authors copy

AUTHORS COPY

This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and education use, including for instruction at the authors institution and sharing with colleagues. Other uses, including reproduction and distribution, or selling or licensing copies, or posting to personal, institutional or third party websites are prohibited. In most cases authors are permitted to post their version of the article (e.g. in Word or Tex form) to their personal website or institutional repository. Authors requiring further information regarding Elsevier’s archiving and manuscript policies are encouraged to visit: http://www.elsevier.com/copyright

Author's personal copy Forest Ecology and Management 261 (2011) 570–581

AUTHORS COPY Contents lists available at ScienceDirect

Forest Ecology and Management journal homepage: www.elsevier.com/locate/foreco

Wildfire and salvage harvesting effects on runoff generation and sediment exports from radiata pine and eucalypt forest catchments, south-eastern Australia Hugh G. Smith a,b,∗ , Gary J. Sheridan a,b , Patrick N.J. Lane a,b , Leon J. Bren a,b a b

Department of Forest and Ecosystem Science, The University of Melbourne, Victoria 3010, Australia Cooperative Research Centre for Forestry, Sandy Bay, Tasmania 7005, Australia

a r t i c l e

i n f o

Article history: Received 27 August 2010 Received in revised form 5 November 2010 Accepted 10 November 2010 Available online 8 December 2010 Keywords: Wildfire Salvage harvesting Pinus radiata Eucalypt forest Runoff Erosion Sediment export

a b s t r a c t This study examined the effect of wildfire and salvage harvesting on runoff generation and sediment exports from three small forest catchments in south-eastern Australia. In 2006, wildfire burnt a radiata pine catchment and two adjacent natural eucalypt forest catchments which formed part of a long-term hydrological research project. Subsequently, only the pine plantation catchment was salvage harvested. The combined effect of fire and salvage harvesting in the pine catchment caused a substantial increase in runoff compared to the burnt eucalypt forest catchments and pre-fire conditions, particularly in response to high intensity, short duration summer storms. Post-fire maximum suspended sediment concentrations from fixed-interval sampling greatly exceeded pre-fire values for both eucalypt and pine catchments, while sediment (suspended and bedload) exported from the pine catchment exceeded each of the eucalypt catchments by a minimum of 180 and 33 times. However, the export increase was probably closer to 320 and 71 times based on a survey of eroded channels in the pine catchment combined with measured post-survey exports. Notably, seven summer storm events accounted for approximately 80% of the pine catchment sediment yield. Hillslope process measurements indicated that the highest runoff velocities occurred in log drag-lines formed by cable harvesting, while soil water repellency was more extensive in the harvested pine catchment than in the adjacent eucalypt catchment. The latter effect probably resulted from higher burn severity in the pines combined with reduced soil moisture due to less shading after harvesting. Runoff modelling indicated that the log drag-lines acted as an extension to the drainage network and increased peak flows at the harvested catchment outlet by 48% for a high intensity summer storm event, while substantial reductions in modelled runoff were achieved through increasing the hillslope surface roughness coefficient. It is recommended that post-fire salvage operations should avoid the formation of log drag-lines when using cable harvest techniques and maximise surface cover to limit increases to runoff, erosion and catchment sediment exports. © 2010 Elsevier B.V. All rights reserved.

1. Introduction Wildfires are an important disturbance in forest ecosystems that can cause substantial hydrological and geomorphological change (Shakesby & Doerr, 2006). Following fire, salvage harvesting may be undertaken to recover burnt timber resources. However, post-fire salvage harvesting generally occurs at a time when the forest landscape is most vulnerable to increased runoff, erosion and impacts on downstream water quality (e.g. Kunze & Stednick, 2006; Lane et al., 2006; Reneau et al., 2007; Sheridan et al., 2007; Smith &

∗ Corresponding author at: Department of Forest and Ecosystem Science, The University of Melbourne, 221 Bouverie St., Parkville, Victoria 3010, Australia. Tel.: +61 3 83440676; fax: +61 3 93494218. E-mail address: [email protected] (H.G. Smith). 0378-1127/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.foreco.2010.11.009

