Numerical modelling of wildland fire spread on the windward and leeward sides of a ridge

Wildland fires are occasionally observed to spread rapidly across steep leeward slopes, in a direction that is approximately transverse to the background wind. Laboratory experiments and remote sensing of wildland fires suggest that this atypical lateral fire spread can occur regardless of whether the fire is ignited on the leeward slope or advances onto it from the windward slope. The lateral rate of fire spread is typically greatest on the leeward slope close to the ridge, and can considerably affect the subsequent development of the fire, particularly through spotting downwind of the slope. Until now, numerical modelling of this atypical lateral fire spread has considered only the scenario in which a wildland fire is ignited on the leeward slope, with no fire initially present on the windward slope. However, such a scenario will not always occur in reality, and it is fairly common for a wildland fire to cross over from a windward slope to a leeward slope, due to the combined effects of the slope and background wind on the fire spread. It is therefore of interest to establish how a wildland fire can behave as it crosses from a windward to a leeward slope, for a variety of different fire environment conditions, in particular the background wind speed and terrain slope. The aim of this study is therefore to conduct a series of idealised numerical simulations of wildland fire spread, starting from an ignition on a windward slope and with the fire subsequently crossing onto a steep leeward slope. In particular, the analysis focuses on the occurrence of lateral widening of the fire front as it progresses from the windward to the leeward slope, and the relationship of this lateral widening to the background wind speed. As in previous numerical modelling of atypical lateral fire spread, the WRF numerical weather predicition model is used in a large eddy simulation configuration and coupled to the WRF-Fire wildland fire physics module, allowing for direct modelling of the two-way coupled atmosphere-fire interactions. However, in contrast to previous modelling efforts, we amend the way the combined effects of wind and slope on fire spread are modelled in WRF-Fire. This amendment is made to better account for downslope fire spread in particular, though it also offers advantages in other respects. It has previously been demonstrated that to resolve the fire whirls and turbulent atmospheric eddies that are predominantly responsible for driving the atypical lateral fire spread, is necessary to implement the simulations at high spatial and temporal resolution. In this study we use a spatial resolution of 30 m and a time step of 0.04 seconds. Simulations were conducted with the fire to atmosphere coupling enabled and disabled to examine the relative effect of coupled fire-atmosphere feedbacks on the lateral fire spread. The non-coupled simulations failed to produce any significant lateral spread on the windward or leeward slope for any of the background wind speeds considered. In contrast, the coupled simulations exhibited significant lateral spread as the windward slope fire crossed the ridge onto the leeward slope for background wind speeds above 10 m s−1. In the coupled simulations the occurrence of lateral spread was strongly associated with the formation of pyrogenic vortices, otherwise known as fire whirls. This is in accordance with previous results concerning the modelling of atypical lateral spread of leeward slope fires. The modelling results were used to examine the relationship between lateral enhancement of fire spread on the leeward slope and the background wind speed. Raposo et al. (2015) found a power law relationship between the lateral spread enhancement on the leeward slope and the background wind speed in a series of laboratory experiments, as well as in three wildland fire cases. The coupled simulations exhibited a general increase in the lateral enhancement of fire spread for all but the highest background wind speed, though no obvious power law relationship was evident. A number of reasons for the differences in these findings with previous work are briefly discussed.