CN114266392A - Forest fire early warning model construction method based on time attenuation precipitation algorithm - Google Patents
Forest fire early warning model construction method based on time attenuation precipitation algorithm Download PDFInfo
- Publication number
- CN114266392A CN114266392A CN202111573401.2A CN202111573401A CN114266392A CN 114266392 A CN114266392 A CN 114266392A CN 202111573401 A CN202111573401 A CN 202111573401A CN 114266392 A CN114266392 A CN 114266392A
- Authority
- CN
- China
- Prior art keywords
- data
- forest fire
- precipitation
- early warning
- forest
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000001556 precipitation Methods 0.000 title claims abstract description 58
- 238000010276 construction Methods 0.000 title claims abstract description 10
- 238000000034 method Methods 0.000 claims abstract description 22
- 238000012549 training Methods 0.000 claims abstract description 16
- 238000007781 pre-processing Methods 0.000 claims abstract description 7
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 12
- 238000004364 calculation method Methods 0.000 claims description 9
- 239000011159 matrix material Substances 0.000 claims description 5
- 230000001419 dependent effect Effects 0.000 claims description 4
- 238000013507 mapping Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 238000005315 distribution function Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000002310 reflectometry Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
Images
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a forest fire early warning model construction method based on a time attenuation precipitation algorithm, which comprises the following steps: acquiring precipitation data and calculating a comprehensive precipitation value based on a time attenuation algorithm; acquiring historical fire data and converting forest fire points into forest fire point densities based on a two-dimensional Gaussian convolution function; acquiring meteorological data, remote sensing data and elevation data, and preprocessing the data to obtain preprocessed data; training an SVM regression model by taking the comprehensive rainfall value, the forest fire point density and the preprocessed data as a training set to obtain a trained SVM regression model; predicting forest fire density data based on the trained SVM regression model; and dividing fire grades according to the density data of the forest fires, and combining the trained SVM regression model to obtain a forest fire early warning model. By using the method and the device, the accuracy of forest fire early warning can be improved. The method can be widely applied to the field of fire risk prediction.
Description
Technical Field
The invention relates to the field of fire risk prediction, in particular to a forest fire early warning model construction method based on a time attenuation precipitation algorithm.
Background
In most forest fire early warning models constructed by predecessors, precipitation is selected as an important characteristic factor, and meanwhile, the top row of the characteristic factors indicates that precipitation values play an important role in forest fire early warning. However, in most models of predecessors, the precipitation calculation was done as a monthly average precipitation or a regional average precipitation with equal weight averages. The average precipitation is used as the precipitation value, the attenuation effect of precipitation at different time nodes is not considered, and only the average weight is considered. The method for calculating the rainfall value cannot well express the influence effect of the rainfall value on the forest fire.
Disclosure of Invention
In order to solve the technical problems, the invention aims to provide a forest fire early warning model construction method based on a time attenuation precipitation algorithm.
The first technical scheme adopted by the invention is as follows: a forest fire early warning model construction method based on a time attenuation precipitation algorithm comprises the following steps:
s1, acquiring precipitation data and calculating a comprehensive precipitation value based on a time attenuation algorithm;
s2, acquiring historical fire data and converting forest fire points into forest fire point densities based on a two-dimensional Gaussian convolution function;
s3, acquiring meteorological data, remote sensing data and elevation data, and preprocessing the data to obtain preprocessed data;
s4, taking the comprehensive rainfall value, the forest fire point density and the preprocessed data as a training set, and training an SVM regression model to obtain a trained SVM regression model;
s5, predicting forest fire density data based on the trained SVM regression model;
and S6, dividing the fire grades according to the forest fire density data, and combining the trained SVM regression model to obtain a forest fire early warning model.
Further, the calculation formula of the comprehensive precipitation value is as follows:
in the above formula, DPrecipitationRepresenting the comprehensive precipitation value, A representing undetermined constant, n representing the previous n days, t representing time, DtRepresenting the precipitation value t days ago.
