There was no single consistent effect of fire on stream physicochemistry or habitat. While variables did respond to fire in a consistent way, the magnitude and longevity of impacts varied by the factors whether the site or upstream catchment was burnt and alpine or montane environment. Even accounting for these factors there was still much variability suggesting that there are a) site differences not accounted for in this study that influence response to fire b) differences in local fire intensity and behaviour that affect impacts. These points aside, in the alpine zone the effects tended to be longer lasting postfire than those observed in the montane zone. Effects of the fire on alpine sites were evident up to 8 years following fire, with for example, % riparian grass cover increasing but % riparian shrub cover decreasing at burnt sites relative to unburnt sites (Table 2, Supplementary Figure S1.25 and Figure S1.26). While in the montane sites % riparian grass cover decreased for 2.5 years post-fire but pH increased for 2 years post-fire. However, for 90% of response variables within both alpine and montane site (33 at alpine sites, and 12 at montane sites), there was no evidence of a statistically significant interaction term between site type and time since fire beyond 2 years, e.g., see habitat score at alpine sites (Supplementary Figure S1.33) and see pH at montane sites (Supplementary Figure S2.3).
Montane vs. alpine sites:
As predicted, the effects of fire on water physicochemical and habitat variables differed between the alpine and montane sites. This comparison was not planned when the data was collected, and consequently there are not matched pairs of similar sites at different elevations. There is also only one ‘replicate’ of each elevation type. These points aside, for the three physicochemical water measured in both regions (EC, pH and turbidity), fire effects were greater at montane sites than alpine sites (Supplementary Figures S1.3 to S1.5, and Supplementary Figures S2.1 to S2.3), but slightly longer lasting at alpine sites (2.5 years vs. 2 years) (Table 2). Where riparian vegetation variables were affected, these variables decreased at both montane and alpine sites (% shrub cover and % grass cover) following the fire, but recovery was slower at alpine sites compared to montane sites (up to 8 years versus 4 years). Multiple reasons could explain the differential responses at montane and alpine sites. Fire does not burn evenly throughout the landscape because many factors affect the intensity, severity and spatial extent of wildfire, and these factors vary within the landscape. Firstly, The subalpine woodlands found at our alpine sites contain different communities of plants to those found in montane forests at lower elevations (Costin et al. 1979, Adams et al. 2013), plants in alpine zones grow slower than those at lower elevations (Atkin et al. 1996), and Snow Gums (Eucalyptus pauciflora) which are found at alpine sites are more fire sensitive relative to lower elevation Eucalyptus species (Barker 1988, Green and Osborne 1994). In general, the vegetation in the alpine zone is lower and sparser (Costin et al. 1979) potentially leading to lower fuel load, but see Adams et al. (2013). Secondly, there are substantial differences in catchment use within each of the two elevational zones. Our sites in the alpine zone are located within largely natural areas with minimal development relative to our sites at lower elevations. While some sites in the montane zone are also located in largely natural areas, many are located in, or adjacent to agricultural and/or urban areas. Thirdly, there are differences in water availability in the two elevational zones. Precipitation is substantially different, with mean annual precipitation in the alpine zone 1750–2200 mm, but about 630 mm in the montane zone (www.bom.gov.au). Further differences in water availability arise because of evapotranspiration rates, which tend to decrease as elevation increases (Bruijnzeel and Veneklaas 1998, Lüttge 2007, Stoutjesdijk and Barkman 2015, cited in Gallardo-Cruz et al. (2009). Thus, it is probable the wildfire was less intense in alpine areas relative to montane areas in the current study which was reflected greater proportion of variables effected by the fire in the montane zone relative to the alpine zone, but the effects in alpine zones were generally more severe and longer lasting relative to the montane zone. The longer lasting effects likely related to the slower growth of terrestrial vegetation in the alpine relative to the montane zone (Atkin et al. 1996).
