1. Introduction

Fires and floods are a regular part of the landscape in mountainous regions around the world. For example, in the Rocky Mountains of the western United States, the climate ranges from arid and semiarid to intensely alpine to monsoonal (Western Regional Climate Center 2006). The Victorian Alps of southeastern Australia are, by contrast, more temperate. Rainfall has historically been more predictable and less seasonal in character, the terrain less rugged, and drought more rare (Australian Alps Liaison Committee 2005). Nevertheless, particularly in the context of increased development and anthropogenic climate change, in recent years the perceived vulnerability of the inhabitants of the region to extremes, such as fire and flood, has been increasing (W. Raynor, Alpine Shire resident, 2005, personal communication).¹

In fact, the relationship between climate variation and the occurrence of fire and flood respectively is a complex one. In the case of wildfire, studies such as Schoenberg et al. (2003), Keeley et al. (2004), and Abramson et al. (2007, manuscript submitted to Int. J. Wildland Fire, hereafter AEA) suggest that local geography and the details of seasonal and intraseasonal variations have a sufficiently strong influence that general conclusions are difficult to draw. Fires in the state of Victoria are generally associated with a small range of synoptic environments that support strong northerly winds (Long 2006). AEA have found that across Victoria, there is some evidence over recent decades that local warming and drying has led to an earlier start to the fire season and a larger total area burned, primarily through increasing the flammability of fuel. This is consistent with the findings of, for example, Westerling et al. (2006) who noted in the Rocky Mountains of the United States that longer fire seasons were associated with a general trend of warming summer and spring temperatures. However, in the Victorian alpine region specifically, there is little evidence of drying in the fire season, little evidence of warming, and no statistically significant trend in fire season or severity. In the record, there have been around² 19 major fires in the Victorian Alps in the last 70 yr, but reduced precipitation in autumn is offset by generally wetter summers. Further, while the daily maximum temperature has increased very slightly in all seasons, the daily minimum temperature has decreased by a larger amount, particularly in autumn.

Similarly, the source of the perception of Victorian alpine residents and decision makers of increased vulnerability to flooding is difficult to characterize. In other parts of the world, there is evidence of increasing weather hazards. In a report to the Association of British Insurers (Climate Risk Management 2005), it was noted that between 1990 and 2004, there have been at least 20 weather events around the world annually that were severe enough to be classified by reinsurers as significant natural catastrophes. In contrast, between 1970 and 1990 only three years experienced more than 20 such events. However, this observed increase brings little evidence for a general increase in weather-related mortality, at least until Hurricane Katrina. For example, the total number of flood events in Europe has increased since 1974, while the number of deaths per flood event has decreased, likely due to improved forecasting, improved warning systems, and a greater awareness of risks (McMichael et al. 2003). This decreased mortality rate is despite increased vulnerability due to nonclimatic factors, such as population growth in areas at risk of flooding. In the Alpine Shire, although there have been 15 significant floods (over the minor flood level, including riverine and flash flooding) in the last century, there is no evidence of an upward trend in the frequency of floods or mortality associated with them.

In the Alpine Shire of Victoria (Fig. 1a), the fire season ranges from early spring to late summer, and fires are frequently located in areas of rugged terrain in the national park and state forest, which have limited access. The heavily forested area typically provides a large fuel supply. Combined with a relatively small population, this can make containing fires difficult. The typical flooding season is in early spring, associated with snowmelt at higher elevations. The year 2003 saw an unusual conjuncture in the Alpine Shire that combined both of these types of extremes in the form of a postfire flash flood. On the evening of 7 January 2003 (a day of total fire ban across the entire state of Victoria), a weather system generating many thunderstorms swept across eastern Victoria and southern New South Wales (Fig. 1a). Lightning associated with these thunderstorms started over 80 individual fires in Victoria, which formed the inception of Victoria’s largest bushfire since the fire of 1939 (known as “Black Friday”), in which 71 lives were lost and over 1000 homes were destroyed. Major fire episodes in Victoria, like those in 1939 and 2003, are generally associated with long-term rainfall deficiencies and prolonged drought. Even in summer the gully environments tend to be moist; however, streamflows during the most severe dry spells reduce significantly so that these natural barriers to fire spread are nonexistent (P. Billing, Department of Sustainability and Environment, 2007, personal communication). In 2003, the fires could not be fully controlled for a period of nearly 60 days. No lives were lost as a direct result of the 2003 fires, in which 75 000 ha of farmland, 241 buildings, and 110 000 head of stock were destroyed. However, a firefighter lost her life in Alpine Shire when storms generated localized flash flooding in late February, causing her utility truck to be swept away as she attempted a bridge crossing.

