Vegetation response to a natural gas pipeline rupture fire in Canada’s montane cordillera

Pipelines are critical for energy distribution, but incidents causing rupture fires are hazardous. While wildland fires are a natural disturbance, rupture fires are a potential risk and novel disturbance given the greater heat yield constants for fossil fuels, fuel volume, and flaming concentration and duration. We quantified vegetation response to a 2018 rupture fire case study in the montane cordillera of Canada. Plant species, functional groups, ground cover, and live vegetation height were sampled in 2018, 2019, 2020, and 2021 [0, 1, 2, and 3 years since fire (YSF)] in permanent plots stratified by burn severity and compared to the unburned reference plots sampled in 2019. Woody plant species and forb cover in burned plots recovered to levels similar to unburned plots. Litter and bare soil changes relative to YSF suggest trajectories to return to levels similar to unburned plots within 3 to 5 years post-rupture. Plant species richness, evenness, and diversity had also recovered to levels statistically similar to unburned comparisons by the final year of sampling in this study. Plots closest to the rupture epicenter that experienced ‘extreme’ burn had greater botanical dissimilarity from other burn severities or unburned comparisons. Vegetation structure showed significant (p < 0.0001) recovery with additional growth expected as the overstory re-establishes. The multiple metrics of ecological recovery on 3–5 year trajectories are comparable to published responses to wildland fire in the literature for this ecosystem’s response to fire. The recovery of conifers and soil microbiota should be assessed in the next decade.


Introduction
Energy development and distribution is critical for society globally with many ecological, legal, political, and social aspects that must be considered (Sovacool 2006;Strogen et al. 2016;Tong et al. 2016;Coleman 2018Coleman , 2019Hunsberger and Larsen 2021). Pipelines are one such means of energy distribution and are specifically used for the distribution of liquid and gaseous fossil fuel material (Balaji and Rabiei 2020). While pipeline transport has been regarded as a safer transport alternative compared to other modalities such as trains and trucks, the use of pipelines still has risks of unplanned failures and ruptures. These unplanned events can be attributed to numerous operational issues including external interference, incorrect operation, natural force damage, equipment failure, and corrosion/ cracking (Ossai et al. 2015;Bubbico 2018). Pipeline ruptures are concerning as they are point source releases of energy that can negatively affect environmental quality, life, and property (Hansen et al. 2006;Richards 2013; Bariha et al. 2016). The rupture and subsequent release of material is not the only concern however, as escaping fuel can ignite and the subsequent effects of the resulting sustained fire are also a potential hazard characteristic of pipelines (Jo and Ahn 2002;Alawode and Ogunleye 2011). A rupture can also have a negative effect on future property values due to fear of another rupture and possible cascading effects on other utilities in the same transmission corridor (Simons 1999;Foroughi et al. 2019). While much of the research on pipeline ruptures is concerned with urban settings and individual risk to life and property, there is a paucity of information about the effects on natural resources in the context of optimization (Mishra and Wehrstedt 2015;Ríos-Mercado and Borraz-Sánchez 2015;Bubbico 2018;Shin et al. 2018;Froeling et al. 2021).
The effects of fire, regardless of ignition source, on vegetation and ground cover is an indicator of environmental impact and is visually apparent to the public. However, fire is both a natural and anthropogenic disturbance, and has been suggested as one of the predominant agents of natural disturbance processes structuring the patterns of North American ecosystems (Rowe and Scotter 1973;Wright and Bailey 1982), including the montane cordillera ecozone of Canada which includes critical conservation resource values (Ireland and Petropoulos 2015;Wang et al. 2022). Consequently, differential intensity and spatiotemporal occurrence of wildfires are the cause of landscape and successional heterogeneity (Johnson et al. 1998) which is critical to maintaining mosaics of habitat types for a diverse suite of dependent wildlife species and ecological functions (Agee and Huff 1987;Burton et al. 2009;Leverkus et al. 2017). Moreover, Canadian First Nations historically used fire to manipulate vegetation and forest succession for a diversity of vegetation altering purposes (Gottesfeld 1994;Miller et al. 2010). While fire can facilitate different vegetation successional trajectories, such as in boreal ecosystems where fire can facilitate the early successional phase of rapid deciduous shrub growth (Viereck 1970;Rowe and Scotter 1973;Gillingham and Parker 2008), fire also creates opportunities for erosion (Mataix-Solera et al. 2011) and the invasion of undesirable and/or exotic plants (Fornwalt et al. 2010). Such negative consequences can reduce the abundance of native plants and alter plant community diversity, structure, and function (Harrod and Reichard 2000).
