scholarly journals Prediction of Forest Fire Spread Rate Using UAV Images and an LSTM Model Considering the Interaction between Fire and Wind

2021 ◽  
Vol 13 (21) ◽  
pp. 4325
Author(s):  
Xingdong Li ◽  
Hewei Gao ◽  
Mingxian Zhang ◽  
Shiyu Zhang ◽  
Zhiming Gao ◽  
...  

Modeling forest fire spread is a very complex problem, and the existing models usually need some input parameters which are hard to get. How to predict the time series of forest fire spread rate based on passed series may be a key problem to break through the current technical bottleneck. In the process of forest fire spreading, spread rate and wind speed would affect each other. In this paper, three kinds of network models based on Long Short-Term Memory (LSTM) are designed to predict fire spread rate, exploring the interaction between fire and wind. In order to train these LSTM-based models and validate their effectiveness of prediction, several outdoor combustion experiments are designed and carried out. Process data sets of forest fire spreading are collected with an infrared camera mounted on a UAV, and wind data sets are recorded using a anemometer simultaneously. According to the close relationship between wind and fire, three progressive LSTM based models are constructed, which are called CSG-LSTM, MDG-LSTM and FNU-LSTM, respectively. A Cross-Entropy Loss equation is employed to measure the model training quality, and then prediction accuracy is computed and analyzed by comparing with the true fire spread rate and wind speed. According to the performance of training and prediction stage, FNU-LSTM is determined as the best model for the general case. The advantage of FNU-LSTM is further demonstrated by doing comparison experiments with the normal LSTM and other LSTM based models which predict both fire spread rate and wind speed separately. The experiment has also demonstrated the ability of the model to the real fire prediction on the basis of two historical wildland fires.

2009 ◽  
Vol 18 (6) ◽  
pp. 698 ◽  
Author(s):  
Paulo M. Fernandes ◽  
Hermínio S. Botelho ◽  
Francisco C. Rego ◽  
Carlos Loureiro

An experimental burning program took place in maritime pine (Pinus pinaster Ait.) stands in Portugal to increase the understanding of surface fire behaviour under mild weather. The spread rate and flame geometry of the forward and backward sections of a line-ignited fire front were measured in 94 plots 10–15 m wide. Measured head fire rate of spread, flame length and Byram’s fire intensity varied respectively in the intervals of 0.3–13.9 m min–1, 0.1–4.2 m and 30–3527 kW m–1. Fire behaviour was modelled through an empirical approach. Rate of forward fire spread was described as a function of surface wind speed, terrain slope, moisture content of fine dead surface fuel, and fuel height, while back fire spread rate was correlated with fuel moisture content and cover of understorey vegetation. Flame dimensions were related to Byram’s fire intensity but relationships with rate of spread and fine dead surface fuel load and moisture are preferred, particularly for the head fire. The equations are expected to be more reliable when wind speed and slope are less than 8 km h–1 and 15°, and when fuel moisture content is higher than 12%. The results offer a quantitative basis for prescribed fire management.


2016 ◽  
Vol 25 (4) ◽  
pp. 456 ◽  
Author(s):  
Nuria Prat-Guitart ◽  
Guillermo Rein ◽  
Rory M. Hadden ◽  
Claire M. Belcher ◽  
Jon M. Yearsley

The consumption of large areas of peat during wildfires is due to self-sustained smouldering fronts that can remain active for weeks. We studied the effect of peat moisture content and bulk density on the horizontal propagation of smouldering fire in laboratory-scale experiments. We used milled peat with moisture contents between 25 and 250% (mass of water per mass of dry peat) and bulk densities between 50 and 150 kg m–3. An infrared camera monitored ignition, spread and extinction of each smouldering combustion front. Peats with a bulk density below 75 kg m–3 and a moisture content below 150% self-sustained smouldering propagation for more than 12 cm. Peat with a bulk density of 150 kg m–3 could self-sustain smouldering propagation up to a critical moisture content of 115%. A linear model estimated that increasing both moisture content and bulk density significantly reduced the median fire spread rate (which ranged between 1 and 5 cm h–1). Moisture content had a stronger effect size on the spread rate than bulk density. However, the effect of bulk density on spread rate depends upon the moisture content, with the largest effect of bulk density at low moisture contents.


