KR102108752B1 - Method for realtime forest fire danger rating forecasting in north korea - Google Patents

Abstract

The present invention comprises the steps of obtaining a satellite image of the inaccessible area; Extracting an ignition point from the acquired satellite image in pixel units; Removing diffusion ignition points extracted from the independent ignition points in consideration of the degree of diffusion of wildfires and the vector of wind by time series; Obtaining weather information, clinical information, and topographical information at a corresponding time for a unit grid including an independent ignition point obtained for each time series for each unit grid; Deriving each function for calculating a weather risk index, a clinical risk index, and a terrain risk index from the obtained weather information, clinical information, and terrain information; Calculating current weather risk index, clinical risk index, and terrain risk index by inputting weather information, clinical information, and terrain information for each unit grid obtained in real time to each obtained function; And extracting the wildfire risk index for each unit grid from the calculated current weather risk index, clinical risk index, and terrain risk index, and displaying the result in a identifiable manner on the map for each unit grid to display on the user terminal. It provides a real-time forest fire risk forecasting method for inaccessible areas characterized by including.

Description

Real-time forest fire risk forecast method in inaccessible areas {METHOD FOR REALTIME FOREST FIRE DANGER RATING FORECASTING IN NORTH KOREA}

The present invention relates to a method for predicting a forest fire risk, and more specifically, it is possible to quickly and accurately predict a probability of a forest fire occurring in an inaccessible area in the near future or in the near future. Real-time forest fire risk prediction method for inaccessible areas that can significantly reduce the likelihood of occurrence.

The fire in the inaccessible areas, including North Korea, is not simply a problem that arises in the territory occupied by a group of heterogeneous systems called North Korea, and North Korea is a nation like ours, and in terms of Korea's territorial concept prescribed by the Constitution Recognizing that it constitutes part of the fire, fires in North Korea are also required to have the same degree of problem recognition as fires in Korea.

It is very important to prevent and manage massive national losses such as large forest fires in inaccessible areas in advance, as North Korea's territory is also the basis for our descendants to live after unification and a common asset inherent in our eternal nation. .

In North Korea, it is difficult to collect accurate data on the time and location of wildfires that occur annually due to the system's rain or fall, and it is impossible to quickly and accurately predict the risk of forest fires because there is no system to interpret it.

Moreover, in North Korea, the fire extinguishing equipment is not modern, and the difficulty of extinguishing is made up mainly by manual work by manpower, and the prevention of forest fire is very important and urgent along with the countermeasures against the fire due to the weight of human life caused by the fire. .

Currently, in order to prevent forest fires, the excluded countries operate their own developed forest fire risk forecasting system. In the United States, forest fire research began in 1913, and the National Forest Fire Risk Forecast (NFDRS) began in 1978. In Canada, forest fire research began in 1925, and the Canadian forest fire risk assessment system was developed and put into practical use in 1970 (CFFDRS). ). In Korea, the study of forest fire risk forecasting systems using fuel humidity measuring rods began in 1994-1996 and has been put into practical use in 23 regions including Seoul since 1997. However, these systems are merely systems for predicting forest fire risk by acquiring weather information and the like as basic information at accessible points.

Accordingly, an inaccessible area that can quickly and accurately predict the probability of a forest fire occurring in the inaccessible area at present or in the near future, and can significantly reduce the likelihood of a large forest fire by warning the risk of forest fire in advance. There is an urgent need for a real-time forest fire risk prediction method and system.

The present invention has been devised to solve the problems of the prior art as described above, the purpose of which can be predicted quickly and accurately with respect to the probability of the risk of wildfires occurring in the near future or in the near future. It is to provide a method for real-time forest fire risk prediction in inaccessible areas that can significantly reduce the possibility of large-scale forest fires by warning in advance the danger of forest fires.

The technical problem of the present invention as described above is achieved by the following means.

(1) obtaining a satellite image of an inaccessible area; Extracting an ignition point from the acquired satellite image in pixel units; Removing diffusion ignition points extracted from the independent ignition points in consideration of the degree of diffusion of wildfires and the vector of wind by time series; Obtaining weather information, clinical information, and topographical information at a corresponding time for a unit grid including an independent ignition point obtained for each time series for each unit grid; Deriving each function for calculating a weather risk index, a clinical risk index, and a terrain risk index from the obtained weather information, clinical information, and terrain information; Calculating current weather risk index, clinical risk index, and terrain risk index by inputting weather information, clinical information, and terrain information for each unit grid obtained in real time to each obtained function; And extracting the wildfire risk index for each unit grid from the calculated current weather risk index, clinical risk index, and terrain risk index, and displaying the result in a identifiable manner on the map for each unit grid to display on the user terminal. Real-time forest fire risk prediction method for inaccessible areas, characterized in that it comprises.

