Changes in DOM and alum doses for two rivers of contrasting catchments after intense wildfires

Wildfires are a feature of the Australian landscape (Mannik et al. 2013; Nolan et al. 2021) that occur following extended periods of dry weather conditions and where fuel loads are high and moisture levels are low. Events such as the 2019–2020 Australian wildfires have been attributed to the effects of anthropogenic climate change driving increased temperature extremes with long-term warming trends (Nolan et al. 2021; van Oldenborgh et al. 2021). The impacts of large-scale wildfires or megafires can be severe, including widescale destruction of native fauna and flora, their habitats, loss of human life, and detrimental socioeconomic outcomes (Richards et al. 2020; Neris et al. 2021).

The impacts of wildfires on water quality have been widely reported, including investigations on the dynamics of turbidity, major ions, metals and heavy metals, polyaromatic hydrocarbons, nutrients (N and P), and organic matter quantified as total (TOC) or dissolved organic carbon (DOC) or as chemical oxygen demand (Quill et al. 2010; Sherson et al. 2015; Mansilha et al. 2019; Biswas et al. 2021; Johnston & Maher 2022). This also includes various studies that have applied fire severity and vegetation indices derived from satellite imagery data to assess the impacts of wildfires on water quality and streamflow dynamics. The normalized burn ratio (NBR) is a spectral index that provides an indication of photosynthetic biomass and soil and foliar moisture (Miller & Thode 2007; Chuvieco et al. 2020). Low (or negative) NBR values generally correspond to sparsely vegetated areas, such as bare ground or recent burns. The difference between the pre- and post-fire NBR products (dNBR) therefore provides an effective way to assess the degree of burn severity, and a general interpretation of dNBR values is given in Supplementary Table S1. Hogue & McCray (2018) utilized the dNBR index, derived from Landsat imagery, to map the impact of 159 fires across 153 burnt watersheds in the western United States. Their investigation aimed to discern post-fire water quality responses, revealing significant increases in nutrient, major-ion, and metal concentrations within the first 5 years after a fire event. Joehnk et al. (2020) also utilized dNBR to assess burn severity in the Upper Murray and the Mitta Mitta catchments, examining its ramifications on hydrology, sediment, nutrient transport, and subsequent water quality alterations in the receiving waters. Their findings highlighted severe implications for the water quality of River Murray and Lake Hume from the fires (Biswas et al. 2021).

Dissolved organic matter (DOM) is a ubiquitous component of fresh surface waters originating from natural and anthropogenic allochthonous sources in the catchment and by autochthonous production and degradation processes (Sachse et al. 2005; Young et al. 2005). Organic compounds that comprise DOM are derived from decomposition and biodegradation of plant and animal matter, synthetic organic matter, bacteria, and algae. These compounds vary highly in their characteristics including reactivity, molecular size, hydrophobicity/hydrophilicity, structural and functional groups, and elemental constituents (Mostofa et al. 2013; Wang et al. 2013). DOM is commonly measured as DOC and comprises humic substances (humic and fulvic acids) and non-humic compounds such as carbohydrates, proteins, amino acids, lipids, phenols, alcohols, organic acids, and sterols (Mostofa et al. 2013; Wang et al. 2013). DOM levels in fresh waters are influenced by a wide range of factors, including amounts of organic matter (e.g. soil litter layers) present in catchments, soil sorption of DOM (Kaiser & Zech 1997), vegetation types and loadings, topography, seasonality, and climate conditions (Awad et al. 2015, 2017; Huang et al. 2015). Concentrations of DOC in fresh waters can range widely between different water sources (from <1 to ≫10 mg/L) and vary highly among water sources depending on factors such as preceding rainfall intensities, vegetation types, soils, and topography within a catchment (Awad et al. 2015, 2016). DOM present in fresh waters can be significant based on it being a substrate for microbial growth fuelling microbial-based food webs (Kreutzweiser & Capell 2003; Young et al. 2005), transporting metals including heavy metals (Kaiser & Zech 1997), and at high concentrations, it can cause water to be highly coloured, attenuating light penetration that consequently limits photosynthetic activity in the water body (Mostofa et al. 2013). Conversely, DOM can attenuate the penetration of UV irradiance into the water body protecting aquatic organisms (Mostofa et al. 2013). High loadings of DOM can occur when catchments become inundated with surface water flows following heavy rainfalls, causing the transport of terrestrial particulate organic matter and DOM into rivers and streams (Mosley et al. 2021), potentially leading to elevated heterotrophic bacterial activity leading to oxygen depletion within the water body and then death of the aquatic organisms. Although negative impacts can occur on aquatic organisms, these events have also been noted as being of benefit in terms of supplying food and nutrients, increasing river productivity and benefiting waterbirds, native fish, and other aquatic organisms (DAWE 2023).

