On the Meteorological and Hydrological Mechanisms Resulting

Australian Meteorological and Oceanographic Journal 61 (2011) 31-42

On the meteorological and hydrological mechanisms resulting in the 2003 post-fire flood event in Alpine Shire, Victoria Jillian Gallucci1, Lee Tryhorn2, Amanda Lynch3 and Kevin Parkyn4 1Department of Sustainability and Environment, Victoria, Australia. 2Department of Earth and Atmospheric Sciences, Cornell University, Ithaca, New York, USA. 3School of Geography and Environmental Science, Monash University, Victoria, Australia 4Australian Bureau of Meteorology

(Manuscript received May 2007, revised November 2010)

This paper describes an analysis of the factors surrounding a post-fire flooding event that occurred in the Alpine Shire, Victoria, Australia in late February 2003. Flash flooding occurred because of highly localised thunderstorms and was enhanced by the burned landscape. In addition, our analysis of the resulting situation suggests that other contributing factors included the storm cells likely being pulse wet microburst, cell regeneration over the same area, and the steepness of the Buckland River Catchment. The synoptic conditions surrounding the event suggest that the major drivers of the extreme rainfall event were the high levels of precipitable water in the atmosphere and enhanced atmospheric instability (including associated high CAPE values) from increased surface heating due to the reduction in surface albedo and soil moisture of the recently burned fire surface. In addition, weak upper atmospheric winds during the flash flood event contributed to slow storm movement. Hydrological modelling of the flash flooding event further indicated that fire-induced soil hydrophobicity was likely to have caused an enhancement of the hydrological response of the catchment. With an increase in frequency of fires expected in the Alpine Shire associated with anthropogenic climate change, the relationship between fire and flood, even for rare events, has implications for emergency managers and Alpine Shire residents. It is intended that the results of this study will assist in emergency preparedness and thereby contribute to a reduction in the vulnerability of fire-affected communities to post-fire flash floods.

Introduction February 2003 saw a post-fire flash flood event occur in the Alpine Shire of Victoria (Fig. 1). This flash flood followed the Flash flood events arise from high intensity rainfall events longest campaign to that time in the recorded history of fire in regions where the hydrological conditions are favourable fighting in Victoria, with fires that could not be controlled (Doswell et al. 1996). In order for heavy precipitation to for nearly 60 days. On 7 January 2003, a day of Total Fire occur, high rainfall rates must be sustained. Slow movement Ban across the entire state of Victoria, a weather system of rain-producing systems is required to maintain high generating many thunderstorms swept across eastern rainfall rates over a long duration. Flash floods represent Victoria and southern New South Wales. Lightning associated an acute problem in fire-prone mountainous regions, such with these thunderstorms started over 80 individual fires in as the Victorian Alps of southeastern Australia. For areas Victoria. Most of these were quickly controlled, but those in recently burned by fires, flash floods can be caused by lower remote, inaccessible terrain came together to form Victoria’s rainfall rates than would normally be required, even to the largest bushfire since the fire of 1939. No lives were lost in extent of a ten-year rainfall event causing a 100- to 200-year Victoria as a direct result of the 2003 fires, but 75 000 hectares flood event (Conedera et al. 2003). of farmland, 241 buildings and 110 000 head of stock were destroyed. Significantly, localised flash flooding immediately following the fires in the Dingo Creek area of Alpine Shire Corresponding author address: Jillian Gallucci, Department of Sustain- caused a firefighter to lose her life when her utility truck was ability and Environment, Locked Bag 3000, Box Hill, Vic 3128, Australia Email: [email protected] swept away as she attempted a bridge crossing.

31 32 Australian Meteorological and Oceanographic Journal 61:1 March 2011

Fig. 1 The three-nested model domains used to produce the Cerro Grande, New Mexico 2000 (Cannon et al. 2001b); and MM5 simulations. (a) Domains 1 (bold solid line) and Buffalo Creek, Colorado 1996 (Chen et al. 2001c; Elliot and 2 (dashed line) and (b) Domain 3 (bold dashed). The Parker 2001). In the Alpine Shire specifically, while both fires extent of the 2003 bushfires are shown in grey and the Alpine Shire boundary is shown as a solid line. and floods are not uncommon, a comprehensive survey of all the data available, including local newspapers, interviews, 145°0'0"E 150°0'0"E historical records, and precipitation and streamflow records, (a) failed to identify another definitive post-fire flash flood event over the past five decades. However, the poor data quality in the Alpine Shire leaves open the possibility that other similar New South Wales events in the region have gone unrecorded. For example, the 1939 bushfires were followed by state wide rain, resulting in erosion along the Macalister River that community members 35°0'0"S 35°0'0"S report is still visible today. Flash floods may occur immediately following a fire or be

Victoria delayed by several weeks, but nonetheless can be linked to the fire event through two categories of mechanism. First, Tasman Sea bushfires have the potential to affect the hydrogeological response of catchments. This occurs through the destruction 40°0'0"S Bass Strait 40°0'0"S of vegetation, which leads to decreases in canopy interception, surface roughness, transpiration and ground Tasmania litter storage (Scott 1997; Liu et al. 2004). The response can also be affected by the alteration of soil structure and properties, leading to increased hydrophobicity (Krammes