(2) In the above (1),

When a forest fire is detected in a time series at a specific pixel, a real-time forest fire risk forecasting method for an inaccessible area, characterized in that the fire point of the first detected time point is determined as the ignition point and the ignition time.

(3) In the above (1),

A method for real-time forest fire risk prediction in inaccessible areas, characterized in that individual forest fires are considered when forest fires of different periods are detected in the same pixel.

(4) In the above (1),

A method for real-time forest fire risk prediction in an inaccessible area characterized in that a fire point appearing within a 3 × 3 pixel area centered on the fire point is excluded as a diffuse ignition point on the first day after the ignition point.

(5) In the above (4),

A method for real-time forest fire risk prediction in an inaccessible area characterized in that a fire point that appears within a 5x5 pixel area centered on the fire point is excluded as a diffuse fire point from 2 days after the ignition point.

(6) In the above (1),

A real-time forest fire risk forecasting method for inaccessible areas, characterized in that the function for calculating the weather risk index is represented by Equation 1.

[1 + exp {-(2.745+ (0.0905 * highest temperature per day))-(0.0517 * lowest relative humidity per day) + (0.0334 * effective humidity) + (0.1283 * average wind speed))} ^-1 ] ^-1

According to the present invention, it is possible to quickly and accurately predict the probability of a fire that may occur in the inaccessible area in the near future or in the near future, and predict the risk of forest fire in advance, thereby remarking the possibility of a large forest fire in the inaccessible area. Can be reduced.

1 is an example of a process diagram of a forest fire risk forecasting process according to the present invention.
Figure 2 is an example of the overall configuration of a forest fire risk forecasting system according to the present invention.
3 is an example of a dot filtering process according to the present invention.
4 is a result of re-estimation of wildfire ignition points using year-by-year (2011-2015) wildfire occurrence information
5 is an FMI index generated through clinical classification
Figure 6 shows the tendency of forest fires by altitude section and slope using reanalysis data
7 is a TMI index generated through topographic classification
8 is a weather data DB construction of a fire spot by reanalysis
9 is a correlation matrix between input data of probability of forest fires caused by weather.
10 is a logistic regression model analysis result of meteorological factors affecting the occurrence of forest fires
11 is a verification result of the weather risk index (DWI) model
12 is a real-time weather information connection and data processing, forest fire risk forecast analysis system diagram
13 shows weather information on April 15, 2014 (15:00), hourly weather risk index and fire risk rating
14 shows weather information on April 25, 2014 (15:00), hourly weather risk index, and fire risk rating.
Figure 15 is the weather information on April 27, 2014 (15:00), hourly weather risk index and forest fire risk rating
16 is a flowchart of an automated real-time forest fire risk prediction model.
Fig. 17 shows the details of the forest fire risk forecast pilot system in the inaccessible area
18 is a real-time forest fire risk forecast pilot system manager page in inaccessible areas.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. DETAILED DESCRIPTION The following detailed description, together with the accompanying drawings, is intended to describe exemplary embodiments of the present invention, and is not intended to represent the only embodiments in which the present invention may be practiced. The following detailed description includes specific details to provide a thorough understanding of the present invention. However, one of ordinary skill in the art to which the present invention pertains knows that the present invention may be practiced without these specific details.

In some cases, in order to avoid obscuring the concept of the present invention, well-known structures and devices may be omitted, or block diagrams centered on the core functions of each structure and device may be illustrated.

Throughout the specification, when a part "comprising or including" a certain component, this means that other components may be further included instead of excluding other components, unless otherwise specified. do. In addition, terms such as “… unit”, “… group”, and “module” described in the specification mean a unit that processes at least one function or operation, which may be implemented by hardware or software or a combination of hardware and software. have. In addition, "a (a or an)", "one (one)," "the (the)" and similar related terms in the context of describing the present invention (especially in the context of the following claims) is different herein. It may be used in a sense including both singular and plural unless indicated or clearly contradicted by context.

In describing embodiments of the present invention, when it is determined that a detailed description of known functions or configurations may unnecessarily obscure the subject matter of the present invention, the detailed description will be omitted. In addition, terms to be described later are terms defined in consideration of functions in an embodiment of the present invention, which may vary according to a user's or operator's intention or practice. Therefore, the definition should be made based on the contents throughout this specification.

