Analysis of the uncertainty of fuel model parameters in wildland fire modelling of a boreal forest in north-east China

2019 ◽  
Vol 28 (3) ◽  
pp. 205 ◽  
Author(s):  
Longyan Cai ◽  
Hong S. He ◽  
Yu Liang ◽  
Zhiwei Wu ◽  
Chao Huang
Keyword(s):  
Monte Carlo ◽  
Wildland Fire ◽  
Time Lag ◽  
Fire Spread ◽  
Model Parameters ◽  
Spread Rate ◽  
North East ◽  
Fire Modelling ◽  
Fuel Model ◽  
Model Parameter Uncertainty

Fire propagation is inevitably affected by fuel-model parameters during wildfire simulations and the uncertainty of the fuel-model parameters makes forecasting accurate fire behaviour very difficult. In this study, three different methods (Morris screening, first-order analysis and the Monte Carlo method) were used to analyse the uncertainty of fuel-model parameters with FARSITE model. The results of the uncertainty analysis showed that only a few fuel-model parameters markedly influenced the uncertainty of the model outputs, and many of the fuel-model parameters had little or no effect. The fire-spread rate is the driving force behind the uncertainty of other fire behaviours. Thus, the highly uncertain fuel-model parameters associated with spread rate should be used cautiously in wildfire simulations. Monte Carlo results indicated that the relationship between model input and output was non-linear and neglecting fuel-model parameter uncertainty of the model would magnify fire behaviours. Additionally, fuel-model parameters have high input uncertainty. Therefore, fuel-model parameters must be calibrated against actual fires. The highly uncertain fuel-model parameters with high spatial-temporal variability consisted of fuel-bed depth, live-shrub loading and 1-h time-lag loading are preferentially chosen as parameters to calibrate several wildfires.

Parametric uncertainty quantification in the Rothermel model with randomised quasi-Monte Carlo methods

2015 ◽  
Vol 24 (3) ◽  
pp. 307 ◽  
Author(s):  
Yaning Liu ◽  
Edwin Jimenez ◽  
M. Yousuff Hussaini ◽  
Giray Ökten ◽  
Scott Goodrick
Keyword(s):  
Monte Carlo ◽  
Uncertainty Quantification ◽  
Monte Carlo Methods ◽  
Wildland Fire ◽  
Parametric Uncertainty ◽  
Monte Carlo Sampling ◽  
Model Parameters ◽  
Quasi Monte Carlo ◽  
Control Variate ◽  
Fuel Model

Rothermel's wildland surface fire model is a popular model used in wildland fire management. The original model has a large number of parameters, making uncertainty quantification challenging. In this paper, we use variance-based global sensitivity analysis to reduce the number of model parameters, and apply randomised quasi-Monte Carlo methods to quantify parametric uncertainties for the reduced model. The Monte Carlo estimator used in these calculations is based on a control variate approach applied to the sensitivity derivative enhanced sampling. The chaparral fuel model, selected from Rothermel's 11 original fuel models, is studied as an example. We obtain numerical results that improve the crude Monte Carlo sampling by factors as high as three orders of magnitude.

High-resolution infrared thermography for capturing wildland fire behaviour: RxCADRE 2012

2016 ◽  
Vol 25 (1) ◽  
pp. 62 ◽  
Author(s):  
Joseph J. O'Brien ◽  
E. Louise Loudermilk ◽  
Benjamin Hornsby ◽  
Andrew T. Hudak ◽  
Benjamin C. Bright ◽  
...  
Keyword(s):  
Wildland Fire ◽  
Radiant Energy ◽  
Fire Spread ◽  
Fire Effects ◽  
Fire Behaviour ◽  
Oblique View ◽  
Fire Modelling ◽  
Radiative Power ◽  
Fire Front ◽  
Research Experiment

Wildland fire radiant energy emission is one of the only measurements of combustion that can be made at wide spatial extents and high temporal and spatial resolutions. Furthermore, spatially and temporally explicit measurements are critical for making inferences about fire effects and useful for examining patterns of fire spread. In this study we describe our methods for capturing and analysing spatially and temporally explicit long-wave infrared (LWIR) imagery from the RxCADRE (Prescribed Fire Combustion and Atmospheric Dynamics Research Experiment) project and examine the usefulness of these data in investigating fire behaviour and effects. We compare LWIR imagery captured at fine and moderate spatial and temporal resolutions (from 1 cm2 to 1 m2; and from 0.12 to 1 Hz) using both nadir and oblique measurements. We analyse fine-scale spatial heterogeneity of fire radiant power and energy released in several experimental burns. There was concurrence between the measurements, although the oblique view estimates of fire radiative power were consistently higher than the nadir view estimates. The nadir measurements illustrate the significance of fuel characteristics, particularly type and connectivity, in driving spatial variability at fine scales. The nadir and oblique measurements illustrate the usefulness of the data for describing the location and movement of the fire front at discrete moments in time at these fine and moderate resolutions. Spatially and temporally resolved data from these techniques show promise to effectively link the combustion environment with post-fire processes, remote sensing at larger scales and wildland fire modelling efforts.

