Fire Size Research Articles

Fire Size and Response Time Predictions in Underground Coal Mines Using Neural Networks

Fire Size and Response Time Predictions in Underground Coal Mines Using Neural Networks

Future wildfire extent and frequency determined by the longest fire-conducive weather spell

Great efforts have been made to understand the impacts of a changing climate on fire activity; however, a reliable approach with high prediction confidence has yet to be found. By establishing linkages between the longest duration of fire-conducive weather spell and fire activity parameters, this study projected annual area burned (AAB), annual number of fires (ANF), and annual maximum fire size (MFS) into the future. We found that even though the rates of change differ, the spatial pattern of changes for all three parameters are similar by Canadian ecozone. Areas with the lowest fire activity may see higher rates of change in comparison to high fire activity areas. By end of the century, the changes of AAB and MFS for the study area are projected to be about four and five times that of the baseline respectively, while ANF could almost double.

First Spatial Reconstruction of Past Fires in Temperate Europe Suggests Large Variability of Fire Sizes and an Important Role of Human-Related Ignitions

The spatial component of past forest fires in temperate Europe has been little studied, despite the value of such data in quantifying human and natural factors driving fire activity and associated forest dynamics. Changes in fire regimes reported across a range of ecosystems call for a better understanding of variability in historic fires and may help define reference points that can be relied upon when discussing climate change effects. We provide the first dendrochronological reconstruction of historical fire sizes in Central Europe and analyze the minimum extent of fires during the last four centuries in a 9.2 km2 (920 ha) conifer-dominated section of Białowieża Forest, one of the largest continuous lowland forests of the subcontinent. We recorded 82 fires between 1666 and 1946, using 275 sample trees, while 92% of fires (76 out of 82) spread beyond the studied area. Fires varied considerably in size, from events recorded at only one site (1–200 ha) to fires recorded in more than half of the studied area, thus exceeding 500 ha in size. The fire cycle was 11 years over the whole study period, with three distinct periods revealed by the regime shift analysis. In the years 1670–1750, the fire cycle averaged 12 years. It shortened to 7 years between 1755–1840 and increased to 22 years over the 1845–1955 period. In comparison with present day data, the reconstructed fire density of 3.2 fires per 100 km2 (10 000 ha) and year exceeded lightning ignition density by one to two orders of magnitude, suggesting a significant contribution of human-related ignitions. Our results highlight the important role of fire disturbance in Białowieża Forest and provide critical baseline information to design biological conservation strategies for European forests.

Building a machine learning surrogate model for wildfire activities within a global Earth system model

Abstract. Wildfire is an important ecosystem process, influencing land biogeophysical and biogeochemical dynamics and atmospheric composition. Fire-driven loss of vegetation cover, for example, directly modifies the surface energy budget as a consequence of changing albedo, surface roughness, and partitioning of sensible and latent heat fluxes. Carbon dioxide and methane emitted by fires contribute to a positive atmospheric forcing, whereas emissions of carbonaceous aerosols may contribute to surface cooling. Process-based modeling of wildfires in Earth system land models is challenging due to limited understanding of human, climate, and ecosystem controls on fire counts, fire size, and burned area. Integration of mechanistic wildfire models within Earth system models requires careful parameter calibration, which is computationally expensive and subject to equifinality. To explore alternative approaches, we present a deep neural network (DNN) scheme that surrogates the process-based wildfire model with the Energy Exascale Earth System Model (E3SM) interface. The DNN wildfire model accurately simulates observed burned area with over 90 % higher accuracy with a large reduction in parameterization time compared with the current process-based wildfire model. The surrogate wildfire model successfully captured the observed monthly regional burned area during validation period 2011 to 2015 (coefficient of determination, R2=0.93). Since the DNN wildfire model has the same input and output requirements as the E3SM process-based wildfire model, our results demonstrate the applicability of machine learning for high accuracy and efficient large-scale land model development and predictions.

Deep Learning and Transformer Approaches for UAV-Based Wildfire Detection and Segmentation.