The effects of a pipeline rupture fire must be considered unique from wildland fire because the fire physics are fundamentally different than a wildland fire and subsequently facilitate a novel disturbance (Bragin and Molkov 2011;Twidwell et al. 2013;Zhang et al. 2018). In a wildland fire, a primary driver of fire effects on vegetation recovery is fire intensity. Fire intensity, for which fire temperature can be thought of as a proxy, is the heat energy release rate per unit of time and per unit of area. Fire intensity can be quantified as a simple function of heat yield constant for natural fuels, the amount of fuel consumed, and the rate of spread as quantified by Byram's equation for fire line intensity commonly used in wildland fires (Keeley 2009). Pipeline rupture fires are different than wildland fires because in such situations the fuel and fire variables would be considered as point-source heat rather than a wind drive flame front moving across the landscape, and high intensities are focused in a discrete area often referred to as the High Consequence Area (HCA) (Jo and Ahn 2002). Thus, an understanding of pipeline rupture fires in terms of fire intensity must incorporate the diameter, operating pressure, and amount of fuel remaining in the available length of the pipeline in relation to the size of the HCA to understand impact distance (Jo and Ahn 2002;Stephens et al. 2002;Shan et al. 2020). This simplified HCA concept is then suggested to be inversely and proportionally related to the length of the pipeline and the decay coefficient (Jo and Ahn 2002;Bariha et al. 2016). Finally, buried pipelines that rupture and ignite manifest uniquely as a jet fire in a pit (JFP) with unique flaming characteristics (Zhou et al. 2022) and variable effects depending upon soil type (Cleaver and Halford 2015).
While both fire types, wildland fire and pipeline rupture fires, have a time-dependency factor, these function differently based on movement of wildland fire through a landscape which is often referred to as 'residence time' for wildland fire versus the duration of escaping and flaming gas from a rupture. Another difference would be the heat yield constant for fossil fuels are higher than natural fuels. For example methane (CH 4 ) has a heat yield ranging from 50 to 55 MJ/kg that are 2 to 3 times greater than wood fuel or dry peat (* 21 MJ/kg and * 15 MJ/kg, respectively) (WNA 2022). Thus, a stationary pipeline rupture fire consuming fossil fuels would result in a much hotter, more intense, and more concentrated fire than a wildland fire consuming limited natural fuels moving across the landscape-a pyric event considered as a 'sustained jet fire' (Bragin and Molkov 2011). Stephens et al. (2002) conceptualized such a 'sustained jet fire' using the heat of combustion for a specific gas, combustion efficiency, emissivity, the number of point sources, effective gas release rate, and the radial distance from heat source to the potentially affected area of interest. Finally, the cumulative effects of such a fire on the land would also be contingent upon weather conditions and fuels and the relative ignitability of the surrounding vegetation (Weir and Scasta 2014).
Given that the intensity of a fire influences burn severity and vegetation recovery, but that pipeline rupture fires are a novel anthropogenic fire disturbance for which a paucity of research on post-rupture effects on vegetation is available, we sought to characterize the vegetation response to a recent pipeline rupture fire case study (Ireland and Petropoulos 2015). Specifically, we have monitored the vegetation within the HCA of a natural gas pipeline that ruptured and ignited in 2018 in the montane cordillera region of British Columbia, Canada. From a post-rupture environmental assessment perspective, we sought to understand plant biodiversity and plant functional group response as indicators of ecological function and environmental damage. Because first-order fire effects, specifically the heat-induced chemical processes such as plant injury and mortality, fuel consumption, smoke production, and soil heating, are direct and occur during or immediately after a fire; and these drive the second-order fire effects on plants, specifically plant reproduction, productivity, and composition and physical structure, over a longer time frames from days to months to decades, our results also have the ability to inform additional monitoring and reclamation.