A real-time forest fire risk prediction method for an inaccessible area according to the present invention, as shown in FIG. 1,

Obtaining a satellite image of the inaccessible area; Extracting an ignition point from the acquired satellite image in pixel units; Removing diffusion ignition points extracted from the independent ignition points in consideration of the degree of diffusion of wildfires and the vector of wind by time series; Obtaining weather information, clinical information, and topographical information at a corresponding time for a unit grid including an independent ignition point obtained for each time series for each unit grid; Deriving each function for calculating a weather risk index, a clinical risk index, and a terrain risk index from the obtained weather information, clinical information, and terrain information; Calculating current weather risk index, clinical risk index, and terrain risk index by inputting weather information, clinical information, and terrain information for each unit grid obtained in real time to each obtained function; And extracting the wildfire risk index for each unit grid from the calculated current weather risk index, clinical risk index, and terrain risk index, and displaying the result in a identifiable manner on the map for each unit grid to display on the user terminal. Includes.

In order to implement the above-described real-time forest fire risk prediction method for the inaccessible area, the real-time forest fire risk prediction system according to the present invention is as shown in FIG.

A weather information providing server 100 that provides weather information for each unit grid divided into unit grids nationwide;

A forest fire risk forecast analysis server 200 that derives a forest fire risk index using the weather information received in real time from the weather information providing server, and clinical information and terrain information in a unit grid, and transmits a forecast according to the risk index to a client Including,

The forest fire risk prediction analysis server includes a satellite image acquisition means (201) for acquiring a satellite image of an inaccessible area; A dot extraction means 202 for extracting a dot point in pixels from an image obtained from the satellite image acquisition means; Fire point filtering means 203 which diffuses from the independent ignition point and removes the extracted ignition point by considering the diffusion degree of forest fire and the wind vector by time series; A basic information acquisition means (204) for acquiring weather information, clinical information, and topographical information at each point in time for a unit lattice including independent ignition points obtained for each time series; Function deriving means (205) for deriving respective functions for calculating weather risk index, clinical risk index, and terrain risk index from weather information obtained from the basic information acquisition means, clinical information, and terrain information; And inputting meteorological information, clinical information, and topographical information for each unit grid obtained in real time to each of the obtained functions to calculate a current meteorological risk index, a clinical risk index, and a topographical risk index, and the calculated current meteorological risk index and clinical risk. It includes a forest fire forecasting agent 210 including a forest fire risk index calculation unit 206 for calculating a fire risk index and a risk class for each grid from the index and the terrain risk index.

Hereinafter, a method for real-time forest fire risk prediction in an inaccessible area according to the present invention will be described in more detail with reference to FIGS. 1 and 2.

The weather information providing server 100 preferably receives current weather information (eg, relative humidity, effective humidity, wind speed, temperature, etc.) from meteorological observation stations installed in all parts of the country at real time to constant time intervals, and analyzes the risk of forest fire risk forecast. (200).

Preferably, the weather information may be obtained for each unit grid when the whole country is divided using a certain number of unit grids (for example, 5 km × 5 km).

The clinical information can be provided, for example, from a server that provides a numerical clinical scale of 1 / 25,000 scale issued by the National Forest Research Institute, and provides clinical information of the inaccessible area to the forest fire risk forecast analysis server 200. Clinical information is not changed in a short time, so pre-stored information can be applied as it is.

The topographic information can be provided from a server that provides a numerical topographic map of 1 / 25,000 scale issued by the National Geographic Information Institute, and provides the topographic information of the inaccessible area to the forest fire risk forecast analysis server 200. Like clinical information, terrain information is not changed in a short period of time, so pre-stored information can be applied as it is.

The forest fire risk prediction analysis server 200 of the present invention derives a forest fire risk index using real-time weather information received from the weather information providing server 100, and clinical information and terrain information in a unit grid, and according to the risk index, The forecast is sent to the client.

In the present invention, the wildfire risk index is calculated by objectively converting the wildfire risk into a numerical value and calculated by the wildfire risk prediction analysis server 200. It is a value calculated by calculating the clinical risk index and the topographic risk index and assigning a predetermined weight to each of these indices.

The “climatic risk index” refers to the numerical value of the risk of forest fire using the weather information according to the weather classification, and the “clinical risk index” means that the forest fire may occur using clinical information according to the clinical classification. It means that the level of risk is numerically indexed, and the "topographical index" means that the risk of wildfires can be numerically indexed using terrain information according to the terrain classification.

As a preferred embodiment of the present invention, as a function for calculating weather risk index, clinical risk index and topographic risk index in an inaccessible area are as follows.

(1) Meteorological risk index (DWI) calculation function

According to the present invention, as the weather information, temperature (eg, average temperature, maximum temperature, minimum temperature), humidity (eg, effective humidity, maximum humidity, minimum humidity), wind speed (eg, average wind speed, maximum wind speed, minimum wind speed) Can be used. In addition, it is possible to use other information such as precipitation. It is preferable to receive such weather information from the weather information providing server 100 24 times a day for one hour. The weather information is recorded and stored in a storage device (not shown).