Dragovich, 2008; Moody & Martin, 2009; Smith et al., in press). In addition, ecological consequences of salvage harvesting in natural forests may include the removal of critical habitat, impaired ecosystem recovery, and loss of soil organic matter and nutrients with implications for soil biota and plant growth (Karr et al., 2004; Lindenmayer et al., 2004). As a result, there has been considerable debate over the merits of post-fire salvage harvesting in terms of the economic benefit and environmental consequences (McIver & Starr, 2000; Beschta et al., 2003; Karr et al., 2004). Impacts on runoff and erosion have been identified as one consequence of post-fire salvage harvesting (Karr et al., 2004). However, very few studies have quantified this effect in excess of burning alone (Silins et al., 2009). To date, this limited research has found that impacts from fire and harvesting generally result in only minor increases in sediment exports compared to burnt sites that have not been harvested (Van Lear et al., 1985; Cornish & Binns, 1987;

Author's personal copy H.G. Smith et al. / Forest Ecology and Management 261 (2011) 570–581

571

AUTHORS COPY

Fig. 1. (a) Location of Cropper Creek in Victoria and (b) hillshade DEM based on LiDAR survey of the three catchments.

Silins et al., 2009). In contrast, initial observations made in the present study suggested salvage harvesting after fire can have substantial impacts on runoff and erosion, indicating that the previous research cannot be considered generally representative of post-fire salvage harvesting effects for all natural and plantation forest systems. There is a need for studies of post-fire salvage harvesting effects on runoff, erosion and catchment sediment exports across different fire-prone forest types which quantify both the magnitude of impacts and identify specific harvesting effects that may exacerbate the post-fire response. The present study examines post-fire salvage harvesting in a plantation forest of radiata pine (Pinus radiata) in south-eastern Australia. It is the only investigation of post-fire salvage harvesting impacts in wildfire-burnt plantations known to the authors. While salvage harvesting of plantation forests may present less concern about onsite habitat effects than native forests, the potential for reduction in soil productivity as well as downstream impacts on aquatic habitat and water quality are important forest management issues. Therefore, post-fire salvage harvesting in plantation forests requires an informed management approach that recognizes potential impacts and adopts mitigating strategies. Wildfires in 2003, 2006–2007 and 2009 burnt large areas of forest across south-eastern Australia, which in Victoria alone exceeded three million hectares (Victorian Department of Sustainability and Environment, 2009). This included substantial areas of state forest and plantation, totalling 1.4 million and 28,750 ha, respectively, with most of the plantation forest (both softwood and eucalypt) burnt in 2009 (VAFI, 2009). In response to these fire events, salvage harvesting operations have increased with 3500 ha of burnt native forest salvaged since 2003 in Victoria, and a further 1000 ha expected to be salvaged in 2010 (VicForests, 2010). Specific information on recent post-fire salvage harvesting in plantation forests is not available, although large-scale salvage operations commenced following the 2009 fires in an effort to recover as much burnt plantation timber as possible. This study is focused on the Cropper Creek research catchments, which are comprised of two natural eucalypt forest catchments and a third catchment planted with radiata pine. These catchments have been the subject of long-term hydrological monitoring which has yielded numerous publications on the hydrology of plantation and native forest catchments (Bren & Turner, 1979, 1980; Hopmans et al., 1987; Bren & Papworth, 1991; Bren & Hopmans,