Further, the formula of the two-dimensional gaussian convolution function is expressed as follows:
in the above equation, σ represents a convolution range, x represents an abscissa difference from the center point, and y represents an ordinate difference from the center point.
Further, the preprocessed data includes a current day relative humidity, a current day maximum wind speed, a current day maximum air temperature, a normalized vegetation index, a vegetation water supply index, and an altitude.
Further, the step of acquiring meteorological data, remote sensing data and elevation data and preprocessing the data to obtain preprocessed data specifically comprises:
s31, acquiring meteorological data, remote sensing data and elevation data of a forest area;
s32, interpolating the meteorological data based on a Krigin interpolation method to obtain uniform grid meteorological data in the forest region;
s33, calculating a normalized vegetation index and a vegetation water supply index according to the remote sensing data;
and S34, obtaining altitude data according to the elevation data.
Further, the step of training the SVM regression model by using the comprehensive rainfall value, the forest fire point density and the preprocessed data as a training set to obtain the trained SVM regression model specifically comprises the following steps:
s41, training the SVM regression model by taking the comprehensive rainfall value, the uniform grid meteorological data, the normalized vegetation index, the vegetation water supply index and the altitude data in the forest area as independent variables and the forest fire point density as dependent variables and combining with an objective function to obtain the trained SVM regression model;
s42, the SVM regression model is used for normalizing the independent variable, and meanwhile, a selected Gaussian kernel is used.
Further, the formula of the objective function is expressed as follows:
ξ1i≥0,ξ2i≥0(i＝1,2,…m)
in the above equation, ω represents a weight vector of the output space,
representing a non-linear mapping, ξ, from the input space to the output space1i,ξ2iRepresenting the relaxation variable, b the offset value, epsilon the insensitive range of the loss value, m the dimension, T the transpose of the matrix.
Further, the step of dividing the fire grade according to the forest fire density data and obtaining a forest fire early warning model by combining the trained SVM regression model specifically comprises the following steps:
s61, dividing the forest fire density data into five forest fire danger areas with different levels based on a natural breakpoint method;
s62, judging that the forest fire danger area is higher than a preset grade, and sending out early warning;
and S63, obtaining a forest fire early warning model by combining the trained SVM regression model.
The method has the beneficial effects that: according to the method, a comprehensive rainfall index is calculated through a time attenuation algorithm, other characteristic factors are combined to serve as independent variables of a forest fire early warning model, discrete forest fire points are convolved into forest density to serve as dependent variables of the model, and after data preprocessing is carried out, an SVM regression model is combined to construct a forest fire early warning model with high accuracy.
Drawings
FIG. 1 is a flow chart of steps of a forest fire early warning model construction method based on a time attenuation precipitation algorithm.
Detailed Description
The invention is described in further detail below with reference to the figures and the specific embodiments. The step numbers in the following embodiments are provided only for convenience of illustration, the order between the steps is not limited at all, and the execution order of each step in the embodiments can be adapted according to the understanding of those skilled in the art.
Referring to fig. 1, the invention provides a forest fire early warning model construction method based on a time attenuation precipitation algorithm, which comprises the following steps:
s1, acquiring precipitation data and calculating a comprehensive precipitation value based on a time attenuation algorithm;
specifically, in the time attenuation model, the change amplitude of the numerical value is in direct proportion to the time of the numerical value, and the influence degree of rainfall on forest fire occurrence on different days is met. In the time attenuation model, the above mapping relationship is converted into a mathematical formula which can be expressed as:
in the above formula, N represents a numerical value, t represents time, and γ represents a decay constant.
Applied to this patent, the solution to the time decay equation can be expressed as:
N(t)＝Dte-γt (2)
in the above formula, N (t) is a quantity related to time t, representing the influence of precipitation t days ago, DtRepresenting the precipitation value t days ago.
