Forested catchments are critical for water supply globally and provide ~60% of the water supplies for the world’s 100 largest cities and 2/3 of all water supplies, including drinking water for ~180 million people in the U.S. In Alberta, Canada, approximately 2/3 of the population’s drinking water comes from the eastern slopes of the Rocky Mountains. Ironically, the high quality and quantity water from these forested regions makes these supplies particularly vulnerable to the deleterious impacts of climate change and associated landscape disturbances. Wildfire has the potential to be the most catastrophic of these disturbances. It can produce significant changes in the quantity, timing, and quality of water originating in these settings. Notably, it also may necessitate significant increases in costly drinking water treatment infrastructure, operations and maintenance.

Aquatic natural organic matter (NOM) is typically evaluated by measurement of dissolved organic carbon (DOC) and is one key water quality parameter that drives the design of drinking water treatment infrastructure. Changes in the amount and quality of DOC can increase the need for and cost of water treatment infrastructure because of increased chemical coagulant dosing requirements and the potential for formation of several currently regulated disinfection by-products (DBPs), such as trihalomethanes (THMs) and haloacetic acids (HAAs). They can also result in increased membrane fouling and microbial regrowth in the distribution system. While many proxy indicators (DOC, UV₂₅₄, specific UV absorbance [SUVA], fluorescence index [FI], fluorescence excitation-emission matrices [FEEMs], other NOM fractions, etc.) have been suggested for inferring drinking water treatability implications of changes in NOM, clear guidance regarding the most informative proxy indicators and the reliability of their connectivity to drinking water treatability assessment is still lacking.

The overall goal of this research was to compare and improve upon available strategies for characterizing challenges and threats to drinking water treatability arising from wildfire and forest harvesting disturbance-associated changes in DOC. Potential increases in regulated DBP formation potential (i.e., DBP-FP) were focused upon because infrastructure and operations implications;​ relative potential implications of these disturbances to membrane fouling and microbial regrowth in distributions systems were also evaluated. Of course, other impacts such as those on coagulant demand are equally important, though more site- or treatment configuration-specific. While it is generally believed that coagulant demand during drinking water treatment may increase after severe disturbance as a result of elevated and increasingly variable turbidity and/or changes in source water DOC, the implications of wildfire to membrane fouling and microbial regrowth potential in distributions systems have not been reported to date. Similarly, forest harvesting impacts on DBP-FPs have not been reported and elevated DBP-FPs resulting from wildfire have been suggested, but only recently demonstrated at the watershed-scale with consideration of hydro-climatic variability. Despite these critically foundational, but nascent linkages, clear guidance regarding optimal approaches for informing drinking water treatability in response to landscape disturbance-associated changes in source water quality is currently unavailable. Thus, to advance the broad goal of informing strategies for characterizing challenges and threats to drinking water treatability arising from potential wildfire- and forest harvesting-associated changes in NOM, five phases of research were conducted.

In Phase 1, the most common methods of NOM characterization and their relationship to drinking water treatability (including limitations) were reviewed, particularly as related to the formation of regulated carbonaceous DBPs. These methods include DOC, UV₂₅₄, and SUVA metrics, as well as resin fractionation, liquid chromatography-organic carbon detection (LC-OCD), fluorescence excitation-emission matrices, and other techniques. The review demonstrated that no universal proxy indicators for NOM reactivity with oxidants such as chlorine have been identified to date, thereby underscoring the need to advance approaches for evaluating NOM reactivity in a manner that links different source watershed settings and disturbance impacts to treatability challenges.

In Phase 2, a comprehensive DOC characterization investigation was conducted throughout the treatment process at a conventional water treatment plant (WTP) with aerobic biofiltration. This work is among the first studies in which NOM removal during conventional treatment and biofiltration has been evaluated concurrently using several metrics of NOM concentration and character—this enabled direct confirmation of which of these parameters might be the most useful as proxy indicators for drinking water treatability when characterizing changes in source water quality. Samples were collected from the WTP intake and at different treatment stages (post-sedimentation, post-ozonation, and GAC biofilter effluents) at the Mannheim WTP, in Kitchener, Ontario. As would be expected, the coagulation/flocculation/sedimentation process (after which post-clarification samples were obtained) efficiently removed aromatic compounds (UV₂₅₄, hydrophobic organic carbon as measured by resin fractionation [HPO %], and the humic substances [HS] fraction as measured by liquid chromatography with organic carbon detection [LC-OCD]) and THM- and HAA-FPs. Further removal of these compounds was observed during biofiltration, highlighting that aromatic compounds (removed by chemical pre-treatment) were the main contributors to the THMs, though some smaller DOC fractions (removed by biofiltration with GAC) also played a role in the formation of THMs. Changes in post-treatment THM- and HAA-FP were generally comparable—this was expected given that they share common precursors. Higher molecular weight fractions contributed more to the formation of HAAs than THMs. Overall, metrics indicative of aromatic compounds were shown to be good proxy indicators of DOC reactivity and formation of regulated DBPs. These quantitative results were consistent with the qualitative results obtained using fluorescence excitation-emission matrices [FEEMs]. Utilization of LC-OCD had the additional advantage of detecting changes in medium to low molecular weight (LMW) fractions of DOC (e.g. building blocks and LMW neutrals) throughout treatment.