Incident description
On October 9th 2018, a 36 inch (914.4 mm) natural gas pipeline ruptured due to stress corrosion cracks approximately 13 km northeast of Prince George, British Columbia, Canada ( Fig. 1) (TSBC 2018). This pipeline was completed in 1972 and hydrostatically tested for strength and leakage prior to approval by the National Energy Board as part of the routine integrity analysis and management program. The pipeline had an average depth cover in the general area of occurrence of 1.34 m and was transporting sweet natural gas within the maximum operating pressure of 6453 kPa. An estimated 140.72 million standard cubic feet (MMscf) of natural gas was released and combusted by the rupture (TSBC 2018).
It was not possible to directly attack and suppress the pipeline rupture fire; however the fire extinguished itself after the rupture once the pipeline was isolated by shutting valves at upstream and downstream compressor stations and vented (TSBC 2018). The ambient temperature was 0°C at the time of the rupture and the risk of wildfire spreading beyond the immediate area was low due to the wet ground. At the time the pipeline fire was confirmed to have extinguished it was noted that some spot fires and smoke continued. As a result of the rupture, a crater 4095 m 3 (35 m long, 13 m wide, and 9 m maximum depth) resulted. The escaping natural gas ignited, burning a total of 5.2 ha of land surrounding the rupture point consisting of provincial Crown land with no damage to private land noted (TSBC 2018). While the TSBC report notes that sampling in 2018 and 2019 indicated there were no lingering contamination sources or offsite risks in the soil or water, there is no mention of the effects on the vegetation.

Vegetation and ground cover sampling
In the fall of 2018, 2019, 2020, and 2021 vegetation and ground cover were assessed at permanent plots established at the rupture site in this case study . This study is a control-impact design where measurements prior to the unexpected event were not possible and assessment was in response to the event. This event occurred within the montane cordillera ecozone and more specifically within the sub-boreal spruce biogeoclimatic zone in which certain site associations are known to have a moderately developed shrub layer (Meidinger et al. 1991). Weather, site access, and access development rendered some plots unuseable. Plots span topoedaphic variation of slope and aspect and were stratified by disturbance severity (n = 9) with 3 classes including: Extreme (n = 2), High (n = 6), and Moderate (n = 1) (Fig. 2a). The unburned reference plots (n = 4) that did not experience active flaming and vegetative fuel combustion were identified immediately after the fire but were not sampled until the following year (2019) due to plant dessication due to cold temperatures and to ensure they were not affected by the radiative and convective heat from the fire. Technicians randomly threw a 1 m 2 frame within different disturbance zones (Kirkman 1981;Underwood 1998). Metrics assessed included the proportion of cover for the following 6 functional groups and ground cover classes: Woody plants (tree ? shrub species), Grasses, Forbs, Moss, Bare soil, and Litter [cumulatively including coarse woody debris (CWD), litter, and needles]. In addition, the height of the tallest living plant in each quadrat was also measured. In unburned plots, diameter breast height (dbh) was measured of the largest tree and used to extrapolate tree height following Larsen and Hann (1987) and Lynch et al. (2012).

Statistical analyses
For plant functional groups and ground cover data that was proportional, an arcsine transformation was applied prior to analyses. In three instances, the total woody plant cover exceeded 100% and was thus capped at 100% to meet the assumptions of the transformation. To compare the effect of years since fire, we used a mixed effects analysis of variance (ANOVA) model with the main fixed effect coded with as either unburned or burned, and for burned plots with the elapsed years since fire (specifically: Unburned, Burned*0YSF, Burned*1YSF, Burned*2YSF, or Burned*3YSF) and burn severity as a random effect. For the main effect we assessed least squares post hoc tests at a 0.1 for each pairwise comparison. We then used linear least squares regression to estimate recovery trajectories for litter and bare ground responses relative to years since fire using both linear and quadratic models. Then a Principal Components Analysis (PCA) was run to understand the main plant functional groups and ground cover classes of the plots at the rupture site using an oblique simplimax rotation method to optimally identify informative patterns (Kiers 1994). Eigenvalues and proportions of explained variance for each component were assessed and interpreted with the aid of a path diagram to understand how each variable worked in the model. For plant species data, we calculated species richness, evenness, and diversity (using the Shannon Wiener diversity index (H 0 )) and used the same mixed effects ANOVA model explained above for the analyses of plant functional groups. We then ran a constrained multivariate analyses at the plant species level using a redundancy analysis with years since fire (YSF), distance from the rupture, and severity as environmental variables. We calculated the proportion of variation explained and used a Monte-Carlo test to calculate p-values for the first axes and then all axes combined. Finally, we analyzed maximum vegetation height using a one-way ANOVA model with Burned/Unburned*YSF as the fixed effect at a 0.05.