And then, adding the precipitation influence values of the previous n days to obtain a comprehensive precipitation value, wherein the calculation formula of the comprehensive precipitation value is as follows:
in the above formula, DPrcipitationRepresenting the integrated precipitation value, A representing the undetermined constant, t representing the time, DtRepresenting the precipitation value t days ago. It is now necessary to determine the value of the undetermined constant a and to select the value of the precipitation n days before. Mutual information can represent the amount of information one random variable contains by another random variable, can be used to measure the correlation between them, and is not limited to a linear relationship. The formula for calculating mutual information is:
where p (X, Y) is the joint probability distribution function of X and Y, and p (X) and p (Y) are the edge probability distribution functions of X and Y, respectively. And selecting the constant term A value with the maximum mutual information value of the comprehensive precipitation data and the forest fire point density and the previous n-day precipitation value according to the comprehensive precipitation data obtained by different combinations of the undetermined constant term A value and the previous n-day precipitation value. After iterative calculation, when A is 0.1657 and n is 6, the calculated comprehensive precipitation value can represent the influence of the precipitation value on the fire prediction.
S2, acquiring historical fire data and converting forest fire points into forest fire point densities based on a two-dimensional Gaussian convolution function;
specifically, the historical fire data adopted by the invention is from a global forest fire historical data website provided by NASA. The data provided by the website is low latency fire activity data recorded from satellite images provided by the U.S. national aerospace and department of agriculture, department of forestry. The data form of the forest fire point is a scatter diagram in the forest area, and in the method, the scattered forest fire points are converted into continuous forest fire point density through two-dimensional Gaussian convolution, so that the classification problem is converted into the regression problem. Each discrete forest fire point represents the situation of forest fire at the periphery, and after two-dimensional Gaussian convolution, the higher the density value of the forest fire points is, the more serious the forest fire at the periphery is.
The expression of the two-dimensional gaussian convolution is as follows:
in the above equation, σ represents the convolution range. The convolution radius is taken as ten kilometers in the present invention.
S3, acquiring meteorological data, remote sensing data and elevation data, and preprocessing the data to obtain preprocessed data;
s3.1, acquiring meteorological data, remote sensing data and elevation data of a forest area;
specifically, the present invention also selects the current day relative humidity, the current day maximum wind speed, the current day maximum air temperature, the normalized vegetation index (NDVI), the vegetation water supply index (VSWI) and the altitude of the forest area as the characteristic factors. Wherein the relative humidity of the current day, the maximum wind speed of the current day and the maximum air temperature of the current day are obtained through data provided by a local meteorological office.
S3.2, interpolating the meteorological data based on a Krigin interpolation method to obtain continuous meteorological data in the forest region.
Specifically, the meteorological data are discrete and discontinuous in the forest region; the invention needs to obtain continuous meteorological data in the forest region, so that linear interpolation needs to be carried out on the data, each data point in the forest region has the meteorological data, and the invention utilizes a kriging interpolation algorithm to carry out interpolation on the meteorological data.
S3.3, calculating a normalized vegetation index and a vegetation water supply index according to the remote sensing data;
the vegetation coverage of the forest is represented by NDVI model data, and NDVI is a normalized vegetation index which reflects the growth state and the vegetation coverage of the vegetation. The NDVI is calculated by using the reflection value of the near infrared band and the reflection value of the red light band, and in this paper, the NDVI is calculated by using the reflectivity data of the bands 1 and 2 in the MOD021KM remote sensing data, and the formula is as follows:
the value of NDVI is [ -1,1], when it is negative, it means that the ground is covered by cloud, water, snow, etc., and has high reflection to visible light; when positive, it indicates vegetation coverage and increases with increasing coverage.
The VSWI index is also called as vegetation water supply index and can reflect the growth condition of crops and the water content of vegetation to a certain extent. The formula is as follows:
NDVI is a normalized vegetation index, Ts represents the earth surface temperature, and Ts data is obtained from MOD11A2 remote sensing data which records the comprehensive earth surface temperature value for eight days. The greater the VSWI value, the higher the vegetation water content.