Postfire floods may be associated with several different meteorological mechanisms and either may occur immediately following the fire or may be delayed by several weeks or more. During or immediately following a fire, rain-bearing clouds known as pyrocumulus can form, drawing on the aerosols, water vapor, and intense heat released from fires (Potter 2005; Damoah et al. 2006; Fromm et al. 2006). Alternatively, a fire started by lightning strikes ahead of an advancing cold front may then be extinguished by frontal precipitation that is extreme enough to cause flooding. Delayed floods are more likely to be caused by surface modifications reducing infiltration, with precipitation due to either a large-scale drought break or localized thunderstorms. In combination, these processes can create an enhanced potential for a severe flooding event.

Postfire mudflows and flash floods represent a particularly acute problem in mountainous regions. Burned catchments are at an increased hydrological risk and respond faster to rainfall than unburned catchments (Meyer et al. 1995; Cannon et al. 1998; Wilson 1999). Bushfires affect the hydrogeological response of catchments through the destruction of vegetation and the alteration of soil structure and properties. An intense fire modifies the surface soil structure such that the soil becomes hydrophobic (Conedera et al. 1998; Huffman et al. 2001; Letey 2001; Martin and Moody 2001; Shakesby and Doerr 2006), and there is an increase in runoff of both precipitation and snowmelt (Giovannini 1994; Marxer et al. 1998; Robichaud and Brown 1999; Giovannini et al. 2001). The loss of vegetation can also contribute to increased runoff and erosion through decreases in plant interception. Increased runoff can lower the threshold of intensity and the amount of precipitation necessary to cause a flood event and exacerbate the impact of the precipitation, even to the extent of a 10-yr rainfall event causing a 100–200-yr flood event (Conedera et al. 2003). Combined with steep slopes, this can create the potential for flash floods. The nonmeteorological factors leading to postfire flash flooding will be the subject of a subsequent hydrological study (Gallucci et al. 2008).

Residents of the Alpine Shire have expressed concern over the 2003 postfire flood event, and in interviews many wanted to know whether they should typically expect a flood after a fire. We have been able to document that this type of postfire flash flood event has not occurred particularly frequently in the context of the regular bushfire regime and alpine spring flooding events. A comprehensive survey of all the data available, including local newspapers, interviews, historical records, and precipitation and streamflow records, has failed to identify another definitive case in this region over the past 5 decades. However, the data quality in the Alpine Shire is extremely poor, and combined with a low population it is possible that other similar events have gone unrecorded. Further, studies performed in other mountainous regions indicate that floods that occur soon after fires have been common events (Morris and Moses 1987; Meyer et al. 1995; Cannon et al. 1998; Elliot and Parker 2001). Other examples of postfire floods include: Storm King Mountain, Colorado, in 1994 (Cannon et al. 2001b); Buffalo Creek, Colorado, in 1996 (Chen et al. 2001; Elliot and Parker 2001); Riale Buffaga, Switzerland in 1997, (Conedera et al. 2003); Cerro Grande, New Mexico, in 2000 (Cannon et al. 2001a); San Bernardino County, California, in 2003 (Barkley and Othmer 2004); and Licola, Victoria, Australia, in 2007 (Houghton 2007). It is perhaps significant that these events have generally occurred in more arid regions, with a climate more typical of the last few decades in Alpine Shire.

This paper describes an analysis of some of the candidate mechanisms leading to this extreme rainfall event in Alpine Shire, Victoria. It is known from the limited available observations that the flood event was preceded by very intense, localized rainfall totals. Previous work on modeling extreme rainfall has shown that the representation of a burned fire area in a model causes an enhancement of the convection over that area (Chen et al. 2001). Hence, the candidate mechanisms we focus upon in this study are associated with surface modifications and their impact on the meteorological conditions preceding the flooding event. Changes in the partitioning of the surface energy balance could affect boundary layer depths, clouds, convection, and precipitation.