The weather information may use one or more information correlated with the occurrence of forest fires depending on the region. In the present invention, the weather risk index is calculated according to a function f (atmospheric temperature, humidity, wind speed) that can be determined experimentally using such weather information as a variable, and an example of calculating the weather risk index is presented as a preferred embodiment below.

DWI = [1 + exp {-(2.745+ (0.0905 * highest temperature per day))-(0.0517 * lowest relative humidity per day) + (0.0334 * effective humidity) + (0.1283 * average wind speed))} ^-1 ] ^-1

(2) Function for calculating the clinical risk index (FMI)

FMI = ΣP _i I _i (In the above formula, Pi is a type i clinically distributed within a unit lattice, and Ii represents a risk index of a type i clinical.)

For example, as shown in FIG. 5, 10 is given to deciduous broad-leaved trees, 4 for evergreen conifers, and 2 for mixed forests. According to this, if the number of deciduous broad-leaved trees occupies 87 places (69%), the evergreen conifers 18 places (14.3%), and the mixed forests 21 places (16.7%), the clinical risk index is calculated as follows.

Clinical risk index :

0.690 × 10 + 0.143 × 4 + 0.167 × 2 = 7.806

(3) Functions for calculating the Terrain Risk Index (TMI)

TMI = ΣP _{i, elevation} I _{i, elevation} + ΣP _{j, orientation} I _{j, orientation}

In the above equation, Pi represents the distribution ratio of the elevation or orientation of i-type present in the unit grid, and Ii represents the hazard index of elevation or orientation of i-type.

As shown in Fig. 7, the topographical risk index is presented in the example of the highest 5 in the north, 4.5 in the northeast / southeast, 4.0 in the east / northwest, and 3 in the southwest. These figures can be obtained based on empirical statistics related to forest fires. In addition, with respect to altitude, 5, 550 to 1044 m, the highest of which is 550 m or less, are given 3.5, 1,537 m or more to 0.5.

According to this, if there are 80 places (80.0%) in 550m, 18 places (18.0%) in 550 ~ 1044m, and two places (2.0%) in 1,537m or more, the topographic risk index for elevation is as follows: Can be calculated.

Elevation Terrain Risk Index :

0.80 × 5 + 0.18 × 3.5 + 0.02 × 0.5 = 4.64

Similarly, within the unit grid, 10 places (10.0%) in the north, 30 places in the northeast (30.0%), 20 places in the northeast (20.0%), 10 places in the southeast (10.0%), 30 places in the southwest (30.0%) If exists, the topographic risk index for defense is as follows.

Defense Terrain Risk Index :

0.10 × 5 + 0.30 × 4.5 + 0.20 × 4.0 + 0.10 × 4.5 + 0.30 × 3.0 = 4.00

Overall Terrain Risk Index :

Therefore, according to the example given above, the terrain risk index within the unit grid is 4.64 + 4.00 = 8.64.

Preferably, the meteorological risk index, clinical risk index, and topographic risk index are extracted for each unit grid when the inaccessible area is divided using a certain number of unit grids (eg, 5㎞ × 5㎞). Can be.

For specific calculation examples in the inaccessible areas of the meteorological risk index, clinical risk index, and terrain risk index, reference will be made to embodiments of the present invention described later.

The forest fire risk forecast analysis server 200 according to the present invention includes a forest fire forecast agent 210 for calculating the current weather risk index, clinical risk index, and terrain risk index in an inaccessible area.

The forest fire forecasting agent 210 includes a satellite image acquisition means 201, a fire point extraction means 202, a fire point filtering means 203, a basic information acquisition means 204, and a function extraction means 205.

The satellite image acquisition means 201 is for acquiring a satellite photographed image of an inaccessible area, and the satellite photographed image may be, for example, satellite information obtained using a MODIS system (eg, pixel resolution of 250, 500, 100m). have.

The fire point extraction means 202 extracts the fire point in units of pixels from the satellite image obtained by the satellite image acquisition means 201, and uses the color information of each pixel to extract a fire spot determined to have occurred in the fire. . At this time, the fire point may be an independent ignition point that is an initial ignition point, or a diffusion ignition point that is diffused by the influence of wind or the like from the independent ignition point.

The point filtering means 203 removes the diffusion ignition point from among the points extracted by the point extraction means 202 in consideration of the degree of spread of the forest fire and the wind vector by time series, and performs a filtering process to obtain only the point as an independent ignition point. .

In the present invention, preferably, when a picture point is extracted in a time series from a specific pixel, only the picture point at the time of the first extraction is specified as an independent ignition point.

In addition, when a different point of time is extracted from the same pixel, each wildfire, that is, each point is regarded as an independent ignition point.

In the present invention, as shown in FIG. 3, in order to determine the diffusion ignition point, a flash point appearing within a 3x3 pixel area centered on the corresponding point on the first day after the ignition point is not counted as an independent ignition point except for the diffusion ignition point. This is because it is difficult to see it co-fired in a nearby area over a short period of time, and it is more appropriate to see it as a fire spread by the influence of wind.