And S3.4, acquiring altitude data according to the elevation data.
Altitude data is acquired from the SPTMDEM 90M elevation data.
S4, taking the comprehensive rainfall value, the forest fire point density and the preprocessed meteorological data as a training set, and training an SVM regression model to obtain a trained SVM regression model;
specifically, the comprehensive precipitation value and the preprocessed meteorological data are used as independent variables, the forest fire point density is used as a dependent variable, then a Gaussian kernel is selected as an SVM kernel function, the data in the independent variables are normalized, in addition, a Gram matrix is used in SVM model training, and the meaning of the Gram matrix is that when two samples are close to each other, the value is made to be 1, and when the distance is far away, the value is 0. And the kernel scale is a scaling parameter used for evaluating the scaling size of a proper Gram matrix, the kernel scale is specified to be 3 after screening calculation, and training is carried out in the SVM regression model by using training set data. And then, using the trained SVM regression model in the test set to obtain forest fire density data in the prediction set. The objective function expression of the SVM regression model is as follows:
s5, predicting forest fire density data based on the trained SVM regression model;
and S6, dividing the fire grades according to the forest fire density data, and combining the trained SVM regression model to obtain a forest fire early warning model.
S6.1, dividing the forest fire density data into a plurality of forest fire danger areas with different levels based on a natural breakpoint method;
after forest fire density data in the prediction set are obtained, dividing the forest area into five fire early warning levels by a natural breakpoint method, wherein the higher the level is, the higher the risk of fire occurrence in the area is. The natural breakpoint method calculates the variance of each class, then calculates the sum of the variances, and compares the classification quality by using the variance sum, and the minimum value is the optimal classification result. The calculation method is as follows:
GVF＝(SDAM-SCDM)/SDAM (9)
wherein GVF is the goodness of fit of the variance, SDAM is the sum of the squares of the deviations of the mean, and the calculation is as follows:
wherein
Is a mean value, xiIs the density of forest fires. And SCDM is the sum of the squared deviations of the class means. The calculation is as follows:
by using the natural breakpoint method, the present invention divides forest fire danger regions into five different levels according to the following numerical regions, and divides forest fire density (— infinity, 0) into level one, [0,0.08) into level two, [0.08,0.27) into level three, [0.27,0.45) into level four, [0.45, ∞) into level five.
S6.2, judging that the forest fire danger area is higher than a preset grade, and sending out early warning, wherein the preset grade is grade three;
and S6.3, obtaining a forest fire early warning model by combining the trained SVM regression model.
And when the forest fire early warning model is constructed, using the forest regions with the fire danger level of more than three for predicting the occurrence of the forest fire, wherein the prediction accuracy rates of the forest regions in the test set are 90.13%, 93.04% and 87.5% respectively. If the average precipitation value is used for representing the influence of precipitation on the forest fire, and other factors are unchanged, the prediction accuracy of the forest region in the test set is 85.17%, 89.50% and 84.90% respectively. By comparing the comprehensive rainfall value obtained by calculating the time attenuation algorithm and taking the comprehensive rainfall value as the characteristic factor, the accuracy of the forest fire early warning model can be improved by about 5%. The model result is shown to be capable of better predicting the forest fire risk situation.
While the preferred embodiments of the present invention have been illustrated and described, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (8)
1. A forest fire early warning model construction method based on a time attenuation precipitation algorithm is characterized by comprising the following steps:
s1, acquiring precipitation data and calculating a comprehensive precipitation value based on a time attenuation algorithm;
s2, acquiring historical fire data and converting forest fire points into forest fire point densities based on a two-dimensional Gaussian convolution function;
s3, acquiring meteorological data, remote sensing data and elevation data, and preprocessing the data to obtain preprocessed data;
s4, taking the comprehensive rainfall value, the forest fire point density and the preprocessed data as a training set, and training an SVM regression model to obtain a trained SVM regression model;
s5, predicting forest fire density data based on the trained SVM regression model;
and S6, dividing the fire grades according to the forest fire density data, and combining the trained SVM regression model to obtain a forest fire early warning model.