145°0'0"E 150°0'0"E and DeBano 1965; Conedera et al. 1998; Huffman et al. 2001; Legend Letey 2001; Martin and Moody 2001; Shakesby and Doerr CCoastlineoastline 2006). In the context of a rainfall event, these changes can SStatetate B oBoundaryundary AAlpinelpine Sh iShirere Bou nBoundarydary lead to increases in runoff and erosion and hence lower 22003003 bu Bushfireshfire the amount and rate of precipitation necessary to cause a DDomainomain 1 1 DDomainomain 2 2 flood event (Giovannini 1994; Marxer et al. 1998; Robichaud DDomainomain 3 3 and Brown 1999; Giovannini et al. 2001; Conedera et (b) al. 2003). Second, precipitation can be linked with fire- associated meteorological mechanisms. Surface heating has the potential to enhance the development of localised thunderstorms, which can be of sufficient intensity to initiate a flood event (Tryhorn et al. 2008). In combination with the steep slopes of mountainous regions and hydrogeological Myrtleford modifications, this process can create an enhanced potential Mt Buffalo Harris Lane Chalet Bright for severe flash flooding. Halls The aim of this study was to use both a meteorological and Falls Creek Mt Selwyn Mt Hotham a hydrological model to provide the Alpine Shire community Mt Hotham Airport with information on the basic mechanisms behind flash Alpine Shire 2003 Bushfire flooding immediately after fire. Overall, this research was Domain 3 River/Creek Buckland C’ment motivated by community need and aims to support their common interest. This paper describes an analysis of two of 0 10000 40000 60000metres the mechanisms contributing to this flash-flooding event. It is In interviews conducted by Amanda Lynch and Lee known from the limited available observations that the flood Tryhorn of the Monash Regional Climate Group in 2005, event was preceded by very intense, localised rainfall totals. residents of the Alpine Shire expressed concern over the Previous work on modelling extreme rainfall and surface 2003 post-fire flood event and many wanted to know whether hydrology has shown that the appropriate representation they should typically expect a flood after a fire. Concerns of a burned fire area in a model can cause an enhancement were exacerbated by a similar event in nearby Licola in 2007. of the convection (Chen et al. 2001c; Tryhorn et al. 2008) or This type of post-fire flash flood event has occurred in an increase in runoff (Beeson et al. 2001). The candidate other mountainous regions around the world including: San mechanisms focused upon in this study are associated with Bernardino County, California 2003 (Barkley and Othmer both of these sets of mechanisms: the surface modifications 2004); Riale Buffaga, Switzerland 1997 (Conedera et al. 2003); and their impact on the meteorological conditions preceding Storm King Mountain, Colorado 1994 (Cannon et al. 2001a); the flooding event, and the hydrological response during Gallucci et al.: Meteorological and hydrological mechanisms resulting in the 2003 post-fire flood event 33

and after the flooding event. UTC 27 February 0000 UTC 28 February 2003. This system The introduction has provided an overview of the produced widespread rainfall across northeast Victoria, motivation for this research and the linkages between including 21 mm at Falls Creek (24 hour total to 2200 UTC 27 recently burnt areas and flash flooding. The section that February), the majority of which fell between 1700 and 1900 follows (‘The observed event’) describes the observed flood UTC. Other rainfall totals in the surrounding area included event and a second rainfall event that occurred 36 hours 13.4 mm at Mount Hotham Airport, 26.6 mm at Mt Buffalo later. Data and model descriptions and methodology discuss Chalet, 20.8 mm at Bright, and 20 mm at Halls. The response the models and methodology used in this study. The results on the Buckland River at Harris Lane streamflow gauge was section details an analysis of the results. The conclusions of minor compared to the flash flood event (Tryhorn et al. 2008). this investigation can be found at the end of this paper. Broad synoptic conditions In the days leading up to the 2003 Buckland Valley flash The observed event flood event, the mean sea-level pressure analysis showed a persisting blocking high pressure system over the Tasman This section provides a detailed overview of the observations Sea and a monsoonal low pressure trough extending from associated with the flash-flood event. The section is organized Queensland toward Victoria (Fig. 2(a)). The mean sea-level as follows: ‘Rainfall and streamflow’ which describes the analysis at the time of the flood, valid at 0600 UTC 26 February observed rainfall and streamflow associated with both the 2003, shows an easterly trough extending from western New flash-flood event and the second rainfall event that followed, South Wales into eastern Victoria (Fig. 2(b)). The analysis also ‘Broad synoptic conditions’ characterizes the broad synoptic shows the persistent blocking high over the Tasman Sea, conditions, ‘Radar observations’ provides an overview of which is linked to a subtropical ridge that extends across the radar observations, ‘Modified soundings’ describes Tasmania toward Bight waters. A cold front associated with soundings modified to represent the atmosphere in the a deep low pressure system is evident over the Southern vicinity of the storms and ‘Pulse thunderstorms’ describes Ocean, well south of the subtropical ridge. Satellite imagery our hypothesis about the pulse nature of the convection. (not shown) during the morning of the flash flood event indicated relatively cloud-free conditions over southeastern Rainfall and streamflow On 26 February 2003, several thunderstorms developed Australia. The weak surface pressure gradient was supported during the afternoon over the Great Dividing Range to the by observations from automatic weather stations in the northeast of Melbourne. The thunderstorm responsible for region, indicating light winds prevailed over eastern Victoria the Buckland Valley flash flood developed at the top of Dingo during the afternoon, with the direction largely dominated by Creek near Mt Selwyn between 0530 and 0630 UTC. In less topography in the form of anabatic winds. than an hour, several kilometres downstream, the Buckland The mean sea-level pressure analysis six hours later at 1200 River at Harris Lane streamflow gauge (Fig. 1(b)) rose 1.8 UTC 26 February (Fig. 2(c)) shows a similar synoptic situation, metres, with flow rates increasing from 14 megalitres per however by this time the thunderstorm activity over Buckland day to over 6000 megalitres per day, subsequently reaching River Catchment had ceased. At 0000 UTC 28 February 2003, minor flood levels according to the Victorian Flood Class 36 hours after the event (Fig. 2(d)), the blocking high had Levels. It was around this time that a firefighter’s life was lost moved further east, while a low pressure trough had crossed in an attempt to cross a bridge over Dingo Creek. In addition, the State to be positioned over eastern Victoria. large quantities of debris (mud, fallen trees, boulders) were Several reanalysis fields of atmospheric parameters swept into Dingo Creek, and then into the Buckland and were obtained from the National Centre for Environmental Ovens Rivers, leading to water quality problems for several Prediction Reanalysis 2 global data set (hereafter denoted large towns downstream in the following weeks. as NCEP-R2) to gain an appreciation of the synoptic-scale There are no rain gauges within close proximity to environment in which the thunderstorm cells developed. Dingo Creek; consequently it was not possible to obtain Upper winds were correspondingly light (Fig. 3), suggesting an accurate measurement of precipitation that fell in the northwesterly winds of 5–10 knots in the vicinity of Mt Selwyn. catchment during the afternoon of 26 February 2003. The The combination of light surface winds and light upper winds closest Automated Weather Station (AWS) to Dingo Creek indicate weak vertical shear between the surface and 600 hPa, (within a 40 km radius) that provided a continuous record suggesting that any deep convection would be buoyancy of rainfall during the afternoon was Mount Hotham Airport, dominated with very little horizontal displacement of the which registered 37.0 mm in less than an hour between 0430 column with height. Upper-level temperature gradients over and 0530 UTC from a separate thunderstorm cell (and 37.2 Victoria were very weak. The NCEP-R2 temperature charts at mm in the 24 hour period to 2200 UTC 26 February 2003). 500hPa and 700hPa for 0000 UTC and 0600 UTC are shown in A rain event followed the flash flood approximately 36 Fig. 4((a)–(d)). The upper temperature analyses indicate little hours later. This study also includes the hydrologic modelling change between 0000 UTC and 0600 UTC in the vicinity of Mt of that event. A rain bearing cloud band associated with Selwyn. the passage of a cold front affected Victoria between 1200 34 Australian Meteorological and Oceanographic Journal 61:1 March 2011

Fig. 2 The mean sea level pressure analysis (hPa) over the time of the forecast. (a) 1200 UTC 23 February, (b) 0600 UTC 26 Febru- ary, (c) 1200 UTC 26 February, and (d) 0000 UTC 28 February.