Current approaches to modelling the spread of wildland fire: a review

1998 ◽  
Vol 22 (2) ◽  
pp. 222-245 ◽  
Author(s):  
G. L.W. Perry
Keyword(s):  
Information Technologies ◽  
Spatial Information ◽  
Wildland Fire ◽  
Fire Risk ◽  
Fire Spread ◽  
Physical Models ◽  
Accurate Estimation ◽  
Fire Event ◽  
Fire Modelling ◽  
Modelling Techniques

This review considers the development of some of the models and modelling approaches designed to predict the spread and spatial behaviour of wildland fire events. Such events and their accurate prediction are of great importance to those seeking to understand and manage fire-prone ecosystems. The key problem which fire modelling seeks to address is outlined. Models predicting the rate of fire spread may be classified as physical, semi-physical or empirical according to the nature of their construction. The benefits and shortcomings of each type of model are considered with reference to specific examples of each type. It is shown that there are problems with current operational models which restrict their effective use. However, the development of rigorous physical models as replacements is impeded by conceptual and practical difficulties. Accurate estimation of the rate of spread and the intensity of a fire allows prediction of the final shape and area of a fire event. The modelling techniques used to estimate the shape and area of a fire are considered including the development of sophisticated computer-based simulations of fire spread. Spatial information technologies such as remote sensing and geographic information systems (GIS) offer great potential for the effective modelling of wildland fire behaviour. While such spatial information technologies have been frequently used in the evaluation of fire danger risk, their use for the simulation of the spatiotemporal behaviour of wildland fire is not common. The way in which spatial information technologies and decision-support systems are used for fire risk evaluation and fire spread simulation is discussed. Two research areas of great importance if fire modelling techniques are to improve are a better understanding of fire-dependent phenomena and the development of a ‘new generation’ of fire spread models; current trends in these areas of research are evaluated.

scholarly journals Building Rothermel fire behaviour fuel models by genetic algorithm optimisation

2015 ◽  
Vol 24 (3) ◽  
pp. 317 ◽  
Author(s):  
Davide Ascoli ◽  
Giorgio Vacchiano ◽  
Renzo Motta ◽  
Giovanni Bovio
Keyword(s):  
Goodness Of Fit ◽  
Fire Spread ◽  
Validation Dataset ◽  
Model Parameters ◽  
Initial Population ◽  
Fire Behaviour ◽  
Continuous Space ◽  
Fuel Model ◽  
Spread Model ◽  
Fuel Models

A method to build and calibrate custom fuel models was developed by linking genetic algorithms (GA) to the Rothermel fire spread model. GA randomly generates solutions of fuel model parameters to form an initial population. Solutions are validated against observations of fire rate of spread via a goodness-of-fit metric. The population is selected for its best members, crossed over and mutated within a range of model parameter values, until a satisfactory fitness is reached. We showed that GA improved the performance of the Rothermel model in three published custom fuel models for litter, grass and shrub fuels (root mean square error decreased by 39, 19 and 26%). We applied GA to calibrate a mixed grass–shrub fuel model, using fuel and fire behaviour data from fire experiments in dry heathlands of Southern Europe. The new model had significantly lower prediction error against a validation dataset than either standard or custom fuel models built using average values of inventoried fuels, and predictions of the Fuel Characteristics Classification System. GA proved a useful tool to calibrate fuel models and improve Rothermel model predictions. GA allows exploration of a continuous space of fuel parameters, making fuel model calibration computational effective and easily reproducible, and does not require fuel sampling. We suggest GA as a viable method to calibrate custom fuel models in fire modelling systems based on the Rothermel model.

scholarly journals Resolving vorticity-driven lateral fire spread using the WRF-Fire coupled atmosphere-fire numerical model

2014 ◽  
Vol 2 (5) ◽  
pp. 3499-3531 ◽  
Author(s):  
C. C. Simpson ◽  
J. J. Sharples ◽  
J. P. Evans
Keyword(s):  
Spatial Resolution ◽  
High Spatial Resolution ◽  
Wildland Fire ◽  
Fire Spread ◽  
Lateral Spread ◽  
Grid Spacing ◽  
Spread Rate ◽  
Model Framework ◽  
Fire Model ◽  
Horizontal Grid