2. The method for constructing the forest fire early warning model based on the time attenuation precipitation algorithm is characterized in that the calculation formula of the comprehensive precipitation value is as follows:
in the above formula, DPrecipitationRepresenting the comprehensive precipitation value, A representing undetermined constant, n representing the previous n days, t representing time, DtRepresenting the precipitation value t days ago.
3. The method for constructing the forest fire early warning model based on the time attenuation precipitation algorithm, as claimed in claim 2, wherein the formula of the two-dimensional Gaussian convolution function is represented as follows:
in the above equation, σ represents a convolution range, x represents an abscissa difference from the center point, and y represents an ordinate difference from the center point.
4. The method for constructing the forest fire early warning model based on the time attenuation precipitation algorithm according to claim 3, wherein the preprocessed data comprise current day relative humidity, current day maximum wind speed, current day maximum air temperature, normalized vegetation index, vegetation water supply index and altitude.
5. The forest fire early warning model building method based on the time attenuation precipitation algorithm as claimed in claim 4, wherein the step of obtaining meteorological data, remote sensing data and elevation data and preprocessing the data to obtain preprocessed data specifically comprises:
s31, acquiring meteorological data, remote sensing data and elevation data of a forest area;
s32, interpolating the meteorological data based on a Krigin interpolation method to obtain uniform grid meteorological data in the forest region;
s33, calculating a normalized vegetation index and a vegetation water supply index according to the remote sensing data;
and S34, obtaining altitude data according to the elevation data.
6. The forest fire early warning model construction method based on the time attenuation precipitation algorithm is characterized in that the SVM regression model is trained by taking the comprehensive precipitation value, the forest fire point density and the preprocessed data as a training set to obtain the trained SVM regression model, and the method specifically comprises the following steps of:
s41, training the SVM regression model by taking the comprehensive rainfall value, the uniform grid meteorological data, the normalized vegetation index, the vegetation water supply index and the altitude data in the forest area as independent variables and the forest fire point density as dependent variables and combining with an objective function to obtain the trained SVM regression model;
s42, the SVM regression model is used for normalizing the independent variable, and meanwhile, a selected Gaussian kernel is used.
7. The method for constructing the forest fire early warning model based on the time attenuation precipitation algorithm is characterized in that the formula of the objective function of the SVM regression model is as follows:
ξ1i≥0,ξ2i≥0(i＝1,2,…m)
in the above equation, ω represents a weight vector of the output space,
representing a non-linear mapping, ξ, from the input space to the output space1i,ξ2iRepresenting the relaxation variable, b the offset value, epsilon the insensitive range of the loss value, m the dimension, T the transpose of the matrix.