Fig.3 NCEP Re-analysis 2 plot of 600hPa wind direction and Fig.4 NCEP Reanalysis 2 plot of: (a) 500hPa temperature speed at 0600 UTC 26 February 2003. Isotachs shown at 0000 UTC 26 February, (b) 500hPa temperature at 5 knot intervals. at 0600 UTC 26 February, (c) 700hPa temperature at 0000 UTC 26 February, (d) 700hPa temperature at 0600 UTC 26 February. Isotherms shown at 1 °C intervals. Gallucci et al.: Meteorological and hydrological mechanisms resulting in the 2003 post-fire flood event 35

Radar observations The closest Automatic Weather Station (AWS) to Mt Radar data are typically accessed to gather information Selwyn is Mt Hotham at a height of 1849 metres, directly 22 about a thunderstorm’s structure and intensity. At the time km to the east. At 0200 UTC, Mt Hotham AWS reported five of the flash flood event, the Melbourne Weather Radar was knots of wind from the west to northwest with a temperature a 1° beam width C-band radar with a 10-minute scanning of 17.1 °C and a dewpoint of 11.4 °C. Mt Hotham AWS frequency. The radar is located at Laverton, which is ceased reporting after 0200 UTC. Hotham Airport AWS, at approximately 210 km southwest from the thunderstorm cell 1296 metres, is a similar height to Mt Selwyn, and is located that developed near Mt Selwyn. Consequently, the distance 34 km to the east-southeast on the southern aspect of the between the radar and the thunderstorm makes it difficult to Great Dividing Range. At 0400 UTC, Hotham Airport AWS infer a great deal about thunderstorm structure. What can be reported three knots of wind from the southeast (typical inferred from radar imagery is that a high-intensity echo was anabatic wind direction during summer afternoons) with a first identified at 0530 UTC in the vicinity of Mt Selwyn. Two temperature of 22.7 °C and a dewpoint of 9.2 °C. Falls Creek further higher-intensity echoes in excess of 50 dBZ to a height AWS is at 1771 metres, and is located 38 km to the east of Mt of at least seven kilometres were evident on subsequent radar Selwyn. At 0300 UTC, Falls Creek AWS reported no wind scans prior to the cell dissipating at 0630 UTC (Yeo 2003). with a temperature of 18.8 °C and a dewpoint of 11.6 °C. An interpolated range height indicator scan (vertical cross The temperature sounding from Melbourne Airport, section) of the thunderstorm cell near Mt Selwyn at 0610 UTC valid for 0000 UTC 26 February 2003 is illustrated in Fig. on 26 February 2006 is illustrated in Fig. 5. 6. Melbourne Airport is located less than 200 km to the southwest of Mt Selwyn. The sounding shows a strong Modified soundings low-level inversion near 1000 metres and a higher, weaker To assist in assessing the potential for a flash-flood- inversion near 4000 metres. The sounding also indicates producing thunderstorm the temperature sounding from uniform moisture between the surface and 1200 metres, Melbourne Airport was modified to be representative of the with dewpoint values of 12–15 °C in that layer. Several days environment in which the thunderstorm near Mt Selwyn of mostly east to northeasterly flow in the boundary layer developed. A representative temperature sounding enables brought warm, moist, subtropical air from the north, which the calculation of thunderstorm diagnostics such as the is likely to have contributed to the presence of moisture in Convective Available Potential Energy (CAPE). An analysis the lowest levels of the atmosphere. of pre-thunderstorm surface and upper conditions was The sounding was modified to reflect an elevated undertaken to determine the most appropriate modifications, environment at 1400 metres in the vicinity of Mt Selwyn such that a sounding could be constructed that best during the afternoon when thunderstorm initiation took represented the vertical profile of the atmosphere in which place (Fig. 7) . Based on observations from nearby automatic the thunderstorm near Mt Selwyn developed. Observations weather stations a surface air parcel at 1400 metres was from three elevated AWSs close to Mt Selwyn were used assumed to have an air temperature of 23 °C and a dewpoint to modify the Melbourne Airport temperature sounding to of 12 °C. The Melbourne Airport temperature profile above reflect an elevated environment around 1400 metres. This the boundary layer was unmodified and used to approximate approach is similar to that undertaken by Schwartz et al. the upper temperature structure in the vicinity of Mt Selwyn. (1990) who analysed the Minneapolis flash flood that took The decision not to modify the upper temperature profile place on 23 July 1987. was based on the 700 hPa and 500 hPa temperature analyses in Fig. 4, which suggest that there was very little difference Fig.5 Interpolated RHI scan from Melbourne radar of the in the temperature (at these levels) between Melbourne and thunderstorm cell near Mt Selwyn at 1710 hrs (0610 Mt Selwyn, and also little change at these locations between UTC) 26 February 2003 (Bureau of Meteorology, as 0000 UTC and 0600 UTC. adapted in Yeo, 2003). Rainfall intensity is directly related to the rate at which a thunderstorm can process water vapour (Doswell et al. 1996). Hence, for a given amount of moisture ingested by a thunderstorm, the water vapour mass flux increases with stronger updrafts. The updraft within a thunderstorm can be calculated from the amount of CAPE, which from the sounding constructed for Mt Selwyn is approximately 1600 J kg–1. This value of CAPE is well above the threshold of 1000 J kg–1 that is referred to by meteorologists in the Victorian Regional Forecast Office as a threshold indicating the environment has the potential to support thunderstorms capable of producing flash flooding. The maximum potential updraft due to buoyancy alone is calculated as 56 m s–1, a vertical velocity that is capable of processing a significant 36 Australian Meteorological and Oceanographic Journal 61:1 March 2011

Fig.6 Aerological diagram from Melbourne Airport at 0000 Fig.7 Constructed aerological diagram for the vicinity of Mt UTC 26 February 2003. Selwyn for 0530 UTC 26 February 2003. The shaded area represents positive buoyancy or CAPE above the lifted condensation level.