Abstract. Fire channelling is a form of dynamic fire behaviour, during which a wildland fire spreads rapidly across a steep lee-facing slope in a direction transverse to the background winds, and is often accompanied by a downwind extension of the active flaming region and extreme pyro-convection. Recent work using the WRF-Fire coupled atmosphere-fire model has demonstrated that fire channelling can be characterised as vorticity-driven lateral fire spread (VDLS). In this study, 16 simulations are conducted using WRF-Fire to examine the sensitivity of resolving VDLS to spatial resolution and atmosphere-fire coupling within the WRF-Fire model framework. The horizontal grid spacing is varied between 25 and 90 m, and the two-way atmosphere-fire coupling is either enabled or disabled. At high spatial resolution, the atmosphere-fire coupling increases the peak uphill and lateral spread rate by a factor of up to 2.7 and 9.5. The enhancement of the uphill and lateral spread rate diminishes at coarser spatial resolution, and VDLS is not modelled for a horizontal grid spacing of 90 m. The laterally spreading fire fronts become the dominant contributors of the extreme pyro-convection. The resolved fire-induced vortices responsible for driving the lateral spread in the coupled simulations have non-zero vorticity along each unit vector direction, and develop due to an interaction between the background winds and vertical return circulations generated at the flank of the fire front as part of the pyro-convective updraft. The results presented in this study demonstrate that both high spatial resolution and two-way atmosphere-fire coupling are required to reproduce VDLS within the current WRF-Fire model framework.

A radiation-driven model for crown fire spread

2004 ◽  
Vol 34 (8) ◽  
pp. 1588-1599 ◽  
Author(s):  
B W Butler ◽  
M A Finney ◽  
P L Andrews ◽  
F A Albini
Keyword(s):  
Closed Form ◽  
A Priori ◽  
Radiant Energy ◽  
Fire Spread ◽  
Crown Fire ◽  
Spread Rate ◽  
Fire Modelling ◽  
Crown Fires ◽  
Environment Variables ◽  
Fire Front

A numerical model for the prediction of the spread rate and intensity of forest crown fires has been developed. The model is the culmination of over 20 years of previously reported fire modeling research and experiments; however, it is only recently that it has been formulated in a closed form that permits a priori prediction of crown fire spread rates. This study presents a brief review of the development and structure of the model followed by a discussion of recent modifications made to formulate a fully predictive model. The model is based on the assumption that radiant energy transfer dominates energy exchange between the fire and unignited fuel with provisions for convective cooling of the fuels ahead of the fire front. Model predictions are compared against measured spread rates of selected experimental fires conducted during the International Crown Fire Modelling Experiment. Results of the comparison indicate that the closed form of the model accurately predicts the relative response of fire spread rate to fuel and environment variables but overpredicts the magnitude of fire spread rates.

Two methods for calculating wildland fire rate of forward spread

2020 ◽  
Vol 29 (3) ◽  
pp. 272 ◽  
Author(s):  
Jim S. Gould ◽  
Andrew L. Sullivan
Keyword(s):  
Wind Tunnel ◽  
Field Experiments ◽  
Prediction Models ◽  
Wildland Fire ◽  
Model Development ◽  
Laboratory Data ◽  
Fire Spread ◽  
Accurate Estimation ◽  
Spread Rate ◽  
Eucalypt Forest

Accurate estimation of a wildland fire’s progression is critical for the development of robust fire spread prediction models and their validation. Two methods commonly used to determine spread rate are the cumulative spread rate, calculated as the total distance travelled by a fire divided by the total time of travel, and the interval spread rate, calculated using the minimum time and maximum distance between observations. This paper analyses the differences between these two methods using experimental fires conducted in dry eucalypt forest leaf litter in either a combustion wind tunnel or large (4ha) field sites. Fires were ignited from a point, 400-mm and 800-mm line ignitions in the wind tunnel, and point and 120-m line ignitions in the field experiments. A total of 312 and 397 observations of distance travelled and time taken were made during the laboratory and field experiments respectively, along with associated environmental variables. Mean spread rates and standard deviations were significantly greater for the interval method than those of the cumulative method for all the laboratory data and the field point ignition fires, and the difference between them varied with distance and time since ignition. These findings have important implications for fire spread and acceleration model development.