Wildfires are a worldwide natural disaster causing important economic damages and loss of lives. Experts predict that wildfires will increase in the coming years mainly due to climate change. Early detection and prediction of fire spread can help reduce affected areas and improve firefighting. Numerous systems were developed to detect fire. Recently, Unmanned Aerial Vehicles were employed to tackle this problem due to their high flexibility, their low-cost, and their ability to cover wide areas during the day or night. However, they are still limited by challenging problems such as small fire size, background complexity, and image degradation. To deal with the aforementioned limitations, we adapted and optimized Deep Learning methods to detect wildfire at an early stage. A novel deep ensemble learning method, which combines EfficientNet-B5 and DenseNet-201 models, is proposed to identify and classify wildfire using aerial images. In addition, two vision transformers (TransUNet and TransFire) and a deep convolutional model (EfficientSeg) were employed to segment wildfire regions and determine the precise fire regions. The obtained results are promising and show the efficiency of using Deep Learning and vision transformers for wildfire classification and segmentation. The proposed model for wildfire classification obtained an accuracy of 85.12% and outperformed many state-of-the-art works. It proved its ability in classifying wildfire even small fire areas. The best semantic segmentation models achieved an F1-score of 99.9% for TransUNet architecture and 99.82% for TransFire architecture superior to recent published models. More specifically, we demonstrated the ability of these models to extract the finer details of wildfire using aerial images. They can further overcome current model limitations, such as background complexity and small wildfire areas.

Topography, Climate and Fire History Regulate Wildfire Activity in the Alaskan Tundra

AbstractAlthough the link between climate change and tundra fire activity is well‐studied, we lack an understanding of how fire, vegetation, and topography interact to either amplify or dampen climatic effects on these tundra fires at Pan‐Arctic scale. This study investigated the relative influence of fire history, climate, topography and vegetation on fire occurrence and size in Alaskan tundra (1981–2019) and the potential for self‐reinforcing/limiting fire behavior. Regime shift analysis identified a step increase in fire frequency after 2010 with increased average annual area burned (+96%) and area reburned (+61%) over the 2010–2019 period, consistent with climatic thresholds in fire activity being crossed. Correlation analysis shows variation in fire frequency was positively and significantly related to mean summer temperatures. The competing roles of fire history and bio‐climate were investigated via random forest models using (a) environmental conditions and (b) environmental conditions and fire history. Fire occurrence was primarily driven by topographic complexity and elevation, suggesting that areas at 50–200 m elevation with gently rolling terrain such as the Arctic Foothills of the Brooks Range or the Kotzebue Lowlands in northern Alaska will likely become hotspots of fire activity. In contrast, fire size was affected mainly by fire history and winter‐spring climate, which demonstrates the importance of both fuel limitations and self‐reinforcing (e.g., rapid fuel regrowth following smaller‐sized fires) fire‐vegetation interactions in regulating tundra fire activity. Future modeling studies need more nuanced representations of fire‐terrain and fire‐vegetation interactions to accurately project how tundra ecosystems may respond to climatic warming.

Large fires or small fires, will they differ in affecting shifts in species composition and distributions under climate change?

Climate change is expected to increase fire activity, which has the potential to accelerate climate-induced shifts in species composition and distribution in the boreal-temperate ecotone. Wildfire can kill resident trees, and thus provide establishment opportunities for migrating tree species. However, the role of fire size and its interactions with tree species with varied life-history attributes in driving climate-induced shifts is not understood. Future fire regimes could be characterized by many small fires or a few large fires. Large and small fires create and regulate distinct burn patterns, which may influence tree-species responses and post-fire successional trajectories. Here we investigated the effects of future fire-regime variability on the boreal-temperate ecotone of northeastern China under climate change using a coupled forest dynamic model (LANDIS PRO) and ecosystem process model (LINKAGES). We simulated fire regimes using the LANDIS PRO fire module. We designed two fire scenarios (frequent, small fires and infrequent, large fires) to represent different fire regimes in terms of fire size. Results showed fire-catalyzed, climate-induced transitions from boreal to pioneer and temperate forest communities. Frequent, small fires resulted in 13% and 23% higher increases in pioneer and temperate species respectively, relative to infrequent, large fires. Therefore, species composition shifts were faster following frequent, small fires than infrequent, large fires. The results can help policymakers and forest managers determine tradeoffs among strategies to mitigate or adapt to climate change under altered fire regimes.