In Phase 3, changes in DOC concentration and character, and their relationships to regulated DBP-FPs (THM-FPs and HAA-FPs), were comprehensively characterized using multiple NOM characterization techniques during a two-year period following severe wildfire in the eastern slopes of the Rocky Mountains in south-western Alberta. Several NOM fractions also were characterized by LC-OCD during the latter of those years. This work was conducted as part of an ongoing (>​9 years, at the time) watershed-scale study of wildfire and post-fire salvage logging impacts on hydrology, water quality, and aquatic ecology (i.e., the Southern Rockies Watershed Project). In that work, samples collected from multiple unburned (reference), burned, and post-fire salvage logged watersheds during dominant regional streamflow regimes (baseflow, snowmelt freshet, and stormflow) demonstrated that DOC concentration and hydrophobicity increased after wildfire and even more so after post-fire salvage-logging, especially during high discharge events in headwater streams. These changes in aquatic DOC in streams draining disturbed watersheds were concurrent with increases in THM- and HAA-FPs. Contributing to, and building on that investigation, the work presented herein is the first to report that the mass of HS, biopolymers, and building blocks fractions of DOC also increased significantly in streams draining wildfire and post-fire salvage logged watersheds, thereby suggesting that these disturbances may have significant implications for carbonaceous DBP-FP, coagulant demand, and membrane fouling. In contrast, the mass of the LMW neutrals fraction of DOC, which contributes to microbial regrowth in the distribution system, was not significantly different in streams impacted by either wildfire or post-fire salvage logging. This work was also the first to comprehensively demonstrate wildfire-associated changes in DOC character (by measuring HPO %, UV₂₅₄, SUVA, FI, and FEEMs) and related DBP-FPs, at the watershed-scale and over multiple flow regimes. The disturbance impacts indicated by all of these quantitative, DOC-associated metrics were all statistically significant, except for FI. Qualitative FEEM results were consistent with these significant shifts. Notably, despite the continued development and promotion of various proxy indicators, UV₂₅₄ offered the most precise linear correlation with THM-FP, with a coefficient of determination (R²) of 0.6 (in contrast to values of 0.47, 0.42, and 0.39 for DOC, SUVA, and HPO %). Thus, changes in the proxy indicators were related to changes in THM-FP;​ however, they could not adequately explain response variability, thereby demonstrating the need to 1) better understand relationships between disturbance-associated changes in DOC and their implications to DOC reactivity and 2) advance modeling approaches for describing these relationships. While the mass of various DOC fractions obtained using LC-OCD and HAA-FPs was not analyzed in this manner because of the limited size of the data sets, similar relationships were suggested. Overall, these data suggest that severe wildfire may lead to significant DOC-associated drinking water treatability challenges and that post-fire salvage logging may further exacerbate them—notably, UV₂₅₄ is unequivocally the best available tool for monitoring these potential impacts at present.

THM-FP is generally understood to be linearly correlated with aromatic NOM as measured by UV₂₅₄ and/or SUVA. In Phase 4, simple strategies for enhancing the prediction of THM-FPs using NOM-associated proxy indicators were investigated. Specifically, the relationship between NOM aromaticity (HPO %, HS, UV₂₅₄, and SUVA) and THM-FP was examined. Then, HPO and HS were re-analyzed after weighting by mass (DOC concentration)—this appreciably enhanced their prediction performance. This improvement was especially evident for HS, for which the coefficients of determination (R²) increased from 0.10 and 0.26, to 0.85 and 0.88 (Phase 2 and 3 data, respectively). Thus, data processing and reporting are critical to anticipating NOM reactivity;​ absolute quantities have superior prediction performance. Notably, regardless of these improvements, the relationships between DBP-FP and NOM proxy indicators can be quite variable spatially and temporally, and frequently site specific. More work is required to link source water quality to DBP-FP and drinking water treatability more broadly.

In Phase 5, changes in DOC concentration and character and their relationships to regulated DBP-FPs were comprehensively characterized using multiple NOM characterization techniques in the two years during and immediately after forest harvesting in the eastern slopes of the Rocky Mountains in south-western Alberta. Several NOM fractions also were characterized by LC-OCD to inform the relative potential for membrane fouling and microbial regrowth in distribution systems. Like Phase 3, this work was conducted as part of the ongoing SRWP in which two watersheds that served as unburned-reference watersheds in Phase 3 were studied. They were fully calibrated for climate, streamflow, and water quality for 11 years [2004-2014]). Three sub-watersheds within one watershed were harvested using clear-cut with patch retention, strip-shelterwood cut, and partial cut. All possible best management practices (BMPs) were followed to minimize disturbance impacts on water quality. Samples were collected during the dominant regional streamflow regimes. Notably, no substantial impacts of forest harvesting on water quality and treatability were observed during the harvest and first post-harvest years. Thus, this work suggests that forest harvesting with careful implementation of BMPs for erosion control may mitigate the potentially catastrophic impacts of wildfire on drinking water treatability without significantly compromising it.