2. Model description and methodology

This analysis was performed using the fifth-generation Pennsylvania State University–National Center for Atmospheric Research (PSU–NCAR) Mesoscale Model (MM5). MM5 is a limited-area, nonhydrostatic model capable of simulating meso- and synoptic-scale atmospheric circulations (Grell et al. 1994). The model has been used in many other studies of precipitation events in mountainous areas and has been shown to perform adequately for this type of application (Colle and Mass 2000; Rotunno and Ferretti 2001; Lin and Chen 2002; Das 2005; Galewsky and Sobel 2005; Li et al. 2005; Monaghan et al. 2005). To analyze the meteorological processes associated with the extreme precipitation event preceding the flash flood, from the sample space of initial conditions, a five-member ensemble of 5-day forecasts was performed from 1200 UTC 23 February to 1200 UTC 28 February 2003. The flash flood occurred at around 0600 UTC 26 February 2003. The initial and boundary conditions for the model were created using the National Centers for Environmental Prediction (NCEP)–NCAR reanalysis data at a grid spacing of 2.5° latitude × 2.5° longitude. To achieve sufficient horizontal resolution, a series of nested domains were configured. Domain 1 (Fig. 1a) is centered at 36.7°S latitude and 147.8°E longitude and has an extent of 1800 km × 1550 km, with a grid resolution of 50 km and 23 vertical levels. Domains 2 and 3 (Figs. 1a,b) have resolutions of 10 and 2 km, respectively, with 35 vertical levels. The simulations were all nested one way.

The topography for all domains was derived from U.S. Geological Survey terrain data at resolutions of 55, 9.25, and 3.70 km. The Noah land surface model was used. The land surface in MM5 for the area burned in 2003 was changed to resemble a recently burned fire surface, with decreased albedo, lowered surface roughness (Z₀), and reduced soil moisture. The Buckland River Catchment is 320 km² in size and approximately 85% of the catchment was burned (North East Water 2003). Based on the interviews with Shire residents and photos during and after the event, it was apparent that the fire was of great intensity. The values chosen reflect this observation and were based on previous work on modeling fire scars (J. Beringer et al. 2003; Görgen et al. 2006; C. Wendt et al. 2007). The albedo was reduced from 0.20 to 0.08, and the roughness length was from 2.65 to 0.10 m. The soil moisture for the uppermost 10-cm layer was initialized at 0.05 m³ m⁻³, in contrast to standard values of between 0.25 and 0.45 m³ m⁻³.

3. The observed event

At 1200 UTC 23 February, at the time of forecast initialization, the operational analysis showed a large blocking high in the Tasman Sea and a monsoonal trough extending south from the Cloncurry heat low [Fig. 2a(i)]. This blocking situation had persisted for several weeks and was associated with a “dry slot” extending from northern Australia toward the southeast, as evident in Geostationary Meteorological Satellite-5 (GMS-5) imagery showing water vapor (Fig. 2b) (Mills 2005). A cold front originated in a low in the Southern Ocean was advancing, but to the southeast.

At the time of the flood, at approximately 0600 UTC 26 February 2003, the blocking high in the Tasman Sea had persisted, the cold front had passed to the south of the continent, and a subtropical ridge extended through the Bass Strait [Fig. 2a(ii)]. A new cold front originating in the Southern Ocean was advancing toward Victoria. Moisture levels had increased after several days of mostly east to northeasterly flow, bringing warm, moist subtropical air from the north (Figs. 2a,b), and values of precipitable water were high (above 30 kg m⁻²; Yeo 2003). No large-scale precipitation was recorded across the state at this time. The Bureau of Meteorology forecast possible isolated thunderstorms over the ranges (i.e., the alpine regions in northeast Victoria) in the early morning (2200 UTC) public weather forecasts on 26 February (Yeo 2003).