DOM present in drinking water sources can impart colour and comprise taste and odour compounds impacting on the aesthetic quality of water. Organic compounds that are recalcitrant to removal by conventional water treatment processes, i.e. coagulation and flocculation, can react with chlorine during disinfection, leading to halogenated disinfection by-products that can be of health concern (Teksoy et al. 2008; Golea et al. 2017), as well as being a substrate for microbial regrowth in drinking water distribution systems (Niquette et al. 2001). To minimize the formation of halogenated disinfection by-products in drinking waters, raw waters may be treated using enhanced coagulation, which practically maximizes the removal of DOM prior to chlorination, as required in the United States by the United States Environmental Protection Agency (USEPA) (Edzwald & Tobiason 1999) and applied elsewhere.

In this paper, we report the results of an investigation on the dynamics of DOM of two river waters from contrasting catchments following intense wildfires and the alum doses needed for enhanced coagulation in drinking water supply. These alum doses were derived using a model recently reported by Daraei et al. (2023a) where input data are DOC and turbidity. A further alum dose prediction model, which is based on input data of UV absorbance at 254 nm (UV@254 nm), true colour, and turbidity, is also described. These were developed for source waters with wide-ranging qualities including those impacted by the wildfires.

The Middle River catchment (Figure 1) has an area of ∼14,500 ha located in the north of Kangaroo Island, South Australia. Prior to the fires, land uses in the catchment were dry land agriculture (mainly pasture, 50%), plantation forestry (pine and blue gums plantations, 12%), and conservation and natural environments (native vegetation, 38%) (DEWNR 2013). A land-use map for the Middle River catchment is shown in Supplementary Figure S1. The waters of the Middle River catchment flow into the Middle River Reservoir, which holds about 540 ML. The reservoir water, known for its high DOC concentrations and high colour levels, is treated using the Magnetic Ion Exchange (MIEX)^® technology along with conventional treatment processes for drinking water supply.

Severe wildfires occurred in Australia during the 2019–2020 wildfire season affecting the ACT, New South Wales, Queensland, South Australia, Tasmania, Victoria, and Western Australia, with reported areas burnt of over 17 million hectares (ha) (Huff et al. 2020; Richards et al. 2020). The year 2019 was the warmest year on record in Australia up to that point, with record-low levels of rainfall over large areas of Australia. In most parts of tropical Western Australia and the Northern Territory, rainfalls had been well below average in the 2018–2019 wet season, resulting in very low soil moisture levels over most of the continent leading into December 2019. In December 2019, Australia experienced 11 days with a national area-averaged maximum of 40 °C or above, and on 18 December 2019, Australia experienced the hottest averaged day of 41.9 °C. In 2019, Australia had the highest December accumulated forest fire danger index (FFDI) since recording began in 1950 (BOM 2020).

The Orroral Valley fire started on 27 January 2020 and in the first 24 h, the fire had spread ∼2,580 ha and then at 4,500–8,000 ha per day over the subsequent 3 days into the remote parts of the Namadgi National Park. Forest fuel loads were exceptionally dry and approximately 78% (over 82,700 ha) of the Namadgi National Park was burnt during the fire (ACT Government 2020). Kangaroo Island has a total area of over 440,000 hectares of which approximately a third is a nature reserve. During the 2019/2020 wildfires, about 38% of the island was burnt, of which 83% was burnt at high to very high severity (DCCEEW 2022). The Middle River catchment was extensively burnt in the summer fires.

The Orroral River water sampling site (35.66512S, 148.98929E) is located in the Namadgi National Park, Canberra, and the Middle River site (35.7349S, 137.1032E) is located on Kangaroo Island South Australia. Water samples were collected from the Orroral River as grab samples and from the Middle River, predominantly using a composite sampler. The composite sampler was programmed to pump 500 mL of uniform increments of flow into an unrefrigerated container. The water samples were collected in PET bottles, transported chilled to the laboratory, and stored at ∼4 °C until analyses were performed. After the fires, sampling was undertaken from February 2020 through to November 2021, when access to sampling sites of the burnt catchments was logistically feasible. In February and March 2020, soon after the fires had occurred, grab samples were collected from both sites.

