@article {bnh-8335, title = {Investigating the suitability of aviation tracking data for use in bushfire suppression effectiveness research}, year = {2022}, month = {04/2022}, institution = {Bushfire and Natural Hazards CRC}, address = {Melbourne}, abstract = {

Aircraft are an important part of bushfire suppression and their use is increasing. They were used heavily during the 2019/20 {\textquotedblleft}Black Summer{\textquotedblright} bushfire season in NSW and several inquiries have highlighted the need for research into their effectiveness.

Tracking equipment is becoming routinely deployed on aircraft and there is increasing availability of high-quality ancillary data such as aerial imagery and fire severity mapping. These allow detailed analyses of aircraft activities. However, the usefulness of the data needs to be evaluated, and the analysis needs to be informed by information about the tasking objectives of the aircraft and whether those objectives were met.

This project provides an initial investigation into the process of evaluating aerial suppression using these new data sources and interviews with personnel involved in the suppression activities.

Firebombing event data (drops/fills) from the 2019/2020 bushfire season in NSW from the National Aerial Firefighting Centre (NAFC){\textquoteright}s Arena database was provided by the NSW Rural Fire Service. This data included ~70000 aircraft suppression drop locations and times from aircraft that included helicopters (mainly large and medium helitaks), Single-Engine Air Tankers and Large Air Tankers. As an initial step, we examined the data for completeness, accuracy and errors, and described the data contents. This data was missing for most of the aircraft known to be dropping on the fires, especially the smaller ones. The type of drop (gel, water, retardant) was unknown in most cases, the quantity dropped was unknown in 45\% of cases, and the location for the end of drops was often unreliable.

We then tested methods to identify drop objectives based on relationships between drops data and other spatial data including building locations and weather. Using a combination of automated pattern matching and manual checking, the data can be used to identify cases where the objective was initial attack, extinguishing spot fires, asset protection, pre-emptive laying of retardant lines and direct attack. There were a few cases where the success or failure of the objective could be assessed purely with the spatial data. We also explored two particular analytical methods for determining objectives. First, we compared the distribution of Forest Fire Danger Index (FFDI, fire weather) during a fire and for the drops within that fire. This identified several fires for which a large proportion of the drops were more likely to be during extreme fire weather even though extreme weather was rare in that fire. Second, we compared the distribution of distance to houses between all parts of the fire and the drops at that fire. Here we found many fires where the drops were clustered closer to houses than if the drops were (hypothetically) spread evenly across the fire ground. These analyses are preliminary but show great potential.

We conducted 10 interviews with personnel who worked as Air Attack Supervisors during the 2019/20 season. Interviewees were knowledgeable and experienced, and expressed the view that the aerial program could be improved with further knowledge sharing and training. They provided a lot of general information about objectives, how they learned during the season, their views on limitations in aerial suppression, and their own capacity to document the process.

The interviews also highlighted several operational issues that warrant more investigation using a large number of aviation specialists and more specific questions. Chief among these are:

We conducted eight detailed case studies where there were interesting features in the drop data and insightful comments from the interviewees. These were particular days at a particular part of a fire. They included one example with multiple objectives playing out as one failed and the fire spread changed, several where property protection was the dominant objective (largely successful), one on spot fires, and two initial attacks, of which one succeeded and the other failed.\  The case studies demonstrate the power of the approach where spatial data and interview interpretation are combined.

The air drop data has the potential to enable deep analyses of aircraft use and effectiveness during real bushfire responses, especially when combined with other contextual information, such as objectives and environmental conditions. This will require more matching of the data to interviews to determine whether the drop data can be used in this way. We have started this process in this report, identifying clear clusters of activity related to weather and distance to houses, and cross-checking with interviews in the case study, and in some of these cases, the success could be judged. In order to realise the full potential of this approach, the completeness and accuracy of the drop data should be improved and interviews should become a routine part of the seasonal review process.