Furthermore, in the present invention, from 2 days after the ignition point, it is desirable to exclude the point of view in the 5 × 5 pixel area centered on the point of view, as a diffusion ignition point when considering the degree of spread of the fire and the wind vector. This is because after 2 days, the probability that the measured point in the 5 × 5 pixel region is statistically a diffuse point derived from the central independent ignition point is high.

The basic information acquisition means 204 acquires weather information, clinical information, and topographic information for each unit grid for a unit grid including an independent ignition point obtained by the dot filtering means 203 for each unit grid. At this time, the method for obtaining the weather information, clinical information, and topographic information is as described above.

The function deriving means 205 according to the present invention derives each function for calculating the weather risk index, the clinical risk index, and the terrain risk index from the weather information, clinical information, and terrain information obtained from the basic information acquisition means 204.

The forest fire risk index calculation unit 206 calculates the current weather risk index, clinical risk index, and terrain risk index by inputting weather information, clinical information, and terrain information for each unit grid obtained in real time to each of the obtained functions. Calculate the forest fire risk index and risk grade for each grid from the current weather risk index, clinical risk index, and topographic risk index.

In the present invention, preferably, after assigning a weight according to each risk index, the forest fire risk index is finally extracted. For example, in a preferred embodiment of the present invention, the weather risk index is multiplied by 0.6 to 0.8, preferably 0.7, as a weight, and the clinical risk index and topographic risk index are multiplied by 1.2 to 1.8, preferably 1.5, respectively. In total, this is extracted for each unit grid as a forest fire risk index, and the value is provided to the forest fire risk prediction analysis server 200. It was experimentally confirmed that accurate prediction is possible when the above weight range is applied.

The risk grade can be displayed by dividing it into four grades, for example, interest, caution, caution, and severity, by classifying the fire risk index calculated as described above.

As described above, the forest fire risk forecast in the inaccessible area provided by the forest fire risk analysis server 200 may be provided in units of the unit grid as described above, and through a predetermined format according to the request of the client 300 To be displayed.

Hereinafter, the contents of the present invention will be described in detail with reference to more specific embodiments. However, the exemplified examples below are only presented to help the understanding of the present invention and should not be construed as limiting the scope of the present invention.

[Example 1] Extraction of wildfire information from inaccessible land and DB re-analysis in the last 5 years

(A) Extraction of information from wildfire points

Based on the results of extracting a total of 3,637 fire spot points using the MODIS fire spot (MOD14) data from the last 5 years (2011-2015), the following hypothesis was established to spread the fire ignition points. Duplicate forest fire information was removed.

1) When a forest fire is detected in a time series at a specific pixel, the fire spot at the first detection time is regarded as the ignition point.

2) If a forest fire with a difference of more than 5 days is detected in the same pixel, it is regarded as an individual forest fire.

3) Identify the presence or absence of forest fire spread from the point of ignition to 4 days after the point of ignition

ㅇ Remove fire spot after setting 3 × 3 pixel area from the time of ignition to after 1 day

ㅇ After 2 days, check the 5 × 5 pixel area

ㅇ In case of missing data due to clouds, the data of the fire spot afterwards is grasped.

4) Using the UV vector information of the wind, it is possible to remove the pixels judged to be the spread of forest fire

(B) Reanalysis of wildfire information

As a result of analyzing the wind speed and wind direction information for the fire spot to extract the fire ignition point, it was analyzed that the wind of the west wind series including the northwest wind and the southwest wind was blowing strongly in the range of the wind speed of 10m / s ~ 30m / s. In addition, when the wind speed was 10 m / s or less, the main wind direction was varied, but the frequency of the southwest wind series was relatively high.

As a result of re-analyzing the MODIS fire spot information detected from 2011 to 2015 by the hypothesis of extracting the above-mentioned fire ignition point, about 60% of the previously detected 3,637 pixels overlap with forest fire diffusion. It was estimated as the location of forest fires. Therefore, a total of 1,469 wildfire pixels were finally extracted by removing the overlapping points by forest fire diffusion estimation. Looking at the fire ignition points extracted by year, 368 in 2011, 151 in 2012, 172 in 2013, 554 in 2014, and 224 in 2015 were extracted. In particular, in 2015, about 71% (fire spot 538) out of 762 fire spots were removed as the diffusion point, and it was analyzed that the spread of wildfire was the most severe, and in 2013, about 39% (111 fire spots) was removed. It was analyzed that the forest fire spread was the least (Fig. 4).