Effects of fire on nutrients:
Increases in nutrient, i.e., N and P, concentrations are frequently reported post wildfire, across different environments (e.g., Smith et al. 2011a, Verkaik et al. 2013, Sherson et al. 2015, Verkaik et al. 2015, Collins et al. 2019). TN at alpine sites in our study increased as predicted (Table 1) for 18 months post-fire at burnt sites only, but not catchment burnt sites. Increases in nitrogen elsewhere generally return to pre-fire levels within 5 years (e.g., Lane et al. 2008, Mast and Clow 2008), but see Rhoades (2019) that lasted 14 years post-fire). Recovery of TN has been linked to recovery of hillslope and riparian vegetation (Rhoades et al. 2019). The largest influxes of nitrogen to waterways are typically associated with erosion after fire (Lane et al. 2008), often occurring within 12 months of the fire (Bladon et al. 2008). However, seasonal runoff patterns at alpine sites are different to sites at lower elevations because of the seasonal snowpack and its melt at alpine sites, resulting in different rates and patterns of nutrient movement (Mast and Clow 2008). The recovery of TN is likely related to recovery of grasses (discussed below) that slow overland water flow, and in doing so reduce ash and sediment movement into waterways that would otherwise increase TN. It is possible that recovery of grasses is more important than recovery of shrubs in terms of reducing or controlling TN concentrations post-fire.
Contrary to our predictions, TP decreased at alpine catchment burnt sites for one month following fire. Others have generally observed TP to increase following fires (Son et al. 2015, Emelko et al. 2016), especially after rainfall (Son et al. 2015), but reports of TP concentrations decreasing following fire exist (e.g., Noske et al. 2010). It is logically possible that fewer people visited and used the areas immediately post-fire because visitor access was restricted, or because fewer people recreated in these areas immediately post-fire. Thus, there may have been a reduced load on sewerage treatment facilities and therefore lower TP in the burn areas. Likewise, we could also expect fewer wildlife e.g., wombats in burnt areas because of mortality or movement to unburnt areas which could lower TP.
pH and Alkalinity:
The response of pH in waterways following fire is determined by the buffering capacity of the stream, and the acidity or alkalinity of stream inputs (Paul et al. 2022). Thus, a stream with high buffering capacity would be expected to resist changes in pH in response to ash inputs, but a stream with low buffering capacity would be expected to change pH in response to fire. Both increases (e.g., Son et al. 2015) and decreases (Dahm et al. 2015, Sherson et al. 2015) in pH following fire have been reported. Indeed, we detected both increases and decreases in pH. Post-fire pH was unchanged at alpine burnt sites but decreased at alpine catchment burnt sites for 12 months post-fire. In contrast, at montane sites, pH increased at both catchment burnt and site burnt sites which is a change in the opposite direction to alpine sites, persisting for 24 months at burnt sites and 3 months at catchment burn sites. Further, alkalinity in the montane zone at burnt sites increased for 12 months following fire, but because this variable was not measured at alpine sites, we cannot compare or contrast changes in the buffering capacity of streams with those at our montane sites.
The up to 2 year duration of the responses observed in pH are consistent with those reported elsewhere (e.g., Rhoades et al. 2011, Lydersen et al. 2014) and likely linked to movement of ash. The amount of ash produced by wildfire and its characteristics depend on the mass and type of fuel burned, the completeness of combustion (Bodi et al 2014), and these factors vary spatially within areas burnt. Accordingly, the depth of the ash layer ranges from a thin layer (e.g., < 5mm) from a grassland fire with low fuel load and high combustion completeness (Bodí et al. 2014), to a thick layer (e.g., up to 200 mm) from a dense forest that contains a higher fuel load (Gabet and Sternberg 2008). Ash is often removed quickly by wind or water (Bodí et al. 2014), sometimes within days or weeks (Pereira et al. 2015).