amount of water vapour (Doswell 1993). It is also pertinent in the atmosphere it is likely that the thunderstorm that to note that the constructed sounding for Mt Selwyn is void developed near Mt Selwyn was a pulse wet thunderstorm. of any convective inhibition creating an extremely unstable Radar imagery was used to identify three short-lived environment. high-intensity elevated echoes in the same location over a 60-minute period. An interpolated RHI radar scan of the cell Pulse thunderstorms at 0610 UTC on 26 February 2006 indicates a high-intensity The amount of rain produced by a thunderstorm is elevated echo with negligible horizontal displacement with dependent on numerous interrelated atmospheric processes height (Fig. 5). Pulse thunderstorms tend to be short-lived that occur from the micro to the global scale. Precipitable due to the precipitation core cutting off its low-level inflow; water is a measure of the amount of moisture in the however in high humidity environments the low-level atmosphere and has been found in some studies to have outflow can be weak enough to allow new cells to develop a correlation with rainfall amount (Brenner 2004) and also close to existing cells. In the case of the thunderstorm flash flooding associated with thunderstorms (Doswell near Mt Selwyn, the weak upper-level northwesterly et al. 1996; Schwartz et al. 1990; Pontrelli et al. 1999). It is wind directed the collapsing cells toward the southeast, constructive to think of precipitable water as the amount of leaving Dingo Creek Valley (to the north of the developing rain (in millimetres) reaching the surface if all the moisture convection) free to receive further insolation supporting in the atmosphere was precipitated as rain. A stationary the anabatic wind. It is possible that the northerly anabatic thunderstorm has the capacity to convert available wind acted as a mechanical lifting mechanism, supporting moisture in the atmosphere into rain reaching the surface, ‘back building’ (or new cell development on the upwind although it is recognised that factors such as evaporation, moisture convergence and amount of instability also play Fig.8 NCEP Reanalysis 2 plot of precipitable water at 0600 UTC 26 February 2003. Precipitable water shown in an important role. The precipitable water value from the 2mm intervals. 0000 UTC sounding at Melbourne Airport is calculated to be 27 mm, which is similar to the NCEP-R2 value at 0600 UTC (Fig. 8). A value of 27 mm for precipitable water is not climatologically remarkable (February average precipitable water is approximately 20 mm for Melbourne), although it is conceivable that a near-stationary thunderstorm with high precipitation efficiency (in a weakly sheared environment) could deposit close to, and potentially more than, the precipitable water value in a saturated environment. Doswell (1993) contends that a precipitation rate of 25 mm h–1 is considered marginally heavy and capable of producing flash flooding. The magnitude of the flooding in the Buckland River valley suggests considerably more than 25 mm fell in an hour or less. Given the weakly sheared environment and instability Gallucci et al.: Meteorological and hydrological mechanisms resulting in the 2003 post-fire flood event 37

side) of the thunderstorm on its northern flank. This process flash-flood forecasting. Instead, the model selection was potentially explains why more than one high intensity echo based upon the need to find a simple approach to providing was observed on radar during the thunderstorm’s lifecycle. decision-making support and enhancing local knowledge The interaction between thunderstorm advection and in the community. The HBV-96 model, created by the propagation leading to a quasi-stationary thunderstorm, Swedish Meteorological and Hydrological Institute (SMHI), such as the one near Mt Selwyn on 26 February 2003, is is classified as a conceptual semi-distributed model and described by Doswell et al. (1996). requires input climatological (temperature, precipitation Radar imagery suggests that three high-intensity pulses and potential evapotranspiration) and streamflow data for were observed over the same area near Mt Selwyn within calibration (Lindström et al. 1997). Catchments are divided a 60-minute period. Theoretically, if the thunderstorm was into sub-basins as single hydrological units, and each such capable of efficiently converting the available precipitable basin is divided into zones according to topography and land water into rainfall then the combination of the three high- use/vegetation. The HBV-96 model contains functionality intensity pulses, due to a quasi-stationary thunderstorm that for soil moisture accounting, snowmelt and accumulation developed as a result of the process identified in the previous and runoff response (generation and routing). Detailed paragraph, could have combined to deliver in excess of 50 descriptions and diagrams of the model can be found in mm and perhaps as much as 80 mm. SMHI (undated) and Lindström et al. (1997). The most relevant model processes to this study are the soil moisture accounting and runoff response subroutines. Data and model descriptions Within the soil moisture accounting procedure, precipitation Mesoscale Model Version 5 that reaches the surface is either evaporated or infiltrated The analysis of the meteorological conditions associated into the soil moisture storage. The soil moisture storage is with the event was performed using the Pennsylvania controlled by the maximum possible soil moisture storage State University/National Center for Atmospheric Research (field capacity), the limit for potential evapotranspiration (PSU/NCAR) Mesoscale Model Version 5 (MM5). MM5 is a and an empirically specified coefficient that controls the limited-area, non-hydrostatic model capable of simulating amount of precipitation or snowmelt that is partitioned into meso- and synoptic-scale atmospheric circulations (Grell et runoff. Once the maximum possible soil moisture storage is al. 1994). The model has been used in many other studies reached, any additional precipitation becomes excess soil of precipitation events in mountainous areas (Colle and moisture (also termed effective precipitation), and is routed Mass 2000; Rotunno and Ferretti 2001; Lin and Chen to the runoff response function. 2002; Das 2005; Galewsky and Sobel 2005; Li et al. 2005; The runoff response takes this effective precipitation Monaghan et al. 2005) and therefore is appropriate for this and initially stores it in the upper soil column (upper type of application. The initial and boundary conditions for response box). This water then percolates over time into the model were created using the NCEP-R2 data at a grid the groundwater reservoir (lower response box) according spacing of 2.5° latitude by 2.5° longitude. In order to achieve to an empirically determined percolation rate. If the upper sufficient horizontal resolution, a series of nested domains soil column becomes saturated, the excess water from were configured. Domain 1 (Fig. 1(a)) is centred at 36.7°S this reservoir becomes stream discharge, the quickflow latitude and 147.8°E longitude and has an extent of 1800 km x component of the hydrograph. The groundwater contributes 1550 km, with a grid resolution of 50 km and 23 vertical levels. to the baseflow component of the hydrograph. Domains 2 and 3 (Fig. 1(a) and (b)) have resolutions of 10 km Input climatological data was sourced from the Bureau of and 2 km respectively with 35 vertical levels. The simulations Meteorology, streamflow data from Theiss Services Pty. Ltd., were all one-way nested. The topography for all domains and topography data from the Department of Sustainability was derived from U.S. Geological Survey terrain data at and Environment. The simulated catchment encompasses resolutions of 55 km, 9.25 km, and 3.70 km. The simulations the entire Buckland River Catchment (including Dingo Creek) all employed the NOAH (National Center for Environmental in addition to the section of the Ovens River (Wangaratta) Prediction, Oregon State University, Air Force, Hydrologic Catchment, which lies between the Harris Lane streamflow Research Lab) land surface model (Chen and Dudhia, 2001a, gauge (the only streamflow gauge on the Buckland River b), the Schultz moisture scheme (Schultz 1995), the Blackadar with a long-term data-set) and the northern border (directly planetary boundary layer scheme (Zhang and Anthes, 1982), downstream) of the Buckland River Catchment (Fig. 1(b)). and the Grell cumulus parameterization scheme (Grell, 1993), The studied catchment has an area of approximately 460 km2, with the Buckland River Catchment approximately 320 2 The hydrological model km in size, the catchment immediately surrounding Dingo 2 The hydrological response of the Buckland River Catchment Creek approximately 41 km and the portion of the Ovens 2 to fire-induced surface modifications was analysed using the River (Wangaratta) Catchment approximately 140 km . Integrated Hydrological Modelling System (IHMS) version 4.5 (hereby known as HBV-96 model). This model was not selected on the basis that it could be used operationally for 38 Australian Meteorological and Oceanographic Journal 61:1 March 2011