Thunderstorms in the Buckland River valley (Fig. 1b) were first identified on the Melbourne radar at 0530 UTC. These storms are clearly visible on the water vapor image at this time [Fig. 2b(ii)] and are confined to a small region in the northeast of Victoria. These thunderstorms produced short bursts of heavy rainfall. Severe thunderstorm advice,³ containing warnings of possible flash flooding, was issued at 0619 UTC. Redevelopment of cells in the same area meant that one small area of the Buckland River Catchment, between Mount Hotham and Mount Selwyn, experienced intense inundation for several hours. The Mount Hotham Airport Automatic Weather Station (AWS) recorded 30 mm in 20 min and a 24-h precipitation total of 37.2 mm. This was the only recorded precipitation in the vicinity of the flood itself—unfortunately there are no rain gauges located in the Dingo Creek area where the flood occurred. Over the time the Mount Hotham Airport AWS has been in operation (since July 2000), no other event of this intensity has been recorded. Figure 3 shows the rainfall recorded at the nearest available stations in the Alpine Shire during the period of interest. One of the factors that could not be fully addressed was rainfall rate, because of the paucity of contextual data (i.e., a climatology of rates and floods) and limitations in model performance. The precipitation resulted in Dingo Creek, which feeds into the Buckland River, flooding rapidly. It was around this time that the firefighter’s life was lost in an attempt to cross a bridge over Dingo Creek. The streamflow gauge on the Buckland River shows a substantial peak at the time of the event (Fig. 4); however, according to the official Victorian flood classifications, this was only a minor flood (over 1.8 m and less than 2.5 m). Minor floods are defined as floods in which low-lying areas next to watercourses are inundated and may require the removal of stock and equipment. Minor roads may be closed and low-level bridges submerged. The recurrence interval for the observed peak flow (over 6000 MLday⁻¹) is 2 yr (Ian Drummond and Associates 1999). The bushfires that had devastated the area in previous weeks meant that the valley where Dingo Creek flows was filled with a high quantity of debris, including mud, fallen trees, and boulders. As mentioned previously, approximately 85% of the Buckland River Catchment was burned, with much ash and sediment lying on the ground. The resulting lack of vegetation produced greater runoff and loosening topsoil, and lead to the movement of weakened trees (Yeo 2003; North East Water 2003). The severe storm caused major landslips, river flooding, and deposition of a massive load of ash and sediment into the water. Large quantities of debris were swept into the Buckland River and thence into the Ovens River, which supplies several large towns in the region and downstream with drinking water. In the weeks after the event this led to water quality problems in northeast Victoria, with large amounts of sediment blocking the waterways. Sampling of the water by North East Water, showed turbidity levels in the Buckland River of up to 120 000 nephelometric turbidity units (NTU); normal turbidity levels are around 1–2 NTU (North East Water 2003). To measure the turbidity level, samples had to be diluted 1000 times in order to register a reading. By 7 March 2007 the sludge had traveled 110 km downstream to the Wangaratta. The filters at North East Water’s treatment plant could not cope with the thick sludge, and consequently their daily output of clean water was cut to 18 ML from the normal 40 ML. The turbidity of 3–5000 NTU far exceeded the normal flood event levels of 180 NTU that the plant had been designed to treat.

The alpine region is a favorable location for thunderstorm development because of the ability of the ranges to provide enhanced uplift, act as a blocking mechanism to horizontal airflow, and allow the formation of afternoon anabatic breezes. Given the synoptic situation it is likely that anabatic breezes played a significant part in thunderstorm initiation, which eventually allowed convective inhibition to be overcome. The lateness in the afternoon meant that it is also highly probable that moisture had been allowed to build up in the boundary layer. Once convective inhibition had been overcome it is likely the convective column would have expanded significantly in the vertical direction.

The analysis 6 h later at 1200 UTC 26 February [Fig. 2a(iii)] shows the large-scale situation persisting. The front continued to advance slowly on Victoria, but had not reached the continent. There were no large-scale features present that would account for the precipitation in the Alpine Shire. Two days after the event, at 0000 UTC 28 February 2003, the blocking high had moved on and the cold front had reached eastern Victoria [Fig. 2a(iv)]. This cold front brought with it further precipitation across the Alpine Shire (Fig. 2a) and the state of Victoria.

4. Large-scale simulation

The results presented in this and the following sections are calculated using the ensemble mean. The sea level pressure simulation for domain 1 accurately captured the features of the observed sea level pressure evolution over the 5-day forecast (Fig. 5a). The forecast initialization at 1200 UTC 23 February [Fig. 5a(i)] shows a high centered in the Tasman Sea, and this remained in place for much of the forecast, similar to that observed. There is northerly flow over the Alpine Shire region and a dry slot extending from the north (not shown).

At the time of the flood (0600 UTC 26 February), the high remains over the Tasman Sea, with northeasterly flow over most of the forecast area [Fig. 5a(ii)]. There was no cloud or precipitation simulated in domain 1 in the area of the Alpine Shire. The large-scale water vapor pattern in the model [Fig. 5b(ii)] does not show the storms over northeast Victoria that are present in the satellite images [Fig. 2b(ii)]. In both the observations and the model there appears to be some moisture inflow from the Tasman Sea. The 1200 UTC 26 February simulation shows the blocking situation persisting, but with precipitation enhanced east of the Great Dividing Range, with the northeasterly flow off the Tasman Sea [Fig. 5b(iii)].

Toward the end of the forecast, the cold front swept across the state, as can be seen in the simulation for 0000 UTC 28 February [Fig. 5b(iv)]. The front was delayed in the model by several hours; however, the large-scale precipitation in the model associated with the front agreed well with observations. Overall, the evolution of synoptic-scale events was well simulated, apart from the absence of thunderstorm activity in northeast Victoria.