1988 ◽  
Vol 18 (4) ◽  
pp. 391-397 ◽  
Author(s):  
Ralph M. Nelson Jr. ◽  
Carl W. Adkins

Data for the behavior of 59 experimental wind-driven fires were extracted from the literature for use in determining a correlation among several variables known to influence the rate of forest fire spread. Also included in the correlation were unpublished data from six field fires. This information consisted of behavior measurements on small-scale burns of artificial fuels in the laboratory and measurements on field fires in diverse fuels such as grass and logging slash. Fire intensities ranged from about 40 to 4600 kW/m. Dimensional analysis was used to derive three variables governing the fire spread process. These variables, rearranged into a dimensionless rate of spread and a dimensionless wind speed, are strongly correlated and lead to a simple expression for fire spread rate in terms of fuel consumption, ambient wind speed, and flame residence time.


1995 ◽  
Vol 5 (3) ◽  
pp. 143 ◽  
Author(s):  
RS McAlpine

It has been theorized that the amount of fuel involved in a fire front can influence the rate of spread of the fire. Three data sets are examined in an attempt to prove this relationship. The first, a Canadian Forest Service database of over 400 experimental, wild, and prescribed fires showed a weak relationship in some fuel complexes. The second, a series of field experimental fires conducted to isolate the relationship, showed a small effect. The final data set, from a series of 47 small plots (3m x 3m) burned with a variety of fuel loadings, also show a weak relationship. While a relationship was shown to exist, it is debatable whether it should be included in a fire behavior prediction system. Inherent errors associated with predicting fuel consumption can be compounded, causing additional, more critical, errors with the derived fire spread rate.


2002 ◽  
Vol 11 (2) ◽  
pp. 153 ◽  
Author(s):  
Ralph M. Nelson, Jr.

In previous descriptions of wind-slope interaction and the spread rate of wildland fires it is assumed that the separate effects of wind and slope are independent and additive and that corrections for these effects may be applied to spread rates computed from existing rate of spread models. A different approach is explored in the present paper in which the upslope component of the fire's buoyant velocity is used with the speed and direction of the ambient wind to produce effective values of wind speed and direction that determine the rate of spread vector. Thus the effective wind speed can replace the ambient wind speed in any suitable fire spread model and provide a description of the combined effects on the fire behavior. The difference between current and threshold values of the effective wind speed also can be used to determine whether fire will spread in a given fuel type. The model is tested with data from experiments reported by Weise (1993) in which fire spread was in response to variation in both wind speed and slope angle. The Weise spread rate data were satisfactorily correlated using dimensional methods and the observed spread rate was reasonably well predicted with an existing rate of spread model. Directional aspects of the model were not tested because the Weise (1993) study did not include winds with a cross-slope component.


2005 ◽  
Vol 8 ◽  
pp. 851-862 ◽  
Author(s):  
E. Koo ◽  
P. Pagni ◽  
S. Stephens ◽  
J. Huff ◽  
J. Woycheese ◽  
...  

2020 ◽  
Vol 99 (3) ◽  
pp. 54-61
Author(s):  
V.A. Perminov ◽  
◽  
K.O. Fryanova ◽  

Currently, methods of mathematical modeling are used to study processes in emergency situations. Forest fires are extremely complex and destructive natural phenomena which depend on availability of fuel, meteorological and other conditions. Mathematical model of forest fire is based on an analysis of known experimental data and using concept and methods from reactive media mechanics. In this paper the theoretical study of the problems of crown forest fire spread in windy condition and their thermal impact on the wooden building were carried out. The research was based on numerical solution of two-dimensional Reynolds equations. The boundary-value problem is solved numerically using the method of splitting according to physical processes. A discrete analogue for the system of equations was obtained by means of the control volume method. A study of forest fire spreading made it possible to obtain a detailed picture of the change of the component concentration of gases and temperature fields in forest fire and on the wall of building with time. It let to determine the limiting distances between forest fire and building for possibility of wooden walls ignition for different meteorology conditions, size of building and intensity of fire impact.