Examination of the wind speed limit function in the Rothermel surface fire spread model

2013 ◽  
Vol 22 (7) ◽  
pp. 959 ◽  
Author(s):  
Patricia L. Andrews ◽  
Miguel G. Cruz ◽  
Richard C. Rothermel
Keyword(s):  
Wind Speed ◽  
Wildland Fire ◽  
Fire Spread ◽  
Speed Limit ◽  
Limit Function ◽  
Surface Fire ◽  
Rate Of Spread ◽  
Fire Modelling ◽  
Spread Model ◽  
Fire Spread Model

The Rothermel surface fire spread model includes a wind speed limit, above which predicted rate of spread is constant. Complete derivation of the wind limit as a function of reaction intensity is given, along with an alternate result based on a changed assumption. Evidence indicates that both the original and the revised wind limits are too restrictive. Wind limit is based in part on data collected on the 7 February 1967 Tasmanian grassland fires. A reanalysis of the data indicates that these fires might not have been spreading in fully cured continuous grasslands, as assumed. In addition, more recent grassfire data do not support the wind speed limit. The authors recommend that, in place of the current wind limit, rate of spread be limited to effective midflame wind speed. The Rothermel model is the foundation of many wildland fire modelling systems. Imposition of the wind limit can significantly affect results and potentially influence fire and fuel management decisions.

Large eddy simulation of atypical wildland fire spread on leeward slopes

2013 ◽  
Vol 22 (5) ◽  
pp. 599 ◽  
Author(s):  
Colin C. Simpson ◽  
Jason J. Sharples ◽  
Jason P. Evans ◽  
Matthew F. McCabe
Keyword(s):  
Wildland Fire ◽  
Average Rate ◽  
Mass Density ◽  
Flow Characteristics ◽  
Fire Spread ◽  
Lateral Spread ◽  
Flame Height ◽  
Ridge Line ◽  
Eddy Simulation ◽  
Fire Modelling

The WRF-Fire coupled atmosphere–fire modelling system was used to investigate atypical wildland fire spread on steep leeward slopes through a series of idealised numerical simulations. The simulations are used to investigate both the leeward flow characteristics, such as flow separation, and the fire spread from an ignition region at the base of the leeward slope. The fire spread was considered under varying fuel type and with atmosphere-fire coupling both enabled and disabled. When atmosphere–fire coupling is enabled and there is a high fuel mass density, the fire spread closely resembles that expected during fire channelling. Specifically, the fire spread is initially dominated by upslope spread to the mountain ridge line at an average rate of 2.0kmh–1, followed by predominantly lateral spread close to the ridge line at a maximum rate of 3.6kmh–1. The intermittent rapid lateral spread occurs when updraft–downdraft interfaces, which are associated with strongly circulating horizontal winds at the mid-flame height, move across the fire perimeter close to the ridge line. The updraft–downdraft interfaces are formed due to an interaction between the strong pyro-convection and the terrain-modified winds. Through these results, a new physical explanation of fire channelling is proposed.

scholarly journals A Deep Learning Approach to Downscale Geostationary Satellite Imagery for Decision Support in High Impact Wildfires

Forests ◽  
2021 ◽  
Vol 12 (3) ◽  
pp. 294
Author(s):  
Nicholas F. McCarthy ◽  
Ali Tohidi ◽  
Yawar Aziz ◽  
Matt Dennie ◽  
Mario Miguel Valero ◽  
...  
Keyword(s):  
Deep Learning ◽  
Decision Support ◽  
Real Time ◽  
Forest Fires ◽  
Wildland Fire ◽  
Fire Spread ◽  
Low Earth Orbit ◽  
Practical Implementation ◽  
Spatiotemporal Resolution ◽  
Active Fire

Scarcity in wildland fire progression data as well as considerable uncertainties in forecasts demand improved methods to monitor fire spread in real time. However, there exists at present no scalable solution to acquire consistent information about active forest fires that is both spatially and temporally explicit. To overcome this limitation, we propose a statistical downscaling scheme based on deep learning that leverages multi-source Remote Sensing (RS) data. Our system relies on a U-Net Convolutional Neural Network (CNN) to downscale Geostationary (GEO) satellite multispectral imagery and continuously monitor active fire progression with a spatial resolution similar to Low Earth Orbit (LEO) sensors. In order to achieve this, the model trains on LEO RS products, land use information, vegetation properties, and terrain data. The practical implementation has been optimized to use cloud compute clusters, software containers and multi-step parallel pipelines in order to facilitate real time operational deployment. The performance of the model was validated in five wildfires selected from among the most destructive that occurred in California in 2017 and 2018. These results demonstrate the effectiveness of the proposed methodology in monitoring fire progression with high spatiotemporal resolution, which can be instrumental for decision support during the first hours of wildfires that may quickly become large and dangerous. Additionally, the proposed methodology can be leveraged to collect detailed quantitative data about real-scale wildfire behaviour, thus supporting the development and validation of fire spread models.

Sign in / Sign up

Export Citation Format

Share Document