Mountain waves and extreme fire behaviour

By Dr Jeff Kepert. This article first appeared in Issue Two 2017 of Fire Australia.

One of the most challenging situations in fire management is when relatively non-threatening weather conditions are expected, but a severe fire eventuates. These situations can result in significant loss of property or even life.

Identifying the cause of such incorrect expectations can help to prevent them from recurring in the future. Through analysis of recent bushfires, Bushfire and Natural Hazards CRC research has identified three cases where a weather phenomenon known as ‘mountain waves’ have contributed to severe fire behaviour.

Mountain waves are atmospheric oscillations that occur due to air flowing over hills or mountains. They can result in particularly disastrous consequences during a fire, and can arise in several different ways—some more predictable than others. Often, mountain waves cause strong downslope winds on the lee slope of a hill or mountain. They are extremely complex to predict because their existence and amplitude is sensitive to the atmospheric temperature structure and vertical variation of the wind.

Modelling weather conditions

As part of our research into bushfire meteorology, the team uses the Bureau of Meteorology operational numerical weather prediction system, the Australian Community Climate and Earth-System Simulator (ACCESS), to conduct case studies into severe fire weather events. We configure ACCESS in research mode, to simulates the weather at very high resolution, with a grid spacing of around 440 m. After running the model, the simulation is verified against the observed weather from the event. If the simulation proves to be sufficiently accurate, we can assume it is a good representation of what actually occurred.

In many cases, the team has uncovered instances of fine-scale meteorology—too small in scale to be resolved by the operational models or captured by available observations—that would have contributed to more severe fire behaviour. Often these features would either not have been depicted in the traditional fire-danger parameters of surface temperature, humidity and wind, or they would have been filtered out in a broader-scale depiction of the data.

After analysing several significant bushfires, we uncovered mountain waves as a factor in three fires: the NSW State Mine fire in 2013, the Margaret River fire in WA in 2011, and the Victorian Aberfeldy fire in 2013. This suggests that the impact of mountain waves on fire is a reasonably common problem—one that we need to learn more about.

State Mine fire, NSW

In October 2013, large and destructive fires burnt through the Blue Mountains, destroying more than 200 homes.

A detailed case study of the fires focused on 17 October, when the majority of the damage occurred. On this day, the State Mine fire grew from 1,036 to 12,436 hectares in around 10 hours. This was severe fire behaviour by any definition—especially occurring so early in the fire season, even though preceding conditions had been dry.

Modelling of the weather during the fire showed that a band of strong winds extended downwards towards the surface in the vicinity of the fire. Looking at the vertical motion of the model, alternating bands of ascent and descent are present. Together, these bands are the characteristic features of mountain waves. While wind speeds at higher altitudes are often significantly stronger than those at the surface, here the mountain waves have provided a mechanism to bring these strong winds downwards to where they can directly impact the fire.

The modelling also showed a marked ‘dry slot’ of drier air passing over the fireground during the day. A dry slot, in the context of fire weather, is a relatively long, narrow band of dry air often associated with a wind change. The slot can cause sudden drops in humidity and increases in wind speed if it mixes down to the surface. The onset of dry air can reduce fine-fuel moisture and thereby elevate the fire risk.

Margaret River fire, WA

In 2011, a prescribed burn in the Margaret River region escaped overnight. Strong winds on the following day resulted in the fire destroying 39 homes in the communities of Prevelly and Gnarabup.

While the fire activity was reasonably consistent with the fuels and weather on the day, the behaviour overnight was not. For the days preceding the escape, the fuels had been reluctant to burn—to the extent that it was decided to leave the fire overnight. Early the following morning, fire crews returned to find that the fire had dramatically intensified, and was in the process of crossing control lines. As it spread into heavily inaccessible terrain they were unable to contain it burning into Prevelly and Gnarabup.

The area of the prescribed burn included the southern slopes of a small hill about 200 m in height. Modelling of the event showed that overnight, as the wind tended northerly, strong downslope winds developed on this slope (see Figure 1 overleaf). This reinvigorated the fire, pushing it towards the containment line. In this particular case—and in contrast to the State Mine fire—strong near surface atmospheric stability due to a nocturnal temperature inversion were crucial to the development of the mountain waves. A further meteorological contributor to the unexpected fire behaviour was the dry continental air that moved over the fire earlier in the night, making the fuels more flammable.

Aberfeldy fire, Victoria

In January 2013, a fire that ignited in Aberfeldy, Victoria, tragically took one life and destroyed dozens of houses. The fire experienced unexpected activity on the night of 17 January. Like the Margaret River fire, this was against the usual diurnal trend. In contrast to the Margaret River fire, however, the Aberfeldy fireground was at a much higher elevation, being on the southern slopes of the Great Divide overlooking the Latrobe Valley.

Modelling showed clear evidence that mountain waves and strong downslope winds developed overnight. These winds would have directly increased the fire intensity and spread, as well as contributed to firebrand transport. However, other factors also likely contributed. One influence was the steep and rugged topography. Another was that the fireground, being elevated, was in the warm, dry air above the nocturnal inversion. This would have limited overnight recovery of the fuel moisture.

Forecasting in the future

So what do mountain waves mean for fire management? The ingredients that led to the Margaret River fire are well known: nocturnal cooling, reasonable strong synoptic flow, gentle upwind slope and steeper downwind. Other mountain wave forecasting cases are more difficult, because mountain wave activity is sensitive to the atmospheric wind and temperature structure, to the shape of a particular hill or mountain, and to the topography upwind. Forecasters could be reasonably confident of some activity, but unsure whether it is strong enough to cause a serious problem.

Theory helps in some cases, but not all. For example, attempts to fit the Blue Mountains case study into one of the existing theoretical paradigms were not successful. On the other hand, sufficiently high-resolution modelling can capture at least some of these events, and the Bureau of Meteorology’s new ‘city domain’ versions of ACCESS have a grid spacing of 1.5 km, which should suffice in many circumstances. However, these are only available in those regions covered by those models—in most cases, a roughly 1,000- km square centred on the state capitals.

A further problem is that the area affected by mountain waves is often comparatively small. In the case of the Margaret River fire, for instance, it was only a few kilometres across. District-level forecasts may be too broad-scale to capture the effect, as will forecasts based on anything but the finest-resolution numerical weather prediction.

Even if numerical weather prediction could reasonably correctly resolve the occurrence and amplitude of mountain waves, the information still needs to reach the fire manager. A request for a spot forecast from the fire manager mentioning the precise location and perhaps the possible concern regarding mountain waves, might be necessary. But this approach could rapidly become impractical when there are many fires, and fire managers might be reluctant to request a spot forecast when they expect the fire behaviour to be benign. Both meteorologists and fire managers might require training in the potential impacts of mountain waves on fires.

These studies show the usefulness of high-resolution numerical weather prediction in diagnosing the cause of unexpectedly severe bushfires. This ability should translate into skill in a forecast situation. But high-resolution numerical weather prediction contains a wealth of fine-scale three-dimensional detail—only a small part of which will be relevant to a particular situation. Teasing out the useful information, and avoiding swamping the user with unnecessary detail, are challenges that will become harder as forecast capabilities increase.

These studies also illustrate the value of research. As we become more aware of the subtle ways that meteorology, topography and fire can interact, we can learn from adverse outcomes and be better prepared into the future.