Results
Woody cover the year of the burn was significantly lower than woody cover 1 YSF, 2 YSF, and 3 (all pairwise pvalues \ 0.05), but was not different than the unburned reference plots (p [ 0.05). Woody cover 1, 2, and 3 YSF was statistically similar to the unburned reference plots (all pairwise p-values [ 0.05) (Fig. 3a). The covariance parameter estimate for burn severity and residual were 0.1929 and 0.1277, respectively. Forb cover was marginally different (p-value = 0.0955; Fig. 3c) with forb cover 2 and 3 YSF significantly higher than forb cover the year of the burn but not different than 1 YSF or the unburned reference plots. The covariance parameter estimate for burn severity and residual were 0.0075 and 0.1205, respectively. Ground cover of litter, coarse woody debris, and needles (henceforth 'litter') the year of the burn and 1 YSF (p [ 0.05) were significantly lower than 2 YSF and 3 YSF (all pairwise p-values \ 0.05) (Fig. 3c). Litter cover in all burned plots were lower than the unburned reference plots with a p-value for the pairwise comparison between 3 YSF and the unburned reference plots of 0.0568 (Fig. 3c). The covariance parameter estimate for burn severity and residual were 0.0378 and 0.0631, respectively. Bare soil the year of the burn was significantly higher than 1 YSF, 2 YSF, 3 YSF, or the unburned reference plots (all pairwise p-values \ 0.05). Bare soil 1 YSF was similar to 2 YSF (p [ 0.05) but higher than 3 YSF or the unburned reference plots (p \ 0.05). Bare soil in all burned plots were lower than the unburned reference plots with a p-value for the pairwise comparison between 3 YSF and the unburned reference plots of 0.0501 (Fig. 3d). The covariance parameter estimate for burn severity and residual were 0.0124 and 0.0618, respectively. Graminoid cover, including grasses, sedges, and rushes, was not detected in any of the plots. Moss cover did not vary significantly in the study and was not different for any comparisons (all pairwise p-values [ 0.05). Moreover, moss cover was low in all stratified groups, with means always \ 1.5% (data not shown).
In the PCA path diagram for plant functional groups and ground cover, the first component (RC1) is explained by the positive response of litter and woody plant cover and the negative responses of bare soil while the second component (RC2) is explained by the positive response of moss and forb cover and the negative response of woody cover (Fig. 4). The first component is more informative (eigenvalue = 2.252, proportion of variance explained = 0.450) than the second (eigenvalue = 1.134, proportion of variance explained = 0.227).
When bare ground and litter cover are plotted relative to YSF, and assumed to be linearly related to time, the data suggests that bare ground will recover to levels similar to the unburned reference plots in 3 to 4 years and litter will recover in 4 to 5 years (Fig. 5). In this case, the linear assumption results in a more conservative projection for litter (i.e., longer time to recovery) and a liberal projection for bare ground (i.e., shorter time to recovery) whereas quadratic functions would suggest more rapid or slower responses, respectively (Fig. 5) (Mack et al. 2008). The statistical significance of linear fit trendlines (both p-values \ 0.05) compared to the statistical insignificance of quadratic fit trendlines (both p-values [ 0.05) suggest the ecological response is following a linear response up to this point in time (Fig. 5).
The dominant plant species noted in the plots reveal that this system is now dominated by shrubs and forbs with a minor component of trees (Table 1). However, it is important to note that trees and shrubs co-dominate the unburned reference plots. For shrub species, thimbleberry (Rubus parviflorus Nutt.) was a dominant species in the final year of sampling occurring in 78% of burned plots with a mean cover of 58% across all the burned plots and a mean cover of 75% in the burned plots in which it occurred. Fireweed (Chamerion angustifolium (L.) Holub) and Populus species were the next two most frequent and abundant species occurring in 22% and 33% of burned plots, respectively, with a mean cover of 14% and 2% across all the burned plots with a range of cover values (data not shown).