Water samples were filtered using 0.45 μm polyether-sulfone membrane filters (Sartorius, Germany) for DOC, ultraviolet–visible (UV–Vis), and colour analyses. DOC analysis was performed using either a TOC analyser (O I Analytical, Aurora 1030 – wet oxidation) for samples collected prior to the Kangaroo Island wildfires (by the Australian Water Quality Centre, SA Water Corporation, a NATA registered laboratory) and using a Sievers InnovOx Laboratory TOC analyser (GE Instruments, USA) at the University of South Australia for samples collected after the wildfires had occurred. For DOC analyses of water samples, a series of DOC standards were prepared from potassium hydrogen phthalate stock solutions (100 mg C/L) and analysed with the water samples. UV light absorbance at 254 nm (A254) was determined using a UV–Vis spectrophotometer, Agilent Technologies Cary 60, for samples collected prior to the Kangaroo Island wildfires at the Australian Water Quality Centre, SA Water Corporation, and by Model UV-120 (Miostech, Australia) equipped with a 1 cm quartz cuvette for grab samples collected after the wildfires had occurred. Sample measurements were taken after the instrument was set to zero with high-purity Milli Q water high-purity Milli Q water in the same cuvette. The colour of the samples was measured at 456 nm using a 5 cm glass cuvette. A 50 Hazen unit (HU) colour standard (platinum-cobalt) was used for calibration after the instrument was set to zero with high-purity Milli Q water, high-purity Milli Q water, as reported by Bennett & Drikas (1993).

Turbidity (FNU) was measured using a YSI EXO2 multi-parameter sonde with turbidity probe (YSI, Yellowsprings, USA) or otherwise by a benchtop turbidity meter (Model 2100n, Hach, USA). The YSI EXO2 turbidity probe was calibrated according to the manufacturer's instructions, using standard turbidity solutions (<0.5, 100, and 500 NTU, analytical reagent (AR) grade, Rowe Scientific/Aim Scientific, South Australia). The Hach turbidity meter was calibrated using Hach 2659505 StablCal Turbidity standards (<0.1, 20, 200, 1,000, and 4,000 NTU), also according to the manufacturer's instructions. The turbidity of water samples and standards were cross-checked between the YSI EXO2 turbidity probe (FNU) and the Hach turbidimeter (NTU) and calibrated to ensure the readings were directly comparable.

The total dissolved solids (TDS) parameter was determined from the electrical conductivity of water measured using an EXO EC/temperature probe (Xylem Australia) after calibration using a 1413 micro Siemens standard solution. In this study, the coefficient of variation (CV) values for analyses of water quality parameters were as follows: for A254 nm/m 1.3%, true colour (HU) 8.0%, DOC (mg/L) 3.4%, turbidity (FNU) 0.85%, and TDS (mg/L) 0.1%. Chlorophyll a (Chl a) was measured using a YSI EXO2 Sonde (Xylem, USA) multi-meter equipped with a Chl a probe (excitation/emission paired wavelengths (ex/em) of 470 [±15]/685 [±20] nm, EXO Chl a probe) along with other probes (including pH, turbidity, temperature, and conductivity) as previously described (Daraei et al. 2023b). The Chl a probe was calibrated using a standard Rhodamine B solution, according to the manufacturer's instructions.

The NBR images were produced for each acquisition close to a water sampling date, where possible, and geographic information system (GIS) boundaries of the water catchments were used to restrict the analysis to values within the catchments. Imagery dates for the Orroral River and Middle River catchments are given in Supplementary Table S2. Post-fire vegetation recovery within the catchments was implied by the recovery of NBR towards pre-fire values. This was indicated by examining the shift through time in the ratios of the burnt, unburnt, and regrowth pixel areas.

The correlation coefficients, r, between the enhanced alum doses (determined by jar tests) and the corresponding model fitted doses using the aforementioned Equations (2)–(4) are 0.9663, 0.9858, and 0.9745, respectively. Enhanced alum doses were based on ΔDOC/ΔAlum = −0.015 mg/mg, as previously described (van Leeuwen et al. 2005; Daraei et al. 2023a). These models were developed for predictive estimation of coagulant doses where, in application at water treatment plants, verification of dose suitability by jar testing is recommended. In this study, model-based alum doses were considered for comparative assessment of coagulant (alum) requirements for water treatment of river waters from local and contrasting catchments impacted by extreme wildfire events and, subsequently, with catchment recoveries. Direct comparisons of the model-determined alum doses with coagulant use at operational drinking water treatment plants were not possible. This was due to their large distances downstream (and with inflows from other catchments) from the sampling point in the case of the Orroral River. In the case of the Middle River, this river flows into a storage reservoir, which then provides source water to a MIEX treatment plant.