Methodology between simulated precipitation at Mount Hotham Airport and the maximum simulated precipitation total, a value can There are two parts to the methodology. First, the experiments be postulated to approximate the decrease in precipitation using the mesoscale model are described. Second, details of from the most intense part of the storm and Mount Hotham the hydrological modelling simulations are provided. Airport. By adding this difference to the actual observation at Mount Hotham Airport, the likely maximum precipitation MM5 methods total for the thunderstorm can be estimated. The MM5 The methodology used to simulate the thunderstorms is precipitation ensemble maximum was, as noted, 44.4 mm, outlined in Tryhorn et al. (2008) but relevant details will be with 9.8 mm simulated at Mount Hotham Airport. This gives given here. From the sample space of initial conditions a a difference of 34.6 mm, which when added to the actual five member single-model time-lagged ensemble of five-day observation at Mount Hotham Airport (37.2 mm) gives a forecasts was performed from 1200 UTC 23 February – 1200 maximum precipitation estimate of 71.8 mm. Hence, 71.8 UTC 28 February 2003 using MM5. The model dynamics, mm was used as the input rainfall data for the HBV-96 model. physical parameterizations, and numerics were all the same. Evapotranspiration was calculated using the Thornthwaite Differences among ensemble members resulted from the method (Thornthwaite 1948). More detail on this can be differences in initialization time. The results presented in found in Gallucci (2007). the results section are the ensemble mean. The flash flood Analysis of the hydrological response of the Buckland occurred at around 0600 UTC on 26 February 2003. The land River catchment surrounding the flash flood event required surface in MM5 for the area burned by the 2003 fire was two sets of simulations. The first set of simulations used changed to resemble a recently burned fire surface, with parameters obtain by calibrating the HBV-96 model over long decreased albedo (to resemble a blackened surface), lowered timescales using observational data. This set of parameters surface roughness (reduced vegetation) and reduced soil is described as the climatological (or unburned) parameters. moisture. Based on the interviews with shire residents and Daily streamflow data at the Harris Lane streamflow gauge photos during and after the event, it was apparent that the on the Buckland River is available for the period 1972– fire was of great intensity. The chosen values for the relevant 2005. The HBV-96 model was calibrated over a 28-year model parameters reflect this observation and were based period (1972–1999). The remaining five years of streamflow on previous work on modelling fire scars (Görgen et al. 2006; data (2000–2005) were used for model verification. The Wendt et al. 2007.) The albedo was reduced from 0.20 to HBV-96 model was calibrated using Nash and Sutcliffe’s 0.08, the roughness length from 2.65 m to 0.10 m, and the (1970) efficiency criterion and sensitivity testing of the soil moisture for the uppermost 10 cm layer was initialised climatological parameter set was also completed. This at 0.05 m3/m3, in contrast to standard values of between 0.25 testing was undertaken to ensure that the climatological and 0.45 m3/m3. The ensemble was then compared with parameter set was (1) capable of simulating typical (or a simulation covering the same time period in which no average) conditions in the Buckland River catchment, and surface modifications were made (known as the unburned (2) responsive to extreme precipitation events. Using the simulation). calibration parameters, the model was found to be capable of simulating typical conditions and responsive to extreme HBV-96 methods precipitation events (not shown). The HBV-96 model was run over the period encompassing In the second set of simulations, parameter values that the flash-flooding event (1300 UTC 24 February 2003 – 1300 represent typical post-fire conditions were used (Scott and UTC 4 March 2003) using observed daily mean temperature Van Wyk 1990; Robichaud and Hungerford 2000; Huffman et data, estimated values of mean monthly potential al. 2001). Soil hydrophobicity was represented by reducing evapotranspiration, and a combination of estimated and the maximum possible soil moisture storage to a minimum observed daily precipitation totals. Precipitation estimates (10 mm) and reducing the percolation rate between the were created using the inverse distance squared method upper and lower response boxes to zero. These parameters (Thompson, 1999). As noted in ‘The observed event’ section, were increased to standard (climatological) values post- the radar imagery indicates that the highest intensity of the flood in order to test whether the hydrophobic conditions storm cell was located over the Buckland River catchment, were subsequently restored immediately after a flood event not over Mt Hotham Airport. However, accurate rainfall (as is suggested by DeBano 1975 and Robichaud 2000). The estimates cannot be made from the radar images due to maximum soil moisture storage was 10 mm before and the distance of the thunderstorms from the radar sites at during the flood event (24 February 2003 – 26 February Laverton and Yarrawonga, which are approximately 200 km 2003), then 250 mm post-flood (27 February 2003 onwards). and 130 km from the Buckland River catchment respectively. Percolation values were 0 mm/day before and during the In order to approximate a rainfall total for the head of the flood event, then 4.4 mm/day post-flood with a canopy catchment, output precipitation data from the MM5 over interception storage capacity of 0 mm. The timing of this the Buckland River catchment and the observations at change in parameter values was chosen to achieve the highest Mount Hotham Airport were used. By finding the difference correlation between observed and simulated streamflow. Gallucci et al.: Meteorological and hydrological mechanisms resulting in the 2003 post-fire flood event 39

The interception canopy storage capacity was also reduced Fig.9 Satellite picture of water vapour over southeastern to zero in order to represent the post-fire decrease in canopy Australia taken by Geostationary Meteorological Satellite 5 (GMS-5) on Channel 3 (6.5–7.0 µm) using interception. Post-fire changes in evapotranspiration a Visible Infrared Spin Scan Radiometer (VISSR) At- and surface roughness were not accounted for in these mospheric Sounder (VAS), 0530 UTC 26 February. simulations. For the former, while it is recognised that during a severe fire vegetation will likely be destroyed, leading to a decrease in evapotranspiration, the amount by which this value decreases has not been adequately quantified in field observation. Regarding the latter, there isno parameter in this model that represents surface roughness, and hence the effects of this quantity could not be tested. Because of uncertainties in the appropriate values of model parameters, which generally do not have strictly observable counterparts, the post-fire conditions described here were adjusted so that the simulated hydrograph represented that of the observed in terms of the peaks, trough, and timing. To explore the possible impact of each parameter on post-fire flooding, sensitivity simulations were completed in which the parameters were returned, in turn, to their unburned (climatological) values.