Species richness was significantly lower the year of the burn and the first year following the burn when compared to the unburned reference plots. However, by two years Fig. 2 a Post-rupture aerial image of the West Coast gas pipeline rupture site and sampling plots with colors based on severity (red = Extreme, orange = High, yellow = Moderate, green = Unburned; Photo credit to D. Tisseur * 4:30 pm October 10, 2018). b and c Postrupture image of the West Coast gas pipeline rupture site and felled trees (Photo credit to S.E.R. Leverkus October 20, 2018). Note the footprint of the rupture fire and construction to enhance the access road and repair the 4095 m 3 crater after the incident which resulted in the loss of some study plots during the course of the assessment after the burn, species richness had recovered to levels similar to the unburned reference plots (Fig. 6a). Species evenness and diversity followed similar patterns as species richness, with diversity recovering to levels similar to the unburned reference plots by the second year ( Fig. 6b and  c).
Multivariate analyses revealed the role of burn severity in explaining species composition and dissimilarity (Fig. 7a). The first axis explained 52.95% of the fit variation significantly (p = 0.001) and displays the separation of extreme burn severity plots to the right of the biplot ( Fig. 7a and b). The second axis explained an additional 21.64% of the fit variation and was explained by time since fire, distance from the rupture epicenter, and differences between unburned, moderate, and burn severity plots. All axes combined were also significant (p = 0.001).
The general recovery through time, as indicated by the shrub and forb response was demonstrated in the time series photos in Fig. 8a-c for three plots. The response of thimbleberry is shown in Fig. 8a and the response of thimbleberry, fireweed, and aspen are shown in Fig. 8b and c. However, the impact of extreme burn severity, and the  stronger persistence of bare ground is apparent in plots with greater burn severity and closer to the rupture epicenter ( Fig. 9a-d).
Maximum vegetation height of live plants also showed a significant and positive trend through the years after the incident (p \ 0.0001; Fig. 10). Immediately after the incident there was no remaining aboveground vegetation structure within the highest disturbance zone. One year after the incident the maximum vegetation height was 60.5 cm (± 7.4), two years after the incident the maximum vegetation height was 134.1 cm (± 16.0), and 3 years after the incident the maximum vegetation height was 171.9 cm (± 23). The unburned reference plots however continued to be significantly taller with maximum vegetation height of the overstory estimated to be 2475.0 cm (± 2.5).

Discussion
In this pipeline rupture case study in a montane cordillera ecological site, our results suggest that the plant functional groups and ground cover classes we measured have recovered, or are on a trajectory to recover, to levels observed in the unburned reference plots. While there has been a shift in composition, particularly to a shrub and forb dominated plant community, this is in line with results from repeated severe wildfires over time in similar study areas (Kelsall et al. 1977). Moreover, the montane cordillera ecozone is considered to be fire prone based on dendrochronological studies indicating annual probabilities of fire ranging from 2 to 5% annually producing 20 to 60 year fire return intervals which included moderate-high fire severity synchrony (Chavardès et al. 2021). Specifically for ground cover classes and based on the temporal sampling scheme and the values measured at the 3 YSF sampling interval, litter and bare ground are expected to recover within 3 to 5 years since the rupture fire. From an ecological perspective, the recovery of litter (including herbaceous, coarse woody debris, and needles) provides a critical ecological function. Specifically, the litter layer in this system and many North American terrestrial systems, represents a crucial pathway for nutrient return to the soil (Krishna and Mohan 2017). In the burned plots, regardless of YSF, litter had not recovered to levels observed in unburned plots but was on a trajectory to do so within a 3 to 5 year timeframe. Relatedly, bare ground displayed a somewhat similar trajectory with projected recovery within a similar timeframe. Such an initial increase in litter, and decrease in bare ground, may be attributed to reduced litter decomposition rates, which increase significantly with time since fire, as driven by the longer time for soil microbiological recovery (Holden et al. 2013).
Woody plant species cover also responded positively in burned plots as shrub cover at 1, 2, and 3 YSF were similar to the unburned reference plots statistically, with mean values exceeding 40% cover. This woody plant response, particularly for resprouting shrubby species, is similar to results shown in boreal ecosystems generally where fire is Fig. 5 Litter and bare ground cover classes plotted relative to an elapsing years since fire (YSF) time horizon with both linear (black lines) and quadratic functions (gray lines) displayed. Error bars represent 1 standard error known to facilitate the early successional phase and rapid deciduous shrub growth (Viereck 1970;Rowe and Scotter 1973;Gillingham and Parker 2008). This post-fire shrub response is likely associated with alterations to the forest tree canopy and greater availability of resources such as light (Hart and Chen 2008;Liu et al. 2020) and nitrogen (Driscoll et al. 1999). Positive responses of woody plants is also attributed to below-ground meristematic tissue (Mack et al. 2008). Shrubs and trees dominating the burned plots include thimbleberry (Rubus parviflorus Nutt.) and Populus species all of which are documented to positively respond to fire (Wan et al. 2014;Ortiz et al. 2019). Woody plants, particularly shrubs, are important for habitat including concealment and browse for large ungulates.