As a result of analyzing the annual and monthly forest fire occurrence frequency of the reanalyzed forest fire ignition point, South Korea showed the highest occurrence of forest fire in April, 43.0% (2013) to 70.7% (2011) in the spring period. It was analyzed to be the same as the intensive fire season in. Next, it was analyzed that the occurrence of forest fires concentrated in May from 7.6% (2011) to 47.0% (2012). As a result, it was about one month late compared to South Korea where forest fires were concentrated in March and April. Forest fires tended to occur in April and May.

[Example 2] Development of clinical risk index model for inaccessible areas

As a result of analyzing the clinical characteristics in order to calculate the clinical risk index for the forest fire point in the inaccessible area, the deciduous broad-leaved forest occupied the highest ratio at 59.4%. Next, it was analyzed that the evergreen coniferous forests accounted for 21.7% and the mixed forests 18.9%. The results of this study showed that forest fires were high in deciduous broad-leaved forests in inaccessible areas, as opposed to the results of the evergreen coniferous forests (69.0%), which showed the highest rate of forest fire in South Korea. When the results of reanalysis (total 1,469 fire spot points) compared to before reanalysis (total 3,637 fire spot points), the ratio of deciduous forests increased by 3.1% after reanalysis, but about 1.8% for evergreen coniferous forests and mixed forests Showed a decrease of about 1.3% (Table 1). Therefore, the calculated data was analyzed by frequency of each fire condition, followed by indexing the frequency of forest fire occurrence by weight in steps of 1 to 10 to develop a forest fire risk index (FMI, Fuel Model Index) according to clinical characteristics ( Fig. 5).

Comparison of changes in clinical characteristics for forest fires division Clinical proportion of forest fires (%) Before reanalysis After reanalysis Rate of change after reanalysis compared to before reanalysis Deciduous broad-leaved forest 56.3 59.4 +3.1 Evergreen coniferous forest 23.5 21.7 -1.8 Honhyorim 20.2 18.9 -1.3

[Example 3] Development of terrain risk index model for inaccessible areas

In order to calculate the Topography Model Index (TMI) considering the terrain information, the elevation distribution and the slope direction of the wildfire point were calculated by using the DEM (Digital Elevation Model) data of 30 m spatial resolution for 1 second.

Using the average and standard deviation statistics obtained through altitude analysis for forest areas in North Korea, the distribution range of altitude is 5 sections (550m or less, 550 ~ 1,044m, 1,044 ~ 1,537m, 1,537 ~ 2,030m, 2,030m or more) As a result of analyzing a total of 1,469 forest fire occurrence frequencies, the highest intensity of forest fire occurrence was 828 (about 56.3%) at the point where the altitude information was less than 550m, followed by 542 from 550 ~ 1,044m ( About 36.8%) showed a high frequency of wildfire occurrence (Fig. 6). The average altitude for a total of 1,469 wildfire points was about 796m, and there was no difference in the altitude distribution ratio before and after the reanalysis of the wildfire area.

As a result of analyzing the trend of occurrence distribution by classifying into 8 directions to calculate the slope direction for the re-analyzed ignition estimation points (1,469 fire spot points) in the last 5 years, the rate of forest fire occurrence on the southeast slope was the highest at about 13.9%. It showed the lowest rate of wildfire occurrence at about 11.6% on the northeast slope and the verb side, and there was no significant difference within the 8th defense range. That is, it is considered that the slope direction of the inaccessible area does not have a significant influence in the early stage of forest fire. Therefore, the calculated data was analyzed by frequency (0.5 each for altitude and slope) for each terrain condition of the forest fire point, and then indexed in 1 to 10 stages to develop the forest fire risk index (TMI) based on terrain characteristics (Figure 7). ).

[Example 4] Development of weather risk index model for inaccessible areas

In order to calculate the Daily Weather Index (DWI) for inaccessible areas, a database of weather information, such as temperature, humidity, and wind speed, which are weather factors related to the occurrence of forest fire, is targeted for the fire ignition points extracted through reanalysis. It was rebuilt (Fig. 8). In this embodiment, the dependent variable and the independent variable are used to determine whether or not the fire is incurred and the meteorological factors related to the occurrence of the wildfire by using a logistic regression model to identify meteorological factors affecting the occurrence of the fire. Was set to The selection of weather input variables for the probability model of forest fires caused by meteorology is high temperature (average, highest, lowest), relative humidity among the weather elements that can be obtained from 5km reanalysis data from the National Weather Service through correlation analysis. (Average, minimum), effective humidity, wind speed (average, maximum), and precipitation were used, and the correlation results between the input data of the probability of forest fires caused by the weather were shown in FIG. 9.

The probability of daily wildfire occurrence (DWI) due to the weather in an inaccessible area was calculated by using a logistic regression model from time-series weather data at the location of the wildfire and information on the presence or absence of the forest fire. In the case of input data, 10,283 were used as information on the number of days of wildfire occurrence and non-occurrence date between 3 days before and after occurrence (Table 2).