Results under-prediction of precipitation. Spatially the rain during this second event was highly variable, both in the model and MM5 simulations the available observations (see Fig. 6 in Tryhorn et al. 2008). The forecast initialisation at 1200 UTC on 23 February (not The simulated CAPE during the storm was calculated to shown) includes a high pressure system similar to the be 1546 J/kg, indicating an unstable environment capable analysis, centred over the Tasman Sea, and this remained of producing deep convection and agreeing well with the in place for much of the forecast until a cold front swept value estimated in ‘The observed event’ section. The surface across the state, as simulated for 0000 UTC 28 February. At energy balance for the control ensemble mean (taken from the time of the flood (0600 UTC 26 February), there was no an average of nine grid boxes, not shown) indicated a large cloud or precipitation simulated in Domain 1 in the area of sensible heat flux, caused primarily by the reduction in soil the Alpine Shire. The large-scale water vapour pattern in the moisture (Tryhorn et al. 2008). This led to increased lower model does not simulate the broader scale rain processes in boundary layer heating which destabilised the atmospheric northeast Victoria that are present in the satellite images (Fig. column. This in turn led to an increase in convective clouds 9). Overall the evolution of synoptic-scale events was well and an increase in precipitation compared to the unburned simulated, apart from the (crucial) absence of thunderstorm simulation (not shown). The unburned simulation produced activity in northeast Victoria, (seen in the ‘smaller scale’ little rain at the time of the flood event (3.9 mm)—this was moisture plumes that identify the thunderstorms (Fig. 9)). A well outside the range of variation of the control ensemble, more detailed analysis of the synoptic-scale events can be hence we consider the response to be significant. In this case, found in Tryhorn et al. (2008). we suggest that the extreme rainfall event was enhanced by In contrast, the high-resolution Domain 3 simulations the fire-induced modifications. However, we recognise that generate intense convective rainfall in the correct location. many different adjustments to the model may be considered The Domain 3 simulation in the 24 hours before 2200 UTC acceptable and the uniqueness of a particular event may 26 February, the time period associated with the flood, not be unequivocally represented by a particular model produced an ensemble mean of 31.7 mm and an ensemble configuration. Interpretation of the values of the individual maximum of 44.4 mm. This is in good agreement with the variables must therefore be made carefully. closest observed rainfall of 37.2 mm, but as noted previously, this is likely to be an underestimate. The 6-hour totals of HBV-96 simulations rainfall (0000–0600 and 0600–1200 UTC 26 February) in When the parameters determined from the climatological Domain 3 at the time of flood reveals significant falls across calibration were used, the HBV-96 simulation yielded an the southern and southwestern areas of the domain. This underestimation of streamflow at the Harris Lane gauge on difference from the large-scale simulation suggests a highly 26 February and an overestimation of streamflow when the localised event. Following the thunderstorm event, the front passed through on 28 February 2003 (Fig. 10). Without simulated precipitation associated with the passage of the alterations to the climatological parameter values, the flash front agreed well with observations, although the simulated flood event could not be replicated. front was delayed by several hours, causing an apparent 40 Australian Meteorological and Oceanographic Journal 61:1 March 2011

When the HBV-96 model was run using the post-fire incoming water begins to infiltrate through the hydrophobic parameter values the simulated and observed hydrographs layer, alleviating hydrophobic conditions, and pre-fire were almost identical in shape and magnitude (Fig. 11) and infiltration conditions return. In this case, if the soil moisture an efficiency criterion value of 0.96 was achieved. Using storage in the Buckland River catchment did not return to this configuration, the flash-flooding event was successfully pre-fire conditions after the flood, it is likely that a second simulated using the HBV-96 model. flood would have occurred on 28 February 2003 when the Changing the value of maximum soil moisture storage next rain event took place (Fig. 12). Adjusting the percolation had the most significant effect on the efficiency criterion. value yielded a similar but smaller impact on the hydrograph The return of this quantity to pre-fire values post-flood response, supporting the hypothesis that during flooding support the results of DeBano (1975) and Robichaud (2000), the hydrophobic layer ceases to obstruct infiltration, and who found that following the onset of precipitation, the hence post-flood, percolation to the lower response box

Fig.10 Results of the simulation for the period 0000 AEDT 25 February 2003 to 0000 AEDT 5 March 2003 using the parameters used in long-term calibration and the approximated precipitation value of 71.8 mm. The computed streamflow (m3/s) is shown as the dashed line, and the observed streamflow (m3/s) is shown as the solid line.

Fig.11 Results of the simulation for the period 0000 AEDT 25 February 2003 to 0000 AEDT 5 March 2003 using the optimal parameters. The computed streamflow (m3/s) is shown as the dashed line, and the observed streamflow (m3/s) is shown as the solid line.

Fig.12 Results of the simulation for the period 0000 AEDT 25 February 2003 to 0000 AEDT 5 March 2003 using the optimal parameters and FC=10 mm for the entire period. The computed streamflow (m3/s) is shown as the dashed line, and the observed streamflow (m3/s) is shown as the solid line. Gallucci et al.: Meteorological and hydrological mechanisms resulting in the 2003 post-fire flood event 41