Thimbleberry had the greatest frequency and cover of any species in the burned plots in 3 YSF. Thimbleberry is a native erect rhizomatous perennial shrub that can grow to 0.5 to 2 m in height, has a wide distribution across North America, and has cultural significance for food and medicinal purposes (Tuner 1997). Thimbleberry is a clonal shrub that can reproduce sexually by seed and asexually by basal bud propagules (Maxwell 1990). Plants establishing from seed are considered genets but plants establishing from vegetative resprouts are often called ramets. It is these ramets that arise from the same parent clone but potentially display morphological independence that enable such shrubs to vigorously resprout after disturbance . From a functional morphology perspective, of the common Rubus species, thimbleberry has the highest leaf-area ratio and cane tissue density but the lowest specific cane length (Caplan and Yeakley 2013). Haeussler and Coates (1986) state that thimbleberry is common in ''burned-over forest land '' in British Columbia. Francis (2004) goes on to state that thimbleberry is fire-resistant, benefits from all kinds of disturbances, and may dominate understories of logged areas for up to 5 years-dominance that is important as protective cover for regenerating conifer species. Recovery of conifers is of concern in this ecosystem and the response and function of thimbleberry is encouraging. Other authors have also concluded that severe fires in spruce dominated forest systems induce a strong positive response of other woody species such as aspen and birch (Betula species L.) to repeated severe wildfires fires (Kelsall et al. 1977). The post-fire forb response in this study included fireweed which is a recognized fire-adapted plant species that positively responds to fire as indicted by its common name. Fireweed also has Indigenous descriptions as a ''healer of burns'' and for medicinal values (Rogers 2014). Increases in response to fire have been shown for fireweed in the montane cordillera ecozone (Sachro et al. 2005;Sittler et al. 2019) and beyond (Bartos and Mueggler 1981;Turner et al. 1997). Density of fireweed has also been positively correlated with burn severity with sprout densities being the greatest in severe surface and crown fire areas (Turner et al. 1997)-responses which have relevance for the intense and severe pipeline rupture fire in this study.
The response of vegetation height is a measure of structure which in turn relates to habitat for diverse wildlife species. Heterogeneity, the critical root of biological diversity (Tamme et al. 2010), is achieved across a site when there is variable vertical structure and species  diversity, but can be difficult to robustly encapsulate in a control-impact design study such as this. Additional limitations to this study include a short time span, limited spatial range, and a constrained range of potential risk factors. However, our results suggest that while the structure of the disturbance has not achieved pre-disturbance heights (i.e., the height of the tree canopy in the surrounding area), forbs and woody shrubs are growing rapidly and gaining height each year. Vegetative structure across landscapes is important in order to support multiple species (Leverkus et al. 2017). While a small scale in comparison with large landscapes, this site is behaving or trending in a way that supports multiple species. We observe this through the tracks, scat, and presence of wildlife on site through ocular observation over 4 years. Following disturbance from the rupture of the pipeline and the wildfire event, there has been rapid recovery of habitat structure on site. Wildlife present on this section of the pipeline was noted during all field sampling from 2018 to 2021 and included visual observations of multiple bird species (hawks, songbirds, and woodpeckers), paw and hoof tracks (canids, deer, elk, and moose), and scat (bears, canids, deer, elk, rabbits, and multiple ungulate species)all indicators of the regaining of ecological function postdisturbance. Flowering plants and berries on the location are browsed, grazed, and fed up on by these species and  The positive post-fire responses of litter, shrubs, some trees, and forbs is an indication of recovery and is supported by other studies of wildfire that indicate plant biomass and productivity over a 116 year wildfire chronosequence can achieve ''recovery'' in as little as 4 years (Mack et al. 2008). However, the aboveground vegetation response is also related to the belowground soil response. While such features were not part of this particular project, it is important to note that soil features such as aggregate stability and microbiology are known to be sensitive to fire temperatures and can take longer to recover. For example, soil microorganisms have specific lethal temperatures starting as low as 50°C with the direct effect of soil heating a microbial biomass reduction (Mataix-Solera et al. 2011). Recovery of some of these features can take decades as for example fungal hyphal length took up to 24 years to recover with implications for litter decomposition and soil respiration (Holden et al. 2013). Other features such as soil organic matter and aggregate stability can demonstrate more variable responses to fire and heating, particularly as it relates to fire severity, soil texture, and principle binding agents (Mataix-Solera et al. 2011).