Logistic regression statistics Fire Total Frequency 0 8,814 One 1,469 Probability modeled is Fire = '1' -2 Log likelihood 8,350.885 X ² value 1,207.6659 % Forecast 66.6

As a result of analyzing the logistic regression model for the inaccessible area, it was found that meteorological factors affecting the occurrence of forest fire were significant at 99% confidence level at the highest daily temperature, the lowest relative humidity, the effective humidity, and the average wind speed (FIG. 10). . As a result of analyzing the relationship between forest fire occurrence and meteorological factors, the daily minimum relative humidity showed a negative relationship, and the maximum daily temperature and average wind speed showed a positive relationship while the probability of forest fire occurrence was low and the relative humidity was low. It was found that the higher the temperature and the higher the wind speed, the higher. The suitability for the estimation model was found to be significant at 99% confidence level with a maximum daily temperature of 13.4266, a minimum relative humidity of 84.1784, an effective humidity of 25.8035, and an average wind speed of 22.1316, and the predictive power of the estimated model in the sample was analyzed to be 66.6% (Table 3).

Forest fire risk model formula and in-sample predictive power for inaccessible areas estimated by logistic regression model Model (Pr) (%) predict value [1 + exp {-(2.745+ (0.0905 * highest temperature per day))-(0.0517 * lowest relative humidity per day) + (0.0334 * effective humidity) + (0.1283 * average wind speed))} ^-1 ] ^-1 66.6

The forest fire risk index for the inaccessible area was estimated by using the total forecast obtained through this example to estimate the percentile for each 10% interval, and the predicted probability interval was indexed and set (Table 4). As a result of comparing the daily weather risk index (DWI) frequency of the estimated wildfire occurrence and non-occurrence days by the model, it was finally confirmed that the forest fire risk prediction in the inaccessible area was well performed (FIG. 11).

DWI forest fire estimation probability interval setting Ratio section DWI Estimated probability interval 10% One [.00000 ~ .11138] 20% 2 [.11139 ~ .16559] 30% 3 [.16560 ~ .21032] 40% 4 [.21033 ~ .25141] 50% 5 [.25142 ~ .29238] 60% 6 [.29239 ~ .33452] 70% 7 [.33453 ~ .37828] 80% 8 [.37829 ~ .43005] 90% 9 [.43006 ~ .49471] 100% 10 [.49472 ~ .10000]

[Example 5] Real-time weather information transmission and reception system establishment

In order to establish a real-time forest fire risk forecasting system for inaccessible areas, we prepared a system for transmitting and receiving reanalytical weather information, which is a very short-term situation produced and distributed by the Korea Meteorological Administration. It was constructed as a system that stores meteorological information in real time through a national dedicated network between the Korea Meteorological Administration and the National Forest Science Institute for reanalyzed meteorological information at 5 km resolution produced every hour. The weather information stored in the collection server was transmitted to the real-time forest fire risk forecast analysis server in inaccessible areas and used as input data to calculate the forest fire risk index through pre-processing such as coordinate conversion. This series of processes was programmed to automate the analysis of forest fire risk forecast in real time (FIG. 12).

[Example 6] Pilot system for forest fire risk forecast in inaccessible areas

The automation system of the process to create a forest fire risk map (FFDRI) by assigning weights to weather risk index (DWI), clinical risk index (FMI), and topographic risk index (TMI) developed for inaccessible areas I prepared it. Also, in order to verify the predicted accuracy of the probability of forest fire occurrence for the actual forest fire case, a case analysis study was conducted in April 2014 when the forest fires were heavy. In particular, analysis and verification were conducted on April 15 (51 cases) and 25 (70 cases), where forest fires occurred the most, and April 27 (1 case), where forest fires occurred the least.

On April 15, 2014, the forest fire risk index of the point where 51 fires were estimated to have occurred was generally high, and as a result of evaluating the accuracy of the forest fire ignition point by the predicted forest fire risk grade in the model, 'very high' 13.7% in the region, 60.8% in the 'high' region, 11.8% in the 'normal' region, and 13.7% in the 'low risk' region. At this time, the actual fire occurrence prediction accuracy was 86.3% when the fire risk rating was 'normal (more than 51)' (Fig. 13). On April 25, 2014, when it was estimated that a total of 70 simultaneous forest fires occurred, the forest fire risk class by forest fire occurrence point was 'very high' of 7.1%, 'high' of 60.8%, 'normal' of 24.3%, and 'low risk' 'Was 8.6%, and the model accuracy in the normal grade was 91.4% (FIG. 14). On the other hand, on April 27, 2014, when rainfall was detected in the inaccessible area, it was estimated that only one forest fire occurred, and the forest fire risk index at the forest fire occurrence point was 58, and the forest fire risk level was normal (FIG. 15).