(groundwater) is re-established. in the addition of Flood Watches to Bureau of Meteorology Lastly, canopy interception was found to have an forecasts for recently fire-affected areas in the Victorian insignificant effect on the short-term simulation. This was Alps. Alteration of land surface properties for recently not surprising for two reasons. First, long-term individual burned areas within forecast models could potentially parameter sensitivity testing indicated that streamflow is improve forecasting of these events. In this context, the relatively insensitive to canopy interception. Second, during study indicates that the parameters that should be altered in an extreme rainfall event, a (typical) 1 mm canopy storage atmospheric and hydrological models are those relating to capacity would be negligible. albedo, soil moisture, and soil infiltration rates. Community awareness of the risks to recently burned areas from flooding and the level of preparedness of local government Conclusions and emergency managers will also have an important role in An analysis of the meteorological and hydrological factors decreasing the vulnerability of communities to these events. leading to the post-fire flood event in the Buckland Valley, Victoria, Australia in February 2003 has been conducted. Acknowledgments It has been demonstrated that flash flooding occurred because of highly localised thunderstorms and was likely to This work would not have been possible without the have been enhanced by the burned landscape. In addition, participation, support and interest of the people of Alpine our analysis of the resulting situation suggests that other Shire. This work has been supported by the Australian contributing factors included the storm cells likely being Research Council though FF0348550, by Monash University pulse wet microburst, cell regeneration over the same area, through the postgraduate scholarship program, and by the and the steepness of the Buckland River catchment. CSIRO Division of Marine and Atmospheric Research. We The synoptic conditions surrounding the event suggest would also like to thank Rebecca Abramson, Klaus Görgen that the major drivers of the rainfall event were the high and Petteri Uotila for helpful comments and assistance. Also, levels of precipitable water in the atmosphere, enhanced two anonymous reviewers’ and Dr Graham Mills’s valuable surface heating due to reduced surface albedo and soil suggestions helped to improve the manuscript. moisture, and strengthened atmospheric instability from surface heating. It is possible that deep convection and heavy rainfall would have occurred regardless of the burned References landscape. However, our results show that the modified Barkley, C.M. and Othmer Jr. E.F. 2004. Response to post-fire flood threat: surface conditions increased the degree of instability and California 2003. Southwest Hydrol., 3, 17. provided a greater chance that convective inhibition could Beeson, P.C. Martens, S.N. and Breshears, D.D. 2001. Simulating overland flow following wildfire: mapping vulnerability and landscape distur- be overcome. Hence, our modelling study suggests that the bance. Hydrol. Process., 15, 2917-30. intensity of the event was enhanced by the preconditioning Brenner, I. 2004. The relationship between meteorological parameters caused by the fire. and daily summer rainfall amount and coverage in west-central Flori- da. Weather and Forecasting, 19, 286-300. The results of the hydrological modelling suggest that Cannon, S., Bigio, E. and Mine, E. 2001a. A process for fire-related debris the fire-induced soil hydrophobicity further contributed to flow initiation, Cerro Grande fire, New Mexico. Hydrol. Processes, 15, the flash-flooding event. It should be recognised that there 3011–23. are many different parameter sets that are, in some sense, Cannon, S., Kirkham, R. and Parise, M. 2001b. Wildfire-related debris flow initiation processes, Storm King Mountain, Colorado. Geomorphol- acceptable in describing and predicting the responses ogy, 39, 171–88. of the system. This is an issue inherent to all hydrological Chen, F. and Dudhia, J. 2001a. Coupling an advanced land-surface/hy- modelling. Nevertheless, hydrological models have drology model with the Penn State/NCAR MM5 modeling system. proven to be useful tools in simulating environments that Part I: Model implementation and sensitivity. Mon. Weath. Rev., 129, 569-85. cannot be measured, elucidating physical mechanisms, Chen, F. and Dudhia, J. 2001b. Coupling an advanced land-surface/hy- and in operational hydrological forecasting. The different drology model with the Penn State/NCAR MM5 modeling system. parameter sets produce similar results in different ways, Part II: Preliminary model validation. Mon. Weath. Rev., 129, 587-604. thus, there is the opportunity to propose testable hypotheses Chen, F., Warner, T. and Manning, K. 2001c. Sensitivity of orographic moist convection to landscape variability: A study of the Buffalo based on process implications. Creek, Colorado, flash flood case of 1996. J. Atmos. Sci., 58, 3204–23. Overall, these findings point to an increased risk of flash Colle, B.A. and Mass, C.F. 2000. The 5-9 February 1996 flooding event over flooding after a severe fire. With an increase in fires expected the Pacific Northwest: sensitivity studies and evaluation of the MM5 precipitation forecasts. Mon. Weath. Rev, 128, 593-617. in the Alpine Shire associated with anthropogenic climate Conedera, M., Marxer, P., Ambrosetti, P., Della Bruna, G. and Spinedi, F. change (Hennessy et al. 2005; Williams et al. 2001), this causal 1998. The 1997 forest fire season in Switzerland. International Forest relationship, even for a rare event, has implications for Fire News, 18, 85-8. emergency managers and residents in alpine regions. This Conedera, M., Peter, L., Marxer, P., Forster, F., Rickenmann, D. and Re, L. 2003. Consequences of forest fires on the hydrogeological response has become particularly relevant after a recent event in Licola of mountain catchments. A case study of the Riale Buffaga, Ticino, where, following bushfires, residents were flooded with Switzerland. Earth Surf. Process. Landf., 28, 117-29. mud, ash, and debris (Houghton 2007). This event resulted 42 Australian Meteorological and Oceanographic Journal 61:1 March 2011