2007; Hopmans & Bren, 2007). Following a wildfire in 2006, which completely burnt all three catchments, the pine catchment was salvage harvested. In response, a research program was initiated with the following objectives: (1) to quantify the effect of wildfire and salvage harvesting on runoff and sediment exports from the radiata pine catchment compared to pre-fire data for all three catchments and post-fire data for the eucalypt forest catchments and (2) to identify the hydrological processes contributing to any observed post-fire and harvesting changes in runoff and sediment exports. 2. Study area and project background The Cropper Creek research catchments are located in northeast Victoria, Australia (Fig. 1). Two of the catchments, Ella (113 ha) and Betsy (44.3 ha), are covered by native forest comprised of predominantly broad-leaf peppermint (Eucalyptus dives Schauer), narrow-leaf peppermint (Eucalyptus radiata Sieb.) and brittle gum (Eucalyptus mannifera) on hillslopes and candlebark (Eucalyptus rubida Labill) in riparian areas. The third catchment, Clem (46.4 ha), was cleared in 1980 and replanted with radiata pine, while a 30 m undisturbed native forest riparian buffer strip (2.7 ha) was retained. The catchment geology is comprised of late Ordovician sandstones and shales, which are considered weak and erodible. Elevation ranges from 431 to 786 m, while soils are highly porous clay-loams and slopes range from 10 to 30◦ with most 20–25◦ (Bren & Papworth, 1991). Annual average rainfall is approximately 1412 mm, most of which falls in winter and spring as low intensity (5–20 mm h−1 ), long duration events compared to summer rainfall, which occurs mostly in high intensity (20–60 mm h−1 ), short duration convective storms (Bren & Hopmans, 2007). Streamflow in the catchments is highest in winter and spring, with recession in late spring to the low flow period over summer and autumn. Ella and Betsy cease to flow over prolonged dry periods in summer, whereas Clem has flowed throughout the period of long-term monitoring (Bren & Hopmans, 2007). The high infiltration capacity of soils resulted in little overland flow generation in the unburnt state, with storm hydrographs generated by sub-surface outflows (Bren & Turner, 1979). The Cropper Creek research project was established in 1975 to investigate the hydrology of the three catchments and the effect on water yield and water quality of the subsequent conversion of Clem to radiata pine (Hopmans et al., 1987; Bren & Papworth, 1991). Fol-

Author's personal copy 572

H.G. Smith et al. / Forest Ecology and Management 261 (2011) 570–581

AUTHORS COPY tion for these events. Peak flows were estimated using Manning’s lowing this first phase of research, a second phase (1997–2003) was equation, in which: initiated to evaluate the effect of management actions (thinning and fertilization) in the 17-year-old plantation (Hopmans & Bren, AR2/3 S 1/2 2007). This included collection of fixed-interval (weekly) and stageQ = (1) n triggered automatic water samples to quantify stream exports of where Q is discharge (m3 s−1 ), A is cross-sectional area of flow total suspended solids (TSS) and various solutes (N, P, S, C, Na, K, (m2 ), R is hydraulic radius (m), S is slope (m m−1 ), and n is ManCa, Mg) (Hopmans & Bren, 2007). ning’s roughness coefficient. The weir cross-section and average The most recent phase of the Cropper Creek research project bed slope were surveyed, as was the average depth of bedload was initiated in 2005 to examine the effect of plantation harvestmaterial after each infill event, which allowed calculation of the ing on water yield and quality. However, wildfire in December 2006 fill height above the v-notch. The pressure probe was encased in completely burnt the three catchments and the research plan was a PVC tube and therefore continued to measure stage during the adapted in response. Following the fire, the radiata pine plantation six infill events. Given the very flashy hydrographs with time to in Clem was salvage harvested in January 2007 using a cable harvest peak flows in the order of 5–15 min, it was assumed that infill technique whereby cut trees were dragged down-slope. This prowas complete at the time of peak stage and fill material comduced a radial pattern of log drag-lines (furrows) which converged pletely displaced the flow. On this basis, the difference between towards the base of sub-catchments, with drag-line patterns most peak stage and fill height above the v-notch was calculated and prominent in three sub-catchments located in the upper part of the used to determine the cross-sectional area of flow across the weir. catchment. Peak discharge was then calculated using Eq. (1) with a Manning’s n value of 0.05 for the first two events and 0.04 for subsequent events. The reduced n value reflected a likely decline in channel resistance 3. Methods to flow resulting from these two large events which caused considerable channel erosion, sediment export and removal of coarse 3.1. Streamflow measurement and water sampling debris. Both Manning’s n values were selected from a table of n values and qualitative channel descriptions for mountain streams Streamflow was measured using 120◦ v-notch weirs with water given by Chow (1959). The form of hydrograph recession was based level recorded at 5-min intervals using Hydrological Services preson the measured rate of fall in stage applied to the peak flow estisure probes linked to Datataker loggers at each catchment outlet. mate and extrapolated to the flow level measured after the weir was Post-fire flow monitoring commenced in late November 2007. emptied. Rainfall was measured adjacent to Clem and Betsy catchment outA proportional flow sampling strategy was adopted for mealets and on the upper ridge between Clem and Ella using manual surement of stream suspended sediment exports. Sigma 900 series rain gauges installed in April 2007. These were later combined with automatic water samplers were programmed to take a 0.5 L samautomatic tipping-bucket rain gauges logging at 5 min intervals. In ple for every 100,000 L of discharge, with samples added to a 200 L January 2008 an Automatic Weather Station (AWS) was established insulated tank. A well mixed 1 L sub-sample was taken at weekly near the outlet of Ella measuring rainfall, solar radiation, humidintervals from the tanks. The sampled volume stored in the tank ity, temperature, and wind speed at 30 min intervals. The delay was measured as a check on sampling frequency. Generally, samin establishing post-fire monitoring reflected time taken to repair pled volumes were within ±15% of expected volumes based on fire-damage to outlet monitoring stations and acquire and install cumulative discharge measured during the sampling interval. The new equipment (which was delayed by the processing of insurance proportional sampling approach enabled calculation of loads over claims). weekly measurement intervals by multiplying the sub-sample TSS Streamflow measurement encountered problems caused by the concentration with the total discharge recorded over the interval. In occurrence of flashy flow events with high peak discharges in addition to the proportional load sampling, fixed-interval (weekly) response to summer storms after the fire. As a result of these events, water samples were manually collected from each catchment. This it was necessary to remove bedload material from the weir pool commenced immediately after the fire in November 2006 in Clem, in Clem catchment on 13 occasions after the fire. No infill events whereas sampling began in Ella and Betsy when flow commenced occurred in Ella, while Betsy weir was partially filled after only in mid 2007. TSS concentrations in fixed interval and proportional one storm event. In response, a debris dam was constructed above water samples were measured by vacuum filtration of samples Clem weir in January 2008 and increased in height in March. While through 0.45 ␮m nitrocellulose filters. this reduced the amount of material entering the weir pool, a large rainfall event on 2nd January 2010 filled both the debris dam and 3.2. Hillslope process measurements the weir. Using an excavator, the weir and debris dam were emptied after each infill event and the volume of material recorded. Field measurements were undertaken after the fire and salvage The average bulk density of fill material was calculated using 56 harvesting to obtain data on hillslope runoff and erosion prosamples collected over eight events to determine the total mass of cesses. Rainfall simulation (RFS) experiments (Fig. 2a) were used bedload material trapped. In addition, material deposited immeto measure infiltration-excess runoff, sediment concentrations and diately downstream of the weir was removed and included in the exports from plots, and to obtain parameters for modelling. RFS weir/debris dam total. Due to the magnitude of sediment exports experiments were conducted early in the second year after the fire in Clem, a survey of the main channel and a tributary gully which (February 2008), with two pairs of plots located in Ella and within formed after the fire and harvesting was commissioned by Hancock the General Harvest Area (GHA) in Clem at upper hillslope posiVictorian Plantations (HVP) in May 2008 to determine the volutions with slopes of approximately 20◦ . A nominal rain intensity metric channel change based on estimates of the previous natural surface. The results of this survey were made available to the project of 100 mm h−1 was applied for 30 min by three oscillating sprays by HVP. to bounded plots measuring 1.5 m wide and 2 m long according Six storm events completely filled Clem weir with bedload durto the method outlined by Sheridan et al. (2007). Rainfall intening the period of flow monitoring (November 2007 to January 2010) sity selection was based on previous studies of post-fire erosion and disrupted accurate measurement across the v-notch. This (Benavides-Solorio & MacDonald, 2001; Lane et al., 2004; Sheridan necessitated estimation of peak flows and hydrograph reconstrucet al., 2007). Timed runoff samples (n = 8 per simulation) from RFS


573

AUTHORS COPY

Fig. 2. Runoff and erosion process measurements in Clem: (a) rainfall simulation on the GHA and (b) overland flow measurements in a log drag-line.

plots were collected at regular intervals to measure runoff rate and TSS concentrations. Four sets of overland flow measurements were also completed at the time of RFS measurements (Fig. 2b). This involved applying a known flow rate (3 L s−1 ) metered using a rotameter at multiple hillslope locations to obtain measurements of flow distance relative to time, flow velocities, estimates of infiltration (where steady state was obtained), and an estimate of Manning’s surface roughness parameter (Manning’s n) for confined flow measurements (log drag-lines). The measurements were undertaken in four areas: (i) unharvested burnt stream buffer in Clem (n = 5), (ii) log drag-lines within the GHA in Clem (n = 3), (iii) a burnt non-harvested hillslope in Ella (n = 5), and (iv) an unburnt hillslope in a nearby catchment (n = 5). Slopes at measurement locations were steep, ranging from 26◦ to 37◦ (average 30◦ ). Soil water repellency was measured using a simplified molarity of ethanol droplet test (King, 1981), involving the comparison of water and 2 M ethanol droplet penetration into the soil in 10 s. On the basis of this, water repellency was classified into non water repellent (water drop penetrates in ≤10 s), water repellent (water penetrates in >10 s and 2 M ethanol penetrates in ≤10 s), and strongly water repellent (2 M ethanol penetrates in >10 s) (McDonald et al., 1990). Measurements were made at 1 cm increments between 0 and 5 cm depth at 20 points on hillslopes with N/NW aspects in both Ella and the GHA in Clem, with another set of measurements made in the stream buffer in Clem. A follow-up water repellency survey was undertaken in winter (August 2008) to compare repellency between Ella and the GHA in Clem. 3.3. Modelling storm runoff following wildfire and salvage harvesting Field observations suggested that changes in the magnitude and timing of overland flow during high intensity, short duration summer storms following the fire and harvesting in Clem contributed to the large observed increase in peak flows relative to the

burnt native forest catchments and peak flows prior to the fire. The change in storm flows in Clem resulted in substantial increases in catchment sediment export, main channel erosion and gully development in tributaries. The radial patterns of log drag-lines that converged downslope were considered to act as an extension to the drainage network, thereby routing overland flow more rapidly to the catchment outlet and producing larger peak flows, with high sediment transport capacities and channel scour potential. In addition, reduced infiltration following the fire and harvesting appeared to substantially increase the amount of storm runoff. The potential effect of drainage network extension on peak flows was examined using a runoff model described below, while changes in infiltration properties affecting runoff amounts were assessed using field data. Runoff modelling to quantify the effect of log drag-lines on peak flows for Clem catchment was undertaken using Thales, a physically-based, distributed runoff model designed for application to small catchments (Grayson et al., 1995; Western & Grayson, 2000). In this study, the most recent version of the model as described by Neumann et al. (2010) was used. Thales represents catchments using a grid-based element network with multiple soil layers and allows both surface and sub-surface flow. Infiltration is represented using an approximate solution for the infiltration rate based on constant head conditions and assuming soil water retention and hydraulic conductivity relationships based on the Brooks–Corey model (Neumann et al., 2010). Soil moisture in each layer varies between permanent wilting point and saturation (i.e. porosity) with sub-surface lateral flow determined for saturated conditions using Darcy’s Law and the topographic slope. Overland flow may be generated when an element is saturated or when the water supply rate exceeds the surface infiltration capacity, with flow routing using an implicit solution to the kinematic wave and Manning’s equations assuming sheet flow for hillslopes and a parabolic form for channel elements. Flow is routed to the two steepest adjacent cells using the D∞ method for hillslope elements, which represents flow direction using a single continuous angle and allows flow partitioning between cells (Tarboton, 1997), while the



Table 1 Summary of model parameters used in runoff simulations. Parameter

Surface layer

Canopy capacity/cover Manning’s n Saturated hydraulic conductivity (mm h−1 ) Soil layer depth (m) Initial moisture (m3 m−3 ) Residual moisture (m3 m−3 ) Field capacity (m3 m−3 ) Porosity ET factor Air entry potential Brooks–Corey index

0 0.195 22

AUTHORS COPY Soil layers Layer 1

a b

22 0.1 0.08 0.07 0.35 0.72 0 0.28 1.39

Parameter source Layer 2

22 0.3 0.12 0.09 0.35 0.63 0 0.37 1.59

Layer 3

22 2.1 0.12 0.09 0.35 0.63 0 0.37 1.59

No canopy storage/cover RFS plot measurementa RFS plots/fitting parameter Hopmans (2009) Field measurement Hopmans (2009) Hopmans (2009) Hopmans (2009) Assume negligible ET Rawls (2004)b Estimated from porosity

Manning’s n was calculated from RFS measurements using the method presented by Mohamoud (1992). Mean values from datasets for sandy clay loam (A horizon: layer 1) and clay (B horizon: layers 2 and 3) based on soil classification by Hopmans (2009).

steepest descent method (D8) is used for channels (O’Callaghan & Mark, 1984). Flow does not infiltrate into channel elements and sub-surface lateral flow may exfiltrate to channels. To examine the effect of log drag-lines on runoff it was necessary to generate catchment DEMs with and without the drag-line features. A LiDAR survey of Cropper Creek flown in August 2007 provided data to produce a high resolution DEM (cell size 2 m) of Clem catchment. In addition, a ground-based survey measured the initiation points and convergence locations of drag-lines using a differential GPS, while recording cross-sectional dimensions (n = 24) for the drag-lines. The resolution of the LiDAR-based DEM was insufficient to capture the drag-lines, so their surveyed location and mean depth (0.17 ± 0.11 m) were used to modify the DEM by subtracting the mean depth from the DEM surface. This way, the modified DEM was used to approximate the actual condition of the catchment, while the unmodified DEM represented the catchment without drag-lines. Furthermore, to better represent the measured difference in flow velocities between channelized flow in draglines and sheet flow across the GHA, the drag-lines were defined as channel features in Thales. This was done by setting the channel initiation threshold to 600 m2 for the drag-line DEM and 16,000 m2 for the DEM without drag-lines, with the resulting difference between the two drainage networks approximating the drag-line extent. The model requires a single Manning’ n value to represent the channel network. An n value of 0.05 was used on the basis of overland flow experiments in the drag-lines (median n value from 7 measurements) and the similarity to the n value selected for peak flow estimation in the main channel. Runoff modelling enabled comparison of storm hydrographs and peak flows for the drag-line and the no drag-line DEMs. Modelling was limited to a single summer storm event that generated a large peak flow (one of seven during the monitoring period) but did not fill the weir with bedload material. This allowed comparison of model outputs with peak flow data that did not require estimation using Manning’s equation. The modelled storm event occurred on 22 January 2009 and produced a total rainfall and peak 5 min intensity of 33 mm and 115 mm h−1 , respectively. The parameter values used to run Thales for this event are presented in Table 1. Given that summer storm runoff generation after the fire occurred predominantly as infiltration-excess overland flow in response to high rainfall intensities and reduced infiltration (compared to pre-fire where overland flow was rare), the key model parameters were surface saturated hydraulic conductivity (Ksat ) and Manning’s surface roughness coefficient (n). RFS measurements provided initial input values. However, the RFS data was collected in February 2008 and both parameter values were likely to have increased in response to post-fire recovery effects. Therefore, Ksat was employed as a hydrograph fitting parameter while maintaining a constant n, with

a best fit obtained for a Ksat of 22 mm h−1 , which was consistent with a post-fire recovery trend since the time of the RFS measurements, where mean steady state infiltration was 9 mm h−1 . This allowed the best fit hydrograph generated for the drag-line catchment DEM to be compared to model outputs for the DEM without drag-lines. Subsequently, Manning’s n scenarios (for n values of 0.25, 0.3, 0.4, 0.5, 0.6, 0.7 and 0.8) were run using both the dragline and the no drag-line DEMs to examine the potential effect of surface vegetation recovery and/or management practices which increase hillslope surface roughness. 4. Results and discussion 4.1. Rainfall and streamflow Total post-fire rainfall prior to the commencement of streamflow monitoring (January to November 2007) was 793 mm compared to the longer-term (1985–2009) average of 958 mm for the equivalent period at the nearest Bureau of Meteorology station (Edi Upper), located 13 km north-west of Cropper Creek. Observations of streamflow prior to monitoring indicated that Ella commenced to flow in June, Betsy in July (but ceased again in September), while Clem flowed throughout the year. Between the time of the fire and the start of flow measurement in 2007 five storm events filled or partially filled Clem weir with bedload material. Following the start of flow monitoring, a series of five storm events during summer 2007–2008 (one year after fire) generated large peak flows in Clem (burnt and harvested pine catchment) that substantially exceeded flows in the burnt native forest catchments Ella and Betsy (Fig. 3). Based on rainfall measurements in or near the catchments, estimated Average Recurrence Intervals (ARI) for 30 min rainfall durations were 100 years) which continued to generate peak flows that were much higher than from either Ella or Betsy. In Clem, the estimated or measured peak flows from these seven events (six of which resulted in the weir filling with bedload) considerably exceeded peaks from all other events measured, including all peak flows during winter and spring, historically the period of maximum streamflow (Bren & Hopmans, 2007). Furthermore, all peaks flows from these events exceeded the largest pre-fire peak flow (304 L s−1 ) recorded in response to a 221 mm (over 36 h) rainfall event during the previous monitoring phase from 1997 to 2003 (Hopmans & Bren, 2007). This occurred despite substantially lower rainfall totals for all the post-fire storm events. In contrast, for the same pre-fire storm event the peak flow (474 L s−1 ) in Ella exceeded


575

AUTHORS COPY

Fig. 3. Post-fire streamflow and rainfall measured at 5-min (I5 ) and 30-min (I30 ) intervals from November 2007 to January 2010 at (a) Clem (salvage harvested), (b) Ella and (c) Betsy (note the difference in y-axis scale). Rainfall data is missing from 2 February to 13 March 2009 for Ella AWS, while the Betsy pluvio malfunctioned from June to September 2008 and the data from this period has been excluded.

all peak flows recorded after the fire. Despite the size of the peak flows for these post-fire events, combined they represent only 9% of total flow measured in Clem, with 3% and 8% from Ella and Betsy, respectively. However, the events account for most of the sediment exported during the period of monitoring, indicating their significance for post-fire erosion and sediment transport. Runoff ratios for the seven events for Clem were substantially higher than Ella and Betsy (Table 2). There was also a pronounced decline in runoff ratios for Clem with time since fire, which was not apparent in Ella or Betsy. This indicates that the combined fire and harvesting effect in Clem significantly altered factors controlling storm event runoff generation and these factors have been gradually recovering. In contrast, the absence of a clear recovery signal in the runoff ratios for Ella and Betsy suggests that the initial effect of the fire and subsequent surface vegetation recovery on storm runoff generation was much less significant. Nonetheless, post-fire storm hydrographs in all three catchments were flashier compared to pre-fire conditions. The factors affecting storm runoff generation following fire and harvesting in Clem are considered in subsequent sections. Total water yield during the period of post-fire monitoring was also substantially higher from Clem than Ella and Betsy. By the end of the monitoring period, cumulative flow totalled 1049 mm, 326 mm, and 195 mm from Clem, Ella, and Betsy, respectively. Most of the flow difference in Clem comes from the initial storm events over summer 2007–2008 and higher water yields over winter–spring in 2008 and 2009 (Fig. 4). Post-fire streamflow in Clem also greatly exceeded the pre-fire average annual flow based

on six years of monitoring (1997–2003) reported by Hopmans & Bren (2007), with the average post-fire runoff ratio of 42% (based on two complete years of monitoring) exceeding the pre-fire average of 28%, despite lower post-fire rainfall. In contrast, the average postfire runoff ratio for Ella and Betsy was 10% compared to a pre-fire average of 21% based on 21 years of monitoring (Bren & Hopmans, 2007). This change was associated with a 15% decline in post-fire rainfall compared to the pre-fire average. 4.2. Sediment concentrations and exports Post-fire fixed-interval weekly water sampling provided TSS concentration data for comparison with pre-fire data from Clem

Fig. 4. Post-fire cumulative streamflow from Clem (harvested), Ella and Betsy. Ella cumulative flow may be slightly underestimated due to the absence of flow data over a 5-week period in June and July 2008.



AUTHORS COPY

Table 2 Rainfall, peak and total flows, and runoff ratios for the seven largest post-fire peak flow events recorded in Clem catchment with the same events in Ella and Betsy for comparison. Storm event

Rainfall (mm)a

3 Dec. 07 21 Dec. 07 20 Jan. 08 12 Feb. 08 25 Mar. 08 22 Jan. 09 2 Jan. 10

47 61 41 59 17 33 125

Not recorded 21 16 28 18 66 96


47 61 41 59 17 33 125

Not recorded 21 16 28 18 66 96


47 61 41 59 17 33 125

Not recorded 21 16 28 18 66 96

Max. rainfall intensity (mm h−1 ) (30 min duration)

Peak flow (L s−1 ) c

Total event flow (mm)d