}, issn = {725}, author = {Heather Simpson and Michael Storey and Matt P Plucinski and Owen Price} } @article {bnh-8115, title = {Derivation of a Bayesian fire spread model using large-scale wildfire observations}, journal = {Environmental Modelling \& Software}, year = {2021}, month = {07/2021}, abstract = {

Models that predict wildfire rate of spread (ROS) play an important role in decision-making during firefighting operations, including fire crew placement and timing of community evacuations. Here, we use a large set of remotely sensed wildfire observations, and explanatory data (focusing on weather), to demonstrate a Bayesian probabilistic ROS modelling approach. Our approach has two major advantages: (1) Using actual wildfire observations, instead of controlled fire observations, makes models developed well-suited to wildfire prediction; (2) Bayesian modelling accounts for the complex nature of wildfire spread by explicitly considering uncertainty in the data to produce probabilistic ROS predictions. We show that highly informative probabilistic predictions can be made from a simple Bayesian model containing wind speed, relative humidity and soil moisture. We also compare Bayesian model predictions to those of widely used deterministic ROS models in Australia.

}, keywords = {Wildfire Bushfire Fire behaviour Bayesian Bayesian modelling Rate of Spread}, doi = {https://doi.org/10.1016/j.envsoft.2021.105127}, url = {https://www.sciencedirect.com/science/article/abs/pii/S1364815221001705}, author = {Michael Storey and Bedward, M. and Owen Price and Ross Bradstock and Jason J. Sharples} } @article {bnh-7795, title = {Experiments on the influence of spot fire and topography interaction on fire rate of spread}, journal = {PLOS ONE}, volume = {16}, year = {2021}, month = {01/2021}, abstract = {

Spotting is thought to increase wildfire rate of spread (ROS) and in some cases become the main mechanism for spread. The role of spotting in wildfire spread is controlled by many factors including fire intensity, number of and distance between spot fires, weather, fuel characteristics and topography. Through a set of 30 laboratory fire experiments on a 3 m x 4 m fuel bed, subject to air flow, we explored the influence of manually ignited spot fires (0, 1 or 2), the presence or absence of a model hill and their interaction on combined fire ROS (i.e. ROS incorporating main fire and merged spot fires). During experiments conducted on a flat fuel bed, spot fires (whether 1 or 2) had only a small influence on combined ROS. Slowest combined ROS was recorded when a hill was present and no spot fires were ignited, because the fires crept very slowly downslope and downwind of the hill. This was up to, depending on measurement interval, 5 times slower than ROS in the flat fuel bed experiments. However, ignition of 1 or 2 spot fires (with hill present) greatly increased combined ROS to similar levels as those recorded in the flat fuel bed experiments (depending on spread interval). The effect was strongest on the head fire, where spot fires merged directly with the main fire, but significant increases in off-centre ROS were also detected. Our findings suggest that under certain topographic conditions, spot fires can allow a fire to overcome the low spread potential of downslopes. Current models may underestimate wildfire ROS and fire arrival time in hilly terrain if the influence of spot fires on ROS is not incorporated into predictions.

}, keywords = {air flow, analysis of variance, Bushfire, combustion, fire research, fire suppression technology, fuels, rotors}, doi = {https://doi.org/10.1371/journal.pone.0245132}, url = {https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0245132}, author = {Michael Storey and Owen Price and M. Almeida and Carlos Ribeiro and Ross Bradstock and Jason J. Sharples} } @article {bnh-6874, title = {Analysis of Variation in Distance, Number, and Distribution of Spotting in Southeast Australian Wildfires}, journal = {Fire}, volume = {3}, year = {2020}, month = {04/2020}, abstract = {