In this embodiment, a web-based wildfire risk prediction pilot system has been developed that can analyze and display forest fire risk information for inaccessible areas in real time based on the calculation and verification results of real-time forest fire risk index in inaccessible areas. 16 is a flowchart showing a process flow for calculating a forest fire risk index for an inaccessible area, and calculates a weather risk index (DWI), a clinical risk index (FMI), and a topographic risk index (TMI) every hour, and weights each time It was designed to calculate the forest fire risk index. The forest fire risk index calculated in this way can be published in four stages of forest fire risk level based on the Annex 1 of Article 3 of the Enforcement Decree of the Forest Protection Act. Here, the risk level of forest fire is divided into 4 levels, 'low (less than 51)', 'normal (51 to 65)', 'high (66 to 85)', and 'very high (over 86)' according to the forest fire risk index. . The real-time analysis of forest fire risk is designed to start analysis when weather information is obtained every hour, and at this time, the accuracy of the spatial distribution of weather information is enhanced by applying the effect of temperature reduction according to spatial interpolation and altitude change. The forecast analysis of forest fire risk was set to be possible for analysis up to 72 hours based on the 05 o'clock data released by the Korea Meteorological Administration.

The pilot system consists of the main screen, current forest fire risk index, forest fire risk class by administrative area, detailed forest fire risk information, forest fire risk statistics, and historical data search. Is as follows. On the main screen, real-time weather satellite and ground radar information is displayed along with information on fire risk rating for each administrative area. Currently, the 'Fire Risk Index' page provides wild fire risk index information for each grid that changes every hour, and is structured to represent the spatial change of the fire risk index. In addition, it was designed to express the 'fire risk level' for each administrative area based on the current fire risk index. The 'Detailed Fire Risk Information', composed of the third item, was designed to obtain the current fire risk information by searching the address or coordinates of a desired point on the map. Next, in 'forest fire risk statistics', forest fire risk grades are implemented together with statistical values of the actual conditions and forecasts of forest fire risk index for each administrative area (FIG. 17). The 'Past Data' search item was configured to search for wildfire risk information published in the past. Lastly, a closed 'Administrator page' was constructed for the operation of the pilot system, management of the analysis data, and confirmation of weather data. Additionally, in the administrator mode, a function capable of managing access status information may be used (FIG. 18).

As described above, although described with reference to preferred embodiments of the present invention, those skilled in the art variously modify and change the present invention without departing from the spirit and scope of the present invention as set forth in the claims below. You can understand that you can.

100: weather information providing server
200: wildfire risk prediction analysis server
201: Satellite image acquisition means
202: means for extracting fire
203: fire point filtering means
204: Basic information acquisition means
205: function deriving means
206: wildfire risk index calculation section
210: Forest Fire Forecasting Agent
300: client

Claims (6)

Obtaining a satellite image of the inaccessible area; Extracting an ignition point from the acquired satellite image in pixel units; Removing diffusion ignition points extracted from the independent ignition points in consideration of the degree of diffusion of wildfires and the vector of wind by time series; Obtaining weather information, clinical information, and topographical information at a corresponding time for a unit grid including an independent ignition point obtained for each time series for each unit grid; Deriving each function for calculating a weather risk index, a clinical risk index, and a terrain risk index from the obtained weather information, clinical information, and terrain information; Calculating current weather risk index, clinical risk index, and terrain risk index by inputting weather information, clinical information, and terrain information for each unit grid obtained in real time to each obtained function; And calculating the wildfire risk index and risk rating for each unit grid from the calculated current weather risk index, clinical risk index, and topographical risk index, and displaying the result in a way that can be identified on the map for each unit grid and displayed on the user terminal. Comprising the steps of
When a forest fire is detected in a time series at a specific pixel, the flash point of the first detected time point is determined as the ignition point and the ignition time,
When wildfires of different periods are detected in the same pixel, they are regarded as individual forest fires.
The point of view within the 3x3 pixel area centered on the point on the first day after the point of ignition is excluded as the point of diffusion ignition,
A real-time forest fire risk forecasting method for inaccessible areas characterized in that the ignition points within the 5 × 5 pixel area centered on the ignition point are excluded as the diffusion ignition point from 2 days after the ignition point. delete delete delete delete According to claim 1,
A real-time forest fire risk forecasting method for inaccessible areas, characterized in that the function for calculating the weather risk index is represented by Equation 1.
[1 + exp {-(2.745+ (0.0905 * highest temperature per day))-(0.0517 * lowest relative humidity per day) + (0.0334 * effective humidity) + (0.1283 * average wind speed))} ^-1 ] ^-1

Images