Suspended solids (turbidity)
Although predicted increases in turbidity were not observed at alpine sites, the predicted increases were observed at montane sites within both burnt and catchment burnt site categories for 6 months and 3 months respectively. Within our montane study area, White et al. (2006) attributed the increase in turbidity to intense and localised storms that eroded fire debris and ash from fire affected slopes. The duration of turbidity increases we detected are shorter than the typical 3–5 years reported elsewhere (e.g., Nyman et al. 2011, Rhoades et al. 2011). Elevated suspended solids post-fire can threaten drinking water supply (White et al. 2006, Bodí et al. 2014) not only by placing great demand on water treatment facilities, but sediments and ash can also contain contaminants including metals (Smith et al. 2011b, Abraham et al. 2017, Rust et al. 2018).
Streambed ash, muck and detritus: Streambed ash, muck and detritus data were collected only at alpine sites, and at two spatial scales: the larger ‘reach’ scale i.e., a 100 m stretch of stream, and the smaller 10 m ‘riffle habitat’ scale within the same reach (see Nichols et al. (2000) for details). The higher energy riffle habitat would likely have less deposition relative to the entire reach. We thus expected greater proportional coverage of sedimentary detritus, muck and ash in the reach relative to the riffle, and we expected this increase for a longer duration. Consistent with expectations, we observed a shorter increase in ‘mud and muck’ (the substrate fraction associated with ash inputs) in the riffle habitat of 6 months at catchment burnt sites only, versus four years at the reach scale in both catchment burnt and site burnt categories. This observation of effect lasting four years is similar to the Verkaik et al. (2013) review that concluded that increases in these variables lasted 1 to 4 years in Mediterranean climate streams, and 5–10 years in non-Mediterranean streams.
The amount of detritus observed in alpine waterways increased post fire instead of decreasing, contradicting our predictions. These changes were only evident at the reach scale, but not riffle scale. Increases in reach scale detritus were detected in the alpine zone in both sites categories for up to 6 months following wildfire. Elsewhere, coarse particulate organic matter (CPOM), fine particulate organic matter (FPOM) and leaf litter inputs have typically been reported to reduce in burned catchments after fires and subsequent storms but CPOM recovered quickly (i.e., within 2 to 4 years) at sites where the riparian canopy remained intact relative to where riparian vegetation was burned (Cooper et al. 2015). Where riparian vegetation is burned, recovery of leaf litter inputs and associated CPOM/FPOM ranges from 3 years (Noske et al. 2010) to 5 years (Jackson et al. 2012, Cooper et al. 2015), is of longer duration than the 6 months detected in this study, and changes in the opposing direction to that found in our study.
Streambed geological substrate: As in the preceding sub-section, streambed geological substrate data were collected only at alpine sites, and at two spatial scales: the larger ‘reach’ scale i.e., a 100 m stretch of stream, and the smaller 10 m ‘riffle habitat’ scale within the same reach (see Nichols et al. (2000) for details). At catchment burnt sites we detected two unexpected changes in substrate composition. Firstly, increases in % reach bedrock lasting 5 years. Secondly, decreases in % reach cobble lasting for 18 months. Similarly, Oliver et al. (2012) detected reduced % streambed cobble for two years following a fire, despite not observing any scouring events or large floods post-fire. We offer two non-mutually exclusive mechanisms that could explain our observed increase in reach % bedrock. Firstly, increased surface runoff because of reductions in interception and infiltration of precipitation would tend increase overland flow (Ebel and Moody 2013). Increased overland flow would tend to increase peak discharge and velocity, and shorten periods between precipitation and increased peaks (Shakesby and Doerr 2006). The resultant ‘peaky’ scouring flows that follow would tend to expose bedrock. Secondly, a common short-term response following fires is reduced infiltration leading to reduced groundwater recharge (e.g., Ebel and Moody 2013). If infiltration rates recover before vegetation recovers post-fire, increased recharge of groundwater leads to increased baseflow (Bart and Tague 2017, Poon and Kinoshita 2018), which in turn could contribute to scouring flows thus exposing streambed bedrock. Although changes in these substrate categories were not expected, the site category (catchment burnt) was the area in which we expected to see changes in other substrate size classes.