FIgLib & SmokeyNet: Dataset and Deep Learning Model for Real-Time Wildland Fire Smoke Detection

The size and frequency of wildland fires in the western United States have dramatically increased in recent years. On high-fire-risk days, a small fire ignition can rapidly grow and become out of control. Early detection of fire ignitions from initial smoke can assist the response to such fires before they become difficult to manage. Past deep learning approaches for wildfire smoke detection have suffered from small or unreliable datasets that make it difficult to extrapolate performance to real-world scenarios. In this work, we present the Fire Ignition Library (FIgLib), a publicly available dataset of nearly 25,000 labeled wildfire smoke images as seen from fixed-view cameras deployed in Southern California. We also introduce SmokeyNet, a novel deep learning architecture using spatiotemporal information from camera imagery for real-time wildfire smoke detection. When trained on the FIgLib dataset, SmokeyNet outperforms comparable baselines and rivals human performance. We hope that the availability of the FIgLib dataset and the SmokeyNet architecture will inspire further research into deep learning methods for wildfire smoke detection, leading to automated notification systems that reduce the time to wildfire response.

The Effect of Air Purifier on Smoke Detector Revealed by Fire Dynamics Simulation

In the present study, the relevance of the airflow to the response time of the smoke detector has been investigated regarding the different air purifier locations by Fire Dynamics Simulator (FDS). Particular attention has been devoted to the impact on the smoke detector response with different device types, i.e., top-discharge and front-discharge air purifiers. Additionally, the response delay rate regarding the fire sizes (200 kW and 550 kW) was also evaluated with the top-discharge type purifier. The obscuration plots suggest that the response delay rate generally increases when the air purifier and the smoke detector are in a straight line. The change in the delay rate of the detector was distinct for the top-discharge type device due to the formation of a current barrier by the ceiling jet flow, and a nearby physical barrier (e.g., wall) can further suppress the flow diffusion during a fire event. The delay rate tends to decrease with the increment of the fire intensity.

Effects of blockage on the flame morphologic characteristics in a ventilated tunnel

As a practical but seldom concerned issue, the morphologic characteristics of a turbulence diffusion flame under the effects of upstream blockage in a longitudinally ventilated tunnel was studied based on a series of model-scale experiments. Different blockage sizes, blockage-fire distances, longitudinal ventilation velocities, and fire sizes were considered for research. Numerical simulations were also carried out to calculate the flow field for analyzing the causes of the variety in the flame morphology and determining the length of the near wake. Experimental results showed that the flame morphology can be divided into two stages according to the dimensionless blockage-fire distance normalized by the length of the near wake, namely flame bifurcation and flame tilt. By comparing with previous literature, the critical blockage-fire distance corresponding to the symmetrical bifurcation was achieved which follows a linear relationship with the length of the near wake. On the basic of tilt angle and centerline trajectory, dimensionless models were developed to predict the flame tilt under the effects of blocking ratio and blockage-fire distance. The research outcomes can provide a reference and tool to assess the fire risks of tunnels with blockage effects.