Fire Safety ◽  
2020 ◽  
Vol 35 ◽  
pp. 35-40
Author(s):  
A. Kuzyk ◽  
V. Tovaryanskyi ◽  
K. Drach

Formulation of the problem. The fire hazard of grass ecosystems depends on many factors that determine the proper condition of the combustible material and support the burning process. The most important indicators of danger are the condition and humidity of grass. The main external factors of influence on the occurrence and spread of fires in natural ecosystems are: air temperature and air relative humidity, precipitation, wind speed. The urgent task is to determine the rate of spread of fires in grassy ecosystems. The purpose of this work is to establish the features of occurrence and spread out of grass fires in the forest-steppe conditions   of Ukraine on the basis of fire hazard analysis and experimental studies. Research methods. The studies were conducted in the forest-steppe zone in the Vinnytsia region on November 1-3, 2019. The plots with 10 m in length and 3 m in width were chosen, taking into account the wind direction along the plot. During the studies, the wind speed was in a range of 0 to 10 meters per second, the air temperature varied from 7 to 14 Celsium degrees during the day and from -3 to 12 Celsium degrees at night. The air humidity varied from 52 to 69 per cent during the day and from 72 to 84 per cent at night. The humidity of the grass was 20-22 per cent in the daytime, and 27-30 per cent at night. The fire load on the sites was 4-5 t/ha. The main results of the study. Combustion mostly did not happen since 19 p.m. till 10 a.m. because the grass cover during the night increased the moisture content due to the relative humidity increasing and drop of temperature. After 10 a.m. in clear conditions and in the presence of wind, the humidity of the grasses decreased, which facilitated their ignition and spread of fire. However, in the absence of wind, ignition did not happen. The fire spreading rate depended on wind speed and grass height. For grasses 40 cm high the fire spread rate was from 2.5 m/min (wind speed 1-2 m/s) to 3.5 m/min (wind speed 6-8 m/s). For grasses 60 cm high the fire spread rate was from 3.1 m/min (wind speed 1-2 m/s) to 12.5 m/min (wind speed 6-8 m/s). Conclusions. The rate of fire spread in grassу ecosystems depend on temperature and relative humidity, wind speed, grass humidity, height and spatial location. At night, the high humidity of grasses, caused by high relative humidity and low air temperature, hinders ignition and burning, but the fire may occur in the presence of wind in the presence of several sources of ignition with sufficient energy.


1995 ◽  
Vol 5 (4) ◽  
pp. 237 ◽  
Author(s):  
NP Cheney ◽  
JS Gould

The development of grass fires originating from both point and line ignitions and burning in both open grasslands and woodlands with a grassy understorey was studied using 487 periods of fire spread and associated fuel, weather and fire-shape observations. The largest fires travelled more than 1000 m from the origin and the fastest 2-minute spread rate was over 2 m s-1. Given continuous fuel of uniform moisture content, the rate of forward spread was related to both the wind speed and the width of the head fire measured normal to the direction of fire travel. The head fire width required to achieve the potential quasi-steady rate of forward spread for the prevailing conditions increased with increasing wind speeds. These findings have important implications for relating small-scale field or laboratory measurements of fire spread to predictions of wildfire spread. The time taken to reach the potential quasi-steady rate of spread at any wind speed was highly variable. This time was strongly influenced by the frequency of changes in wind direction and the rate of development of a wide head fire.


Sign in / Sign up

Export Citation Format

Share Document