5. Focus on the Alpine Shire

The precipitation time series over the forecast for a model grid box in domain 3 over Dingo Creek is shown alongside the observed rainfall in Fig. 2. The rainfall in the model in the 24 h before 2200 UTC 26 February, the time period associated with the flood, was 31.7 mm. This is in good agreement with the closest observed rainfall of 37.2 mm. As shown, 7.6 mm fell in the model on the following day associated with passage of a cold front; because of the lateness of the front in the model this is an underprediction compared to the surrounding stations. Spatially the rain was very variable, both in the model [Figs. 6(iii), 6(iv)] and in the observations. Several of the stations in the Alpine Shire were missing data; however, 15 stations did record rainfall varying from 1.2 to 41.2 mm, with an average per station of 20.7 mm. The 6-h totals of rainfall in domain 3 at the time of the flood reveals significant falls across the southern and southwestern areas of the domain [Figs. 6(i), 6(ii)]. This contrasts with the synoptic experiment at that time, in which no rainfall was simulated over the Alpine Shire until the passage of the cold front on 27 February (Figs. 5b, 6(iii) and 6(iv)]. This suggests a highly localized event.

Soundings from the best model forecast at the start of the simulation and during the storm are shown in Figs. 7a,b. The model sounding at 0600 UTC 26 February shows a completely saturated atmosphere at 750 hPa. The CAPE during the storm was calculated to be 1546 J kg⁻¹, indicating significant instability. The nearest observational sounding available was at Wagga Wagga, approximately 300 km from the storm. The upper-air temperatures over the region were very similar to those in the alpine area. To validate the model, the grid box closest to Wagga Wagga was plotted against the observed sounding for the time closest to the storm (Fig. 7c). The model sounding is smoother than that of the observed sounding, as expected from a model grid box of several kilometers in extent with limited vertical levels compared with a point measurement. Nevertheless, the model captures the features of the observed sounding quite well.

This optimal model ensemble included changes to all the surface attributes that may have influenced the evolution of the event, namely, decreased albedo (to resemble a blackened surface), decreased surface roughness (reduced vegetation), and decreased soil moisture. Sensitivity simulations, as single realizations, were performed with these variables returned to standard values to determine the importance of these drivers to the simulation of precipitation (Table 1). A simulation (described as the unburned simulation) was also performed with no alterations made to any variables in the model. The unburned simulation produced little rain (3.9 mm) at the time of the flood; this was well outside the range of variation of the control ensemble, and hence we consider the response to be significant. The surface energy balance for the unburned simulation (taken from an average of nine grid boxes, not shown) had the highest positive latent heat flux of all the model simulations. It was also the only simulation with a negative ground heat flux transfer, indicating that the subsurface in the model was warmer than the surface. This simulation also had the lowest Bowen ratio (0.6), reflecting the abundant water supply.

It is apparent from these simulations that the roughness length is less important than albedo and soil moisture in driving the extreme rainfall event (Table 1). When the decreased soil moisture and albedo were set at unburned values and only the roughness length value was changed, the model produced rainfall of 7.9 mm in the 24 h before 2200 UTC 26 February. This increase, compared to the unburned simulation, appears to be related to increased surface heating due to the reduction in vegetation and associated evapotranspiration. The reduction in vegetation led to a decrease in the latent heat flux and an increase in the Bowen ratio to 1.6. When the roughness length and soil moisture were set to unburned values and only the albedo was changed, the model produced rainfall of 15.1 mm in the 24 h before 2200 UTC 26 February. This increase in rainfall was due to an enhancement of surface heating resulting from the increase in shortwave absorption, a mechanism more efficient in warming the surface than the vegetation change. The surface energy balance for this simulation has the second highest positive latent heat flux transfer of all the runs. In this scenario, more moisture was available from vegetation than in the simulations where the roughness length was reduced. The Bowen ratio for this simulation was 1.8. When the roughness length and albedo were both decreased from unburned values, only 10.4 mm fell in the 24 h before 2200 UTC 26 February. Analysis of the simulations suggests that the reduced near-surface moisture resulting from reduced vegetation was a significant effect in this simulation, which outweighed the enhancement of uplift resulting from surface heating over relatively bare surfaces. In this case the surface energy balance was very similar to the roughness length–only simulation, the only difference being a slightly higher ground heat transfer. These two simulations had the highest ground heat fluxes, indicating that more heat was transferred to the subsurface in these simulations because of the moisture available in the soil. The control ensemble mean simulation included reduced soil moisture in combination with reduced albedo and roughness; this resulted in rainfall of 31.7 mm. This simulation had a greatly increased sensible heat flux, leading to an enhanced Bowen ratio of 16.6 and increased boundary layer heating. This, in turn, led to an increase in convective clouds and an increase in precipitation. This simulation differed from the others in that it was the only one that saw an increase in the sensible heat flux, caused by the reduction in soil moisture. In the simulations with wet soils the surface sensible heat flux was reduced, the boundary layer deepened less rapidly, and convection was not as high. In all cases, the changes resulted in a positive ground heat transfer, a large reduction in the positive latent heat flux transfer, and, hence, an increased Bowen ratio.