Experimental and numerical study of the fire behavior of a tank with oil leaking and burning

Leaking-burning coupling fires caused by oil leakage from tanks cause significant damage to occupants and the equipment. A series of experiments was conducted to investigate the fire behavior of a tank with oil leaking and burning. The tank released n-heptane into the tray below, forming a fire that heated the tank. The tray size and leak diameter varied. Spill fires, steady pool fires, boiling pool fires, and jet fires were observed. The results show that within a specific leak diameter range (2.5–3.5 mm), the maximum pressure increased with decreasing leak diameter. The fuel vapor spread along the bottom of the tank, causing the expansion of the flame. Consequently, the pool fire size was larger than the actual tray size, thereby making the maximum pressure and thermal radiation almost independent of tray size. When the leaking oil began to boil and burn, the pressure increased rapidly and the emitted radiative heat flux was the largest; therefore, fire posed the most serious threat. A computational fluid dynamics (CFD) modeling approach was employed to predict the time when the temperature of the leaking oil reached its boiling point. The predicted results were validated against the experimental ones, which provided useful support for emergency response plans.

Assessing the Accuracy of MODIS MCD64A1 C6 and FireCCI51 Burned Area Products in Mediterranean Ecosystems

The catastrophic impact of wildfires on the economy and ecosystems of Mediterranean countries in recent years, along with insufficient policies that favor disproportionally high funding for fire suppression, demand a more comprehensive understanding of fire regimes. Satellite remote sensing products support the generation of relevant burned-area (BA) information, since they provide the means for the systematic monitoring of large areas worldwide at low cost. This research study assesses the accuracy of the two publicly available MODIS BA products, MCD64A1 C6 and FireCCI51, at a national scale in a Mediterranean country. The research period covered four fire seasons, and a comparison was conducted against a higher-resolution Sentinel-2 dataset. The specific objectives were to assess their performance in detecting fire events occurring primarily in forest and semi-natural lands and to investigate their spatial and temporal uncertainties. Monthly fire observations were processed and analyzed to derive a comprehensive set of accuracy metrics. We found that fire size has an impact on their detection accuracy, with higher detection occurring in fires larger than 100 ha. Detection of smaller (<100 ha) fires was favored by the 250 m FireCCI51 product, but not from MCD64A1 C6, which exhibited less than 50% detection probability in the same range. Their spatial estimates of burned area exhibited a fairly satisfactory agreement with the reference data, reaching an average of 78% in detection rate. MCD64A1 C6 exhibited a more consistent spatial performance overall and better temporal accuracy, whereas FireCCI51 did not substantially outperform the former despite its finer resolution. Additional research is required for a more rigorous assessment of the variability of these burned area products, yet this research provides further insight and has implications for their use in fire-related applications at the local to the national scale.

Full-scale fire tests in the underwater tunnel section model with sidewall smoke extraction

The Hong Kong–Zhuhai–Macau Bridge (HZMB) is a 55-km bridge-tunnel system, including a 6.7-km undersea tunnel that adopts the sidewall smoke extraction system. To evaluate the potential tunnel fire hazards, a 1:1 full-scale HZMB tunnel section model (16 m × 7.2 m × 150 m) was constructed, and eight full-scale tunnel fire tests were conducted with the sidewall smoke extraction. The temperature distribution and smoke movement under different vent arrangements and fire sizes (1.2–6.6 MW) were quantified. Results indicated that the fire HRR was mainly affected by the size of the liquid-fuel pool but insensitive to the arrangement of ventilation. The correlation between HRR and diesel pool-fire area can be fitted by a linear function of HRR=1.24AF-0.87 MW. The sidewall smoke extraction generated a tilted fire plume and non-uniform temperature distribution at the transverse direction, whereas the temperature decay still followed the exponential decay for the far fire field region. The decay factor increases with the increase of the HRR and increases when distributing the ventilation capability into two vent groups. A relatively slow smoke motion (0.8–1.2 m/s) and good smoke stratification were demonstrated in the tests, indicating a robust condition for safe evacuation. This research deepens the understanding of fire and smoke characteristics in tunnels with the sidewall extraction and highlights the importance of full-scale test data in the development of the smart tunnel-fire protection system.