Das, S. 2005. Mountain weather forecasting using MM5 modelling sys- Martin, D. and Moody, J. 2001. Comparison of soil infiltration rates in tem. Curr. Sci., 88, 899-905. burned and unburned mountainous watersheds. Hydrol. Process., 15, DeBano, L.F. 1975. Infiltration, evaporation and water movement as relat- 2893-903. ed to water repellency. Proceedings, Experimental Methods and Uses Marxer, P., Conedera, M. and Schaub, D. 1998. Postfire runoff and soil of Soil Conditioners 15-16 November, 1973, Las Vegas, NV. Soil Science erosion in the sweet chestnut belt of southern Switzerland. Fire Man- Society of America Special Publication Series 7, Madison, WI, USA, agement and Landscape Ecology. L. Trabaud, Ed., International As- pp. 1535-163. sociation of Wildland Fire, 51-62. Doswell, C.A.III. 1993. Flash flood producing convective storms: current Monaghan, A.J., Bromwich, D.H., Powers, J.G. and Manning, K.W. 2005. understanding and research, Proceedings, Spain-U.S. Joint Workshop The climate of the McMurdo, Antarctica, region as represented by one on Natural Hazards 8-11 June 1993, Barcelona, Spain. year of forecasts from the Antarctic Mesoscale Prediction System. Jnl Doswell, C.A. III., Brooks H.E. and Maddox, R.A. 1996. Flash flood fore- climate, 18, 1174-89. casting: An ingredients-based methodology. Weath. Forecasting, 11, Nash, J.E. and Sutcliffe, J.V. 1970. River flow forecasting through concep- 560–81. tual models. Part I – a discussion of principles. J. Hydrol., 10, 282-90. Elliot, J. and Parker, R. 2001. Developing a post-fire flood chronology and Pontrelli, M.D., Bryan, G. and Fritsch, J.M. 1999. The Madison County, recurrence probability from alluvial stratigraphy in the Buffalo Creek Virginia, flash flood of 27 June 1995. Weath. and Forecasting, 14, 384- watershed, Colorado, USA. Hydrol.Processes, 15, 3039–51. 404. Galewsky, J. and Sobel, A. 2005. Moist dynamics and orographic precipi- Robichaud, P.R. and Brown, R. 1999. What happened after the smoke tation in northern and central California during the New Year’s flood cleared: onsite erosion rates after a wildfire in Eastern Oregon. Wild- of 1997. Mon. Weath. Rev., 133, 1594-1612. land Hydrology. D. Olsen and J. Potyondy, Eds., American Water Re- Gallucci, J. 2007. Post-fire flash flooding: using a modelling methodology to sources Association, 419-26. explore the causal link between flash flooding and fire. Honours thesis, Robichaud, P.R. 2000. Fire effects on infiltration rates after prescribed fire Monash University, Clayton, Victoria, Australia. in Northern Rocky Mountain forests, USA. J. Hydrol, 231-2, 220-9. Giovannini, G. 1994. The effect of fire on soil quality. Soil Erosion as a Con- Robichaud, P.R. and Hungerford, R.D. 2000. Water repellency by labora- sequence of Forest Fires. M. Sala and J. Rubio, Eds., Geoforma Edi- tory burning of four Rocky Mountain forest soils. J. Hydrol., 231-2, ciones, 15-27. 207-19. Giovannini, G., Vallejo, R., Luchesi, S., Bautista, S., Ciompi, S. and Llovet, Rotunno, R. and Ferretti, R. 2001. Mechanisms of intense alpine rainfall. J. 2001. Effects of land use and eventual fire on soil erodibility in dry J. Atmos. Sci., 58, 1732-49. mediterranean conditions. Forest Ecology and Management, 147, 15- Schultz, P. 1995. An explicit cloud physics parameterization for opera- 23. tional numerical weather prediction. Mon. Weath. Rev., 123, 3331–43. Görgen, K., Lynch, A.H., Marshall, A. and Beringer, J. 2006. Impact of Schwartz, B., Chappell, C., Togstad, W. and Zhong, X. 1990. The Minneap- abrupt land cover changes by savanna fire on northern Australian cli- olis Flash Flood: Meteorological Analysis and Operational Response, mate. J. Geophys. Res., 111, D19106, doi.10.1029/2005JD006860. Weath. and Forecasting, 5, 3-21. Grell, G.A. 1993. Prognostic evaluation of assumptions used by cumulus Scott, D.F. 1997. The contrasting effects of wildfire and clearfelling on the parameterizations. Mon. Weath. Rev., 121, 764-787. hydrology of a small catchment. Hydrol. Process., 11, 543-55. Grell, G., Dudhia, J. and Stauffer, D. 1994. A description of the fifth-gen- Scott, D.F. and Van Wyk, D.B. 1990. The effects of wildfire on soil wettabil- eration Penn State/NCAR Mesoscale Model (MM5). NCAR Tech. Note. ity and hydrological behaviour of an afforested catchment. J. Hydrol., NCAR/TN-398+STR, 117 pp. 121, 239-56. Hennessy, K., Lucas, C., Nicholls, N., Suppiah, R. and Ricketts, J. 2005. Cli- Shakesby, R. and Doerr, S. 2006. Wildfire as a hydrological and geomor- mate change impacts on fire-weather in south-east Australia. CSIRO, phological agent. Earth Sci. Rev., 74, 269-307. Australia, 88 pp. SMHI IHMS manual undated. IHMS: Integrated Hydrological Modelling Houghton, J. Town caught between fire and flood, The Age, February System, Version 4.5. Swedish Meteorological and Hydrological Insti- 25th 2007. tute, Norrköping, Sweden. Huffman, E., MacDonald, L. and Stednick, J. 2001. Strength and persis- Thompson, S.A. 1999. Hydrology for water management. A. A. Blakema, tence of fire-induced soil hydrophobicity under ponderosa and lodge- Rotterdam, Netherlands. pole pine, Colorado Front Range. Hydrol. Process., 15, 2877-92. Thornthwaite, C.W. 1948. An approach toward a rational classification of Krammes, J.S. and DeBano, L.F. 1965. Soil wettability: a neglected factor climate. Geographical Review, 38, 55–94. in watershed management. Wat. Resour. Res., 1(2), 283-6. Tryhorn, L., Lynch, A., Abramson, R. and Parkyn, K. 2008. On the mecha- Letey, J. 2001. Causes and consequences of fire-induced soil water repel- nisms driving post-fire flash floods: a case study. Mon. Weath. Rev., lency. Hydrol. Process., 15, 2867-75. 136, 1779-91. Li, M.H., Yang, M.J., Soong, R. and Huang, H.L. 2005. Simulating typhoon Wendt, C.K., Beringer, J., Tapper, N.J. and Hutley, L.B. 2007. Local bound- floods with gauge data and mesoscale-modelled rainfall in a moun- ary layer development over burnt and unburnt tropical savanna: an tainous watershed. Journal of Hydrometeorology, 6(3), 306-23. observational study. Bound. Lay. Met., 124, 291-304. Lin, C.Y. and Chen, C.S. 2002. A study of orographic effects on mountain- Williams, A., Karoly, D.J. and Tapper, N. 2001. The sensitivity of Australian generated precipitation systems under weak synoptic forcing. Meteo- fire to climate change. Clim. Change, 49, 171-91. rol. Atmos. Phys., 81, 1-25. Yeo, C. 2003. Severe Thunderstorm Report. Alpine Fires Flash Flood 26 Lindström, G., Gardelin, M., Johansson, B., Persson, M. and Bergström, February 2003. Bur. Met., Australia, pp. 12. S. 1997. Development and test of the distributed HBV hydrological Zhang, D. and Anthes, R.A. 1982. A high-resolution model of the plan- model. J. Hydrol., 201, 272-88. etary boundary layer–sensitivity tests and comparisons with SESA- Liu, S., Leslie, L.M., Speer, M., Bunker, R. and Mo, X. 2004. The effects of ME-79 data. J. Appl. Meteor., 21, 1594-1609. bushfires on hydrological processes using a paired catchment analysis. Meteorol. Atmos. Phys., 86, 31-44.