These results indicate that atmospheric instability was enhanced by the decreased albedo and soil moisture of the recently burned fire surface. Changes in the partitioning of the surface energy balance, most significantly a reduction in latent heat transfer and an increase in ground heat transfer, led to increased surface heating and increased uplift. This combined with a moist air mass from the Tasman Sea created desirable conditions for thunderstorm formation.

6. Conclusions

We have analyzed the factors leading to the postfire flood event in Buckland Valley, Victoria, Australia, in February 2003. Observations suggest that the flood event occurred because of a highly localized thunderstorm not associated with the passage of a large-scale cold front. The results from our model simulations concur with these observations. Analysis of the synoptic conditions surrounding the event suggests that the major drivers of the extreme rainfall event were the above-average precipitable water in the atmosphere, and significant values of CAPE, producing strong updrafts within the thunderstorm, capable of supporting large quantities of suspended water droplets, and thunderstorm cell regeneration in the same area. Atmospheric instability was enhanced by anabatic breezes, above-average boundary layer moisture, and increased surface heating because of the reduced surface albedo and reduced soil moisture of the recently burned fire surface. Flash flooding was also enhanced by 1) the burned landscape, 2) the storm cells likely being pulse wet microbursts, 3) cell regeneration over the same area (very little horizontal movement), and 4) the small catchment size. Given the above-average precipitable water values, the weak upper wind environment, and the large CAPE values, there is a large probability that heavy rainfall leading to flash flooding and deep convection would have occurred regardless of the burned landscape. However, the burned landscape further increased the amount of instability and provided a greater chance that convective inhibition could be overcome. It is likely that a further contributor to the flash flood observed was the reduced infiltration often observed in recently burned catchments. This factor will be addressed in a subsequent study utilizing a hydrological modeling approach (Gallucci et al. 2008).

It is intended that the mechanisms elucidated in this study will assist in emergency preparedness in the Alpine Shire. Given the warmer conditions expected with near-term anthropogenic climate change for the Alpine Shire, and the concomitant increase in fires, this causal relationship, even for a rare event, has implications for emergency managers and Alpine Shire residents. The warning system that is currently in place in Victoria is for severe thunderstorms that produce very heavy rainfall that leads to localized flash flooding; however, the warning is typically issued for thunderstorms in a general area, and the service is not capable of being specific enough to identify particular creeks or streams that may be affected by flash flooding. There are also flood warnings for the larger rivers in Victoria, which are more a result of widespread heavy rainfall. Ash, mud, and debris slides are not specifically warned for in the action statements issued by the State Emergency Service, but a decision has been made recently to include such a statement in severe thunderstorm warnings when severe thunderstorms are expected to develop over a recently burned catchment. This has become an increasingly important issue after a recent event in Licola (a town located in a different region of the Victorian Alps), where following bushfires residents were flooded with mud, ash, and debris (Houghton 2007). This event resulted in the addition of flood watches for recently fire-affected areas in the Victorian Alps to Bureau of Meteorology forecasts. Vulnerability to postfire flooding has already been recognized in southern California, where in autumn 2005 the National Oceanic and Atmospheric Administration (NOAA) and the U.S. Geological Survey (USGS) established a demonstration of a flash flood and debris flow early warning system for recently burned areas (U.S. Geological Survey 2005). Given the danger and potential damage of these events to communities living in fire-prone alpine areas, it is important to improve our understanding of the mechanisms that can cause this type of flood event. Forecasting of these events could potentially be improved by the alteration of land surface properties for recently burned areas within forecast models. Community awareness of the risks to recently burned areas from flooding and the level of preparedness of local government and emergency managers will also have an important role in decreasing the vulnerability of communities to these events.