PREDICTION OF FOREST FIRE USING NEURAL NETWORKS WITH BACKPROPAGATION LEARNING AND EXREME LEARNING MACHINE APPROACH USING METEOROLOGICAL AND WEATHER INDEX VARIABLES

Forest fire is one of important catastrophic events and have great impact on environment, infrastructure and human life. In this study, we discuss the method for prediction of the size of the forest fire using the hybrid approach between Fuzzy-C-Means clustering (FCM) and Neural Networks (NN) classification with backpropagation learning and extreme learning machine approach. For comparison purpose, we consider a similar hybrid approach, i.e., FCM with the classical Support Vector Machine (SVM) classification approach. In the empirical study, we apply the considered methods using several meteorological and Forest Weather Index (FWI) variables. We found that the best approach will be obtained using hybrid FCM-SVM for data training, where the best performance obtains for hybrid FCM-NN-backpropagation for data testing.

Suppression resources and their influence on containment of forest fires in Victoria

Background Wildfire suppression is becoming more costly and dangerous as the scale and severity of impacts from fires increase under climate change. Aims We aim to identify the key environmental and management variables influencing containment probability for forest fires in Victoria and determine how these change over time. Methods We developed Random Forest models to identify variables driving fire containment within the first 24 h of response. We used a database of ~12 000 incident records collected across Victoria, Australia. Key results Response time, fire size at first attack, number of ground resources deployed (e.g. fire fighters), ignition cause, and environmental factors that influence fire spread (e.g. elevation, humidity, wind, and fuel hazard) were key drivers of suppression success within the first 24 h. However, certainty about the factors influencing suppression reduced as the containment period increased. Conclusions Suppression success hinges on a balance between the environmental factors that drive fire spread and the rapid deployment of sufficient resources to limit fire perimeter growth. Implications Decreasing the period between an ignition and the time of arrival at the fire will allow first responders to begin suppression before the fire size has exceeded their capability to construct a control line.

Simulation-based high-resolution fire danger mapping using deep learning

Wildfire occurrence and behaviour are difficult to predict locally for the next day. In the present work, we propose relying on fire spread simulations to provide a fire danger index representative of the potential for fire spread that includes not only weather but also surrounding vegetation and orography. This is achieved using an artificial neural network emulator called DeepFire, trained based on simulated fire sizes. To determine how relevant this index can be in the assessment of next-day fire danger, the application of DeepFire to fire danger mapping using actual weather forecasts is studied. DeepFire forecasts for 13 fairly large fires that occurred in Corsica are analysed and compared with corresponding forecasts using another fire danger index used in operational conditions, highlighting the differences in terms of precision and the expected results of such predictions. The weather forecasts from which the weather inputs of DeepFire are determined have high spatial resolution and high frequency, which also applies to the fire danger predictions. Additionally, input uncertainty is propagated through DeepFire, resulting in ensembles of emulated fire size. Several approaches are proposed to analyse the results and provide fire danger maps and ratings using this new simulation-based prediction system.

Daily Burned Area Mapping for Prescribed Rangeland Burning in the Flint Hills Region Using MODIS Data

Highlights Average burn frequency and fire size were larger in the center part of the Flint Hills region. The maximum daily burned area was 0.2 million ha. Even distribution of fires can reduce smoke impact. Fires were concentrated on days with high solar radiation. Changing this pattern can reduce the impact on O 3 . Abstract. Smoke from prescribed burning in the Flint Hills region impacts air quality and contributes to the exceedances of ozone (O3). The objectives of this study were 1) to map the daily burned area in this region from 2003 to 2019 using the combination of Moderate Resolution Imaging Spectroradiometer (MODIS) highest resolution imagery and standard active fire data from satellite, and 2) to identify the spatial and temporal distributions and patterns of daily fire activities. Red and near-infrared images were combined, and supervised minimum distance classification of the bi-spectral images was applied to map the daily burned area. Results showed that fire activities were highly concentrated in a few days in mid-April. The annual burned percentage (ranged from 11% to 52%) and the average fire size (ranged from 160 to 400 ha) were larger in the central part of the Flint Hills region than in the Southern part and at the edge. On average, the preferred weather conditions by landowners only occurred in around 10 days in one spring burning season. The maximum daily burned area in the Flint Hills study region was as high as 0.2 million ha, about 1/6 of the total burned area in the whole spring burning season. Based on analysis of historical daily burned area and weather data, most fires occurred on days with high solar radiation, favoring O3 production. Changing this pattern or more evenly distributed fires can reduce the smoke impact on daily O3. The results of this study provided a solid foundation for the ongoing development of O3 regression models using daily burned area and meteorological factors as predictor variables. Keywords: MODIS, Ozone, Prescribed burning, Remote sensing, Smoke impact.