Many questions remain unanswered related to several potential hypothetical scenarios under which this rupture could have occurred. For example, what would the extent, intensity, and severity have been with different amounts of debris, droughts, successional stages, maintenance of pipeline roads, etc.?. However, it has been documented that fire behavior is driven by many complex and interacting variables with the primary factors being weather, fuels, and topography (Airey-Lauvaux et al. 2022). Fuel type and moisture content are factors that influence flammability and fire spread once ignition occurs from either natural or anthropogenic causes (Weir and Scasta 2014). When fuels have a higher moisture content, such as following precipitation, they are less flammable. In addition, if the surrounding vegetation is heterogeneous, there may be reduced fire spread and fire intensity with increased ability for suppression (Kerby et al. 2007;Beverly et al. 2020). On the contrary, higher wind speeds with lower relative humidity would increase fire spread and intensity particularly in a homogenous fuel type which has implications for surrounding fuel treatments (Brochez and Leverkus 2022). Moreover, functional group dominance of the vegetation can also be facilitated by the timing of when a fire occurs (Weir and Scasta 2017). Thus, the context of this event is critical as when the rupture occurred in October 2018, the fire environment had shifted into a wetter scenario at the end of the fire season due to recent precipitation with cooler overnight temperatures.

Conclusions
While wildland fires are a naturally occurring hazard that occur broadly across landscapes and to which ecosystems may be adapted to, pipeline rupture fires are a novel anthropogenic pyric disturbance. Pipeline fires have been called a ''sustained jet fire'' due to the greater heat yield constants for fossil fuels compared to natural fuels, relative fuel volume, and concentration and duration of flaming and effects may be limited to relatively short distances from the rupture (Foroughi et al. 2021). With the forecasted demand and movement through continental-scale and global supply chains for natural gas through the next decade (Jeglic 2004;Tong et al. 2016;Vetter et al. 2019;Kan et al. 2020;Lu et al. 2020;Hunsberger and Larsen 2021), a robust understanding of the ecological responses of such failures are necessary. In our assessment of the vegetation response to a pipeline rupture fire in the montane cordillera, we found ecological recovery as indicated by woody plant and forb functional groups and ground cover trajectories of 3 to Fig. 10 Recovery of vegetation structure as indicated by the tallest plant in plots. Error bars represent 1 standard error and different letters indicate significant pairwise differences at the 95% confidence level 5 years for recovery forecasted. These responses are supported by the wildland fire and ecology literature that notes the adaptations of this ecosystem to fire (Chavardès et al. 2021). It is important to note that the spatial extent of such a fire is also contingent upon weather conditions and fuels and the relative ignitability of the surrounding vegetation, which in this case was more limiting than in a dry summer scenario. Regardless, the further recovery of the conifer trees and soil microbiota should be assessed in the next decade for more insight. Practically, effective environmental assessment in industrial environments requires active dialog and adaptive management to ensure monitoring sites are not compromised due to construction activities and minimizing additional disturbance. This underscores the importance of collaborative dialog during an emergency response in an industrial setting and reinforces the need for cooperation between the operations group and their environmental support team to ensure that both baseline and trend data can be gathered and environmental considerations can be integrated into the construction process. While the traditional mindset of doing whatever needs to be done to repair the site has been the standard operating protocol, we provided a unique lens through this work to highlight the importance of minimizing anthropogenic disturbance and conserving the ecological features in close proximity to a novel and rare occurrence. By identifying and remaining within an operationally confined footprint for facilitating repairs to this pipeline, environmental considerations for long-term monitoring of site recovery and conserving ecosystem functions were integrated into the planning for the repair work and provide critical ecological insights about this incident.
Funding JDS, SL, and GL conducted this work as paid consultants to Enbridge. DT is an employee of Enbridge which sponsored this work.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons. org/licenses/by/4.0/.