8. The method for constructing the forest fire early warning model based on the time attenuation precipitation algorithm as claimed in claim 7, wherein the step of dividing the fire grade according to the forest fire density data and obtaining the forest fire early warning model by combining with the trained SVM regression model specifically comprises:
s61, dividing the forest fire density data into five forest fire danger areas with different levels based on a natural breakpoint method;
s62, judging that the forest fire danger area is higher than a preset grade, and sending out early warning;
and S63, obtaining a forest fire early warning model by combining the trained SVM regression model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111573401.2A CN114266392A (en) | 2021-12-21 | 2021-12-21 | Forest fire early warning model construction method based on time attenuation precipitation algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111573401.2A CN114266392A (en) | 2021-12-21 | 2021-12-21 | Forest fire early warning model construction method based on time attenuation precipitation algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114266392A true CN114266392A (en) | 2022-04-01 |
Family
ID=80828298
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111573401.2A Pending CN114266392A (en) | 2021-12-21 | 2021-12-21 | Forest fire early warning model construction method based on time attenuation precipitation algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114266392A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115083130A (en) * | 2022-08-24 | 2022-09-20 | 深圳市博容能源有限公司 | Long-acting distributed emergency monitoring alarm system and method |
| CN115146705A (en) * | 2022-05-27 | 2022-10-04 | 南京林业大学 | Method for recognizing forest lightning stroke fire by combining remote sensing and surface lightning stroke fire environment |
-
2021
- 2021-12-21 CN CN202111573401.2A patent/CN114266392A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115146705A (en) * | 2022-05-27 | 2022-10-04 | 南京林业大学 | Method for recognizing forest lightning stroke fire by combining remote sensing and surface lightning stroke fire environment |
| CN115083130A (en) * | 2022-08-24 | 2022-09-20 | 深圳市博容能源有限公司 | Long-acting distributed emergency monitoring alarm system and method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Belal et al. | Drought risk assessment using remote sensing and GIS techniques | |
| Murphy et al. | Reassessing rainfall in the Luquillo Mountains, Puerto Rico: Local and global ecohydrological implications | |
| TWI769763B (en) | Data processing method and device | |
| CN114266392A (en) | Forest fire early warning model construction method based on time attenuation precipitation algorithm | |
| Kim et al. | Assessment of agricultural drought vulnerability to climate change at a municipal level in South Korea | |
| Newlands et al. | Validation and inter‐comparison of three methodologies for interpolating daily precipitation and temperature across Canada | |
| Ratri et al. | A comparative verification of raw and bias-corrected ECMWF seasonal ensemble precipitation reforecasts in Java (Indonesia) | |
| Gitau et al. | Spatial coherence and potential predictability assessment of intraseasonal statistics of wet and dry spells over Equatorial Eastern Africa. | |
| McRoberts et al. | The use of a high-resolution standardized precipitation index for drought monitoring and assessment | |
| Roznik et al. | Improving agricultural microinsurance by applying universal kriging and generalised additive models for interpolation of mean daily temperature | |
| Glade et al. | Productivity and phenological responses of natural vegetation to present and future inter-annual climate variability across semi-arid river basins in Chile | |
| CN109919515A (en) | Eco-Environmental Synthetic Analyses method and device | |
| Amin et al. | Monitoring agricultural drought using geospatial techniques: a case study of Thal region of Punjab, Pakistan | |
| Gummadi et al. | Evaluation of multiple satellite precipitation products for rainfed maize production systems over Vietnam | |
| Prajapati et al. | PCA–based composite drought index for drought assessment in Marathwada region of Maharashtra state, India | |
| Shin et al. | High-resolution wind speed forecast system coupling numerical weather prediction and machine learning for agricultural studies—a case study from South Korea | |
| Shefine | Analysis of meteorological drought using SPI and large-scale climate variability (ENSO)-A case study in North Shewa Zone, Amhara Regional State, Ethiopia | |
| Martinaitis et al. | The historic rainfalls of Hurricanes Harvey and Florence: A perspective from the Multi-Radar Multi-Sensor system | |
| Lu et al. | A methodology for multiobjective evaluation of precipitation products for extreme weather (in a data-scarce environment) | |
| Woodley | Optimizing the measurement of convective rainfall in Florida | |
| Camberlin et al. | Spatial interpolation of daily rainfall stochastic generation parameters over East Africa | |
| Khoshnazar et al. | Wet-environment Evapotranspiration and Precipitation Standardized Index (WEPSI) for drought assessment and monitoring | |
| Sahilu et al. | Spatiotemporal trends and variability of rainfall across agro-ecologies in East Guji Zone, Southeast Ethiopia | |
| Jozaghi | Improving operational hydrologic forecasting via conditional bias-penalized multi-sensor precipitation estimation and multi-model streamflow prediction | |
| Sanabria et al. | Mapping of australian fire weather potential: Observational and modelling studies |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |