus 430. When the machine-readable instructions are executed by the processor 410, the above-mentioned Figure 1 The specific implementation of the steps of the method for predicting the fire burning area in the method embodiment shown can be found in the method embodiment and will not be repeated here.

[0111] The embodiment of the present application also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the computer program can execute the above-mentioned Figure 1 The specific implementation of the steps of the method for predicting the fire burning area in the method embodiment shown can be found in the method embodiment and will not be repeated here.

[0112] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0113] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. There may be other division methods in actual implementation. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed may be through some communication interface, indirect coupling or communication connection of devices or units, which may be electrical, mechanical or other forms.

[0114] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0115] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0116] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium that is executable by a processor. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0117] It should be noted that similar numbers and letters represent similar items in the following figures. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. In addition, the terms "first", "second", "third", etc. are only used to distinguish the description and are not to be understood as indicating or implying relative importance.

[0118] Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present application, which are used to illustrate the technical solutions of the present application, rather than to limit them. The scope of protection of the present application is not limited thereto. Although the present application has been described in detail with reference to the above-mentioned embodiments, those skilled in the art should understand that any person skilled in the art can modify or easily conceive of changes to the technical solutions described in the above-mentioned embodiments within the technical scope disclosed in the present application, or perform equivalent replacements for some of the technical features thereof. These modifications, changes, or replacements do not deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A method for predicting fire burning area, characterized in that: The prediction method comprises: Obtaining input variable data required for constructing a fire burn area prediction model, and constructing fire characteristic parameters based on the input variable data; wherein the input variable data includes historical ground meteorological variables, historical atmospheric circulation variables, historical ground ecological variables, and historical fire burn areas in historical fire burning seasons; Constructing a prediction sample data set based on the fire characteristic parameters, and establishing the fire burning area prediction model using the prediction sample data set; For each area to be predicted, the meteorological forecast data of the area to be predicted in each prediction time period is input into the fire burning area prediction model to obtain the fire burning area prediction value of the area to be predicted in each prediction time period.

2. The prediction method according to claim 1, characterized in that The fire characteristic parameters include a previous lag factor, a key factor of the burning season, atmospheric circulation variables, potential burning types of vegetation, a composite interaction term, and the burned area of ​​historical fires; the composite interaction term is determined by the following steps: The product of atmospheric thermal intensity and air dryness, the product of relative humidity and saturated water vapor pressure difference, and the product of fire spreading power and soil moisture in the key factors of the burning season are used as the composite interaction terms.

3. The prediction method according to claim 2, characterized in that The constructing of a prediction sample data set based on the fire characteristic parameters includes: For each historical time period, the previous lag factor, burning season key factor, atmospheric circulation variable, vegetation potential combustion type, and composite interaction term corresponding to the historical time period are obtained from the fire characteristic parameters as input features, and the historical fire burned area corresponding to the historical time period is obtained from the fire characteristic parameters as output features; Based on the input features and output features corresponding to each historical time period, sample data corresponding to each historical time period is obtained, and based on the sample data of each historical time period, a predicted sample data set is obtained.

4. The prediction method according to claim 1, wherein: After obtaining the predicted value of the fire burning area of ​​the to-be-predicted area in each prediction time period, the prediction method further includes: A fire prediction trend graph corresponding to the area to be predicted is generated based on the fire burning area prediction value of the area to be predicted in each preset time period.

5. The prediction method according to claim 1, wherein: After obtaining the fire burning area prediction value of each to-be-predicted region in each prediction time period, the prediction method further includes: For each preset time period, the fire risk level corresponding to each to-be-predicted area is determined based on the fire burning area prediction value of each to-be-predicted area within the preset time period; Obtain a regional spatial distribution map, generate a regional fire risk map based on the fire risk level corresponding to each area to be predicted and the regional spatial distribution map, and display the fire risk levels of different areas to be predicted within the preset time period in the regional fire risk map.

6. The prediction method according to claim 1, characterized in that The method of constructing the fire burning area prediction model using the prediction sample data set includes: Using a cross-validation algorithm, a plurality of training sample sets and validation sample sets are determined based on the prediction sample data set; Using multiple sets of training sample sets to perform model training to train multiple fire burning area original prediction models; The verification sample set is used to verify each trained original fire burning area prediction model in turn, and the fire burning area original prediction model that passes the verification is determined as the fire burning area prediction model.

7. A fire burning area prediction device, characterized in that: The prediction device comprises: A characteristic parameter determination module is used to obtain input variable data required to construct a fire burn area prediction model and construct fire characteristic parameters based on the input variable data; wherein the input variable data includes historical ground meteorological variables, historical atmospheric circulation variables, historical ground ecological variables, and historical fire burn areas in historical fire burning seasons; A model building module, configured to build a prediction sample data set based on the fire characteristic parameters, and establish the fire burning area prediction model using the prediction sample data set; The prediction module is used to input the meteorological prediction data of each area to be predicted in each prediction time period into the fire burning area prediction model to obtain the fire burning area prediction value of the area to be predicted in each prediction time period.

8. The prediction device according to claim 7, characterized in that The fire characteristic parameters include a previous lag factor, a key factor of the burning season, atmospheric circulation variables, potential vegetation combustion types, a composite interaction term, and the historical fire burning area; the characteristic parameter determination module is further configured to determine the composite interaction term through the following steps: The product of atmospheric thermal intensity and air dryness, the product of relative humidity and saturated water vapor pressure difference, and the product of fire spreading power and soil moisture in the key factors of the burning season are used as the composite interaction terms.

9. An electronic device, characterized in that: include: A processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor and the memory communicate via the bus. When the processor runs the machine-readable instructions, the steps of the method for predicting the fire combustion area as described in any one of claims 1 to 6 are executed.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method for predicting the fire burning area according to any one of claims 1 to 6 are executed.

Images