The fitting of linear regression lines, determinations of linear equations, coefficients of determinations, and correlation coefficients, reported at 95% confidence were acquired using the Data Analysis add-in package function in Microsoft Excel.

Figure 3 shows high percentages of burnt catchment areas (as imagery pixels) directly after the fires for both catchments. Over the following 2 years, the Middle River catchment showed a cyclic pattern where these percentages reduced, along with vegetation regrowth, to ∼40 to 50% in the late autumn to spring seasons and then increased again in the following summer (Figure 3(a)). The increases in the summer of 2020/2021 are likely a result of the Middle River catchment containing extensive pastures, the curing of grasses, and their reductions as feed for livestock. In the case of the Orroral River catchment, there was a consistent decline from the very high initial percentages of burnt areas directly after the wildfires, over the 2 years of monitoring (Figure 3(b)). In both catchments, revegetation is indicated from dNBR data, but this occurred more consistently in the Orroral catchment where vegetation is predominantly of native forests that are less susceptible to seasonality (than pastures). In a study of the Orroral Valley fire, reported by Gale et al. (2022), dNBR data demonstrated strong agreement with scorched and fire-consumed understory and canopy vegetation (Gale et al. 2022). The slower % increases of pixels of dNBR < −0.1 for the Orroral River catchment in the first year after the wildfire, compared with the Middle River catchment, may be attributed to slower regeneration of the dominant subalpine and woodland eucalypt-dominated forest types (Caccamo et al. 2015; Gibson & Hislop 2022). By comparison, the greater cover of grasslands in the Middle River catchment may have led to a more rapid increase of % pixels in dNBR < −0.1 in the following autumn to winter, since grasses quickly restore their greenness and photosynthetic capacity after fire and rainfalls (Lu et al. 2016; Fornacca et al. 2018). Nonetheless, the rapid increase in % pixels of dNBR < −0.1 in the Middle River catchment again declined to very low percentages in the following summer and early autumn (2020/2021).

Turbidity and TDS levels were measured over the study period, and data acquired are detailed in Supplementary Figure S5. Turbidity levels of the Middle River tended to be higher than those of the Orroral River but were very low in comparison to the high levels (up to 4,200 NTU) reported by Biswas et al. (2021) for the Jingellic site on the Murray River. Turbidity and river flow rate data acquired in this study indicate that turbidity levels became generally higher as river flow rates increased (Supplementary Figure S6). However, other factors such as the soil and vegetation types and land practices of the catchment where rainfalls occurred would have likely been influential. The TDS levels of Middle River waters were much higher than those of the Orroral River, but in both cases, levels were generally higher when river flow rates were lower (Supplementary Figure S7). This indicates that elevated evaporation rates may have contributed to some extent to the very high concentrations of DOM found in the river waters in the summer of 2020. Alternatively, elevated DOM and TDS may have been the result of the summer rainfalls and associated catchment inflows.

The reductions in DOC, A254 nm/m, and colour from the initial high levels recorded following the fires may have been due to catchment surface and subsurface water inflows to the rivers, and assuming these were lower in concentrations, then diluting these parameters in the river waters, along with their transport downstream. Reductions from the high levels of DOM recorded may also have been due to their degradation by microbial processes. Generally, higher catchment flows result in greater transport of particulate matter and DOM derived from catchment vegetation into waterways (Dalva & Moore 1991; Murshed et al. 2014; Ke et al. 2022) and increases in these parameters following the initial decline are consistent with that characteristic. In a modelling study analysing the water quality impacts of extreme events on an Australian reservoir, a higher risk of poor water quality was predicted for a ‘rain after fire’ compared with a fire-only scenario (Bertone et al. 2016, 2018), thus highlighting the impact of flow-induced DOM transport from the catchment. Sherson et al. (2015) studied nutrient dynamics in a stream following a wildfire (Las Conchas wildfire in June 2011 that burnt ∼36% of the contributing watershed) and during several precipitation events. They found significant increases occurred in dissolved NO₃-N (>6 times the background levels), dissolved phosphate (>100 times the background levels), specific conductance (>5 times the background levels), and turbidity (>100 times the background levels). These increases occurred along with substantial lowering in dissolved oxygen (<4 mg/L) and pH (<6.5). Their study highlighted the rapid rates of change in water quality during flow events following wildfires, lasting for a short duration.

As found in this and similar studies, water quality can change rapidly and dramatically following extreme climate events that impact catchments and consequently their streams and rivers, and subsequently water resources downstream. Further changes can occur over time as the catchments restabilize dependent on vegetation regrowth, topography, soils, and land-use practices. In this study, two rivers with contrasting catchments and management, one of native vegetation of a national park and the other of mixed land management practices (dryland agriculture, silviculture, and conservation) showed different water qualities following extensive wildfires in their catchments; the highest levels of DOM and turbidities were of the catchment with mixed management practices. With predictions of more frequent occurrences of extreme climate events arising from climate change, there are also increased risks to the quality of waters from catchments impacted by such events. Hence, water treatment capabilities that enable reliable responses to rapidly changing and contrasting raw water qualities are highly important to ensure consistent supply of high-quality drinking water. The coagulant dose prediction models described and applied here are designed to encompass water qualities from such extreme climate events. These models can be readily adjusted to account for specific raw water qualities and for meeting particular treated water quality targets. These will be further evaluated in future occurrences of extreme climate events that impact water resources.

We investigated the effects of wildfire on DOM in two contrasting river catchments: (1) Middle River (∼14,500 ha), which is mixed dryland agriculture and native vegetation and (2) Orroral Valley (∼9,000 ha), which is conserved as mostly intact native forest. Sentinal-2 dNBR time-series analysis of the Middle River and Orroral catchments following the 2019/2020 wildfires indicated that both catchments were burnt at high severity, although the fire severity in the Middle River catchment was more variable. After the wildfire, the Middle River catchment revegetated rapidly and subsequently showed seasonal variation due to the prominent presence of pasture grasses, while the recovery of forest vegetation in the Orroral River catchment was more gradual. Soon after the wildfires, DOC (mg/L), A254 nm/m, and colour (HU) were recorded at very high levels when river flow rates were very low. These declined to much lower levels by the end of autumn and the start of winter of the first year after the fires. However, these variations occurred at different rates at the two sites studied. The specific colour of the river waters was comparatively low soon after the wildfires, which may have been due to higher UV irradiance-induced cleavage of chromophoric components of DOM and higher relative abundances of organic compounds derived from algae and cyanobacteria than in the subsequently collected waters. A specific colour was also found to be related to river flow rates, while TDS concentration was found to be inversely related. The very high DOM concentrations in the river waters soon after the wildfires occurred when river flow rates were low likely contributed to by high evaporation rates caused by the wildfires and the subsequent hot summer conditions. Following catchment rainfalls and increases in river flows, DOM concentrations lowered at the sampling sites. Alum doses required in conventional treatment for drinking water supply were investigated using two prediction models that are based on different sets of input data (DOC and turbidity; UV@254 nm/cm, colour, and turbidity). The predicted doses are very high for these waters directly after the wildfires and decline to much lower levels as DOM levels become correspondingly less and when turbidity levels remain low. With the prediction of more frequent occurrences of extreme climate events resulting from climate change, such models might be of benefit in the management and treatment of source waters used for drinking water supply.

This research was supported by the Australian Government through the Australian Research Council (ARC LP160100217), SEQ Water, Melbourne Water, and Xylem Analytics Australia. The authors gratefully acknowledge Mrs Amanda Mussared, SA Water Corporation, for the provision of data on the Middle River water quality, Professor John Boland, UniSA, for advice on statistical analyses, and Professor Rodney Stewart, Griffiths University, for his advice and project support.

H.D.: Methodology, Software, Investigation, Writing – Original Draft. J.A.: Conceptualization, Methodology, Writing – Review and Editing, Supervision, Formal analysis. Adam Leavesley: Investigation, Resources, Writing – Review and Editing. M.A.: Investigation, Resources, Writing – Review and Editing. E.J.: Formal analysis, Writing – Review and Editing. M.G.: Formal analysis, Writing – Review and Editing. K.C.: Resources, Writing – Review and Editing. E.B.: Writing – Review and Editing, Supervision. J.L.: Conceptualization, Formal analysis, Writing – Review and Editing, Supervision, Project administration.

This research was financially supported by the Australian Government through the Australian Research Council under Grant Agreement No [ARC LP160100217].

This research did not involve human participants, their data, or biological material. For this research, no ethical approval or consent to participate is required.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.