Spotting during wildfires can significantly influence the way wildfires spread and reduce the chances of successful containment by fire crews. However, there is little published empirical evidence of the phenomenon. In this study, we have analysed spotting patterns observed from 251 wildfires from a database of over 8000 aerial line scan images capturing active wildfire across mainland southeast Australia between 2002 and 2018. The images were used to measure spot fire numbers, number of {\textquotedblleft}long-distance{\textquotedblright} spot fires (\> 500 m), and maximum spotting distance. We describe three types of spotting distance distributions, compare patterns among different regions of southeast Australia, and associate these with broad measures of rainfall, elevation, and fuel type. We found a relatively high correlation between spotting distance and numbers; however, there were also several cases of wildfires with low spot fire numbers producing very long-distance spot fires. Most long-distance spotting was associated with a {\textquotedblleft}multi-modal{\textquotedblright} distribution type, where high numbers of spot fires ignite close to the source fire and isolated or small clumps of spot fires ignite at longer distances. The multi-modal distribution suggests that current models of spotting distance, which typically follow an exponential-shaped distribution, could underestimate long-distance spotting. We also found considerable regional variation in spotting phenomena that may be associated with significant variation in rainfall, topographic ruggedness, and fuel descriptors. East Victoria was the most spot-fire-prone of the regions, particularly in terms of long-distance spotting.

}, keywords = {spot fire; spotting distance; spotting distribution; wildfire behaviour}, doi = {https://doi.org/10.3390/fire3020010}, url = {https://www.mdpi.com/2571-6255/3/2/10}, author = {Michael Storey and Owen Price and Ross Bradstock and Jason J. Sharples} } @article {bnh-6754, title = {Drivers of long-distance spotting during wildfires in south-eastern Australia}, journal = {International Journal of Wildland Fire}, year = {2020}, month = {03/2020}, abstract = {

We analysed the influence of wildfire area, topography, fuel, surface weather and upper-level weather conditions on long-distance spotting during wildfires. The analysis was based on a large dataset of 338 observations, from aircraft-acquired optical line scans, of spotting wildfires in south-east Australia between 2002 and 2018. Source fire area (a measure of fire activity) was the most important predictor of maximum spotting distance and the number of long-distance spot fires produced (i.e. \>500 m from a source fire). Weather (surface and upper-level), vegetation and topographic variables had important secondary effects. Spotting distance and number of long-distance spot fires increased strongly with increasing source fire area, particularly under strong winds and in areas containing dense forest and steep slopes. General vegetation descriptors better predicted spotting compared with bark hazard and presence variables, suggesting systems that measure and map bark spotting potential need improvement. The results from this study have important implications for the development of predictive spotting and wildfire behaviour models.

}, keywords = {Fire behaviour, line scan, spot fire}, doi = {https://doi.org/10.1071/WF19124}, url = {https://www.publish.csiro.au/wf/WF19124}, author = {Michael Storey and Owen Price and Jason J. Sharples and Ross Bradstock} } @article {bnh-3128, title = {The role of weather, past fire and topography in crown fire occurrence in eastern Australia}, journal = {International Journal of Wildland Fire}, volume = {25}, year = {2016}, month = {09/2016}, abstract = {

We analysed the influence of weather, time since fire (TSF) and topography on the occurrence of crown fire, as mapped from satellite imagery, in 23 of the largest wildfires in dry sclerophyll forests in eastern Australia from 2002 to 2013. Fires were analysed both individually and as groups. Fire weather was the most important predictor of crown consumption. TSF (a surrogate for fuel accumulation) had complex nonlinear effects that varied among fires. Crown fire likelihood was low up to 4 years post-fire, peaked at ~10 years post-fire and then declined. There was no clear indication that recent burning became more or less effective as fire weather became more severe. Steeper slope reduced crown fire likelihood, contrary to the assumptions of common fire behaviour equations. More exposed areas (ridges and plains) had higher crown fire likelihood. Our results suggest prescribed burning to maintain an average of 10 years{\textquoteright} TSF may actually increase crown fire likelihood, but burning much more frequently can be effective for risk reduction. Our results also suggest the effects of weather, TSF and slope are not adequately represented in the underlying equations of most fire behaviour models, potentially leading to poor prediction of fire spread and risk.

}, url = {http://www.publish.csiro.au/WF/WF15171}, author = {Michael Storey and Owen Price} }