Despite our prediction, we did not detect an increase in substrate fines (sand, silt and clay fractions combined) at either site burnt or catchment burnt sites. Increasing burn severity and extent have been associated with greater interannual variability, rather than perennial increases in sediment loads, likely because of fire and water flow decreasing habitat stability in burned catchments (Arkle et al 2010). However, sediment yields in subalpine streams may be less affected than yields from lower elevation streams because of the slow release rate of spring snow melt (Mast and Clow 2008). Thus, in the absence of fine sediments at alpine sites, we suggest the fires and subsequent precipitation did not result in entrainment and deposition of fine sediment following fire at alpine sites because the fires in this instance lacked the intensity and severity to mobilise fine sediments, and/or because there was insufficient rainfall or snowmelt to mobilise fine sediments at alpine sites after fire.
Stream channel morphology
Observations at our alpine site do not support hypothesised reductions in bank stability, either as changes in bank width or bank height. However, at montane sites we observed reduced bank stability as changed bank width (but not bank height) following wildfire. Contrary to our predictions, bank widths decreased in the first 18 months following fire instead of increasing, occurring only at burnt sites. Channel narrowing following wildfire has been reported (Shakesby and Doerr 2006) resulting from complex responses to destruction of vegetation and litter, and alteration to soil properties. It may be that the burn severity and extent in our alpine region were insufficient to change our sites, or post-fire rainfall intensity was insufficient to move sediments into the streams.
Riparian vegetation
Contrary to other studies where wildfire consistently enhanced exotic vegetation composition while having no effect on native species composition (e.g., Alba et al. 2015), we did not observe change in the ratio of exotic versus native vegetation following wildfire at alpine sites. Further, at montane catchment burnt sites, we found observed increases in proportions of native vegetation at catchment burnt sites for 18 months following wildfire. These changes were not detected at burnt sites in line with our predictions. The increased portion of native vegetation at catchment burnt sites we observed may result from downstream ash redistributions by wind and water, fertilizing native vegetation. Plants can respond positively to ash additions because of nutrient content including Ca2+, Mg2+, K+, P and N (e.g., Bodí et al. 2014, Paul et al. 2022). These nutrient inputs from the redistribution of ash may explain the increased grass cover in the riparian zone at both alpine and montane sites, and at catchment burnt sites in alpine zones. Indeed, some Australian native plants require ash or chemicals associated with fire for reproduction or growth (Enright and Thomas 2008) offering an explanation for increased native vegetation downstream of fire at catchment burnt sites in the montane zone.
Despite our predictions that riparian trees > 10 m, and riparian trees < 10 m, would be reduced in cover following fire, we did not observe this. However, we did observe reduced riparian shrub cover in both alpine sites (for at least 8 years), and at montane zones (for at least 4 years), where sites had been burnt. Alpine plants grow slower than those from lower and warmer elevations (Atkin et al. 1996) providing a likely explanation as to why alpine sites recovered slower than montane sites. Riparian grass in both our alpine and montane zones recovered from decreases over the same period (12 months). Further, at alpine sites after 12 months, the amount of grass cover increased above what was observed before wildfire. In addition to potential fertiliser effects from ash redistributions discussed above, grasses colonised riparian areas that pre-fire were covered by riparian shrubs. In eucalyptus woodlands, recovery of the shrub layer is known to be quick relative to grasses (Dragovich and Morris 2002). We observed the opposite at alpine sites, where the shrub layer remained sparse for much longer while grasses recover relatively quickly, which is broadly consistent with Adams et al. (2013) review of fire in elevated environments in south-eastern Australia.
The recovery times of riparian vegetation is important because recovery of aquatic systems is closely tied to terrestrial recovery (Verkaik et al. 2013, Bixby et al. 2015, Leonard et al. 2017). The vegetation is important in different ways because ground cover will be most effective at reducing, slowing and filtering overland flow, while trees and large shrubby vegetation will have a large effect on transpiration and rainfall interception.