Forest fire and its key drivers in the tropical forests of northern Vietnam

Fire increasingly threatens tropical forests in northern Vietnam as climate changes and human population grows. Understanding fire occurrence patterns may support more effective forest management and reduce fire risk. We investigated spatiotemporal patterns and drivers of wildfire across three provinces in northern Vietnam and assessed the effectiveness of the Modified Nesterov index (MNI) fire danger rating system. We explored fire occurrence and size within and between years and forest types using descriptive analyses and developed spatiotemporal Maximum Entropy (Maxent) models incorporating variables representing potential drivers of fire, including weather, fuel, topography and human activity. Most fires occurred late in the dry season and fires were most common in natural forest. Maxent models successfully predicted fire occurrence (area under the receiver operating characteristic curve (AUC) values 0.67–0.79). While the contributions of drivers varied among provinces, MNI, temperature, elevation and distance to road were consistently important. The model for combined provinces showed that fire probability was greater under higher temperature and MNI, in areas with lower population, farther from roads, at higher elevations and in natural forests. This study suggests that an assessment integrating multiple drivers better predicts fire occurrence than a system based on weather alone and may support improved fire management and education in northern Vietnam.

Future expansion, seasonal lengthening and intensification of fire activity under climate change in southeastern France

Background An increase in fire weather is expected in a warming climate, but its translation to fire activity (fire numbers and sizes) remains largely unknown. Additionally, disentangling the extent to which geographic and seasonal extensions as well as intensification contribute to future fire activity remain largely unknown. Aims We aimed to assess the impact of future climate change on fire activity in southeastern France and estimate changes in spatial and seasonal distributions. Methods We projected future fire activities using a Bayesian modelling framework combined with ensemble climate simulations. Changes in numbers of escaped fires (>1 ha), large fires (>100 ha) and burned area were studied for different emission scenarios or degrees of global warming. Key results Fire activity could increase by up to +180% for +4°C of global warming, with large expansions of fire-prone regions and long seasonal lengthenings. Overall, changes will be dominated by intensification within the historical fire niche, representing two-thirds of additional future fire activity, half of this occurring during the high fire season. Conclusions This study confirms that major changes in fire niches would be expected in Euro-Mediterranean regions. Implications Long-term strategic policies for adapting prevention and suppression resources and ecosystems are needed to account for such changes.

Fire scar characteristics in two tropical montane conifer species from central Mexico

Fire scar analysis is a fundamental tool for reconstructing fire regimes in conifer forests. However, little is known about fire scar properties in tropical montane conifers, where some assumptions limit dendroecological research. These include that fir species do not exhibit external fire scars and that pines without external fire scars have not experienced past fires. This study describes fire scar patterns in two conifer species growing in Mexican temperate forests: sacred fir (Abies religiosa) and smooth-bark Mexican pine (Pinus pseudostrobus). We extracted cross-sections from 110 trees and measured tree age, basal diameter, bark thickness, fire scar size and seasonality, the number of scars per tree, years and basal diameter from pith to the first scar, and the proportion of external and buried fire scars. Most trees had three fire scars, which appeared during the dry season after the first 15 years measured from the pith. Old and large-diameter trees did not have more fire scars, but the time between fire scars influenced fire scar closure in sacred firs. Bark thickness and the proportion of visible and buried fire scars were similar in both species. Our results suggest that the absence of visible fire scars in smooth-bark Mexican pines does not imply the absence of fire, and sacred firs can exhibit external fire scars like pines.